Content uploaded by Alisa Young
Author content
All content in this area was uploaded by Alisa Young on Jan 07, 2016
Content may be subject to copyright.
The Characterization of Deep Convection in the
Tropical Tropopause Layer Using Active and
Passive Satellite Observations
A Thesis
Presented to
The Academic Faculty
By
Alisa H. Young
In Partial Fulfillment of the
Requirements for the
Degree of Doctor of Philosophy in the
School of Earth and Atmospheric Sciences
School of Earth and Atmospheric Sciences
Georgia Institute of Technology
August 2011
The Characterization of Deep Convection in the Tropical Tropopause Layer Using Active
and Passive Satellite Observations
Approved by:
Dr. Judith A. Curry, Advisor
School of Earth and Atmospheric Science
Georgia Institute of Technology
Dr. Robert X. Black
School of Earth and Atmospheric
Science
Georgia Institute of Technology
Dr. Irina N. Sokolik
School of Earth and Atmospheric Science
Georgia Institute of Technology
Dr. John J. Bates
Remote Sensing and Applications
Division
NOAA/NESDIS/NCDC
Dr. Peter J. Webster
School of Earth and Atmospheric Science
Georgia Institute of Technology
Date Approved: June 23, 2011
iii
ACKNOWLEDGEMENTS
I am deeply grateful for all the time, energy, and guidance of my advisors Dr. Judith A.
Curry and Dr. John J. Bates. While the development of this thesis has at times been very
challenging, this body of work is not only the result of my steadfastness but it is also the result of
your own persistence and diligence in carefully guiding me to become a better scientist and
researcher.
I would also like to thank Dr. Irina N. Sokolik, who has also given much of her time and
expertise to me through classroom instruction and memorable group meetings. For guidance
especially in the earlier stages of my thesis work, I would like to thank Zhengzhao (Johnny) Lou,
who provided valuable insight on CloudSat measurements. I also acknowledge Karen Rosenlof,
who graciously provided observations that were needed for several of the analysis presented in
Chapter 5. To Priti Brahma and Yassin Jeilani, thank you both for helping me to work through
many of the professional challenges I have faced. Your wisdom and encouragement has helped
me to better understand and also conquer difficult barriers. I am also indebted to Lei Shi and Ken
Knapp, who were very helpful in the early phases of my thesis work.
Finally, I would like to thank my heavenly father, for the favor that He has given me. I
would also like to thank my beloved family, Alphanso Young, Alphanso Young Jr., and Reagan
M. Young for their love and support while I have worked to complete my doctoral degree. In
addition, I would also like to thank my father and mother, John and Desiree Holley and siblings,
David, Tamar, and Dannon for their continued guidance, love, and support.
iv
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ........................................................................................................ iii
LIST OF TABLES ....................................................................................................................... vi
LIST OF FIGURES ................................................................................................................... viii
LIST OF ABBREVIATIONS .................................................................................................... xii
LIST OF SYMBOLS ................................................................................................................. xiv
SUMMARY ..................................................................................................................................xv
1. INTRODUCTION......................................................................................................................1
1.1 Introduction and Review ........................................................................................................1
1.2 Water Vapor in the Stratosphere ............................................................................................3
1.3 Climate Impacts of Lower Stratospheric Water Vapor Variability ........................................4
1.4 Deep Convection in the TTL ..................................................................................................7
1.5 Exploring Hypothesis ...........................................................................................................12
1.5.1 Overview of CloudSat, Calipso, and Aqua MODIS Data Products .........................12
1.5.2 Characterization of Penetrating and Overshooting Deep Convection ......................13
1.5.3 Evaluation of Traditional IR Techniques .................................................................13
1.5.4 Climatology of Penetrating and Overshooting Deep Convection .............................14
2. OVERVIEW OF CLOUDSAT, CALIPSO, AND AQUA MODIS DATA PRODUCTS ..17
2.1 Introduction ..........................................................................................................................17
2.2 Comparison of TRMM and CloudSat Radar Characteristics ..............................................17
2.3 Products from CloudSat Cloud Profiling Radar (CPR) ......................................................19
2.3.1 CloudSat 2B-GEOPROF product ............................................................................20
2.3.2 CloudSat 2B-CLDCLASS product ..........................................................................20
2.3.3 CloudSat 2B-Lidar-GEOPROF product ..................................................................22
2.4 Products from Aqua MODIS ...............................................................................................25
2.4.1 MODIS-AUX product .............................................................................................26
2.4.2 MODIS L2 Cloud Product (MAC06S0) ...................................................................27
2.4 Summary and Discussion .....................................................................................................28
3. CHARACTERIZATION OF PENETRATING DEEP CONVECTION ..........................36
3.1 Penetrating Deep Convection from Previous CloudSat and TRMM Studies .....................36
3.2 Application of CloudSat, Calipso, and Aqua MODIS data products ..................................40
3.3 Results from CloudSat Observations of Penetrating Deep Convection ..............................42
3.3.1 Evidence of Vertical Extent ......................................................................................42
v
3.3.2 Geographical and Seasonal Distribution ...................................................................44
3.3.3 Passive Sensor Cloud Top Brightness Temperature and BTD Signature ................46
3.3.4 Areal Size Distribution .............................................................................................47
3.4 Summary and Discussion ..................................................................................................48
4. EVALUATION OF TRADITIONAL IR TECHNIQUES ...................................................62
4.1 Introduction .......................................................................................................................62
4.2 Characterization of Penetrating Deep Convection from Spaceborne IR Data ..................67
4.3 Data and Methods ..............................................................................................................70
4.4 Observations from Aqua MODIS .....................................................................................72
4.4.1 Penetrating Deep Convective Clouds Compared with Other High Clouds ..............72
4.4.2 Characterization of Cold Cloud Features and Positive BTD Signatures ..................76
4.4.3 Cold Cloud Features and +BTD Signatures Compared with CloudSat Observations..
............................................................................................................................................77
4.5 Summary and Discussion ..................................................................................................80
5. CLIMATOLOGY OF PENETRATING DEEP CONVECTION .......................................96
5.1 Introduction .......................................................................................................................96
5.2 Data and Methods ..............................................................................................................99
5.3 Results .............................................................................................................................102
5.3.1 Evaluation of Climatological Data..........................................................................102
5.3.2 Penetrating Deep Convection and Lower Stratospheric Water Vapor ...................106
5.4 Summary and Discussion ................................................................................................109
6. CONCLUSIONS ....................................................................................................................124
APPENDIX A .............................................................................................................................133
APPENDIX B .............................................................................................................................137
REFERENCES ...........................................................................................................................141
vi
LIST OF TABLES
2.1 The radar characteristics of CloudSat and TRMM show differences in geographic
coverage, temporal resolution, scan characteristics, radar characteristics, and list
auxiliary instruments in or associated with the payload ....................................................29
2.2 (cf., Wang et al., 2007) Characteristic cloud features for the major cloud types
derived from numerous (midlatitude) studies ...................................................................30
2.3 (cf., Wang et al., 2007) Cloud ID rules based approximately on the properties for
the 98th percentile of data for each cloud type that was sampled during first 6
months while CloudSat was in orbit .................................................................................31
2.4 Approximate vertical and horizontal resolutions of the CloudSat CPR and the
Calipso Lidar ......................................................................................................................32
3.1 Multi-satellite mean statistics for observations of deep convection reaching and
penetrating 14 km and 16.9 km over the regions between 20°N─20°S and
35°N─35°S. The normalized frequency distributions corresponding to these
observations are provided in Figures 3.3a and 3.3b...........................................................54
3.2 Percent of occurrence of deep convection reaching 14 km and 16.9 km for cloud
brightness temperatures at varying brightness temperature thresholds and positive
BTD from 20°N─20°S and 35°N─35°S ............................................................................54
3.3 Framework for areal extent of penetrating tops provided in Figure 3.6 ............................55
3.4 Ratio of Plume Diameter as a function of varying chord lengths ......................................55
4.1 Relevant information for nine of the most popular studies on penetrating deep
convection. These studies consist of passive and active space borne remote sensing .......84
4.2 Mean optical and microphysical properties for various types of high clouds derived
from MODIS L2 Cloud Product along the CloudSat orbital track (MAC06S0) for
January 2007 ......................................................................................................................85
4.3 Percent of occurrence of deep convection reaching 14 km and 16.9 km with cold
cloud features from 20°N─20°S and 35°N─35°S ............................................................85
4.4 Properties of cold cloud feature distributions and penetrating deep convection
reaching 14 km and cold cloud features from 20°N─20°S and 35°N─35°S .....................86
vii
4.5. Percent of BT11 210 K pixels provided according to cloud type and cloud optical
thickness for different latitudinal bands where data was derived from the Aqua
MODIS Level 2 Cloud Product for October 2007 .............................................................87
4.6 Frequency distribution between 35°N─35°S of a) +BTD signatures (with BT11 < 235 K)
and seasonal patterns of +BTD for b) December-February (DJF) and c) June-August
(JJA).
5.1 Linear regression statistics for tropical and regional standardized frequency
anomalies in monthly IRWBT 210 K where the slope/trend is equivalent to the
correlation coefficient, p-value and t-stat are both standards for identifying the
significance of the tropical and regional relationships and are compared to =0.05
and a critical t-value of 1.962 ........................................................................................114
5.2 Time series regression statistics for standardized frequency anomalies in monthly
IRWBT 210 K with number of years (
ˆ
n *
) of monthly data needed to detect
the trend provided for each region at the 95% confidence level as a function of the
autocorrelation(
ˆ
N
) and standard deviation (
ˆ N
) of the noise
(Weatherhead et al., 1998) ..............................................................................................114
5.3 Linear regression statistics for anomalies in monthly frequency of IRWBT 210 K
and monthly anomalies in 82 mb water vapor mixing ratio at lags of 0 to 6 months
for the tropics, western Pacific, central Pacific and Indian Oceans. Again, the
p-value and t-stat is respectively compared to =0.05 and a critical
t-value of 1.962 ..............................................................................................................115
viii
LIST OF FIGURES
1.1. Schematic diagram adapted from of upward mass flux of tropical overshooting
deep convection penetrating from the convective boundary layer (CBL) into the
tropical tropopause layer (TTL), cold point tropopause (CPT) and on into
the lower stratosphere. The diagram also of shows the anvil the deep convective
cloud identified within the deep convection outflow layer located between
11 km and 15 km as well as a shallow outflow layer where shallow convective
clouds are primarily detrained. At ~15 km, the level of zero net radiative
heating is distinguished as Qclear=0. As indicated by the one directional
arrows above and below this level, there is net downward motion associated
with clear sky radiative cooling and subsidence just below this level and net
radiative heating and upward vertical motion just above this level ...................................16
2.1 (cf., Li and Schumaker, 2010) Images of a coincident CloudSat and TRMM
overpass showing a) TRMM PR horizontal cross sections at 2 and 7.5 km for
orbit 55469 with CloudSat track in magenta and vertical cross sections of b)
CloudSat CPR and (c) TRMM Precipitation Radar. The scan time of the images
is around 19:23 local time on 10 August 2007. CloudSat is about 5 minutes in
front of TRMM with the track centered at 19.85 N, 87.93 W. The color bars are
reflectivity in dBZe. ...........................................................................................................33
2.2 Cloud classification in the ISCCP D-series dataset ...........................................................34
2.3 Schematic representation of MODIS-AUX 3 km x 5 km subset associated with
15 pixels that surround and overlap the CPR footprint, which is highlighted
in light blue. .......................................................................................................................35
3.1 Concatenated reflectivity in dBZe of 10,000 of the 736,443 CPR profiles found
with radar-lidar cloud top heights 14 km and radar heights 13 km .............................56
3.2 Reflectivity profiles of 200 of the 736,443 observations of penetrating deep
convection classified by the strength of their surface echoes (∆dBZe) with
strong surface echoes having ∆dBZ > 20dBZe; moderate echoes between
10 dBZe and 20 dBZe; weak echoes between 6 dBZe and 10 dBZe ; and very
weak echoes < 6 dBZe .......................................................................................................56
3.3 Frequency distribution of Cloudat/Calipso cloud tops a) 14 km and b) 16.9 km
over 35°N─35°S for 2007 organized according to 2.5° x 2.5° bins where the total
of each distribution is 100%. .............................................................................................57
3.4 Same as Figure 3.3 except distribution is for a) June-July-August (JJA) and
b) December-January-February (DJF) ...............................................................................58
3.5 Radar-lidar (black) and radar (alternating color) cloud top height verses a)
brightness temperature difference for all 736,443 observations b) cloud
ix
brightness temperatures for all 736,443 observations c) same as a) but with
210 K cloud brightness temperature constraint d) same as b) but with 210 K
cloud brightness temperature constraint e) same as a) but for cloud brightness
temperatures 200 K and f) same as b) but for cloud brightness temperatures
200 K ..............................................................................................................................59
3.6 Areal size distribution in km2 of penetrating deep convection with consecutive
CPR footprints reaching 14 km. The distribution also corresponds to the criteria
provided in Table 3.2 where the area associated with the penetrating deep
convective cloud top is provided according to the maximum number of
consecutive footprints associated with the CloudSat CPR footprint .................................60
3.7 Top) CloudSat cross-section of a penetrating deep convective cloud (labeled A)
with cloud top heights > 15 km and diameter of ~ 20 km. (Bottom left) Aqua
MODIS (L1B) true color image with the CloudSat track corresponding to the
top panel is shown in yellow and (bottom right) Aqua MODIS cloud optical
thickness from the level 2 cloud product with the CloudSat track also shown in
yellow t-value of 1.962 ....................................................................................................61
4.1 To better describe the function of the CloudSat 2B CLDCLASS product, relevant
vertical range gates depicting the classification of the anvil cloud, the
penetrating top, and the main body of the deep convective cloud are shown.
Anvil clouds, which are not classified by the 2B CLDClass product are
determined when deep convective clouds (cloud class─8) are present and have
high cirriform (cloud class─1) connected to them and no cloud (0 - not shown)
beneath them ......................................................................................................................88
4.2. Frequency Distributions of a) BTD, b) IRW (cloud) brightness temperature, c)
optical thickness, d) WV Brightness Temperature e) cloud top pressure and f)
effective particle radius for all the cloud types described in Table 3.2 where the
overlap associated with each cloud property is shown for each cloud type .....................89
4.3. Distribution of IRW brightness temperatures verses cloud optical thickness
for a) Deep convective clouds and b) cirrus clouds sampled from January 2007
statistics. For both profiles a slightly negative slope of the linear line fit given
by the equations a) and b) show that values of brightness temperature generally
decrease with increasing cloud optical thickness ...............................................................90
4.4. CloudSat cross section on October 22, 2006 showing variability of cloud
brightness temperatures (IRWBT), IR water vapor brightness temperatures
(IRWVBT) and +BTD for stratocumulus cloud (cloud A), overshooting deep
convective cloud (cloud B), penetrating deep convective cloud (cloud C), and
a cirrus cloud (cloud D) .....................................................................................................91
4.5. Frequency distribution between 35°N─35°S of all cold cloud features a) 235 K and b)
210 K and seasonal patterns of cold cloud features 210 K for c) December─
x
February (DJF) and d) June-August (JJA) .........................................................................92
4.6. Global frequency distribution of a) all +BTD signatures (with BT11 < 235 K)
and seasonal patterns of +BTD for b) December-February (DJF) and c) June-
August (JJA) ......................................................................................................................93
4.7 Images of a) MODIS L2 Cloud top temperature b) IRW Brightness Temperature
and c) visible true color images corresponding to CloudSat granule ID 2575 on
October 22, 2006 and time stamp 0450 .............................................................................94
4.8. Normalized frequency of cloud optical depth for all pixels between 15°N─15°S
with BT11 210 K for October 2007 .................................................................................95
5.1 GridSat 2.5° x 2.5° annual frequency distributions of annual positive BTD for.
1998─2008 and over 35°N─35°S ....................................................................................116
5.2 GridSat 2.5° x 2.5° annual frequency distributions of cloud brightness temperatures
210K for 1998-2008 and over 35°N - 35°S ..................................................................116
5.3 November 2005 observations for the Maritime continent (15N-15S, 90E-150E)
of cloud top height from a) ISSCP (D1), b) MODIS-Terra c) MODIS-Aqua
(adapted from Russo et al., 2010) and cloud brightness temperatures 210 K
from d) GridSat ................................................................................................................119
5.4 Time averaged diurnal cycle per year for tropical regions of Africa, the
Indian Ocean, South America, and the western Pacific Ocean........................................120
5.5 Normalized Frequency per month given for all seven regions within the tropics
from January 1998 through December 2008. The western Pacific Ocean clearly
has higher frequencies of penetrating deep convection among all other regions.
considered ........................................................................................................................121
5.6 Standardized frequency anomalies for a) the Tropics (15°N─15°S) b) Africa c)
the Indian Ocean d) the western Pacific Ocean e) central Pacific Ocean f)
eastern Pacific Ocean g) South America and h) the Atlantic Ocean ...............................122
5.7 1998-2008 monthly zonal averages of water vapor volume mixing ratio at 82 mb
from 15°N─15°S ..............................................................................................................123
5.8 1998-2008 anomalies in 82 mb lower stratospheric water vapor mixing ratio from
15°N─15°S ......................................................................................................................123
A.1 (cf., Mace et al., 2007) Illustration of lidar hydrometeor fractions,
Cnh, (in red) that occur within a CPR range resolution volume (in blue). Lidar
hydrometeor fractions reported for each horizontal level are reported in
xi
percentages on the right ...................................................................................................139
A.2 (cf., Mace et al., 2007) Conceptual view of CPR-Lidar overlap with
radar footprint in blue and lidar footprints in red. The black (red) solid and dashed
ellipses (circles) represent the 1 and 2 standard deviation pointing uncertainty of the
radar (lidar) ......................................................................................................................139
B.1 Frequency of Liquid versus Mixed Phase states using modified version of
ECMWF [2007]. At temperatures above 0 C the cloud condensate is all liquid
water. Between 0 C and -35 C condensate is a mixture of ice and liquid water.
At temperatures below -35 C the cloud is fully glaciated ...............................................143
xii
LIST OF ABBREVIATIONS
Ac Altocumulus
As Altostratus
AIRS Atmospheric Infrared Sounder
AR1 Assessment Report 1
AR2 Assessment Report 2
AR3 Assessment Report 3
AR4 Assessment Report 4
AVHRR Advanced Very High Resolution Radiometer
AMSU Atmospheric Microwave Sounding Unit (-A)
BT Brightness Temperature
BT6.7 Infrared Water Vapor Brightness Temperatures at 6.7 m
BT11 Infrared Brightness Temperatures at 11 m
BTD Brightness Temperature Difference
GCM Global Climate Model
CALIPSO Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations
CAPE Convective Available Potential Energy
Cb Deep Convective
CBL Convective Boundary Layer
CERES Clouds and the Earth‘s Radiant Energy System
Ci Cirrus
CPR Cloud Profiling Radar (CloudSat)
CPT Cold Point Tropopause
Cu Cumulus
dBZe Decibels of Reflectivity
DJF December-January-February
EL Equilibrium Level
HALOE Halogen Occultation Experiment
HIRS High Resolution Infrared Sounder
IPCC Intergovernmental Panel on Climate Change
IR Infrared
IRW Infrared Window
IRWBT Infrared Window Brightness Temperature
IRWVBT Infrared Water Vapor Brightness Temperature
ISCCP International Satellite Cloud Climatology Project
ITCZ Intertropical Convergence Zone
JJA June-July-August
LIS Lightning Imaging Sensor (LIS)
LNB Level of Neutral Buoyancy
LRT Lapse Rate Tropoppause
LSWV Lower Stratospheric Water Vapor
LW Longwave
LWP Liquid Water Path
MAM March-April-May
MLS Microwave Limb Sounder
xiii
MODIS Moderate Resolution Infrared Sounder
NCDC National Climatic Data Center
NOAA National Oceanic and Atmospheric Administration
Ns Nimbostratus
ppmv parts per million volume
PR Precipitation Radar
Qclear Level of zero radiative heating
QBO Quasi-biennial Oscillation
Sc Stratocumulus
SON September-October-November
St Stratus
SW Shortwave
SWV Stratospheric Water Vapor
SST Sea Surface Temperature
STEP Stratospheric Tropospheric Exchange Project
TOGA COARE Tropical Ocean Global Atmosphere Coupled Ocean Atmosphere
Response Experiment
TMI Tropical Rainfall Monitoring Mission Microwave Imager
TRMM Tropical Rainfall Monitoring Mission
TTL Tropical Tropopause Layer
UARS Upper Atmospheric Research Satellite
UT/LS Upper Tropospheric/Lower Stratospheric
VIRS Visible Infrared Scanner
WS1 Weather State 1
xiv
LIST OF SYMBOLS
Alpha value; a number between 0 and 1 such that p is considered significant
where p is a p-value.
Cloud optical depth
m Micrometer
ˆ
N
Autocorrelation of the noise, N
ˆ
n *
Number of years to detect trend at the 95% significance level according to
Weatherhead et al., [1998]
ˆ N
Standard deviation of noise, N.
p-value Probability of obtaining a test statistic at least as extreme as the one that
was actually observed, assuming that the null hypothesis is true.
xv
SUMMARY
Several studies suggest that deep convection that penetrates the tropical tropopause layer
may influence the long-term trends in lower stratospheric water vapor. This thesis investigates
the relationship between penetrating deep convection and lower stratospheric water vapor
variability using historical infrared (IR) observations. However, since infrared observations do
not directly resolve cloud vertical structure and cloud top height, and there has been some debate
on their usefulness to characterize penetrating deep convective clouds, CloudSat/Calipso and
Aqua MODIS observations are first combined to understand how to best interpret IR
observations of penetrating tops.
The major findings of the combined CloudSat/Calipso and Aqua MODIS analysis show
that penetrating deep convection predominantly occurs in the western tropical Pacific Ocean.
This finding is consistent with IR studies but is in contrast to previous radar studies where
penetrating deep convective clouds predominantly occur over land regions such as equatorial
Africa. Estimates on the areal extent of penetrating deep convection show that when using IR
observations with a horizontal resolution of 10 km, about two thirds of the events are large
enough to be detected. Evaluation of two different IR detection schemes, which includes cold
cloud features/pixels and positive brightness temperature differences (+BTD), show that neither
schemes completely separate between penetrating deep convection and other types of high
clouds. However, the predominant fraction of +BTD distributions and cold cloud features/pixels
210 K is due to the coldest and highest penetrating tops as inferred from collocated IR and
radar/lidar observations. This result is in contrast to previous studies that suggest the majority of
cold cloud features/pixels 210 K are cirrus/anvil cloud fractions that coexist with deep
convective clouds. Observations also show that a sufficient fraction of penetrating deep
xvi
convective cloud tops occur in the extratropics. This provides evidence that penetrating deep
convection should be documented as a pathway of stratospheric-tropospheric exchange within
the extratropical region.
Since the cold cloud feature/pixel 210 K approach was found to be a sufficient method
to detect penetrating deep convection it was used to develop a climatology of the coldest
penetrating deep convective clouds from GridSat observations covering years 1998-2008. The
highest frequencies of the coldest penetrating deep convective clouds consistently occur in the
western-central Pacific and Indian Ocean. Monthly frequency anomalies in penetrating deep
convection were evaluated against monthly anomalies in lower stratospheric water vapor at 82
mb and show higher correlations for the western-central Pacific regions in comparison to the
tropics. At a lag of 3 months, the combined western-central Pacific had a small but significant
anticorrelation, where the largest amount of variance explained by the combined western-central
Pacific region was 8.25%. In conjunction with anomalies in the 82 mb water vapor mixing
ratios, decreasing trends for the 1998-2008 period were also observed for tropics, the western
Pacific and Indian Ocean. Although none of these trends were significant at the 95% confidence
level, decreases in the frequency of penetrating deep convection over the 1998-2008 shows
evidence that could explain in part some of the 82 mb lower stratospheric water vapor variability.
1
CHAPTER 1
INTRODUCTION
1.1 Introduction and Review
According to the cold trap hypothesis, tropical tropopause temperatures largely control
the annual and inter-annual variability of lower stratospheric water vapor. However, studies
using balloon-borne [Oltmans et al., 2000] and satellite observations [Rosenlof et al., 2001;
Randel et al., 2006; Solomon et al., 2010] show multi-decadal trends in lower stratospheric water
vapor from 1980-2009 that are inconsistent with trends in tropical tropopause temperatures for
the same periods. Using radiative transfer models, these trends have been evaluated and were
shown to considerably impact surface climate [Forster and Shine, 2002; Solomon et al., 2010].
These trends are still largely unexplained. It therefore remains important to evaluate cross
tropopause transport via episodes of strong convection, since several studies provide evidence
(e.g., Chaboureau et al., 2007; Corti et al., 2008; Khaykin et al., 2009; Wang et al., 2009) that
this type of deep convection impacts the lower stratospheric water vapor budget and as suggested
by Rosenlof and Reid [2008] may influence long-term trends in lower stratospheric water vapor
variability.
To address the role of deep convection on lower stratospheric water vapor, several studies
have monitored deep convection reaching the gate to the lower stratosphere, known as the
tropical tropopause layer (TTL). While this approach provides useful approximations on
frequency and regional dominance, the only existing observations that may be used to address
the influence of strong convection on long-term (1980─2009) lower stratospheric water vapor
variability are based on historical passive sensor satellite radiances. It has been challenging to
use observations from passive sensors to quantify this relationship for two reasons. Firstly, deep
2
convective clouds penetrating the TTL contain overshooting deep convection with diameters of
1─10 km. It is unclear how well historical observations with a horizontal resolution of 10 km
may resolve these events. Secondly, observations from passive sensors do not directly resolve
cloud vertical structure and cloud top height. When compared with radar observations that do
provide evidence of cloud vertical structure, active and passive observations yield different
results (e.g. Alcala and Dessler, 2002 vs. Gettelman et al., 2002). Addressing both these issues is
important to better understand the information content that the historical observations may
provide.
Given these details, this thesis evaluates deep convection penetrating the TTL using radar
and IR observations to explore the hypothesis that penetrating and overshooting deep convection
has a strong influence on lower stratospheric water vapor variability. More specifically,
CloudSat/Calipso and Aqua MODIS are used to 1) obtain a statistically robust sample of
penetrating deep convection and evaluate areal extent to determine how well penetrating deep
convection may be resolved from IR sensors (Chapter 3); 2) quantitatively compare IR and radar
distributions of penetrating deep convection using traditional IR techniques to determine the
extent to which traditional IR techniques capture penetrating deep convection (Chapter 4) and 3)
considering the uncertainty of traditional IR techniques, this thesis also examines the
spatiotemporal variability of penetrating and overshooting deep convection from 11 years
(1998─2008) of GridSat observations to address the role of penetrating deep convection in lower
stratospheric water vapor variability (Chapter 5).
The following sections of this introductory chapter further frame the thesis hypothesis by
describing known factors that control lower stratospheric water vapor in Section 1.2. In Section
1.3 the climate impacts of lower stratospheric water vapor variability are described. Section 1.4
3
provides details of penetrating deep convection and the TTL, and Section 1.5 gives an outline of
how the hypothesis of thesis work is explored in the remaining chapters.
1.2 Water Vapor in the Stratosphere
Water vapor is a key climate variable. It is the most abundant green house gas and is
largely found in the lowermost part of the atmosphere, known as the troposphere. Water vapor
comprises between 0–0.4% of the atmosphere‘s gaseous composition by volume and exists in
concentrations that vary as a function of height, latitude, and temperature. Unlike other
greenhouse gases that warm Earth‘s atmosphere and surface, the amount of atmospheric water
vapor directly associated with anthropogenic sources is negligible. The predominant source of
atmospheric water vapor is evaporation from land and ocean surfaces. Atmospheric water vapor
concentration is limited by the saturation specific humidity, which is nearly an exponential
function of temperature; as temperature decreases with height in the free troposphere, the mean
concentration of water vapor decreases exponentially.
Water vapor in the region above the troposphere, known as the stratosphere, is therefore
low (e.g., < 5 parts per million by volume (ppmv)). The water vapor content of the stratosphere
is largely a function of tropical tropopause temperatures whereby air passing through the tropical
tropopause is dehydrated to the region‘s local minimum saturation mixing ratio [Brewer, 1949;
Fluegistaler et al., 2005; Sherwood and Dessler, 2000]. Other factors controlling water vapor in
the stratosphere include methane (CH4) oxidation and in the lower stratosphere it is also
controlled by cross-tropopause transport. The variability of water vapor in the lower stratosphere
is of particular concern in this thesis because despite its low concentration, water vapor in this
region, contributes disproportionately to the natural greenhouse effect. To illustrate this
disproportionality, Solomon et al. [2010] show that for a uniform change of 1 ppmv in 1-km
4
vertical layers, the adjusted total radiative forcing of surface climate is maximized between the
tropical upper troposphere and lower stratospheric regions lying between 14 and 18 km. Thus,
small spatiotemporal changes in the concentration of water vapor between 14 and 18 km may
considerably impact surface climate. For the upper tropospheric region, these impacts were
shown by Sohn and Schmetz [2004]. For the lower stratospheric region, where water vapor has a
lifetime of ~ 1 year, these impacts were shown by Solomon et al. [2010], Forster and Shine
[2002,] and Shindell et al. [2005]. Given these details, it is important to better evaluate the
factors that may be controlling long-term variability of lower stratospheric water vapor. More
details of this variability and its impact on climate are provided below.
1.3 Climate Impacts of Lower Stratospheric Water Vapor Variability
While the major processes controlling stratospheric water vapor are understood, the long-
term variability in lower stratospheric water vapor has not been explained. Oltmans et al. [2000]
used a time series of balloon-borne frost point hygrometer measurements over Boulder, CO,
from 1981─2000 and showed upward trends in lower stratospheric water vapor of ~1% per year,
totaling ~ 1ppmv. Rosenlof et al. [2001] report a comparable trend that spanned a 40-year period
using a combination of datasets. Solomon et al. [2010] also showed a decrease of ~ 0.4 ppmv in
lower stratospheric water vapor between 2000 and 2009. However, corresponding trends in
tropical tropopause temperatures were not observed for each of these periods [Randel et al.,
2004; SPARC, 2000].
Solomon et al. [2010] used a line-by-line radiative transfer model to calculate the
radiative impact of the 1980-2009 lower stratospheric water vapor changes. The authors suggest
that when compared with the radiative forcing due to carbon dioxide, aerosols, and other
greenhouse gases, the 1980─2000 increase of ~ 1-ppmv acted to enhance the decadal rate of
5
surface warming by 30%, while post-2000 decreases of ~ 0.4 ppmv may have slowed the rate of
surface warming by 25%.
In the Intergovernmental Panel on Climate Change (IPCC) third and fourth assessment
reports, AR3 and AR4 respectively, the IPCC acknowledge that several studies show long-term
trends in stratospheric water vapor between 1980 and 2000. The IPCC also state that these trends
would have a significant radiative impact. However, in AR3 and AR4, CH4 oxidation is the only
source of the 1980─2000 trend considered to be a radiative forcing component.
In the illustration of Radiative Forcing components provided by AR4 of the IPCC, the
radiative forcing components from natural and anthropogenic sources are provided. according to
that information, the fraction of anthropogenic stratospheric water vapor changes due to methane
oxidation was estimated to contribute a radiative forcing of +0.07 W m–2 (0.02 ─0.12 W m-2)
based on results from Hansen et al. [2005] (Smith et al., 2001). Stratospheric water vapor
changes unattributable to methane oxidation were not considered because there is a low
understanding of the processes that modulate those changes and there is a lack of scientific
consensus regarding its treatment. More specifically, it is not clear if lower stratospheric water
vapor changes are due to anthropogenic impacts or if they are due to natural variability, which
includes changes in tropical tropopause temperatures or dynamics. The latter is also modulated
by anthropogenic warming [Webster et al., 2006].
In comparison to the radiative forcing of +0.07 W m–2 reported by Hansen et al., [2005],
Forster and Shine [2002] show that the 1 ppmv change from 1980─2001 result in a radiative
forcing of +0.29 W m–2. This radiative forcing significantly differs from results reported by
Hansen et al. [2005]. Radiative forcing values reported by Forster and Shine [2002] are also
consistent with Solomon et al. [2010] who report that the 1980-2000 1ppmv change resulted in a
6
radiative forcing of +0.24 W m-2.
Regarding the decrease of ~ 0.4 ppmv between 2000─2009, Solomon et al. [2010]
suggests that this change led to a radiative forcing of -0.098 W m-2. The authors also suggest that
this negative forcing, which effectively cools the surface climate, provides some explanation of
why global surface temperature rise has slowed in the last decade despite a steady increase in
radiative forcing from CO2 emissions. Thus, a better understanding of the physical processes
controlling changes in the concentration of stratospheric water vapor is a vital prerequisite to
accurately project future climate change. It is necessary to question: What other processes impact
lower stratospheric water vapor? Are changes in lower stratospheric water vapor due to global
warming and climate change?
While several different sources likely contribute to lower stratospheric water vapor
variability, the amount from each source is difficult to quantify. A direct source of water vapor
into the lower stratosphere is from aviation [IPCC, 2007]; while other pathways that may alter
lower stratospheric water vapor are related to stratospheric chemistry and include changes to
methane oxidation rates due to changes in the concentration of stratospheric chlorine, ozone, and
the hydroxyl radical [Röckmann et al., 2004]. Other proposed mechanisms relate to changes in
tropopause temperatures or circulation [Stuber et al., 2001; Fueglistaler et al., 2004] and cross-
tropopause transport [Danielsen, 1993; Rosenlof, 2003; Randel et al., 2004, Chaboureau et al.,
2007; Wang, 2009].
The most likely factors associated with cross-tropospheric transport include slow or
gradual ascent of water vapor by large-scale motion and turbulent diffusion and rapid ascent by
strong convection or volcanic eruption [Joshi and Shine, 2003]. The latter pathway and its
association with strong convection is supported by observations of deep convection reaching 18
7
km (e.g., Alcala and Dessler, 2002) and evidence of tropical tropopause layer and lower
stratospheric hydration by ice crystals lofted from deep convection [Corti et al., 2008]. This is
the source of lower stratospheric water vapor variability addressed in this thesis. More details
associated with this pathway are addressed in the following section.
1.4 Deep Convection in the TTL
Overshooting deep convection occurs when cumulonimbus clouds with strong updrafts
protrude their level of neutral buoyancy (LNB). The LNB is the height at which an air parcel is
no longer more buoyant than the environment because the air parcel‘s temperature is equal to the
environmental temperature. The overshooting cloud top (see Figure 1.1), which has enough
momentum to extend above the LNB, appears as a dome shaped structure rising from the anvil
cloud and can occur with any cumulonimbus cloud when atmospheric instability, usually
estimated by convective available potential energy (CAPE), is high. Although the term,
‗overshooting deep convection‘ is thermodynamically dependent on the cumulonimbus cloud‘s
ability to overcome its LNB, overshooting deep convection investigated in many research studies
(e.g., Rossow and Pearl, 2007; Liu and Zipser, 2005; Gettelman et. al., 2002) are also associated
with penetrating deep convection that reach and extend up into the base of the tropical
tropopause layer (~14 km). From a technical perspective it is important to note that not all
penetrating deep convection belong to the class of overshooting tops, but all overshooting tops
do belong to the class of penetrating deep convection.
Penetrating and overshooting deep convection, which are classified in this thesis as deep
convective clouds with cloud tops 14 km, are most often found within the tropics due to
equatorial convergence of the northeasterly and southeasterly trade winds and high surface
temperatures. More specifically, equatorial convergence of the northeasterly and southeasterly
8
trade winds cause low-level air to converge, forming low pressure systems associated with the
thermally direct equatorial branch of the Hadley cell (meridional) and Walker (zonal)
circulations [Riehl, 1979]. These circulations also produce a seasonally migrating band of low
pressure systems that are most commonly referred to as the intertropical convergence zone
(ITCZ) which is positioned slightly south of the equator in January and near 15°N in July.
As shown in Figure 1.1, overshooting deep convection rises from the convective
boundary layer (CBL) into the tropical tropopause layer (TTL), the cold point tropopause (CPT)
and possibly into the lower stratosphere. Air within the TTL has tropospheric and stratospheric
characteristics. As shown by Folkins et al. [1999], the TTL is the region just below the cold point
tropopause where ozone starts to increase due to inefficient mixing of air from deep convection
and it also marks enhanced stratification according to the environmental lapse rate. The TTL is
the transition zone between ~14 km and 17 km where air starts to assume some of the chemical
characteristics of stratospheric air. It is maintained by the interaction of convective transport,
convectively generated waves (i.e., gravity waves), radiation, cloud microphysics, and the large-
scale stratospheric circulation [Gettleman et al., 2008]. As described by Fueglistaler et al. [2009],
the TTL acts as a ―gate to the stratosphere‖ for atmospheric tracers such as water vapor and so-
called very-short-lived substances. By evaluating the frequency of deep convection reaching this
―gate,‖ it may enable us to answer unresolved questions regarding the role of penetrating deep
convective clouds in long-term changes in lower stratospheric water vapor.
Barrett et al. [1973] was one of the first to link deep convection within mesoscale
convective systems with upper tropospheric/lower stratospheric water vapor exchange. Such
linkages were established using water vapor infrared radiometric measurements made from a U-2
aircraft that flew around and over two penetrating thunderstorms in the southwestern United
9
States. Barrett et al. conclude that a significant fraction of thunderstorms in the plains and
southwest of the U.S. penetrate the tropopause and deposit significant amounts of water vapor in
the stratosphere near and downwind of their tops.
Johnston and Solomon [1979] and Danielsen [1982; 1993] hypothesized that an
overshooting top‘s fate as a source or sink to lower stratospheric water vapor strongly depends
on the fate of ice particles, observed to exist in high number densities [Knollenberg et al., 1993],
within the overshoot. It is suggested that these particles are either hydrate (i.e., add water vapor)
or dehydrate (i.e., remove water vapor) the lower stratosphere. In the case of hydration,
moistening occurs when ice-crystal-laden air within the overshoot entrains warmer stratospheric
air that raises the temperature of the overshoot. The ice crystals sublimate and the layer becomes
more hydrated. In the case of dehydration, radiatively driven overturning of ice-crystal-laden air
within the overshoot facilitates the growth of ice crystals through vapor deposition. Once the ice
crystals become large enough, gravitational settling occurs and the final state of the layer is drier
than its initial state.
While most of the early in situ measurements supported the hypothesis that overshooting
tops would hydrate the lower stratosphere, it is still unclear whether hydration or dehydration
globally dominates. Moreover, the existence of upper tropospheric/lower stratospheric sub-
saturated and super-saturated conditions further complicates the problem. Jensen et al. [2007]
found no evidence to support the hypothesis that overshooting deep convection can dehydrate the
tropical tropopause layer when it is initially sub-saturated with respect to ice. The distribution of
supersaturated regions was shown by Gettelman et al. [2006]. However, more observational data
is needed to understand the mechanics of overshooting deep convection that penetrate these
different regions.
10
Boering et al. [1995] argue that the input of significant amounts of overshooting
tropospheric air above the tropical tropopause is unlikely since it would undermine the seasonal
variation in CO2 propagating upward from the tropopause. However, model simulations suggest
that this is not true in all scenarios [Dessler and Sherwood, 2003]. Field experiments with
isotopic measurements strongly indicate overshooting deep convection, advection, and
microphysics all crucial to the stratospheric water budget [Donner et al., 2007]. In addition,
Rosenlof and Reid [2008] focused on sea surface temperatures in the so-called warm pool region
of the western tropical Pacific Ocean. They conclude that the underlying ocean can have a fairly
direct influence on the lower tropical stratosphere. Without quantitative evidence, they also
speculate that the connecting link between lower stratospheric water vapor variability and sea
surface temperatures in the tropical western Pacific Ocean is probably deep convective towers.
Tseilodius et al. [2010] report that analysis of the time series of convective clouds penetrating
into the lower stratosphere did not show any significant long term trends. The authors conclude
that the influence of convection on stratospheric water vapor trends comes from the overall
moistening of the tropical upper troposphere rather than from direct transport of convection
penetrating the lower stratosphere. However, their result is somewhat questionable according to
the spatial resolution of the ISCCP D1 observations used in their study.
While several field experiments [Thunderstorm Project, Stratospheric Tropospheric
Exchange Project (STEP), Tropical Ocean Global Atmosphere Coupled Ocean Atmosphere
Response Experiment (TOGA COARE)] provide detailed information on tropical deep
convective clouds, these localized studies do not offer a comprehensive picture of the annual and
inter-annual variability of penetrating deep convection throughout the tropics. To gain this view,
observations from space-borne passive sensor satellite radiometers are used. For these
11
observations, two approaches are commonly used to evaluate the frequency and distribution of
penetrating/overshooting deep convection from passive sensors. They include cold cloud features
(e.g., Mapes and Houze, 1993; Liu et al., 2007; Rossow and Pearl, 2007) derived from ~11 μm
brightness temperatures, and positive brightness temperature differences (+BTD) between the
water vapor absorption band at ~6.7 μm and the IR window at ~11 μm (e.g., Schmetz et al.,
1997; Soden, 2000; Setvak et al., 2003; 2007; Chung et al., 2008). While these techniques have
been used to show the seasonality and regional dominance of penetrating deep convective
clouds, passive sensor satellite observations do not provide cloud vertical structure and the extent
to which these techniques are able to sample penetrating deep convective clouds with diameters
of 1-10 km is unclear (Fujita, 1982). Furthermore when IR based distributions of penetrating
deep convection were compared with radar observations that do provide evidence of cloud
vertical structure, active and passive observations yield different results (e.g., Alcala and Dessler,
2002 and Gettelman, 2002). Addressing both these issues is important to better understand the
information content that the historical observations may provide.
Given that passive sensor studies use infrared (IR) cloud brightness temperatures and
active sensor studies rely on rather direct measurements of cloud vertical structure, radar
observations are used to better interpret the extent to which deep convective clouds penetrating
the TTL are sampled from IR observations. Liu et al. [2007] provide the first study to evaluate
deep convection penetrating the TTL from radar and IR using the Tropical Rainfall Monitoring
Mission (TRMM) Precipitation Radar (PR). However, this instrument is designed to observe
cloud vertical structure as determined by its sensitivity to precipitation size particles. This
proves problematic for penetrating deep convection reaching the TTL since its occurrence is
critically determined by cloud top structure most often characterized by the distribution of cloud
12
size particles within the cloud top. In contrast to the TRMM PR, the CloudSat Cloud Profiling
Radar is sensitive to cloud particles. Furthermore, the combination of CloudSat with Cloud-
Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO), which provide a more
highly resolved description of cloud top height, and IR observations from the Aqua Moderate
Resolution Infrared Sounder (MODIS) yield retrievals of penetrating deep convective clouds
superior to retrievals from any one or two instruments.
1.5 Exploring Hypothesis
This thesis evaluates deep convection penetrating the TTL using radar and IR
observations to explore the hypothesis that penetrating and overshooting deep convection has a
strong influence on lower stratospheric water vapor variability. An overview of how this
hypothesis is explored is provided below.
1.5.1 Overview of CloudSat, Calipso, and Aqua-MODIS Data Products
Collocated measurements from CloudSat, Calipso, and Aqua-MODIS enable satellite-
based radar, lidar, and passive remote sensing measurements to be used in concert to better
detect and characterize penetrating and overshooting deep convection. However, these
instruments and their corresponding products are relatively new and the benefits of CloudSat vs.
TRMM observations may not be clear. In Chapter 2, the strengths and weaknesses of the
CloudSat Cloud Profiling Radar are compared to the Tropical Rainfall Monitoring Mission
(TRMM) Precipitation Radar to identify the potential insight CloudSat may provide in
comparison to TRMM. In addition, CloudSat/Calipso and Aqua MODIS data products are
described to clearly outline how these datasets advance this research. CloudSat products
described in Chapter 2 include 2B-GEOPROF, 2B-Lidar-GEOPROF, and 2B-CLDCLASS
products. MODIS products include CloudSat developed MODIS-AUX (Aqua MODIS LIB
13
product) and MAC06S0 (Aqua MODIS L2 Cloud Product) products.
1.5.2 Characterization of Penetrating Deep Convection
In Chapter 3, an overview is given of TRMM and CloudSat studies that have
characterized penetrating and overshooting deep convective clouds. After evaluating the
limitations of these studies, CloudSat and Calipso data products are used to capture penetrating
and overshooting deep convection that reach the base of the tropical tropopause layer (~14 km).
The observations are characterized by cloud top height, cloud base height, maximum reflectivity,
areal size distribution, seasonality, and spatial frequency distribution. CloudSat and Calipso
detections of penetrating and overshooting deep convection are also combined with Aqua
MODIS products to evaluate penetrating deep convective clouds using traditional IR techniques.
As described in Section 1.4 these techniques include the cold cloud feature/pixel technique and
the +BTD signature. Results obtained in Chapter 3 provide a statistically robust characterization
of penetrating and overshooting deep convection from radar and IR observations. The
observations of penetrating deep convection provided in this chapter are also used to estimate the
areal size distribution of penetrating deep convective clouds. The areal size distribution of
penetrating tops is needed to determine the extent to which historical IR observations, with 10
km resolution, are able to resolve this particular type of cloud.
1.5.3 Evaluation of Traditional IR Techniques
Observations of cold cloud features and positive BTD signatures are compared with
penetrating deep convective clouds that were gathered in Chapter 3. The comparison of these
three different sets of observations are used to statistically characterize the extent to which
traditional passive sensor approaches are able to exclusively capture overshooting deep
convection and to evaluate, to the extent possible, the percentage of other types of high level
14
clouds (i.e., anvil clouds, pileus clouds, jumping cirrus, lower level deep convection, etc.) that
may be present in the IR distributions. In addition, Chapter 4 also compares the microphysical
and optical properties of penetrating deep convection with other types of high clouds. By
evaluating these properties the capabilities and information content of IR-based distributions of
penetrating deep convection are better understood and their capability to detect penetrating deep
convective cloud tops is revealed. The results gathered in Chapter 4 are then used to
quantitatively characterize the relationship between cold cloud features and observations of
penetrating deep convective clouds. These results are used in Chapter 5 to develop a climatology
of penetrating and overshooting deep convective cloud tops.
1.5.4 Climatology of Penetrating Deep Convection
In Chapter 5, cold cloud pixels and positive BTD signatures are evaluated from
GridSat observations covering the period between 1998 and 2008 to develop a climatology of
penetrating deep convective cloud tops. GridSat observations that have been derived from the
International Satellite Cloud Climatology Project (ISCCP) and developed by the National
Climatic Data Center (NCDC) of the National Oceanic and Atmospheric Administration
(NOAA) are also described. GridSat observations are especially used in Chapter 5 because they
have high temporal resolution, a spatial resolution of ~ 10 km, and unlike Aqua MODIS
observations, they provide observations of penetrating deep convective cloud tops that cover the
onset of a downward trend in lower stratospheric water vapor that occurred around 2000
[Solomon et al., 2010]. While the diurnal, intraseasonal, and interannual variability of
penetrating deep convective cloud tops is evaluated, the major focus of Chapter 5 is to compare
the interannual variability of penetrating deep convection with the interannual variability of
lower stratospheric water vapor for the 1998-2008 period. By performing this evaluation we
15
ultimately evaluate the impact of penetrating deep convective cloud tops on lower stratospheric
water vapor. This evaluation is first carried out over the entire tropics and separate regional
analysis focused on the Indian Ocean, west Pacific, central Pacific, East Pacific, South America,
the Atlantic, and Africa.
16
Figure 1.1: Schematic diagram of upward mass flux of tropical overshooting deep convection
penetrating from the convective boundary layer (CBL) into the tropical tropopause layer (TTL),
cold point tropopause (CPT) and on into the lower stratosphere. The diagram also shows the
anvil of the deep convective cloud identified within the deep convection outflow layer located
between 11 km and 15 km as well as a shallow outflow layer where shallow convective clouds
are primarily detrained. At ~15 km, the level of zero net radiative heating is distinguished as
Qclear=0. As indicated by the one directional arrows above and below this level, there is net
downward motion associated with clear sky radiative cooling and subsidence just below this
level and net radiative heating and upward vertical motion just above this level.
17
CHAPTER 2
OVERVIEW OF CLOUDSAT, CALIPSO, AND AQUA MODIS DATA
PRODUCTS
2.1 Introduction
The TRMM Precipitation Radar has been used to evaluate deep convection reaching the
base of the TTL both independently (e.g., Liu et al., 2005) and in conjunction with collocated
observations from passive IR sensors (e.g., Alcala et al., 2002; Liu et al., 2007). However, as its
name would imply, the TRMM Precipitation Radar is sensitive to precipitation size particles.
This sensitivity prevents the TRMM Precipitation Radar from fully resolving the portion of
cloud vertical structure that is determined by smaller cloud particles typically found at the
boundaries of the cloud, which includes the cloud top. Thus it is important to determine, how
observations from the CloudSat Cloud Profiling Radar may be used to provide more details of
cloud vertical structure in comparison to the TRMM Precipitation Radar. After comparing
CloudSat and TRMM radars in Section 2.2, the remaining sections in this chapter provide details
of CloudSat/Calipso and Aqua MODIS observations. The information provided in this chapter
outlines how these datasets advance this thesis. It identifies how key observations from these
instruments are obtained and clarifies common limitations and concerns.
2.2 Comparison of TRMM and CloudSat Radar Characteristics
The TRMM Precipitation Radar launched in late 1997 and the CloudSat Cloud Profiling
Radar launched in April 2006 are the only radars that have operated in space. The technical
details of both these radars are provided in Table 2.1, where differences in geographic coverage,
temporal resolution, scan characteristics, and radar characteristics are provided.
18
As indicated in Table 2.1, the TRMM PR has a radar sensitivity of 12 dBZ and operates
at a frequency of 13.8 GHz. Given these characteristics, TRMM detects particles with a mean
mass diameter of ~ 275 m [Alcala and Dessler, 2002] and is capable of detecting fairly light to
intense rain rates. As such, TRMM has shown exceptional capabilities in characterizing the
intensity and distribution of rain, rain type, storm depth, and the height at which snow melts into
rain. Although it has also been used to evaluate deep convective clouds reaching the TTL [Alcala
and Dessler, 2002; Luo et al., 2005; 2007], the mean mass diameter detectable by TRMM is
much larger than the mean mass diameter of ice crystals that have been found in the
overshooting tops of deep convective clouds during field studies (e.g., Knollenberg et al., 1993;).
The smaller particles that comprise the cloud top must be detected to truly evaluate cloud vertical
structure and determine when deep convective clouds penetrate the TTL and potentially modify
the lower stratospheric water vapor content.
In comparison to the TRMM PR, the CloudSat Cloud Profiling Radar has a sensitivity of
-28 dBZ and operates at a frequency of 94 GHz. These specifications enable CloudSat to detect
much smaller sized cloud droplets and ice crystals on the order of ~30 µm. To illustrate the
capabilities of CloudSat and TRMM, nearly coincident images of CloudSat and TRMM cross
sections of a deep convective cloud centered at 19.85°N, 87.93°W are provided in Figure 2.1
(cf., Li and Schumaker, 2010). As indicated in Figure 2.1, many features of the vertical cloud
structure provided by CloudSat are unapparent in the TRMM cross section.
Since CloudSat provides more details on cloud vertical structure and especially the
details of the cloud top, it is better suited for penetrating deep convection studies, where cloud
top structure is a key determinant of deep convection entering the tropical tropopause layer
(TTL). On the other hand, CloudSat is not as optimally suited to monitor deep convection within
19
the regional focus of this thesis, which includes the tropics and extratropics. Although
CloudSat‘s high inclination angle allows it to monitor polar regions unobserved by TRMM, this
high inclination angle also prevents the CPR from sampling the tropical atmosphere as often.
Moreover, CloudSat is a nadir pointing radar, while TRMM is a cross-track scanning radar with
a swath width of 247 km. These scanning differences allow TRMM to sample much larger
volumes of the tropical atmosphere in comparison to CloudSat. Finally, CloudSat‘s sun-
synchronous polar orbit, which has an equatorial crossing of 1:30 a.m./p.m., limits CloudSat
from fully resolving the diurnal cycle of deep convection. This issue is more serious over land
than over ocean since Soden [2000] and Liu and Zipser [2005] show that continental deep
convection has a pronounced peak in frequency during late afternoon. This likely impacts the
geographical distribution of penetrating and overshooting deep convection from CloudSat and is
addressed in Chapter 3. Yet despite these limitations, CloudSat has been operating in space for
several years and provides a statistically robust set of observations for most cloud types. A
description of CloudSat Standard Products and collocated Aqua MODIS data products that are
used in this thesis are provided below.
2.3 Products from CloudSat Cloud Profiling Radar (CPR)
CloudSat Standard Data Products are distributed by the CloudSat Data Processing Center
located at the Cooperative Institute for Research in the Atmosphere at Colorado State University.
Process Description/Interface Control Documents for the Standard Data Products may be
obtained at http://www.cloudsat.cira.colostate.edu/dataICDlist.php. A brief description of the
products used and evaluated in this thesis is provided below.
20
2.3.1 CloudSat 2B-GEOPROF Product
The 2B-GEOPROF product is the primary product used in most CloudSat studies.
Version R04 is used in this thesis to provide direct observations of cloud vertical structure,
evaluate cloud top height and cloud base height and validate that penetrating and overshooting
deep convection are indeed being observed. The cloud geometric profile contained in the 2B-
GEOPROF product includes the cloud mask, reflectivity-field, and gaseous absorption provided
on a height grid with 125 vertical range bins that correspond to heights from below sea level up
to ~ 25 km. The 2B-GEOPROF product is developed from range-resolved profiles of
backscattered power obtained from the CPR. These measurements represent a 0.16 second
average (i.e., 540 pulses) of returned power that corresponds to a horizontal resolution of 2.5 km
along track by 1.4 km across track while the instantaneous field of view of the CPR is 1.8 km x
1.4 km. The vertical range resolution of the CPR is 500 m but it is oversampled to generate a
range gate spacing of 250 m. This enables the effective vertical range resolution from CloudSat
to be equivalent to that of TRMM and indicates that any cloud top height provided from
CloudSat has an error of 250 m.
To identify the presence or absence of clouds, the significant echo mask or cloud mask
contains values between 0 and 40 where increasing cloud mask values indicate a reduced
probability of false cloud detection. Values between 30 and 40 that are used in this thesis ensure
that the vertical cloud structure and cloud top heights are associated with the highest confidence
levels of cloud detection.
2.3.2 CloudSat 2B-CLDCLASS Product
In conjunction with the 2B-GEOPROF product, the 2B-CLDCLASS product is used in
this thesis to identify the presence and vertical extent of deep convective clouds and other types
21
of high clouds. According to Sassen and Wang [2007] the 2B-CLDCLASS product is developed
by converting vertical profiles of radar reflectivity into cloud type. To determine cloud type,
algorithms are commonly used to classify clouds based on cloud spectral, textural, and physical
features. Although cloud classification schemes are limited by instrument sensitivity and subject
to misinterpretation, observational features help to organize clouds into categories with unique
characteristics of composition, radiative forcing, and heating/cooling effects [Hartmann et al.,
1992; Chen et al., 2000]. The International Satellite Cloud Climatology Project (ISCCP)
approach to cloud classification [Rossow and Schiffer, 1999] uses the combination of cloud top
pressure and cloud optical depth to classify clouds into either cumulus (Cu), stratocumulus (Sc),
stratus (St), altocumulus (Ac), altostratus (As), nimbostratus (Ns), cirrus (Ci), cirrostratus (Cs),
or deep convective (Cb) clouds.
In contrast to the ISCCP cloud classification, the 2B-CLDCLASS product classifies
clouds by using vertical and horizontal cloud properties, the presence or absence of precipitation,
cloud temperature, and upward radiance from MODIS measurements to identify eight basic
cloud types. These cloud types include cloud types provided from the ISCCP classification
scheme, provided in Figure 2.2 with the exception that cirrus, cirrocumulus, and cirrostratus
represent the class of high clouds. The characteristic cloud features for these eight major cloud
types have been derived from numerous studies and are featured in Table 2.2. According to
Table 2.2, deep convective clouds are categorized as clouds with cloud base heights between
0─3 km, intense shower of rain or hail possible (as suggested by its reflectivity values), a
horizontal dimension of 10 km, a thick vertical dimension, and liquid water path (LWP) > 0.
Although the 2B-CLDCLASS cloud type identification algorithm is based on many of the
characteristics provided in Table 2.2, it more specifically uses role-based classification methods
22
that assign different threshold values to characteristic parameters provided according to Table
2.3. In this approach, results of the radar cloud mask are first used to find cloud clusters that are
present in both horizontal and vertical directions. Once a cloud cluster is found, cloud height,
temperature, maximum dBZ, and occurrence of precipitation are determined. According to Kahn
et al. [2008], cloud types As, Ns, Cb, and Cu (congestus) are well detected and classified with
the radar-only algorithm. However, the class of Ci clouds is well-classified but under-detected
because of the existence of small ice particles that the CPR is unable to detect. While
observations of cirrus clouds become more important to this thesis in Chapter 4, CloudSat
observations of high clouds, which include cirrus, cirrostratus, and cirrocumulus provide suitable
samples to compare with deep convection (e.g., Sassen et al., 2008).
Although deep convective (Cb) clouds addressed in this thesis are well detected from the
2B-CLDCLASS product, adequate characterization of penetrating and overshooting deep
convection from space-borne active and passive remote sensing platforms requires that the view
of the cloud top and corresponding microphysical and optical properties of deep convective
clouds are not contaminated by other cloud types. Thus it is important to account for very thin
cirrus, which may develop over deep convective clouds as a consequence of gravity wave
breaking and other dynamical processes ─ a detail that may have been missed in other studies
using CloudSat observations of penetrating and overshooting deep convection (e.g., Chung et al.,
2008; Luo et al., 2008).
2.3.3 CloudSat 2B-Lidar-GEOPROF product
CloudSat will often miss tenuous cloud condensate at the tops of some clouds or clouds
composed only of very small ice crystals or liquid water droplets. To accurately represent the
condensate at the tops of deep convective clouds and thereby identify the height to which ice
23
crystals may be detrained into the tropical tropopause layer and lower stratosphere, observations
from Calipso, which trails CloudSat by ~15 seconds, are integrated into this analysis via the
CloudSat 2B-Lidar-GEOPROF product.
To develop this product radar and lidar data streams are optimally merged to produce the
most accurate description of hydrometeor layers within the atmospheric column observed by the
CPR. The approximate vertical and horizontal resolutions of the lidar are compared to the CPR
in Table 2.4. As indicated by Table 2.4, the lidar has higher vertical and horizontal resolutions
and advances this thesis work by providing highly accurate estimations of cloud top height
within 30 m.
Combined radar and lidar cloud masks are used to best estimate hydrometeor layers in the
vertical column along the spatial dimension defined by the CPR. The output of this layer
product includes the base and top heights of up to five distinct hydrometeor layers as well as
some indication as to whether those layers were observed by the radar, the lidar, or both the radar
and lidar. For this thesis work, this provision allows observations of penetrating and
overshooting deep convective clouds to be limited to atmospheric columns with only one cloud
layer thereby reducing possible contamination of very thin cirrus which may give a false
estimation of cloud top height (as mentioned in Section 2.2.2). Here vertically connected
CloudSat bins with cloud mask values (≥ 30) define a cloud layer. A layer boundary is defined as
the first encounter of a cloudy range level (of either the radar or lidar) following the occurrence
of a cloud-free range level by the radar or a cloud range level with a lidar hydrometeor fraction <
0.5. More details of this description are provided in Appendix A.
For each cloud layer identified by the 2B-Lidar-GEOPROF product, the following
conventions are defined:
24
Due to its finer vertical resolution, the lidar is always deferred to in reporting a layer
boundary. This point should be considered when interpreting differences between radar
and radar-lidar cloud top heights reported for observations of penetrating deep convection
provided in Chapter 3.
If a layer top is identified by the lidar and the layer is not optically thin (i.e.,
3~
), the
lidar will attenuate before a distinct layer base is identified. Under these circumstances,
the layer base is defined by the radar observations.
For clouds with heavy precipitation, attenuation of the radar may give a cloud-base that is
too high. However for most conditions, the ―cloud-base‖ is too low. This happens
because radar attenuation through the cloud layer is not very strong and the radar
continues to detect precipitation below the cloud; such that the reported ―cloud base‖ is
actually the ―precipitation base‖. On the other hand, when the attenuation is large, there
tends to be a lot of multiple-scattering which yields ―false‖ return power below the cloud
layer. Strong attenuation can be assessed based on the presence or absence of surface
echoes within the radar reflectivity profile as in Battaglia and Simmer [2008]. This will
be addressed in Chapter 3.
For the estimation of cloud top height, no apparent limitations have been provided for the
2B-Lidar-GEOPROF product. Limitations that have been reported relate to the Calipso vertical
cloud-aerosol feature mask, which is used as an input for the 2B-Lidar GEOPROF product. This
limitation suggests that the cloud-aerosol mask correctly identifies a layer as cloud or aerosol
about 90% of the time. The most common misclassification occurs when portions of dense
aerosol layers within the lower troposphere are labeled as cloud. However, this does not appear
to be problematic for the application of the 2B-Lidar GEOPROF product in the detection of
25
penetrating and overshooting deep convection unless observations are reported for atmospheric
regions impacted by volcanic eruptions that produce dense aerosol layers that may be confused
with high level clouds.
2.4 Products from Aqua MODIS
Aqua MODIS is a passive cross-track-scanning imaging radiometer designed to take
measurements in spectral regions adhering to a number of heritage sensors including those
associated with the historical IR observations that will be used in Chapter 5. In addition, Aqua
MODIS provides high radiometric sensitivity (12 bit) in 36 spectral bands ranging in wavelength
from 0.4 µm to 14.4 µm. Bands 1 and 2 are imaged at a nominal resolution of 250 m at nadir,
bands 3-7 and 26 are imaged at 500 m, and the remaining 28 emissive bands are imaged at a
horizontal resolution of 1 km. A ± 55° scanning pattern at an orbit of 705 km achieves a 2,330-
km swath width and provides global coverage every one to two days. This wide coverage also
means that although MODIS has a high horizontal resolution at nadir, its horizontal resolution
increases up to ~12 km x 6 km at the edge of the satellite swath.
Along with the other A-train satellites, Aqua MODIS has a north to south equatorial
crossing of 1 p.m. While the MODIS instrument also flies on the Terra satellite, the Aqua
MODIS instrument was specifically chosen for this thesis work because it trails CloudSat by ~
60 seconds. Given this small temporal degree of separation, the Aqua MODIS instrument
provides collocated passive sensor satellite radiances of penetrating deep convective clouds
detected by CloudSat so that radar and IR observations of penetrating and overshooting deep
convection will be captured near-simultaneously. While high horizontal and radiometric
resolution makes Aqua MODIS very different from conventional satellite radiometers with lower
radiometric resolutions (i.e., 8 bit), it helps us to assess sub-pixel variability not resolved by
26
conventional satellite radiometers, which if necessary may be evaluated at resolutions that are
more consistent with the historical IR observations. Aqua MODIS products used in this thesis
include the Level 1B (MOD02_1KM) radiances and the Level 2 Cloud (MYD06_L2) product
respectively packaged as MODIS-AUX and MAC06S0 for specific use with CloudSat
applications.
2.4.1 MODIS-AUX Product
In this thesis, the CloudSat MODIS-AUX product is used to provide IR based
distributions of penetrating deep convective clouds. It is also used to evaluate the brightness
temperatures corresponding to radar observations of penetrating deep convective clouds and
other types of high clouds that may contaminate IR observations of penetrating and overshooting
deep convection. The MODIS-AUX product contains Level 1B MODIS (MOD02_1KM_L1B)
radiance and cloud mask data that overlap and surround each CloudSat (CPR) footprint.
Operating one CloudSat ray at a time, the subset-to-reference algorithm uses the geolocation of
the CPR footprint as a reference to find the closest MODIS pixel to the CloudSat footprint. A 3-
pixel across-track by 5-pixel along-track grid that is centered on the CloudSat footprint defines
the 15-element vector of MODIS L1B radiances associated with each CPR footprint. The
radiance values are provided with scale factors, offsets, and radiance uncertainty indexes for
Aqua MODIS bands 1-7, 17-20, and 26-36. Figure 2.3 schematically represents the intersection
of the 15─element MODIS vector overlapping a CPR footprint. Additional information on the
MODIS instrument calibration and characterization can be found in the MODIS L1B Algorithm
Theoretical Basis Document.
27
2.4.2 MODIS L2 Cloud Product (MAC06S0)
The MAC06S0 product is used to identify the optical and microphysical properties of
penetrating and overshooting deep convection and other types of high clouds. The MAC06S0
product is the only product used in this study that is not a part of the CloudSat standard suite of
products; it can be obtained from ftp://atrain.sci.gsfc.nasa.gov/data/s4pa/MAC/. The MAC06S0
product provides 3 x 11 pixel subsets corresponding to data from the Aqua MODIS L2 Cloud
product (MYD06_L2) and has a horizontal resolution of 15 km x 10 km for most variables. Data
from the MAC06S0 product consists of cloud optical and microphysical properties derived from
infrared, visible and near infrared radiances, which are provided for various types of high clouds
in Chapter 4. Addressing the microphysical and optical properties of penetrating deep convective
clouds has never been done in any other studies, and may shed further insight on the evaluation
of penetrating deep convective clouds from space-borne passive sensors. Optical and
microphysical properties provided by the MODIS L2 Cloud Product evaluated in this thesis
includes cloud top pressure and temperature, cloud phase (where water=1, ice=2, and mixed=3),
cloud fraction, optical thickness, and effective radius. These particular variables were chosen to
better evaluate the factors controlling how observations of penetrating deep convective clouds
are viewed from passive IR and radar observations.
The Algorithm Theoretical Basis Document provided by [King et al., 1997] can be used
to find more specifics on the retrievals of each variable. However, it is of technical importance to
note that cloud top pressure and temperature are specifically generated using the CO2 slicing
algorithm that corrects for possible cloud semi-transparency. The CO2 slicing method takes
advantage of differing partial CO2 absorption in several of the MODIS infrared channels (33-36)
located within the 15-micron CO2 band. The CO2 slicing method is addressed in Chapter 4 when
28
cloud top temperatures and cloud brightness temperatures are compared. Such comparison
provides a better understanding of how different classes of high-level clouds are represented
from these two perspectives in order to better gauge the extent to which penetrating deep
convective clouds are sampled from traditional IR techniques. The CO2 slicing method has been
used in operational processing of GOES (Geostationary Operational Environmental Satellite) and
HIRS (High resolution Infrared Radiometer Sounder) data, and has been found to have
accuracies of approximately 50 mb for clouds above ~700 mb.
2.5 Summary and Discussion
The TRMM precipitation radar has been used in several investigations of penetrating and
overshooting deep convection [Alcala and Dessler, 2002; Liu and Zipser., 2005; Liu et al., 2007].
However, characteristics of the TRMM PR considerably differ from the CloudSat Cloud
Profiling Radar. As previously described, CloudSat provides more details of the vertical cloud
structure and especially the cloud top. Thus, CloudSat observations provide an opportunity to
more definitively capture and evaluate penetrating and overshooting deep convection entering
the TTL and lower stratosphere.
CloudSat, Calipso, and Aqua MODIS observations that will be used in subsequent
chapters of this thesis have been described. The advantages of each of these three instruments
have been documented and approaches used to obtain various types of measurements (i.e., radar
reflectivity, cloud top pressure, cloud top temperature, etc.) from each instrument have been
provided. As described, the combination of all three data sets provide unparalleled global
sampling of optical and microphysical properties of clouds and is especially well suited for the
detection of penetrating deep convective clouds. While each instrument has limitations, these
limitations do not represent significant issues with regard to the objectives of this thesis.
29
Table 2.1: The radar characteristics of CloudSat and TRMM show differences in geographic
coverage, temporal resolution, scan characteristics, radar characteristics, and list auxiliary
instruments in or associated with the payload.
TRMM PR
CloudSat CPR
Geographic
Coverage
Latitudinal
Coverage
38°N-38°S
90°N-90°S
Longitudinal
Coverage
180 W-180 E
180 W-180 E
Orbit
Equatorial
Polar
Swath Width
247 km
n/a
Temporal
Resolution
Repeat Cycle
16 days
16 days
Orbits per day
16
16
Scan
Characteristics
Scanning Geometry
Cross Track 17
Nadir Pointing
Swath Width
247 km
0
Vertical Res.
250 m
500 m
No. Vertical Range
Gates
80 (0-20 km)
125 (0-25 km)
Horizontal Res.
5 km
1.8 km x 1.4 km
Radar Characteristics
Radar Sensitivity
12 dBZ
-28 dBZ
Operating
Frequency/Band
13.8 GHz/Ku
94 GHz/W
Mean Diameter
Size
~275 µm
~30 m
Satellite Altitude
350 km
705 km
Pulse Width
1.07 s
3.3 s
Antenna Power
500 W
700 W
Instruments in/associated
with Payload
TRMM Microwave
Imager (TMI)
Calipso
Visible Infrared Scanner
(VIRS)
Aqua MODIS
Clouds and the Earth‘s
Radiant Energy System
(CERES)
Aura Microwave Limb
Sounder (MLS)
Atmospheric Infrared
Sounder (AIRS)
Lightning Imaging Sensor
(LIS)
Atmospheric Microwave
Sounder Unit (AMSU-A)
30
Table 2.2: (cf., Wang and Sassen, 2007) Characteristic cloud features for the major cloud types
derived from numerous (midlatitude) studies.
Cloud Class
Cloud Features
High Cloud /Cirriform
(1)
Base
>7.0 km
Rain
no
Horizontal Dimension
103 km
Vertical Dimension
moderate
LWP
=0
As (2)
Base
2.0-7.0 km
Rain
none
Horizontal Dimension
103 km, homogeneous
Vertical Dimension
moderate
LWP
~0, dominated by ice
Ac (3)
Base
2.0-7.0 km
Rain
virga possible
Horizontal Dimension
103 km, homogeneous
Vertical Dimension
shallow or moderate
LWP
>0
St (4)
Base
0-2.0 km
Rain
none or slight
Horizontal Dimension
103 km, homogeneous
Vertical Dimension
shallow
LWP
>0
Sc (5)
Base
0-2.0 km
Rain
drizzle or snow possible
Horizontal Dimension
103 km, inhomogeneous
Vertical Dimension
shallow
LWP
> 0
Cu (6)
Base
0-3.0 km
Rain
drizzle or snow possible
Horizontal Dimension
1 km, isolated
Vertical Dimension
shallow or moderate
LWP
> 0
Ns (7)
Base
0-4.0 km
Rain
prolonged rain or snow
Horizontal Dimension
103 km
Vertical Dimension
thick
LWP
> 0
Deep Convective clouds
(8)
Base
0-3.0 km
Rain
intense shower of rain or hail possible
Horizontal Dimension
10 km
Vertical Dimension
thick
LWP
> 0
31
Table 2.3: (cf., Wang et al., 2007) Cloud ID rules based approximately on the properties for the
98th percentile of data for each cloud type that was sampled during first 6 months while
CloudSat was in orbit.
Type
Zmax
Precipitatin
Length
(km)
Highest Zmax
frequency
Other
Cirrus
<-3 dBZ,
T < -22.5°C
No
2>1000
-25 dBZ @
-40°C
Altostratus
<10dBZ,
-20°<T<-5°C;
=-30 dBZ @
-45°C
No
50 > 1000
-10 dBZ @
-25°C
Altocumulus
<0 dBZ,
-20°<T<-5°C;
=-30dBZ @
-35°C
Yes/No
2 > 1000
-25 dBZ @
-10°C
Ttop > -35°C
St
<-5dBZ,
-15°<T<25°C
Yes/No
50 > 1000
-25 dBZ @
-10°C
(Bright
Band)
Altitude of
Zmax < 2 km
AGL;
Sc
<-5 dBZ,
-15°<T<25°C
Yes/No
2 > 1000
-25 dBZ @
-10°C
(Bright
Band)
Altitude of
Zmax < 2
km AGL;
spatially
inhomogene
ous
Cumulus
< 0 dBZ,
-5°<T<25°C
Yes/No
2-25
-25 dBZ @
-15°C
(Bright
Band)
ΔZ > 2 km
Deep (cb)
< -5 dBZ,
-20°<T<25°C
Yes
10-50
10 dBZ @
5°C
ΔZ > 6 km
Ns
-10 < Z < 15,
-25°<T<10°C
Yes
>100
+5 dBZ @
0°C
ΔZ > 4 km
32
Table 2.4: Approximate vertical and horizontal resolutions of the CloudSat CPR and the Calipso
Lidar.
Cross Track
Along Track
Vertical
CPR
1.4 km
2.5 km
0.25 km
Lidar
0.3 km
1 km- < 8.2 km
0.3 km > 8.2 km
0.03 km < 8.2 km
0.075 km > 8.2 km
33
Figure 2.1: (cf., Li and Schumaker, 2010) Images of a coincident CloudSat and TRMM
overpass showing a) TRMM PR horizontal cross sections at 2 and 7.5 km for orbit 55469 with
CloudSat track in magenta, and vertical cross sections of b) CloudSat CPR and (c) TRMM
Precipitation Radar. The scan time of the images is around 19:23 local time on 10 August 2007.
CloudSat is about 5 minutes in front of TRMM with the track centered at 19.85 N, 87.93 W.
The color bars are reflectivity in dBZ.
34
ISCCP CLOUD CLASSIFICATION
Figure 2.2: Cloud classification in the ISCCP D-series dataset.
50
180
310
440
560
680
800
1000
CIRRUS
CIRROSTRATUS
DEEP
CONVECTION
CLOUD TOP PTESSURE (mb)
ALTOCUMULUS
ALTOSTRATUS
NIMBOSTRATUS
CUMULUS
STRATOCUMULUS
STRATUS
CLOUD OPTICAL THICKNESS
1.3
0
3.6
9.4
23
60
379
35
Figure 2.3: Schematic representation of MODIS-AUX 3 km x 5 km subset associated with 15
pixels that surround and overlap the CPR footprint, which is highlighted in light blue.
36
CHAPTER 3
CHARACTERIZATION OF PENETRATING DEEP CONVECTION
3.1 Penetrating Deep Convection from Previous CloudSat and TRMM Studies
Deep convection that penetrates the tropical tropopause layer (TTL) plays an important
role in affecting the heat budget [Sherwood et al., 2003; Kuang and Bretherton, 2004] and
moisture distributions [Danielsen, 1982; Sherwood and Dessler, 2000] of the tropical upper
troposphere and lower stratosphere.
Although deep convection reaching the TTL has most commonly been monitored from
passive sensor sounders and imagers, these observations do not provide direct measurements of
cloud vertical structure. Instead, cloud vertical structure is derived using combinations of
radiative transfer modeling and a priori assumptions about the surface and atmospheric state. The
advent of space-borne radar has added a complementary view of penetrating deep convection
from passive sensor observations that may be used to better evaluate cloud vertical structure and
cloud top properties inferred from IR data.
Simpson et al. [1998] was the first to observe deep convection extending into the TTL
using the first space-borne radar, the TRMM Precipitation Radar (PR), during Typhoon Paka in
early December 1997. This led to the application of the TRMM PR in studies seeking to evaluate
the characteristics of deep convection reaching the TTL. One of such studies includes Alcala and
Dessler [2002] who defined deep convection reaching the TTL by reflectivity tops with heights
14 km and a minimum depth of 1.5 km where all reflectivities exceed 12 dBZ. From this
definition, the frequency and seasonality of deep convection reaching the TTL was evaluated
over the entire tropics for 4 months (January 1998 and 1999, and July 1998 and 1999). These
observations were then compared with IR brightness temperatures 207.5 K from the TRMM
37
Visible Infrared Sounder (VIRS). The radar and IR observations showed the same inter-seasonal
and inter-annual patterns, including the behavior of the ITCZ and the South Pacific Convergence
Zone (SPCZ). While Alcala and Dessler [2002] also show qualitative differences between radar
and IR observations that are primarily due to the sensitivities of each instrument, no quantitative
comparison between IR and radar observations were made. Although the authors only provide
results for four months, they also show considerable differences in the frequency of penetrating
deep convection when evaluated from tropical deep convective clouds verses tropical deep
convective rain, reported as 5% and 1.5% respectively.
Liu et al. [2007] used observations from the TRMM PR and VIRS between 35 N – 35 S
and over the periods 1998-2001 and 2003-2004. In their study the relative frequency distribution
of 20 dBZ radar echo heights at 6 km, 10 km, and 14 km were compared with the frequency
distribution of cold cloud features, defined by a minimum of 4 VIRS pixels with cloud top
brightness temperatures colder than 235 K and 210 K. Liu et al. [2007] show that the distribution
of cold cloud features 210 K was most highly correlated with 20 dBZ radar echo tops at 6 km
and they report that only 1% of cold cloud features 210 K had 20 dBZ echo top heights > 14
km. Liu et al. [2007] also show that 20 dBZ echoes reaching 14 km are concentrated over the
tropical land regions of central Africa and equatorial South America, rather than over the tropical
west Pacific and Indian Oceans where IR studies typically show that deep convective clouds
reaching the base of the TTL are most frequent. While insight from this study provides statistical
evidence between radar and IR observations of penetrating deep convective clouds, the 20 dBZ
echo top criteria used by Liu et al. [2007] is more rigid than the criteria used by Alcala and
Dessler [2002]. As already discussed in Chapter 2, the TRMM PR is sensitive to larger sized
particles. Given these details and the criteria that penetrating tops must have 20 dBZ echo tops >
38
14 km, it is likely the analysis provided by Liu et al. [2007] does not consider all deep
convection reaching the TTL.
To evaluate the potential insights of penetrating deep convection from CloudSat, Luo et
al. [2008] combined several CloudSat standard data products (ECMWF-AUX, MODIS-AUX,
2B-GEOPROFL, 2B-Lidar-GEOPROF, and 2B-CLDCLASS) to develop a temperature-height
classification scheme for tropical (15 N-15 S) penetrating deep convective clouds observed
during 2007. The height classification scheme revealed three classes of penetrating deep
convection; warm-high, cold-high, and cold-low. The warm-high class is defined by penetrating
deep convective clouds with cloud tops that are warmer and higher than the cold point
tropopause (CPT). The cold-high class is defined by penetrating deep convective clouds that are
colder and higher than the CPT and the cold-low class is defined by penetrating deep convective
clouds with cloud tops that are colder but lower than the CPT.
Given these three classes, Luo et al. [2008] suggest that deep convective clouds with
cloud brightness temperatures colder than the CPT do not always determine the occurrence of
penetrating and overshooting deep convective clouds. The authors also reveal that the warm-high
class dominates as ~ 47% of all tropical deep convection reaching 14 km. They interpret the
warm-high class to be due to a geometrically thick depth of small ice crystals contained in the
upper portions of the deep convective cloud. In this scenario, the IR emission temperature of the
warm-high class comes from deeper within the cloud. This allows the IR brightness temperature
to be warmer than the CPT while the hydrometeor cloud top height, which corresponds to the
height level of the small ice crystals, is higher than the CPT. While Luo et al. [2008] also relate
each cloud class to life cycle stages of a deep convective cloud, their study does not evaluate any
other properties of penetrating deep convective to better determine how well observations of
39
penetrating deep convection from CloudSat are spatially resolved from conventional passive
sensor satellites.
Although observations of penetrating deep convection used for the temperature-height
classification scheme of Luo et al. [2008] were also used to develop a satellite‐based method to
estimate convective buoyancy (B) and entrainment rate [Luo et al., 2010], these observations are
based on snapshots of thunderstorms and their environments. Results of the cloud top
temperature-height classification scheme are consistent with Setvák et al. [2003], Levizzani and
Setvák [1995], Adler and Mack [1986], Heymsfield and Blackmer [1988], and Heymsfield et al.
[1991] who examined storm cloud top structure.
In these earlier studies that only used IR data, cold-warm brightness temperature couplets
occurred in cold U-shape and V-shape thunderstorm cloud top features suggesting that deep
overshooting convection is often marked by relative maximums in brightness temperature.
However, these studies are related to midlatitude systems, and Luo et al. [2008] investigate
penetrating deep convective clouds within the tropics (15°N-15°S) where such structure has not
been reported. While it is clear that most cross tropopause transport occurs across the TTL,
Mulendore et al. [2003; 2005] show that midlatitude systems can reach the lowermost
stratosphere via upward diabatic transport from the midlatitude troposphere. Although
convection in midlatitudes may account for only a small percentage of the mass of tropospheric
air mixed into the lowermost stratosphere, this pathway should not go unchecked.
In this chapter of the thesis, observations of penetrating and overshooting deep
convection from CloudSat/Calipso and Aqua MODIS are obtained to address the following:
1. What additional insight on the characteristics of penetrating and overshooting
40
deep convection does CloudSat provide with regard to cross tropopause transport via
tropical and extratropical deep convection?
2. What is the interpretation of CloudSat observations of penetrating deep
convection when viewed from passive sensor satellite radiometers according to
traditional IR techniques?
3. What is the areal size distribution of penetrating deep convection observed from
CloudSat and what does it tell us about the capabilities of conventional passive sensor
satellite radiometers to distinctly resolve these events?
3.2 Application of CloudSat, Calipso, and Aqua MODIS Data Products
CloudSat/Calipso and Aqua MODIAS data products described in Chapter 2 are used for
the year 2007 and the region between 35°N and 35°S. The CloudSat 2B-CLDCLASS product is
used to determine the locations of deep convection. According to the 2B-CLDCLASS product,
deep convective clouds have cloud base heights between 0-3 km, a horizontal dimension of 10
km, a thick vertical dimension, liquid water path (LWP) > 0 and as suggested by their reflectivity
values (i.e. 10 dBZe at 5 C), an intense shower of rain or hail is possible. For each profile of
deep convection that is detected, the 2B-GEOPROF radar product is used to capture hydrometeor
cloud top height and cloud base height. The vertical range gates where cloud top height and
cloud base heights are detected must have cloud masks values ≥ 30, indicating that relatively
strong radar echoes are present. The cloud‘s vertical boundaries are further characterized using
the 2B-Lidar-GEOPROF product which integrates CloudSat with the characteristics of the
Calipso Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP). The 2B-Lidar-GEOPROF
product provides the most accurate estimate of hydrometeor layer base and top for up to five
layers in each vertical CPR profile, allowing observations of deep convection to be limited to
41
atmospheric columns with only one cloud layer. Penetrating and overshooting deep convective
clouds are obtained from deep convective clouds with radar cloud top heights 13 km and lidar
cloud top heights 14 km.
To assess IR-based methods traditionally used to detect penetrating and overshooting
deep convection, the infrared window (IRW) brightness temperature, which serves as the cloud
brightness temperature, will be examined. To determine if +BTD signatures (introduced in
Chapter 1) occur for penetrating deep convective clouds, the BTD signature is examined
according to the relation:
BTD BT6.7 BT11
where BT6.7 represents brightness temperatures at the 6.7 m infrared water vapor absorption
band and BT 11 represents brightness temperatures in the infrared window (IRW) at 11 m.
Brightness temperatures in the IRW and IR water vapor absorption band are derived using the
CloudSat MODIS-AUX product, which provides a 3 km x 5 km (across track x along track)
subset of Aqua MODIS-L1B data along the narrow dimension of the CloudSat orbit (see Figure
2.3 described in Chapter 2). In accordance with Fujita [1992] and Brunner et al. [2007] who note
that the geometrical size of an overshooting top is usually less than 10 km in diameter, the entire
3 km x 5 km subset provided by the CloudSat MODIS-AUX product is not used. Instead,
brightness temperatures are based on the mean 3 km x 3 km subset of the MODIS-AUX data that
overlaps and surrounds the CPR footprint. Although Aqua MODIS data is given at multiple
wavelengths, IRW and BTD signatures are calculated using Aqua-MODIS radiances from band
31 at ~11 m and band 27 at 6.7 m. In this calculation, radiances are scaled and converted to
brightness temperatures using the inverse of the Planck function. All observations are recorded
with date, time, and geolocation information.
3.1
42
In our final stage of evaluation, observations from the CPR are expanded from their
conventional use in providing cross sectional profiles, to a more focused view of each CPR
footprint. This application of the CPR is adapted here since penetrating deep convective clouds,
which include overshooting tops, are often as small as 1 km2. Thus, the CPR is not just
measuring a vertical cross-section but rather, the complete or near-complete volume of the
penetrating top. Since the CPR has an instantaneous footprint of 1.8 km x 1.4 km (along track x
across track) each observation from the CPR corresponds to a geometrical area of 2.52 km2. By
grouping consecutive CPR profiles of penetrating deep convective clouds and using the CPR
along track dimension of 1.8 km, two approaches are used to estimate the size distribution of
penetrating and overshooting tops. In the first approach, the area of the instantaneous CPR
footprint is used to estimate the areal size distribution of penetrating tops. In the second
approach, we assume that the penetrating top has the geometrical structure of a uniform plume
where the length of its x and y cross-sections is equivalent to the length given by the number of
consecutive footprints.
3.3 Results from CloudSat Observations of Penetrating Deep Convection
CloudSat observations are used to validate that penetrating and overshooting deep
convection are observed, provide evidence of their vertical extent and characterize their
geospatial and seasonal properties, areal size distribution, hydrometeor cloud top height, IRW
brightness temperature, and BTD signature.
3.3.1 Evidence of Vertical Extent
In this analysis, 736,443 CPR profiles were identified as deep convective clouds with
radar-lidar cloud top heights 14 km. To provide evidence of their vertical extent, a
concatenated image of 10,000 of the 736,443 CPR profiles is provided in Figure 3.1. Close
43
inspection of Figure 3.1 shows that many of these profiles reach higher than 15 km. Another
obvious feature of this image is a broken pattern of strong echo returns near the surface that is
due to rather strong attenuation of the radar beam. According to Battaglia and Simmer [2008],
strong attenuation can be assessed by examining each radar reflectivity profile. Some of these
profiles are provided in Figure 3.2 where surface echoes have been divided into four categories
to highlight differences between profiles with strong surface echoes (20 to 30 dBZe), moderate
surface echoes (10 to 20 dBZe), weak echoes (6 to 10 dBZe), and very weak echoes (< 6 dBZe).
Among all profiles, strong surface echoes are present in 47%. Surface echoes ranging from
strong to weak are present in 78.03% and 21.96% have very weak surface echoes (to no surface
echo at all). The presence of a surface echo suggests that the radar beam is not completely
attenuated. Despite multiple scattering, radar profiles with surface echoes can be used to
sufficiently estimate the cloud base height to more fully evaluate the vertical extent of each
penetrating deep convective cloud.
For observations provided in Figure 3.1, Table 3.1 provides multi-satellite statistics
corresponding to cloud base height, cloud top height, cloud brightness temperature, BTD
signature, maximum radar reflectivity (
dBZe
max), and number of cloud layers for CloudSat
observations of deep convection reaching and penetrating 14 km and 16.9 km over 20°N - 20°S
and 35°N - 35°S. As indicated by the number of cloud layers, observations included in this
analysis are uncontaminated by multilayer clouds. This includes thin cirrus distinguishable by
the Calipso lidar. Differences between mean cloud top height and cloud base height show strong
vertical development suggestive of deep convective cores. As indicated by Weisz et al. [2007],
the mean radar-lidar cloud top height is greater than the radar height. This is expected due to the
differences in measuring capabilities between Calipso and CloudSat (see Chapter 2). For
44
observations at 14 and 16.9 km, the total number of observations between 35°N - 35°S are ~
12.6% more than the total number between 20°N - 20°S. While these observations show that
cross tropopause transport occurs not only in the tropics but also the extratropics, the
characteristics of penetrating deep convection in extratropical regions differ negligibly from
most of the statistics between 20°N - 20°S.
Between 20°N - 20°S and 35°N - 35°S, the mean cloud brightness temperature of deep
convective clouds reaching 14 km and 16.9 km is 213.5 K and 197.4 K respectively. The mean
maximum reflectivity,
dBZe
max, from ~ 2 km to ~ 20 km is ~13.25 dBZe. The BTD signature of
deep convective clouds at 14 km is slightly negative at -0.5 while deep convective clouds
reaching 16.9 km show more positive values. When compared with penetrating deep convective
clouds 14 km, deep convective clouds with cloud tops 16.9 km have +BTD values due to
their close proximity to the cold point tropopause. According to Schmetz et al. [1997], +BTD
signatures are likely characteristics of deep convective clouds with rather high cloud top heights
and is supported by this result. However, Schmetz et al. also show that positive BTD depends on
the amount of water vapor above the tropopause region and on the temperature lapse rate in the
stratosphere. For deep convective clouds with very strong updrafts, large BTD values may also
be due to the injection of water vapor into the lower stratosphere.
3.3.2 Geographical and Seasonal Distribution
The geographical frequency distributions of penetrating deep convection at 14 and 16.9
km are provided in Figures 3.3a and 3.3b. As expected, dominant regions in the distribution are
associated with the large-scale dynamical structure of the tropical atmosphere [Webster and
Chang, 1988]. The highest densities are found over the Indian and west tropical Pacific Oceans,
South and central America and Africa. Although only ~ 4% of all penetrating deep convection
45
between 20°N-20°S and 35°N-35°S reaches 16.9 km, the largest contributions to the 16.9 km
distribution occur over the tropical west Pacific Ocean and Indian and Australian Monsoon
regions.
Figure 3.3a differs from Liu et al. [2005, 2007], who used the Tropical Rainfall
Monitoring Mission (TRMM) precipitation radar to evaluate 20-dBZe echo-top heights reaching
14 km and showed that the largest fractions of deep convection reaching 14 km occur over
continental regions including Africa and South America. Differences between CloudSat and
TRMM are likely due to the 1:30 a.m./p.m. equatorial crossing of CloudSat, which prevents the
CPR from sampling the complete diurnal cycle of deep convection. Incomplete sampling of the
diurnal cycle and differences in the detection capabilities of TRMM and CloudSat likely produce
discrepancies between the two datasets. Although the extent to which a full sampling of the
diurnal cycle would modify the results of Figures 3.3a and 3.3b is unclear, the geographical
distribution of penetrating deep convection from CloudSat is consistent with IR studies (e.g.
Gettelman et al., 2002; Rossow and Pearl, 2007) that are based on synoptic observations that do
fully sample the diurnal cycle.
The seasonal distribution of penetrating deep convection from CloudSat is shown for
December-February (DJF) in Figure 3.4a and for June-August (JJA) in Figure 3.4b. The
distribution of penetrating deep convection follows the migration of the ITCZ by shifting south
of the equator near 10°S in January to north of the equator near 15°N in July. By characterizing
the vertical extent and geospatial and seasonal distributions of penetrating deep convection these
observations provide evidence on the nature of the systems contained in this analysis.
46
3.3.3 Passive Sensor Cloud Top Brightness Temperature and BTD Signature
Since penetrating deep convection distinguished from passive sensor satellite radiometers
are based on +BTD signatures and cloud brightness temperatures these observations are provided
with respect to cloud top height in Figures 3.5a and 3.5b. As already described, the mean cloud
brightness temperature of all penetrating deep convective clouds is ~ 213.5 K. However,
relatively warm cloud brightness temperatures that lead to highly negative BTD signatures are
also observed. As shown in Figures 3.5a-3.5b penetrating deep convective clouds with relatively
warm cloud tops produce a range of cloud top brightness temperatures compared with height.
This wide range makes the evaluation of cloud top height with cloud top temperature and BTD
signature obscure. To minimize such obscurities, temperature thresholds are applied in Figures
3.5c─3.5f to delineate clouds that may be colder than the base of the TTL. By applying
brightness temperature thresholds of 210 K (Figures 3.5c and 3.5d) and 200 K (Figures 3.5e and
3.5f) the observations become more consistent with the traditional view that colder clouds reside
at higher levels. After the 210 K brightness temperature threshold is applied the mean cloud top
temperature within the distribution becomes colder with a value of ~ 203 K and BTD signatures
become more positive with a value of 1.1 K. As described in Section 3.3.1 the coupling of
positive BTD signatures and relatively low cloud top brightness temperature thresholds appears
to be a reasonable approach for the detection of penetrating deep convection. Table 3.2 shows
the percent occurrence of penetrating deep convective clouds detected using IRW brightness
temperature thresholds of 235 K, 210 K and 200 K and +BTD signatures. According to Table
3.2, within the tropics (20ºN - 20ºS) 56.6% of all deep convective clouds reaching 14 km have
IRW brightness temperatures between 235 K and 210 K, ~39.89%(10.35%) have IRW
brightness temperatures ≤ 210 K(200 K) and ~59.42% exhibit the +BTD signature.
47
While Chung et al. [2008] show that 96% of all cases of deep convective clouds with
cloud top heights > 14 km and cloud depths > 6 km exhibit the +BTD signature, it is not clear if
warm high clouds were detected in their examination or how these clouds were treated.
However, more positive BTD signatures are obtained for specific observations of penetrating
deep convection when brightness temperature constraints are also applied.
3.3.4 Areal Size Distribution
As described in Section 3.2, two approaches are used to estimate the size distribution of
penetrating and overshooting tops. In the first approach the areal extent is calculated from the
area associated with the number of consecutive CPR footprints. In the second approach, the
penetrating top is assumed to be a uniform plume where the length of its x and y cross-sections
are both equivalent to the length given by the number of consecutive footprints.
The areal size distributions corresponding to these two approaches are provided in Table
3.3. Results provided in Table 3.3 show rather large differences in areal size distributions
calculated using the CPR footprint verses the uniform plume assumption. However, since
penetrating tops are spot-like features according to visible and IR imagery (e.g., Bedka and
Minnis, 2010), the uniform plume assumption is reasonable. According to this criteria, deep
convective clouds detected within 2-4 CPR footprints, where 4 footprints is the maximum, have
an area of 51.84 km2. Penetrating deep convection observed over more than four consecutive
CPR footprints (i.e., 5-7, 8-10, etc.) are calculated similarly. The combination of the information
provided in Table 3.3 with the relative areal size distribution of penetrating tops provided in
Figure 3.6 shows that 35% of all penetrating deep convection are observed over one CPR
footprint while 15.34% are detected within 2-4 CPR footprints. These statistics are provided in
Table 3.3 and gives some indicate that 65-49% of penetrating deep convection are resolvable
48
using historical IR observations with resolutions of 10 km x 10 km. Estimations of the areal size
distribution of penetrating tops provide a key element needed to evaluate the capabilities of
conventional passive sensor satellite radiometers to resolve penetrating deep convective clouds.
However, due to random sampling, it is unlikely that the CPR provides reflectivity profiles
corresponding to the center of each deep convective cloud. It is more likely that the observations
are taken at varying ranges of the outer chord. According to Table 3.4 the estimation of the
plume‘s areal size for lengths corresponding to an outer chord, provides areal sizes even smaller
than the true dimension of the uniform plume. Thus estimations of the areal size distribution
provided here give a lower limit of the areal size of each penetrating top, which further supports
the view that a considerable fraction of penetrating deep convective cloud tops should be
detectable from the historical IR observations.
This information provides a better understanding on the capability of historical IR
observations to detect penetrating deep convection so that the data may be used to develop a
cloud climatology of penetrating and overshooting deep convection events that may have strong
impact on variability of lower stratospheric water vapor.
3.4 Summary and Discussion
Results given in this chapter show that CloudSat provides unprecedented observations on
the vertical extent of penetrating deep convective clouds. To capture deep convection the 2B-
CLDCLASS product has primarily been used. This product classifies deep convective clouds
according to reflectivity profiles with 10 dBZe echoes present at 5 C. When combined with the
restriction that radar-lidar cloud top heights must be 14 km, this product serves well in
providing a database of penetrating deep convection that may be used for further analysis and
study.
49
In the remaining sections of this discussion, the results of this study are evaluated to
outline the general characteristics of penetrating deep convection and specifically address the
questions provided in Section 3.1
What additional insight on the characteristics of penetrating and overshooting deep
convection does CloudSat/Calipso provide with regard to cross tropopause transport via
tropical and extratropical deep convection?
Using CloudSat/Calipso observations deep convection that penetrates the TTL but deep
convection that reaches 14 km has been differentiated from deep convection that reaches even
greater altitudes. From the results presented only ~ 4% of all penetrating deep convection reach
16.9 km. However, the impact these very high reaching deep convective clouds may have on the
moisture and heat distributions of the upper troposphere and lower stratosphere is not quite clear.
While several studies are able to reproduce lower stratospheric water vapor distributions
consistent with direct convectively driven injection of air mass into the TTL followed by slower
radiative ascent during which air moves quasi-horizontally and passes through the tropopause
observational evidence for direct and irreversible injection into the lower stratosphere (e.g.,
Alcala and Dessler, 2002), is not well documented. CloudSat observations of deep convection
with cloud top heights 16.9 km may be used to further evaluate this pathway. They also
provide a quantitative measure regarding the frequency of these events. In addition, observations
of high reaching deep convective clouds may be used to determine the most suitable locations for
regional studies.
Characterization of the seasonal and geospatial properties of penetrating deep convection
provided by CloudSat show that the largest contribution occurs over the western tropical Pacific
and Indian Oceans. This result is consistent with the view of penetrating and overshooting deep
50
convection from IR studies (e.g., Alcala and Dessler, 2002) and also the recent modeling work of
Hoskins et al. [2010]. However, it is unclear how this distribution would change if the diurnal
cycle of deep convection sampled from CloudSat were more fully resolved. Although the high
LNB over the western tropical Pacific likely comes into play when considering deep convective
clouds with cloud tops 14 km, the approach used in this analysis is consistent with other
several other studies evaluating deep convection that reaches the TTL (e.g., Liu et al. 2005;
2007) . Among all observations between 35°N and 35°S ~87% occur within the tropical region
between 20 N-20 S. This result shows the dominance of cross tropopause transport via tropical
penetrating deep convection but also provides a quantitative assessment on the contribution to
cross tropopause transport from the extratropics. This statistic is however likely to decrease if the
14 km height criteria were adjusted to compensate for lower tropopause heights that are
consistent with extratopical regions.
What is the interpretation of CloudSat observations of penetrating deep
convection when viewed from passive sensor satellite radiometers and according to traditional
IR techniques?
Results from the combined CloudSat/Calipso and Aqua MODIS analysis have been used
to characterize penetrating deep convection from passive sensor observations and according to
traditional IR approaches. According to the results shown here, ~40% of all penetrating deep
convective clouds are associated with cold cloud features 210 K and ~60% of penetrating deep
convective clouds exhibit the positive BTD signature. Both techniques sample rather significant
portions of penetrating deep convection justifying their application in many IR studies. The
extent to which cold cloud features 210 K and +BTD signatures represent penetrating deep
convective clouds is still unclear. It is still unknown the BT11 210 K signal represents
51
penetrating deep convection, which may moisten the lower stratosphere, or if the signal primarily
represents cirrus-anvil cloud fractions, which may have an opposing effect on the lower
stratospheric water vapor budget (e.g., Jensen et al., 1996, 2004).
Liu et al. [2007] report that only 1% of cold cloud features 210 K were associated with
penetrating deep convection having 20 dBZ echoes reaching 14 km. However, CloudSat and
TRMM radar characteristics are different and will likely reveal different statistics since the IR
emission cloud top is typically situated above TRMM PR echo tops and below CloudSat echo
tops. If the capabilities of the TRMM PR only allowed Liu et al. [2007] to weakly sample
penetrating deep convective cloud tops than the statistics reported by that study serve a very
limited view and provide a rather incomplete picture.
Fu et al. [1990] evaluated deep convective clouds and cirrus/anvil clouds using satellite
infrared radiances from ISCCP D1 data obtained for January 1984 and July 1983. While the
authors report that it is not possible to select thresholds that admit all deep convective clouds
while filtering out all other cloud types, their study suggests that relatively low cloud top
temperatures is a useful way to isolate deep convective clouds. Their study also suggests that a
larger fraction of clouds with infrared brightness temperatures 210 K are associated with deep
convective clouds which are likely reaching 14 km than suggested by Luo et al. [2007].
However, unlike the latter study, Fu et al. [1990] does not provide a quantitative description on
the fraction of deep convective clouds reaching 14 km. This issue will be further addressed in
Chapter 4.
What is the areal size distribution of penetrating deep convection and what does it
tell us about the capabilities of conventional passive sensor satellite radiometers to distinctly
resolve these events?
52
To evaluate the capabilities of passive sensors to detect penetrating deep convection, it is
not only important to assess traditional IR techniques but it is also important to evaluate how
well penetrating deep convective clouds may be spatially resolved. The latter is especially
important considering that most passive sensor sounders and imagers do not have spatial
resolutions as high as Aqua MODIS, which provides observations in its emissive bands at a
nominal resolution of 1 km.
To evaluate the capability of conventional passive sensor satellite radiometers to spatially
resolve penetrating deep convection, it is assumed that the shape of penetrating and overshooting
deep convective clouds are uniform plumes with equivalent dimensions along x and y axes.
Using this assumption, 49% to 65% of penetrating deep convective clouds have areas of 51.84
km2. This indicates that for the historical IR observations with horizontal resolutions of 10 km x
10 km, which will be used in Chapter 5, a considerable fraction of penetrating and overshooting
deep convective clouds is resolvable. Rossow and Pearl [2007] provide the only other study to
report the size distribution of deep convection entering the TTL. However they use passive
sensor observations that do not directly resolve cloud vertical structure or cloud top height. They
also do not directly identify the areal sizes of penetrating tops but rather the sizes of systems with
pixels defined by BT11 < 220, BT11 < 200 K and BT11 < CPT. Rossow and Pearl [2007] suggests
that penetrating and overshooting deep convective clouds do not occur in systems with radii < 37
km. However, smaller systems occurring over the western tropical Pacific may also be associated
with differences in observations of penetrating tops in the evaluation of CloudSat observations
and TRMM. In Figure 3.7, a CloudSat cross section centered over Indonesia on October 22,
2006 shows an isolated deep convective cloud with diameter of ~ 20 km and radii of ~10 km that
contains an overshooting top.
53
Given these results, the analysis of penetrating deep convective clouds shown here
provides additional evidence on the characteristics of penetrating deep convective clouds
compared with other instruments and studies. Collocated CloudSat/Calipso and Aqua MODIS
observations provide more context regarding the interpretation of penetrating deep convective
from IR sensors. The combined radar/lidar and IR analysis shows that most penetrating deep
convective clouds have IR signatures that are consistent with traditional IR techniques and that
penetrating deep convective clouds have areal extents large enough to be resolved from
conventional spaceborne passive sensor radiometers.
54
Table 3.1: Multi-satellite mean statistics for observations of deep convection reaching and
penetrating 14 km and 16.9 km over the regions between 20°N - 20°S and 35°N - 35°S. The
normalized frequency distributions corresponding to these observations are provided in Figures
3.3a and 3.3b.
†aCTT=Cloud Top Temperature, †bCTH=Cloud Top Height, †cCBH=Cloud Base Height
Table 3.2 Percent of occurrence of deep convection reaching 14 km and 16.9 km for cloud
brightness temperatures at varying brightness temperature thresholds and positive BTD from
20°N - 20°S and 35°N - 35°S.
CloudSat/CALIPSO-MODIS
20°N - 20°S
35°N - 35°S
Cloud Properties
CTH ≥14 km
CTH ≥16.9 km
CTH ≥ 14 km
CTH ≥ 16.9 km
Number of Observations
643,716
26,840
736,443
30,528
DC(IRWBT ≤ 235 K)
96.49
100.0
96.42
99.19
DC(IRWBT ≤ 210 K)
39.89
94.88
38.98
94.76
DC(IRWBT 200 K)
7.8
55.76
10.35
72.65
DC(+BTD)
59.42
91.53
58.96
91.63
20°N - 20°S
35°N - 35°S
Cloud Properties
CTH ≥ 14 km
CTH ≥ 16.9 km
CTH ≥ 14 km
CTH ≥ 16.9 km
# of Observations
643,716
26,840
736,443
30,528
# of Cloud Layers
1
1
1
1
†a (K)
213.47
197.12
213.69
197.38
†b radar-lidar (km)
15.82
17.31
15.77
17.32
†c radar-lidar (km)
0.88
0.91
0.90
0.94
radar (km)
14.59
16.60
14.58
16.62
radar (km)
0.88
0.91
0.90
0.94
BTD (K)
-0.47
1.246
-0.49
1.253
dBZe
max
13.207
14.278
13.254
14.269
55
Table 3.3: Framework for size distribution of penetrating tops provided in Figure 3.6.
# of Consecutive CloudSat
Footprints
CPR Equivalent Area (km2)
Area of Uniform Plume (km2)
1
2.52
3.24
2-4
10.08
51.84
5-7
17.64
158.64
8-10
25.2
324
11-13
32.76
547.56
14.17
42.84
936.36
Table 3.4 Ratio of Plume Diameter as a function of varying chord lengths.
Chord Ratio
Chord 1
Chord 2/Chord1
Chord3/Chord1
Diameter
x
.9x
.75x
56
Figure 3.1: Concatenated reflectivity in dBZe of 10,000 of the 736,443 CPR profiles
found with radar-lidar cloud top heights 14 km and radar heights 13 km.
Figure 3.2: Reflectivity profiles of 200 of the 736,443 observations of penetrating deep
convection classified by the strength of their surface echoes (∆dBZe) with strong surface echoes
having ∆dBZ > 20 dBZe; moderate echoes between 10 dBZe and 20 dBZe; weak echoes
between 6 dBZe and 10 dBZe ; and very weak echoes < 6 dBZe.
Strong Moderate Weak Very weak
57
Figure 3.3: Frequency distribution of Cloudat/Calipso cloud tops a) 14 km and b) 16.9
km over 35°N ─35°S for 2007 organized according to 2.5° x 2.5° bins where the total of
each distribution is 100%.
58
Figure 3.4: Same as Figure 3.3 except distribution is for a) June-July-August (JJA)
and b) December-January-February (DJF).
a)
b)
59
:
Figure 3.5 Radar-lidar (black) and radar (alternating color) cloud top height verses a) brightness temperature difference for all 736,443
observati ons b) cloud brightness temperatures for all 736,443 observations c) same as a) but with 210 K cloud brightness temperature
constraint d) same as b) but with 210 K cloud brightness temperature constraint e) same as a) but for cloud brightness temperatures 200 K
and f) same as b) but for cloud brightness temperatures 200 K.
IRW Cloud Brightness Temperature (K)
IRW Cloud Brightness Temperature (K)
IRW Cloud Brightness Temperature (K)
Brightness Temperature Difference (K)
Brightness Temperature Difference (K)
Brightness Temperature Difference (K)
Cloud Top Height (km)
Cloud Top Height (km)
Cloud Top Height (km)
Cloud Top Height (km)
Cloud Top Height (km)
Cloud Top Height (km)
60
Figure 3.6: Areal size distribution in km2 of penetrating deep convection with
consecutive CPR footprints reaching 14 km. The distribution also corresponds to the
criteria provided in Table 3.2 where the area associated with the penetrating deep
convective cloud top is provided according to the maximum number of consecutive
footprints associated with the CloudSat CPR footprint.
Fraction of Consecutive Cloud Tops
≥ 14 km
61
Figure 3.7: Top) CloudSat cross-section of a penetrating deep convective cloud (labeled
A) with cloud top heights > 15 km and diameter of ~ 20 km. (Bottom left) Aqua MODIS
(L1B) true color image with the CloudSat track corresponding to the top panel is shown
in yellow and (Bottom right) Aqua MODIS cloud optical thickness from the level 2 cloud
product with the CloudSat track also shown in yellow.
A
A
62
CHAPTER 4
EVALUATION OF TRADITIONAL IR TECHNIQUES
4.1 Introduction
Prior to the application of space-borne radar for the evaluation of deep convection
reaching the TTL, several IR studies (e.g., Gettelman et al., 2002, Rossow and Pearl,
2007) evaluated the tropical frequency distribution of deep convection reaching the TTL.
While those studies focused on assessing regional dominance and seasonal patterns (e.g.
Rossow and Pearl, 2007), few studies evaluate the affects of penetrating deep convection
on long-term trends in lower stratospheric water vapor.
To address this relationship, it is first important to assess the climatological
properties of penetrating deep convection. While this is a challenge, it is not impossible.
Visible and infrared observations have been collected and archived for nearly 30 years
from geostationary orbit [Knapp, 2008a]; a wealth of data suited for climatological
studies exist. Yet the potential information that may be obtained from such
measurements is unclear. As shown in Chapter 3, the horizontal resolutions of the
observations obtained from geostationary orbit will often limit the extent to which
penetrating deep convection may be resolved. In addition to this limitation, brightness
temperatures, which are most often used to define deep convection and penetrating deep
convection, do not give an exact measure of cloud top height nor do they provide
evidence of cloud vertical structure. Moreover, traditional IR techniques, which include
cold cloud features and +BTD signatures, have been poorly quantified.
In Chapter 3, penetrating deep convective clouds from CloudSat showed +BTD
signatures when IRW brightness temperature thresholds of 210 K and 200 K were
63
applied. However, there is still much uncertainty associated with the capabilities of
+BTD signatures and the cold cloud feature technique to exclusively sample penetrating
deep convective clouds. As described by Chou and Neelin [1999], cirrostratus and
cirrocumulus are tightly connected with deep convective cloud fractions. When applying
traditional IR techniques that include cold cloud features and +BTD signatures, both
techniques are suggested to largely sample nonraining anvil clouds or thick cirrus in
addition to the convective region. Therefore other approaches have been developed to
detect penetrating deep convection from space-borne radiometers.
Aumann et al. [2011] used hyperspectral observations from the Advanced
Infrared Sounder (AIRS) and the Atmospheric Microwave Sounding Unit (AMSU) to
gain new insights into properties of cold cloud tops by identifying their spectral
differences. The authors determined that the mix of cirrus and deep convective high
clouds can be separated by noting differences between a window (8-11) channel and a
channel with strong CO2 absorption (near 14 m). Aumann et al. found that differences
between channels at these wavelengths distinguish a class of deep convective clouds
characterized by a very high mean rain rate and a local upward displacement of the
tropopause. The authors suggest that improved identification of penetrating cloud tops
could be accomplished by adding one strong CO2 sounding channel on future advanced
geostationary satellites. Yet this approach does little to address the impact of penetrating
deep convection on long-term trends in lower stratospheric water vapor from existing
data.
Other newly developed techniques include objective satellite-based detection
schemes presented by Bedka et al. [2010], which is reportedly less useful for historical IR
64
radiances that have horizontal resolutions 5 km and Berendes et al. [2008], who used
visible radiances that limit the evaluation of penetrating deep convection to daytime
detections only. Given these details, the techniques developed by Bedka et al. [2010] and
Berendes et al. [2008] are considered here to be less appealing to adequately evaluate the
climatological properties of overshooting tops using the historical data.
How then do we reduce and/or quantify the uncertainty associated with
traditional IR methods that are most applicable to the historical observations? To
determine how to best address this point, Alcala and Dessler [2002], Liu et al. [2007],
and Luo et al. [2008] provide useful approaches on how to evaluate penetrating deep
convection from space-borne IR and radar sensors. According to these studies and others
like them, a multisensor view from space-borne sensors is a more accurate way to
evaluate IR-based distributions and techniques used to detect penetrating deep
convection. Such a view incorporates rather direct measures of cloud vertical structure
and cloud top height from radar with passive sensor observations that indirectly provide
cloud top height based on radiative transfer modeling and a priori assumptions of surface
and atmospheric quantities. The benefit of the combined radar-IR approach is that it goes
beyond the traditional top-down view associated with IR-imagers and has already been
shown to yield insightful results (e.g., Luo et al., 2008; Luo et al., 2010; Chung et al.,
2008). However, the scope of IR properties investigated in most radar-IR examinations of
penetrating deep convective clouds limit themselves to IRW brightness temperatures
while other optical and microphysical parameters such as optical depth ( ), cloud top
pressure, effective particle radius, etc., are also available. It is interesting to evaluate
whether the inclusion of additional optical and microphysical properties of cold cloud
65
features/pixels with BT11 210 provides the necessary evidence to better distinguish
between the deep convective cloud core and corresponding anvil/cirrus cloud fractions.
In the assessment of the general macro- and microphysical properties of deep
convection by Yuan and Li [2010], the authors define deep convective clouds with IRW
brightness temperatures less than 243 K and cloud optical thickness greater than 40. Yuan
and Li suggest that these thresholds screen out relatively thick, isolated cirrus clouds not
associated with an active deep convective cloud, but keep portions of the anvil clouds
associated with the deep convective systems. Kubar et al. [2007] suggests that anvil
clouds have cloud tops colder than 245 K and optical depths between 4 and 32. However
these descriptions of anvil clouds are rather arbitrary. Since CloudSat provides vertical
cloud structure, the combination of CloudSat/Calipso and Aqua MODIS observations
may be used to disentangle the description of deep convective cores from the surrounding
cloud fraction to more objectively decipher between anvil and cirrus clouds. Using
CloudSat, Aqua MODIS and the Advanced Microwave Scanning Radiometer for the
Earth Observing System (AMSR-E), Yuan and Houze [2010] quantitatively mapped the
frequency of anvil clouds within mesoscale convective systems (MCS) and non-MCS
anvils over the entire tropics. However, they do not provide a complete range of
microphysical and optical properties with which to distinguish penetrating tops and anvil
clouds from radiometric observations.
Given these details, cold cloud features/pixels and positive BTD signatures are
compared with penetrating deep convective clouds gathered in Chapter 3. The
comparison of these three sets of observations are used to statistically characterize the
extent to which traditional passive sensor approaches are able to exclusively capture
66
overshooting deep convection and to evaluate, to the extent possible, the percentage of
other types of high level clouds (i.e., anvil clouds, pileus clouds, jumping cirrus, lower
level deep convection, etc.) that may be present in the IR distributions. In addition,
Chapter 4 also compares the microphysical and optical properties of penetrating deep
convection with other types of high clouds. By evaluating these properties the capabilities
and information content of IR-based distributions of penetrating deep convection are
better understood and their capability to detect penetrating deep convective cloud tops is
revealed. More specifically, the combination of CloudSat/CALIPSO and Aqua MODIS
observations are used to address the following:
1. To what extent is penetrating deep convection exclusively captured using
traditional IR methods and does this information change the perception of the
frequency distribution of penetrating deep convection as presented in other IR
studies?
2. What are the microphysical and optical properties of penetrating deep convection
in comparison with other high level clouds? How does the incorporation of these
parameters aide in the evaluation of penetrating deep convective clouds using
traditional IR techniques?
To address these questions, section 4.2 provides additional background
information on traditional IR techniques and the studies that have used them to
investigate penetrating and overshooting deep convection. Section 4.3 describes the data
and methods that have been developed to conduct these analyses. In Section 4.4 the
results of the analysis are presented where we focus on the two questions outlined above.
67
Finally, section 4.5 summarizes the chapter and ideas for future work (i.e., a prelude to
Chapter 5).
4.2 Characterization of Penetrating Deep Convection from Spaceborne
IR Data
Much work has been done to detect and characterize the influence of overshooting
deep convection on the tropical tropopause layer and lower stratospheric water vapor
budget (e.g., Jiang et al., 2004; Hong et al., 2005; Heymsfield and Fulton,1988; Cifelli et
al., 2002; Nesbitt et al. 2006; Bedka and Minnis, 2010). While all these studies have
contributed to the understanding of penetrating deep convection, below the focus is on
studies that used observations from space-borne satellite IR imagers/sounders.
Two approaches are commonly used to gauge the frequency and distribution of
penetrating/overshooting deep convection using data from conventional passive-sensor
satellite radiometers. They include, cold cloud features/pixels (e.g., Mapes and Houze
1993; Liu et al, 2007; Rossow and Pearl, 2007) and positive brightness temperature
differences between the water vapor absorption band at ~6.7 µm and the IR window at
~11 µm (e.g., Schmetz, 1987; Soden, 2000; Setvak et al., 2003; 2007; Chung et al.,
2008). While these approaches have been briefly discussed we now offer a more physical
explanation of the two techniques.
The physical basis for the cold cloud feature approach is associated with classical
parcel theory and observations showing that once deep convection overshoots its level of
neutral buoyancy the cloud top is much cooler than its environment [Johnston and
Solomon, 1979]. Consequently, deep convective clouds colder than their level of neutral
68
buoyancy, or some other critical level, have been estimated to have cloud tops that lie at
higher altitudes.
In the second approach, positive brightness temperature differences (+BTD) occur
when deep convective clouds with very strong updrafts force water vapor and ice into the
lower stratosphere. The water vapor that enters the lower stratosphere emits radiation in
the infrared water vapor absorption band (5.7 ─7.1 m) at warmer stratospheric
temperatures while the infrared window (~11 µm) brightness temperature is emitted from
the lower and colder physical cloud top. The +BTD signature associated with
overshooting deep convection was first detected by Fritz and Lazlo [1993] who used the
High Infrared Resolution (HIRS)/TIROS-N Satellite Sounder to monitor very high
clouds. Fritz and Lazlo [1993] also used the atmospheric radiation code (ATRAD) to
make theoretical radiance calculations in effort to physically explain the nature of the
+BTD signature. By placing optically thick clouds at the base of the tropical tropopause
level the authors simulate brightness temperatures in both the IRW (~11 µm) and in the
infrared water vapor absorption band (~6.7 µm) that support their interpretation of the
+BTD signature.
Using cold cloud features/pixels and +BTD signatures from IR observations,
many details of the global frequency distribution of overshooting convection
distinguished as the coldest deep convective cores and cirrus/anvil cloud fractions have
been presented. To disentangle deep convective cores and cirrus/anvil cloud fractions Fu
et al. [1990] used ISSCP radiance data for July 1983 and January 1984 over the
equatorial Pacific. Their results show that when cloud brightness temperature becomes
warmer than 220 K the distribution of deep convective cloud cover shifts to lower
69
reflectance which is indicative of cirrus and anvil cloud fractions. Fu et al. [1990] that an
IRW brightness temperature of 220 K is a good threshold to isolate deep convective
clouds. However, Hong et al. [2006] analyzed the diurnal cycle of deep convective clouds
using the infrared threshold technique and suggest that the peak in BT11 210 K and
235 K distributions has a 2-hour lag behind the maximum TRMM PR deep convective
cloud fractions. The authors suggest that this 2-hour lag is a consequence of anvil cloud
shield expansion, indicating that the 210 K distribution is influenced by cirrus-anvil cloud
fractions.
For the evaluation of penetrating tops within the deep convective core, Gettelman
et al. [2002] used a combination of cold cloud features less than the cold point tropopause
temperature to show that tropical overshooting deep convection occur more frequently
over the tropical oceans compared with continental regions and that the highest
frequencies occur over the western tropical Pacific. Gettelman et al. [2002] also
integrated the global frequency distribution of overshooting events to estimate what
percentage of overshooting tops that occur within the larger distribution of deep
convective clouds. Using observations with a 50 km horizontal resolution, their study
suggests that approximately 0.5% (± 0.25%) of all deep convection fits the description of
overshooting deep convection. Yet it is difficult to apply observations with a 50-km
resolution to evaluate updrafts and penetrating towers with diameters that are typically on
the order of 1–10 km. Gettelman et al. state that differences in the scale of penetrating
and overshooting deep convection verses the horizontal resolution of the data introduce
uncertainties when estimating the cloud fraction of overshooting deep convection from
the global cloud imagery.
70
Rossow and Pearl [2007] used IR data from the ISCCP-DX dataset to conduct a
22-year survey of deep convection penetrating the TTL. They show that land events
exhibit a diurnal cycle while oceanic events do not exhibit this same character. Rossow
and Pearl [2007] also add that penetrating deep convection predominantly occurs in the
larger, organized mesoscale convective systems and that the interannual variations in the
seasonal patterns of overshooting events are ―not very large‖. Using 5 km observations
subsampled to a 30 km horizontal resolution they report that ~ 2% of deep convection
penetrates the TTL. Results of IR and radar studies that have sampled penetrating deep
convection and provide estimates of their fractional coverage are given in Table 4.1. As
indicated, the fraction of penetrating deep convection ranges from values as high as 5%
from radar studies [Alcala and Dessler, 2002] to values as low as 0.5% ( .25%) in IR
studies [Gettelman et al., 2002]. With the exception of Liu et al. [2007] none of the
studies given in Table 4.1 quantify what fraction of the cold cloud features/pixels
correspond to penetrating deep convective clouds.
4.3 Data and Methods
The characteristics of penetrating/overshooting deep convective clouds are
evaluated with high/cirriform, deep convection, and anvil clouds for January 2007 over
35 N to 35 S. The same CloudSat products used in Chapter 3 are used to identify all four
cloud types but with a few modifications. Here, the CloudSat 2B-CLDCLASS product is
used to determine the locations of cirriform and deep convection. These two cloud classes
are identified by the 2B-CLDCLASS product as cloud classes 1 and 8, respectively (refer
to Table 2.2). Anvil and penetrating deep convection are not characterized by the 2B-
CLDCLASS product but are subcomponents of the larger deep convective cell. These
71
two cloud classes are identified by considering the vertical and spatial properties of each
deep convective cloud that is captured along the CloudSat profile (see Figure 4.1).
The criteria for penetrating and overshooting deep convection were presented in
Chapter 3. For anvil clouds it is important to consider the distribution of cloud classes
within the deep convective cell as they relate to each vertical range gate. A schematic
view of the classification of the anvil cloud, the penetrating top, and the main body of the
deep convective cloud are shown in Figure 4.1 with examples of vertical and horizontal
range gates given. As indicated in Figure 4.1 anvil clouds are identified when deep
convective clouds (8) are present with high cirriform (cloud class 1) connected to them
and no cloud (0 - not shown) beneath. For each of the four cloud groups the 2B-
GEOPROF product is used to evaluate cloud top height, cloud base height, and cloud
masks values, where cloud masks values 30 are again required to confirm the presence
of the cloud. Each cloud‘s vertical boundaries are further characterized using the 2B-
Lidar-GEOPROF product with the condition that only 1 cloud layer may be present in the
CPR profile. All observations are recorded with date, time, and geolocation information
in order to evaluate other microphysical and optical properties associated with each cloud
group from MODIS-AUX and MAC06S0 products. Data sampling from the MODIS-
AUX product is the same as in Chapter 3. For the MAC06S0 product, which provides
effective radius, cloud top pressure, particle phase, and optical depth, sampling works
differently. As described in Chapter 2, the MAC06S0 product provides 3 pixels by 11
pixels with an approximate resolution of 15 km x 10 km. Given these details, MAC06S0
pixels that are geographically closest to each CloudSat observation are used to a provide
72
an averaged 5 km x 5 km view of effective radius, cloud top pressure, particle phase, and
optical depth.
In addition to gathering observations of the four cloud classes for January 2007,
IR observations of cold cloud features 235 K and 210 K, and +BTD (with BT11 235
K) for all of 2007 are sampled and compared with observations of penetrating deep
convective clouds. In this part of the approach, all MODIS-AUX observations where the
innermost 3 km x 3 km subset has an average IRW brightness temperature (BT11) 235
K and ≤ 210 K is used to identify cold cloud features 235 K and 210 K. Positive
BTD (+BTD) signatures are similarly evaluated. Observations are recorded with time,
date, and geolocation information. MODIS-AUX observations of cold cloud features and
+BTD are then cross-referenced with CloudSat observations of penetrating deep
convection to determine the extent to which traditional IR techniques exclusively sample
penetrating deep convection.
4.4 Observations from Aqua MODIS
In this section, the optical and microphysical properties of penetrating deep
convection from MODIS is compared with other types of high clouds and distributions of
cold cloud features 210 K and +BTD signatures are also characterized.
4.4.1 Penetrating Deep Convective Clouds Compared with Other High Clouds
Mean optical and microphysical properties listed in Section 4.3 are provided in
Table 4.2 for high/cirriform, deep convection (with radar-lidar cloud top heights < 14
km), anvil clouds, and penetrating deep convection where penetrating deep convection is
evaluated according to cloud top heights 14 km and ≥ 16.9 km observed during
January 2007. As shown in Table 4.2, rather strong distinctions associated with each
73
cloud type occur for cloud top pressure, IRW brightness temperature, BTD signature, and
optical thickness. While the definition of high clouds and deep convection from the 2B-
CLDCLASS product is different from the classification of high clouds (i.e., cirrus,
cirrostratus, and deep convection) provided from the ISCCP scheme, the average
statistics of the four cloud types examined in this study suggest that the sampling
technique is highly useful. With the exception of IRW brightness temperatures, which are
reported here in lieu of cloud top temperature, statistical means reported in Table 4.2 are
consistent with values reported in other studies (e.g., Hong et al., 2007; Rossow and
Schiffer, 1999). Additional information on each cloud class has been gathered by
examining how these microphysical and optical properties are distributed about their
means.
In Figure 4.2, the normalized frequency distribution of a) BTD signature, b) IRW
brightness temperature, c) cloud optical thickness ( ), d) infrared water vapor brightness
temperature e) cloud top pressure and f) effective particle radius is given. According to
the BTD distribution, there is general agreement among deep convection, penetrating
deep convection, and anvil clouds where each of these cloud types has peak BTD values
between ± 2 K. For IRW brightness temperatures, penetrating deep convection 16.9
km has a narrow distribution and a peak value around 193 K while penetrating deep
convection 14 km has a broader distribution of IRW brightness temperatures that range
from 190 K to 230 K. Anvil clouds have a peak IRW brightness temperature of 237 K
suggesting that on average, anvil clouds are warmer than penetrating deep convection
although some anvil clouds do have relatively low temperatures where penetrating deep
convection ( 14 km and 16.9 km) dominates. In comparison to all deep convection,
74
observations of cirrus clouds have relatively warm IRW temperatures (i.e., BT11 > 220 K)
when compared with all of deep convection and anvil clouds.
For each of the four cloud classes the spectrum of optical thickness is skewed
towards low values. For cirrus, anvil clouds, deep convection and penetrating deep
convection the mean optical depth is 6.8 13.7. While low optical thickness is
characteristic of cirrus and even anvil clouds, low optical thickness is not characteristic of
deep convective clouds. This aspect of the observations likely represents a problem with
spatial sampling since optical thickness from the MAC06S0 product is reported at a 5 km
resolution and deep convective clouds typically have optical thickness > 23 (see Figure
2.2). While the 5-km spatial resolution may be too large to accurately identify the
properties associated with relatively small deep convective cells, as in the case of Figure
3.7, optical depth measurements may be very useful in characterizing penetrating deep
convective clouds from IR observations. On the other hand, these rather low optical depth
values may be due to the criteria for deep convective clouds that is utilized in the 2B-
CLDCLASS product (see Chapter 2). Therefore the relationship between IRW brightness
temperature and optical thickness is evaluated in Figure 4.3. This figure shows that IRW
brightness temperatures for both cirrus and deep convective clouds are typically warmer
when optical thickness is low. Approximately 79% of all IRW brightness temperatures >
235 K had optical depths less than 23 while this value is 65% for all IRW brightness
temperatures 235 K. Since these two statistics are similar the oddities in optical depth
may again be due to the 5 km sampling.
The distribution of IR water vapor brightness temperature in Figure 4.4 shows the
lowest peak values for penetrating deep convective clouds. Distributions of cloud top
75
pressure show that the majority of penetrating deep convection ≥ 16.9 km have cloud top
pressures ~120 mb. As in the case of IRW brightness temperature, penetrating deep
convection ≥ 14 km also has a broad distribution in cloud top pressure. The distribution
of effective particle radius is somewhat similar for all cloud types although the largest
values occur for penetrating deep convection ( 14 km and 16.9 km).
To further illustrate the variability IRW brightness temperature, IR water vapor
brightness temperature, and +BTD as a function of cloud type from collocated Aqua
MODIS and CloudSat measurements provided in Figure 3.7 (shown in Chapter 3) is used
again and are provided in Figure 4.4. As shown in Figure 4.4, CloudSat granule ID 2757
is used to highlight observations of a stratocumulus cloud (labeled A), deep convective
clouds (labeled B and C) and a cirrus cloud (labeled D). As indicated, penetrating deep
convective clouds are the only cloud types within the cross section with IRW brightness
temperatures 210 K. Anvil clouds are at levels slightly below 14 km. Although the
anvil portion of the deep convective cloud does not have IRW brightness temperatures
below 210 K, some observations of anvil clouds do have IRW brightness temperatures
that are this low.
The BTD profile corresponding to the CloudSat cross section shows slightly
positive values for the stratocumulus cloud (cloud A) referenced in Figure 4.4. Slightly
positive BTD values show that while +BTD signatures are often associated with deep
convective clouds, +BTD signatures can also occur in the presence of other optically
thick cloud types. This suggests that +BTD signatures explained by an increase in water
vapor entering the lower stratosphere due to the transport of water vapor rich air from the
updrafts of deep convective cores, does not suitable in all cases where +BTD signatures
76
occur. More highly positive +BTD signatures may be related to lower stratospheric water
vapor pumping as in the case of the overshooting deep convection cloud (cloud B)
labeled in Figure 4.4. However, more research is necessary to provide evidence that
supports this explanation.
While results provided in this section are only for January 2007, they have been
used to statistically characterize and investigates IRW brightness temperatures and other
microphysical and optical properties of penetrating deep convection with other types of
high clouds. Characterization of cold cloud features and positive BTD signatures are
provided in the next section.
4.4.2 Characterization of cold cloud features and positive BTD signatures
The frequency distribution of MODIS-AUX cold cloud features 235 K and
210 K are provided for 35 N to 35 S and are given in Figure 4.5 along with the seasonal
properties of the 210 K distribution during December-February and June-August.
Regional similarities in the distribution of cold cloud features 210 K largely agree with
the distribution of deep convection penetrating 14 km (Figure 3.3) provided in Chapter 3
and is also consistent with the distribution of +BTD signatures between 35 N – 35 S
provided in Figure 4.6.
Dominant regions in distributions of cold cloud feature ( 235 K and 210 K)
and +BTD are both associated with the large-scale dynamical structure of the tropical
atmosphere [Webster and Chang; 1988]. Regional patterns in Figures 4.5 and 4.6 are
consistent with IR studies (e.g., Gettelman et al, 2002) that show that the highest densities
of deep convective clouds are found in the Indian and west tropical Pacific Oceans, over
South and central America and Africa. In addition, the seasonal distribution of +BTD and
77
cold cloud features 210 K also show the migration of the ITCZ as in the case of
penetrating deep convective clouds shown in Figure 3.5. Although these results provide
a good basis for qualitative comparison, it is important not to limit the observations to
qualitative comparisons only. Without quantitative information that compares cold cloud
feature and +BTD distributions with the distribution of penetrating deep convective
clouds we limit new possibilities for interpretation that might otherwise go unutilized.
4.4.3 Cold cloud features and +BTD compared with CloudSat observations
The spatial and vertical properties of penetrating deep convection are provided in
Chapter 3. Results reported in Chapter 3 Table 3.3 show that within the tropics (20º N-
20º S) 57% of all deep convective clouds reaching 14 km have IRW brightness
temperatures between 235 K and 210 K, ~40% have IRW brightness temperatures ≤ 210
K and ~59% exhibit the +BTD signature. These statistics identify what fraction of the
complete distribution of deep convection reaching 14 km is observed using each
technique. However, they do not quantify what fraction of all cold cloud features and
+BTD are penetrating deep convection. Quantifying this information is necessary to
determine the extent to which cold cloud features and +BTD exclusively sample
penetrating deep convective events. These results are provided in Table 4.3.
Results from Table 4.3 were determined by cross-referencing observations of
MODIS-AUX cold cloud features 235 K and 210 K and +BTD with CloudSat (radar-
lidar) penetrating deep convection. As shown in Table 4.3, the percentage of cold cloud
features 235 K and ≤ 210 K and +BTD that occur as penetrating deep convection
reaching 14 km is ~ 26%, 66% and 55% respectively. For cold cloud features 210 K,
the value of 66% reported here is much higher than the value of 1% reported by Liu et al.
78
[2007], who provide the only other study that used radar observations to address the
relationship between cold cloud features 210 K and penetrating deep convection. Given
these differences it is important to further investigate the validity of the quantitative
results by evaluating the types of high clouds that we expect to observe at BT11 210 K.
Here, CloudSat granule ID 2575 previously provided in Figure 3.7 (Chapter 3) and
Figure 4.4 is further assessed. However, in this case MODIS L1B (MYD021KM) IRW
brightness temperatures sampled to a 5 km x 5 km resolution and corresponding cloud
top temperatures from the level 2 Aqua MODIS Cloud Product (MYD06_L2) are
evaluated and shown in Figure 4.7.
As shown in the images of cloud top temperature and IRW brightness temperature
provided in Figure 4.7, cloud top temperature and BT11 values 210 K and 235 K have
been distinguished by yellow and red contour shading. The image of cloud top
temperature has ~6% more pixels with temperatures 210 K and 17% more pixels with
cloud top temperatures 235 K. The difference in the percentage of pixels 235 K and
210 K that exists between the two images is largely based on the CO2 slicing method
which is used to derive MODIS L2 cloud top temperature. Using the CO2 method, a
considerable fraction of optically thin cirrus clouds are better represented in the image of
cloud top temperature. This indicates that thin cirrus, which makeup a considerable
fraction of all high clouds, are radiometrically less visible in the IRW region.
According to Figure 4.7 approximately 0.4% of pixels with IRW brightness
temperatures 235 K are also 210 K. When IRW brightness temperatures 210 K are
compared with the corresponding image of optical thickness (provided in Chapter 3
Figure 3.7) where 23, all pixels with IRW brightness temperatures 210 K occur
79
with pixels that have optical depths 23. This relationship suggests that the majority of
clouds with IRW brightness temperatures 210 K are also deep convective clouds. This
relationship has also been evaluated in Figure 4.8, which shows the normalized frequency
distribution of optical depth ranging from 0 to 100 for all observations of BT11 210 K
sampled from 5 km resolution data from the Aqua MODIS L2 Cloud Product during
October 2007. According to Figures 4.7 and 4.8 the most probable cloud class with IRW
brightness temperatures 210 K is deep convection with rather high optical depth. This
finding is further supported by Table 4.5 which shows that ~85% of all pixels with BT11
210 K are associated with deep convective clouds. While anvil clouds may be optically
thick and classified as deep convection, a statistical evaluation of their macro and
microphysical properties (i.e., optical depth, cloud top pressure, and cloud top height)
have not been reported in any other studies. According to the analysis performed in the
previous section anvil clouds have mean IRW brightness temperatures of 237 21.9 K
and have optical thicknesses of 6.8 13.7. These values indicate that some of the cloud
features observed at IRW brightness temperatures 210 K may be associated with anvil
clouds especially since 30.8 % of pixels with BT11 210 K have optical depths le 34.2.
Given these details, the high fraction (i.e., 66%) of cold cloud features 210 K that are
associated with deep convective clouds with (radar-lidar) cloud top heights 14 km is a
reasonable result. Other IR thresholds are also considered in Table 4.4. These results
show that at sufficiently low IR thresholds, BTD signatures are also positive.
80
4.5 Summary and Discussion
There is no precise definition of a deep convective cloud from passive remote
sensing measurements that admits all penetrating deep convective clouds while filtering
out all other cloud types. However, the fraction of cold cloud features/pixels with BT11
210 K and +BTD signatures that are penetrating tops may be quantified to enhance the
information content and applicability of these techniques for climatological studies. For
cold cloud features 210 K the only existing estimate on the fraction of cold cloud
features 210 K is rather low according to Liu et al., [2007] and rather high for positive
BTD signatures according to Chung et al., [2008]. As a consequence, the distribution of
cold cloud features 210 K and +BTD signatures have been evaluated against radar and
IR microphysical and optical cloud properties associated with penetrating deep
convection and other types of high clouds to evaluate extent to which penetrating deep
convection is exclusively captured using traditional IR methods. According to this
evaluation, ~ 66% of cold cloud features/pixels with IRW brightness temperatures 210
K are associated with penetrating deep convective clouds. This conclusion is further
supported by the evaluation of BT11 210 K pixels with cloud optical depth for October
of 2007 which shows that ~ 86% of all clouds with BT11 210 K have optical depths
greater than 23. Since several studies suggest that IR techniques sample considerable
fractions of thick cirrus and non-raining anvil clouds in addition to the deep convective
region a description of microphysical and optical properties of anvil clouds was provided
in section 4.4.1. According to the observations associated with this cloud type, anvil
clouds may have optical depths of 6.7 13.7, which at two standard deviations is =34.
BT11 210 K with 34.1 is 30.7% of the BT11 210 K distribution and still supports,
81
the evaluation that 66% of cold cloud features 210 K is associated with penetrating
deep convective clouds. Yet for the analysis of these statistics it is also important to note
that optical depth provided from MODIS suffers from a poor dynamical range especially
in comparison to ISCCP which characterizes deep convective clouds with optical depths
as high as 379 (see Figure 2.2) since MODIS cloud optical depth retrievals saturate at
100.
The implication of the cold cloud feature result suggest, for example that IR based
studies which propose that penetrating deep convection is 0.5% 0.25% of all deep
convection (cf., Gettelman, 2002), is a rather low estimate that is more so due to the
horizontal resolution of the observations combined with the spatial scale of penetrating
tops, rather than inadequacy of the cold cloud feature technique. However, it has also
been shown that a considerable fraction of penetrating tops do not have cloud brightness
temperatures 210 K according to the results of Chapter 3 which are consistent with Luo
et al., [2008]. These issues prevent a more accurate estimate of the penetrating top cloud
fraction to be revealed from IR observations especially in comparison to radar studies
that report values as high as 5% [Alcala and Dessler, 2002] (see Table 4.1).
To better understand the types of clouds that are observed from cold cloud
features 210 K and positive BTD signatures, four different types of high clouds were
analyzed. This analysis shows that cirrus, anvil clouds, deep convection and penetrating
deep convection, all have distinctions in optical and microphysical properties that could
be used to better separate penetrating deep convection from other high cloud classes and
has been used to further support the interpretation of the cold cloud feature 210 K
distribution. The evaluation of +BTD, show that these signatures are present in cases of
82
optically thick high clouds, higher +BTD thresholds of (>.5K) are evident for penetrating
deep convective clouds and may serve as a more useful application of BTD signatures.
Although each cloud type was observed to have a range of IRW brightness temperatures
penetrating deep convection ( 14 km and 16.9 km) have the lowest mean IRW
brightness temperatures. This result is in agreement with the cold cores outlined by Yuan
and Houze [2010] who quantitatively mapped the areal extent of MCS anvils, by
distinguishing between the deep convective cloud and surrounding cloud fractions.
An interesting outcome in the analysis of high clouds does however show that
some observations characterized by the 2B-CLDCLASS product as deep convection, do
not have corresponding MAC06S0 optical depths > 23. This suggests that in some cases,
either the 5 km x 5 km spatial resolution is too coarse to provide collocated
measurements of optical thickness and other microphysical and optical properties or the
LWP > 0 (see Chapter 2) criteria may not be suitable enough to specifically sample deep
convective clouds characterized by optical depths 23. According to the analysis of
cirrus and deep convective clouds as a function of BT11 and optical depth, we do not
agree with Hong et al. [2006], who suggest that deep convective clouds and cirrus clouds
with similar cloud top altitude cannot be distinguished. Here, it has been shown that in
general, high clouds with varying degrees of optical thickness will have different IRW
emission temperatures. Cirrus clouds typically have rather warm IRW brightness
temperatures, while deep convective clouds, are a bit more complicated due to the
different dynamic and thermodynamic processes.
In the analysis of cold cloud features 210 K with observations of penetrating
deep convective clouds, the evaluations shown in Chapter 4 suggests that 66% of clouds
83
with IRW brightness temperatures 210 K are associated with penetrating deep
convective clouds. While this statistic is much higher than the values reported by Liu et
al., [2007], observations of BT11 210 K have also been analyzed with observations of
optical depth. By using these two parameters, it has been shown that most observations of
BT11 210 K are associated with deep convective clouds, of which the majority have
very high optical depth. Although a few studies characterize the properties of anvil
clouds, the mean optical depth of anvil clouds have not been shown to be associated with
optical depths this high. Given these results, cold cloud features ( 210 K) and +BTD
signatures, are both techniques that provide observations of penetrating deep convective
clouds in comparison to other cloud types. This indicates that the IR threshold technique
is a very useful approach to sampling and characterizing penetrating deep convective
clouds. It is also indicates, that although Aumann et al. [2010] suggests that an additional
CO2 channel be added to geostationary satellites to observe severe weather associated
with penetrating deep convective clouds it is likely unnecessary.
84
Table 4.1: Shows the relevant information for nine of the most popular studies on penetrating deep convection. These studies consist
of passive and active space borne remote sensing.
85
Table 4.2: Mean optical and microphysical properties for various types of high clouds
derived from MODIS L2 Cloud Product along the CloudSat orbital track (MAC06S0) for
January 2007.
Cloud Type
High
Cloud
Deep
Convection
Penetrating
Deep Convection
Anvil
Cloud top height
> 12 km
< 14 km
14 km
16.9 km
> 12
Cloud top Pressure (mb)
215.504
276.177
143.0
112.023
187.66
IRW Brightness Temperature (K)
264.277
240.725
211.001
199.31
237.64
Optical Thickness
2.629
27.241
24.1074
27.23
6.8081
Effective Radius ( m)
24.5843
28.956
29.422
29.396
26.56
Cloud Particle Phase
2.271
2.751
2.008
2.0004
2.069
+BTD (K)
-30.5585
-7.841
-0.1951
1.3701
-11.766
Cloud Fraction (%)
0.9444
0.9993
0.9999
1
0.9925
# of Cloud Layers
1
1
1
1
1
Lidar Cloud Base Height (km)
10.779
0.777
0.945
0.958
4.162
Lidar Cloud Top Height (km)
14.885
11.769
16.218
16.871
15.94
Table 4.3: Percent of occurrence of deep convection reaching 14 km and 16.9 km with
cold cloud features from 20°N─20°S and 35°N─35°S.
a) MODIS – CloudSat/CALIPSO
20°N - 20°S
35°N – 35°S
No. CCF ≤ 235 K
2,397,205
3,248,761
No. CCF ≤ 210 K
422,026
471,416
No. +BTD
699,789
829,973
%CCF(DC14) ≤ 235 K
25.91
22.67
%CCF(DC16.9) ≤ 235 K
1.12
0.93
%CCF(DC14) ≤ 210 K
65.49
60.90
%CCF(DC16.9) ≤ 210 K
6.03
6.14
%+BTD(14)
54.66
52.33
%+BTD(16.9)
3.51
3.37
86
Table 4.4: Properties of cold cloud feature distributions and penetrating deep convection reaching 14 km and cold cloud features from
20°N ─20°S and 35°N ─35°S.
IR Threshold
CPR Height (km)
Lidar Height (km)
MODIS IRW BT (K)
BTD
%POT
%
CCF 235K
min
mean
max
min
mean
max
min
mean
max
min
mean
max
IRWBT 235
5.16
13.61
18.71
6.15
14.89
18.94
178.62
217.24
235.0
-16.52
-0.78
6.15
25.91
100%
IRWBT 230
5.16
13.86
18.71
6.15
15.10
18.94
178.62
214.99
230.0
-14.03
-0.25
6.15
25.14
99.09
IRWBT 225
5.16
14.16
18.71
6.15
15.35
18.94
178.62
212.36
225.0
-12.11
0.22
6.15
28.72
80.3
IRWBT 220
5.17
14.49
18.71
6.15
15.60
18.94
178.62
209.45
220.0
-12.11
0.60
6.15
38.53
51.77
IRWBT 215
5.17
14.84
18.71
7.00
15.85
18.94
178.62
206.34
215.0
-10.37
0.88
6.15
51.86
30.04
IRWBT 210
5.17
15.20
18.71
7.57
16.09
18.94
178.62
203.06
210.0
-10.37
1.07
6.15
65.49
16.35
IRWBT 205
5.17
15.59
18.71
11.90
16.35
18.94
178.62
199.61
205.0
-10.37
1.19
5.69
78.81
7.93
IRWBT 200
5.27
16.02
18.71
13.81
16.64
18.94
178.62
195.96
200.0
-8.719
1.30
5.69
93.67
3.1
87
Table 4.5: Percent of BT11 210 K pixels provided according to cloud type and cloud
optical thickness for different latitudinal bands where data was derived from the Aqua
MODIS Level 2 Cloud Product for October 2007.
Region
Cloud Optical Depth
Cirrus
Cirrostratus
Deep Convection
0
3.6
9.4
20
23
43
63
83
100
15 N-15 S
.12
1.59
10.38
3.64
24.44
14.44
8.62
3.29
33.46
20 N-20 S
.12
1.59
10.25
3.62
24.63
14.92
8.51
3.22
33.12
35 N-35 S
.12
1.53
9.86
3.49
24.22
15.46
8.70
3.25
33.36
88
Figure 4.1: To better describe the function of the CloudSat 2B CLDCLASS product,
relevant vertical range gates depicting the classification of the anvil cloud, the penetrating
top, and the main body of the deep convective cloud are shown. Anvil clouds, which are
not classified by the 2B CLDCLASS product are determined when deep convective
clouds (8) are present and have high cirriform (cloud class 1) connected to them and no
cloud (0 - not shown) beneath them.
89
Figure 4.2: Frequency Distributions of a) BTD, b) IRW (cloud) brightness temperature, c) optical thickness, d) WV Brightness
Temperature e) cloud top pressure and f) effective particle radius for all the cloud types described in Table 3.2 where the overlap
associated with each cloud property is shown for each cloud type.
IRW Brightness Temperatures (K)
90
Figure 4.3: Distribution of IRW brightness temperatures verses cloud optical thickness for a)
Deep convective clouds and b) cirrus clouds sampled from January 2007 statistics. For both
profiles a slightly negative slope of the linear line fit given by the equations a) and b) show
that values of brightness temperature generally decrease with increasing cloud optical
thickness.
y= 0.0313x+11.0625
y= 0.1394x + 54.248
a)
b)
91
Figure 4.4: CloudSat cross section on October 22, 2006 showing variability of cloud
brightness temperatures (IRWBT), IR water vapor brightness temperatures (IRWVBT)
and +BTD for stratocumulus cloud (cloud A), overshooting deep convective cloud (cloud
B), penetrating deep convective cloud (cloud C), and a cirrus cloud (cloud D).
A
B
C
D
92
Figure 4.5 Frequency distribution between 35°N─35°S of all cold cloud features a) 235
K b) 210 K and seasonal patterns of cold cloud features 210 K for c) December-
February (DJF) and d) June-August (JJA).
a)
b)
c)
d)
93
a)
Figure 4.6: Frequency distribution between 35°N─35°S of a) +BTD signatures (with BT11 <
235 K) and seasonal patterns of +BTD for b) December-February (DJF) and c) June-August
(JJA).
b)
c)
94
Figure 4.7: Images of a) MODIS L2 Cloud top temperature b) IRW Brightness Temperature and
c) visible true color images corresponding to CloudSat granule ID 2575 on October 22, 2006 and
time stamp 0450.
95
Figure 4.8: Normalized frequency of cloud optical depth for all pixels between 15 N-15 S
with BT11 210 K for October 2007.
96
CHAPTER 5
CLIMATOLOGY OF PENETRATING DEEP CONVECTION
5.1 Introduction
The regional and seasonal characteristics of penetrating deep convective clouds
have been investigated using radar (e.g., Liu and Zipser., 2005) and IR (e.g., Rossow and
Pearl, 2007) observations. The details of those investigations have already been provided
in previous chapters. Here we specifically note that according to Rossow and Pearl [2007
penetrating deep convective clouds exhibit five major regions of concentration─ over the
maritime continent (east Indian and west Pacific Oceans), in the western end of the South
Pacific Convergence Zone and in the eastern end of the Pacific Intertropical Convergence
Zone, and over South America and Africa. Rossow and Pearl [2007] note that the
distribution of deep convection reaching the TTL does not change when a threshold
relative to the tropical cold point tropopause or the base of the TTL, is used. They also
report that the seasonal and geographic distribution of penetrating convection is very
similar to that provided by Gettelman et al. [2002] and other studies using satellite
infrared radiances. They conclude that such similarities indicate that interannual
variations in these patterns are not very large.
Tselioudis et al. [2010] performed a time series analysis of tropical penetrating
deep convective clouds from July 1983 to June 2008. The authors report that when using
ISCCP-D1 data for this period, deep convective clouds penetrating the lower stratosphere
show no significant long-term trends. However, the authors also evaluated the time
dependent properties of weather state 1 (WS1), which includes the larger mesoscale
convective systems but consist largely of considerable amounts of optically thinner, high,
97
and middle top clouds. It is described by the morphology of relatively high cloud optical
thickness ( ) and cloud top pressure (see details as given by Rossow et al., 2005).
According to the time series analysis of WS1, Tselioudis et al. conclude that deep
convection in the Indian Ocean and the western‐central Pacific regions had increases in
frequency from 1983 to about 2000 and has remained at a nearly constant level after that.
The sharpest increase in deep convection occurred between 1993 and 2000. The authors
conclude that stratospheric water vapor trends come from the overall moistening of the
tropical upper troposphere rather than from direct transport by convection penetrating
into the lower stratosphere.
The ISCCP DX data used by Tselioudis et al. [2010] and Rossow and Pearl
[2007] are both produced from the analysis of infrared (IR) and visible (daytime only)
radiances from weather satellite images with ~ 5 km horizontal resolution that are sub-
sampled to a 30 km horizontal resolution at 3 hr intervals. While it is suggested that the
statistics obtained for the ISCCP-DX data set converge to those obtained from the full 5
km dataset, this is only true if the sample population size is large [Seze and Rossow,
1991]. So, for penetrating deep convection that is reportedly representative of ~ 2% of all
deep convection and largely characterized by relatively small areal extents (see Chapter
3), it is not clear if the ISCCP DX data, with its coarser resolution, provides reliable
results regarding the trends or variability of penetrating deep convective clouds.
Little information is available on the time dependent properties of penetrating
deep convective clouds because of the difficulties associated with techniques and the low
resolution of the available data. The longest history of observations that may be capable
of elucidating the linkages between penetrating deep convection and lower stratospheric
98
water vapor trends occurring from 1980 to 2008 is associated with conventional passive
sensor satellite radiometers analogous to the observations used in the aforementioned
studies. However, while ISCCP-DX data is commonly used to deduce global cloud and
surface properties, the DX data is not the highest resolution of ISCCP data available. A
higher resolution of ISCCP data is available from the GridSat product. This product has
been derived from ISCCP B11 data and provides global coverage for infrared window
(IRWIN) observations dating back to ~ 1980 and infrared water vapor (IRWVP)
coverage dating back to 1998. The horizontal and temporal resolution of this data is
provided at 10 km every 3 hours. Given these details, GridSat observations are likely to
be more useful in analyzing the time dependent properties of penetrating deep convection
compared to Rossow and Pearl [2002] and Tselioudis et al. [2010]. More specifically, the
GridSat observations will be investigated to determine what (if any) linkages exist
between the temporal variability of penetrating deep convective clouds and lower
stratospheric water vapor changes. In addition, since radar and IR studies have shown
differences in regional maxima of penetrating deep convection and unfortunately coarse
temporal sampling of CloudSat observations provided in Chapter 3 do not fully resolve
the diurnal cycle, it is still unclear where deep convection reaching the TTL is most
dominant.
In Chapter 5 of this thesis, the time dependent properties of penetrating deep
convection as obtained from GridSat observations are used to address the following:
1 International Satellite Cloud Climatology Project (ISCCP) B1 data are geostationary imagery from
99
1. How well do positive BTD signatures and cloud brightness temperatures
210 K from GridSat observe annual, interannual and diurnal variability of
penetrating deep convective clouds?
2. When resolving the diurnal cycle of penetrating deep convection, where
do observations show that overshooting deep convection is most prevalent
over the western tropical Pacific or over Africa?
3. Do significant trends or patterns of penetrating deep convection exits? If
so, how do these trends correspond to changes in lower stratospheric
water vapor?
5.2 Data and Methods
Since July 1983, the International Satellite Cloud Climatology Project (ISCCP)
has collected infrared and visible radiances obtained from imaging radiometers carried on
the international constellation of weather satellites2. By 1998, the constellation enhanced
its spectral characteristics with nearly global coverage of upper tropospheric water vapor
at ~ 6.7 μm. Observations from ISCCP have primarily been used to better characterize
how clouds alter the radiation balance of Earth [Schiffer and Rossow, 1983]. However,
ISCCP has also been used to obtain cloud optical and microphysical properties, including
cloud fraction, cloud optical depth, cloud type, cloud top pressure and temperature, etc.
(for the full list see http://isccp.giss.nasa.gov/cloudtypes.html). ISCCP-B3 data is the
primary dataset from which subsequent data products (e.g., pixel level; DX; and gridded
2 The international constellation of weather satellites includes, EUMETSAT for the Meteorological
Satellite (METEOSAT); the Japanese Meteorological Agency (JMA) for the Geostationary Meteorological
Satellite (GMS); the Atmospheric Environment Service of Canada for Geostationary Operational
Environmental Satellite-East (GOES-East); Colorado State University (CSU) for the GOES-west; and the
National Oceanographic and Atmospheric Administration (NOAA) for the polar orbiters.
100
cloud products C1, C2, D1, and D2) originate. It primarily contains reflected and emitted
radiances in regions of the visible (~0.6µm), IR window (~11µm) and IR water vapor
(~6.7µm) channels. However, the ISCCP-B3 data is a reduced resolution product that has
been subsampled to ~ 30 km due to limitations in computing power that existed in the
1980s. After sampling the radiances to reduce data volume, they are calibrated,
navigated, and placed in a common data format (e.g., hdf and netcdf).
In contrast to the ISCCP-B3 data which is calibrated against NOAA-9 Advanced
Very High Resolution Radiometer (AVHRR), ISCCP-B1 data has been calibrated using
the High Resolution Infrared Radiation Sounder (HIRS) as an independent analysis of the
satellite intercalibration performed by (ISCCP) so that biases between ISCCP-B1
observations and HIRS are minimal. These observations are then spatially averaged to
~10 km resolution and nominally referred to as the GridSat product. GridSat provides
synoptic observations of reflected and emitted radiances at the same wavelengths as
ISCCP-B3 data but at a finer resolution. Although the ISCCP-B1 data extends back to
1978 and the GridSat product is new, these datasets have not been frequently used in
research studies. Knapp [2008a] provides more information on the development of
ISCCP-B1 data, and Knapp [2008a; 2008b] outlines the potential uses of ISCCP-B1 data
for climate studies.
GridSat data is used to evaluate the discernable trends or patterns of variability in
the +BTD signature and cold cloud features 210 K. From the application of techniques
described in previous chapters, a climatology of tropical penetrating deep convection is
developed for years 1998─2008. As shown by Knapp [2008b], these years correspond to
the earliest near global coverage (i.e., not including the poles) of ISCCP B1 observations
101
in the IR window region centered at ~11µm and the IR water vapor absorption band
centered at ~6.7 µm. Moreover, several years (i.e., 1999─2001 and 2007─2008) cover
abrupt changes in lower stratospheric water vapor.
Although cold cloud features defined by 3 km x 3 km subsets were used in
Chapter 4, GridSat observations are provided at a horizontal resolution of 10 km. To
resolve penetrating deep convective cloud tops, no additional averaging is performed. All
evaluations made in this chapter of the analysis are based on frequency counts which are
evaluated on the basis of 2.5 2.5 grids between 35 N to 35 S. For each observation,
the frequency per grid is used to evaluate the frequency per region where specific regions
(following Tselioudis et al., 2010) include Africa (0°-45°E), the Indian Ocean (45°E-
105°E), west Pacific Ocean (105°E-150°E), central Pacific Ocean (150°E-160°W), East
Pacific Ocean (160°W-80°W), South America (80°W-45°W), and the Atlantic Ocean
(45°W-0°). Standardized monthly anomalies for each region are determined by
subtracting the long-term monthly mean from the monthly frequency and dividing by the
corresponding monthly standard deviation. By calculating the standardized anomaly, the
influences of seasonal changes in the average monthly frequencies are minimized to show
the year-to-year variability in the monthly frequencies. This procedure is also repeated
for tropical (15 N-15 S) monthly zonal averages of water vapor mixing ratio values at
82.5 mb. Prior to mid-2004 this dataset is based on observations from the Upper Air
Research Satellite Halogen Occultation Experiment (UARS HALOE) but has been
shifted to match the Aura Microwave Limb Sounder (MLS). Observations of 82mb
mixing ratios post-mid-2004 are based on Aura MLS. Similar measurements have been
documented by Rosenlof and Reid [2008], Solomon et al., [2010], and Ray and Rosenlof
102
[2007] with further documentation of Aura MLS data given by Waters et al. [2006] and
further documentation on UARS HALOE observations given by Russell et al. [1993].
5.3 Results
5.3.1 Evaluation of Climatological Data
GridSat is a relatively new dataset. While rigorous calibration testing of the
IRWIN channel has motivated its use in several other studies (e.g., Bain et al., 2010;
Kossin et al., 2007), the calibration of both the IRWIN and IRWVP channels at low cloud
brightness temperatures (i.e., 210K) is questionable due to the limited number of
calibration matchups provided by NOAA HIRS. Since the general characteristics of deep
convection are well documented, it is important to test that observations from GridSat
provide the same general characteristics of deep convection provided in other studies.
To evaluate the signature of penetrating deep convection from +BTD signatures
and cloud brightness temperatures 210 K, their annual frequency distributions are
provided for years 1998─2008 in Figures 5.1 and 5.2, respectively. Comparison of the
observations show that maxima in +BTD signatures most often occur over central Africa
and in the Indian Ocean just west of Indonesia. In comparison, cloud brightness
temperatures 210 K have maxima over Africa, the central Pacific, and west Pacific.
Evaluation of all +BTD annuals provided in Figure 5.1 show expected patterns in
occurrence frequencies from 1998 2001 and 2008 while annual frequency distributions
for 2002─2007 show a lack of positive BTD signatures over Indonesia and the western
tropical Pacific Ocean. In comparison, cloud brightness temperatures 210 K show
annual distributions with expected patterns in occurrence frequencies for all years
between 1998 and 2008. While there is some interannual variability in regions of
103
maximum frequencies within the western tropical Pacific and central Pacific, it is not
clear if this variability has a strong correlation with La Niña (1999, 2000 and 2008) and
El Niño (1998, 2002, 2004, and 2006) years, which could potentially shift maxima in
penetrating deep convection from the western Pacific to the central Pacific.
Since both +BTD signatures and cloud brightness temperatures 210 K should
have similar patterns, the 2002─2007 BTD annuals reveal inconsistencies that are
specifically due to the IRWVP channel and thus dataset driven. While the exact source of
the problem is not clear, the 2002─2007 distributions of +BTD signatures are most
deficient in regions that correspond to GOES-W. Given these results, the remaining
sections of Chapter 5 will not focus on +BTD signatures. Instead, penetrating deep
convection will be evaluated from cloud brightness temperatures 210 K, which appear
to be consistent with the general distribution of penetrating deep convection provided in
other studies.
To further evaluate observations of penetrating deep convective clouds, cloud
brightness temperatures 210 K are also compared with Russo et al. [2010], who used
observations of cloud top height, upper tropospheric/lower stratospheric water vapor, and
surface precipitation estimates to evaluate the performance of nine different models to
represent penetrating deep convection. Among the results presented by Russo et al. we
specifically focus on the evaluation of monthly mean cloud top height in the maritime
continent (15 N-15 S and 90 E-150 E). These results utilize observations of cloud top
height from ISCCP (-D1), MODIS Terra and Aqua for November 2005. These
observations are provided in Figure 5.3(a-c) and compared with the GridSat observations
of brightness temperatures 210 K for the same region and period in Figure 5.3(d). The
104
results of Figures 5.3(a-c) qualitatively show that the ISCCP-D1 data is less capable of
representing cloud top structure compared with both MODIS Aqua and Terra satellites in
terms of regional structure and vertical extent. ISCCP-D1 observations show that for
November 2005, the region does not have cloud top heights > 9 km. On the other hand,
MODIS Aqua and Terra observations show heights up to 12 km. Obviously MODIS
Aqua and Terra monthly mean cloud top structures do not show penetrating deep
convection since these events are rather rare. Thus it is not expected that pronounced
features suggestive of penetrating deep convective cloud tops would be found within the
monthly mean. Although there are differences in the spatial and temporal sampling of
MODIS (Terra and Aqua) and GridSat, in general, they have good agreement. Although
the observations provided in Figures 5.3(a-c) do not show the time-averaged frequency of
deep convective clouds reaching the TTL, these results do show that the most frequent
observations of cloud brightness temperatures 210 K occur in regions that are
associated with higher cloud top heights.
The most fundamental modes of variability of the global climate system are the
diurnal and seasonal cycles. The seasonal cycle of penetrating deep convective clouds has
been addressed in Chapter 3 (Figure 3.3) and Chapter 4 (Figure 4.5 and 4.6) for Northern
Hemisphere winter (DJF) and summer (JJA). GridSat observations of cloud brightness
temperatures 210 K for the same periods are in good agreement with the seasonal
distributions that have already been provided. On the other hand, coarse temporal
sampling associated with the Aqua orbit did not enable the full diurnal cycle of deep
convection to be evaluated from CloudSat and MODIS-AUX observations. In addition, it
was unclear how a full sampling of the complete diurnal cycle would impact the
105
CloudSat distribution, which showed that the highest frequencies of penetrating deep
convection were centered over the Indian Ocean and western tropical Pacific. The
diurnal cycle is therefore evaluated for Africa, the Indian Ocean, South America, and the
western Pacific Ocean in order to compare the variability of penetrating deep convective
clouds over tropical land regions and the tropical oceans.
As shown in Figure 5.4, the land regions of Africa and South America have a
pronounced diurnal cycle that peaks in the late afternoon to early evening while only a
slight diurnal cycle exists over the Indian and western tropical Pacific Oceans, of which
the later also shows peak frequencies during the early morning. These results are
consistent with other studies (e.g., Soden, 2000; Yang and Slingo; 2000) and indicate that
temporal sampling may have an impact on the regional analysis of penetrating deep
convection, especially over land. They also support the interpretation that the BT11 210
K distribution is dominated by deep convective clouds verses other types of high clouds
since deep convective clouds are considerably forced by daytime solar heating. Although
cirrus anvil clouds can make up a third of the BT11 210 K distribution cirrus clouds
have longer lifetimes than deep convection. Thus it is still, unclear whether poorly
resolved temporal sampling may have a strong impact on the regional dominance of deep
convection. To make this evaluation, the regional dominance of deep convection is
evaluated for all regions in Figure 5.5.
According to the monthly normalized frequencies of all cloud brightness
temperatures 210 K occurring within each of the seven regions described in Section 5.2
the largest frequencies of the BT11 210 K signal occurs of the western-central Pacific
and Indian Oceans. The quantitative estimates of the contribution each region makes to
106
the total tropical signal is provided in Table 5.1 with linear regression statistics that
identify the strength and significance of each region when evaluated against the tropical
signal. The highest frequencies of penetrating deep convective clouds are consistently
over the western Pacific, central Pacific and Indian Oceans accounting for 24.2%, 18.3%,
and 17.5% respectively. Following these regions are Africa (14.9%) and South America
(11.0%) and lastly the eastern Pacific (8.9%) and Atlantic Oceans (5.3%), which in
Figure 5.5 show very strong seasonality compared with the western Pacific, central
Pacific and Indian Oceans where frequencies are the highest. Regional analysis according
to linear regression statistics where the autocorrelation3 of the noise (
ˆ
) and standard
deviation of the noise (
ˆ N
) have been considered as in the approach of Weatherhead et
al. [1998] is provided in Table 5.2. According to this information, slightly negative trends
occur in the tropics and most other regions between 1998─2008, with the exception of
Africa, the central Pacific, and the Atlantic Ocean. However, these trends are not
significant at the 95% confidence level and would require much more data to meet this
criteria. The trends are represented in Figure 5.6 for the tropics and individual regions.
Considering the dominance of the western-central Pacific and Indian Ocean and the
significance of their relationships to the tropical signal at the 95% confidence level, a
closer evaluation of these particular regions is made in the following section to evaluate
anomalies in monthly water vapor mixing ratios at 82 mb.
5.3.2 Penetrating Deep Convection and Lower Stratospheric Water Vapor
As stated in Section 1.1 lower stratospheric water vapor mixing ratios are largely
3 Autocorrelation is the cross-correlation of a signal with itself. When the noise within a
time series is said to be autocorrelated, the time series violates the assumption of
independence of errors. Autocorrelation of the noise
ˆ
and a large standard deviation of
the noise
ˆ N
make any trend within a time series more difficult to detect.
107
controlled by the seasonal variation in zonal mean tropical tropopause temperature.
According to Highwood and Hoskins [1998], the average JJA zonal mean in tropical
tropopause temperature is ~ 6 C warmer than the DJF zonal mean. Since tropical
tropopause temperatures impact water vapor mixing ratios in accordance with the
Clausius-Clapeyron relation, the seasonal cycle is the most dominant feature of the 1998-
2008 time series of lower stratospheric water vapor at 82 mb provided in Figure 5.7.
Based on this seasonal variation, water vapor mixing ratios at 82 mb range from about 2
ppmv in DJF to about 5.5 ppmv in JJA. The monthly anomalies in 82 mb water vapor
mixing ratios in Figure 5.8 is associated with decreasing trend of -0.1178 ppmv/month
but would take ~ 15 years to be significant at the 95% confidence level.
Monthly frequency anomalies of cloud brightness temperatures 210 K were
evaluated with monthly anomalies in the water vapor mixing ratios at 82 mb according to
linear regression analysis for the entire tropics, western Pacific, central Pacific, and
Indian Oceans. For this analysis, it is important to note that the Brewer-Dobson
circulation, which lifts air through the tropopause and into the lower stratosphere is quite
slow, progressing at a speed of ~20-30 m/day. Since this is an important dynamical factor
modulating transport from the TTL into the lower stratosphere (e.g., Mote et al., 1995),
anomalies in water vapor mixing ratio at 82 mb (~17.5 km) and BT11 210 K signals
representative of penetrating deep convection with cloud top heights at ~ 15.2 ( 0.91 km)
km (~110 mb) were evaluated at lags of 0 to 6 months. Linear regression statistics for
these evaluations are provided in Table 5.3 and show consistent downward trends for the
tropics, western Pacific, and central Pacific Oceans while trends in the Indian Ocean are
more positive.
108
For each region evaluated, the strongest and most significant relationships were
most consistently observed for lags between 2 – 3 months. At a lag of 3 months, the
combined western-central Pacific had a significant anticorrelation where the largest
amount of variance explained by the combined western-central Pacific monthly
frequency anomalies is 8.25%. This result indicates that among all regions considered
and to the extent that the coldest penetrating deep convection can be determined from the
GridSat record, events within the western-central Pacific dominate the anomalies in the
tropical 82 mb water vapor record. The time lag associated with this result does not
drastically differ from Randel et al. [2004] who evaluated 82 mb H2O anomalies against
10°N-10°S tropopause temperature anomalies and reported the highest correlations at a
lag of 2 months. However, the meaning of the anticorrelated relationship must also be
considered.
Solomon and Reid [2008] evaluated trends in the temperature and water vapor
content of the tropical lower stratosphere and its connection with sea surface
temperatures (SST) in the western-central Pacific. They observed a significant
anticorrelation with SST anomalies that explained ~2% of the variance at lag zero in the
lower stratospheric temperatures for an analysis period of 34 years (1960-1994). In their
evaluation, the authors note that the major response to the variance in lower stratospheric
temperatures is the quasibiennial oscillation (QBO) in the tropical stratospheric winds
and suggests that the QBO signal makes it rather difficult to detect the variance within
the signal that is due to other components. The data also suggests that the 1998-2008
decreasing trend in the monthly frequency of penetrating tops, which explains about
8.25% of the variance in the 82 mb water vapor anomalies, may be influencing the
109
decreasing trend in the 82 mb water vapor values. Removing the portion of the 82 mb
water vapor anomalies that is due to the QBO may however prove to establish a more
direct relationship determined by stronger correlations and a higher percentage of
explained variance.
In comparison to Tselioudis et al. [2010], who have shown the convective
variability of the WS1 series representing organized mesoscale convection and the
presence of considerable amounts of optically thinner, high and middle top clouds, it is
not clear to what extent the WS1 series represents deep convection and more importantly,
there is no corresponding evaluation of changes in convective variability with changes in
lower stratospheric water vapor changes. Hence, it is unclear if similar, stronger, or
weaker correlations exist when we analyze changes in lower stratospheric water vapor
variability against all deep convection, which Tselioudis et al. suggest is a more likely
mechanism driving stratospheric water vapor variability patterns.
5.4 Summary and Discussion
Results given in Chapter 5 show that although +BTD signatures from GridSat
could not be used for further analysis of penetrating cloud tops because of issues with
GridSat BT6.7 observations, cold cloud features 210 K are a useful method to infer
penetrating deep convective cloud activity. GridSat observations are at higher resolution
than ISCCP-D1 observations, which have been used in most of the recent studies
addressing the climatology of penetrating deep convective clouds. As indicated by the
comparison of ISCCP-D1 and Aqua and Terra MODIS observations of cloud top
structure, ISCCP-D1 data are less capable of resolving vertical cloud top structure. On
the other hand, GridSat observations of cloud brightness temperatures 210 K show
110
similar structure as provided by MODIS Terra and Aqua. The limitations of ISCCP (D1)
observations to more accurately resolve cloud top height is an obstacle for any studies
seeking to use the ISCCP (D1) observations to evaluate penetrating deep convection. In
the remaining sections of this discussion, the results of this study are evaluated to
elaborate on the climatological characteristics of penetrating deep convective clouds and
to specifically address the questions in Section 5.1.
How well do +BTD signatures and cloud brightness temperatures 210 K from
GridSat observe annual, interannual, and diurnal variability of penetrating deep
convective clouds? Cloud brightness temperatures 210 K are used to infer the
variability of penetrating deep convective clouds. In general GridSat observations agree
with both the diurnal and annual characteristics of the BT11 210 K signal evaluated in
earlier IR studies. While there is interannual variability especially in the western tropical
Pacific and central Pacific, it is not clear if this variability has a strong correlation with
La Niña (1999, 2000 and 2008) and El Niño (1998, 2002, 2004, and 2006) years since
there is no direct link to shifts in maxima frequencies from the western Pacific to the
central Pacific for El Niño and events. Gettelman et al. [2002] addressed this linkage in
his evaluation of penetrating deep convection from 3.5 years (from October 1986 to
February 1989 and October 1991 to September 1992) of global cloud imagery but the
authors did not resolve a direct linkage between the variability in El Niño and the
interannual variability in penetrating deep convective cloud frequency. However, we
suggests that when considering the variability of penetrating deep convection, it may be
more likely that the La Niña and El Niño phase is not as important as the variability of
strong events vs. weak events. This is an interesting point for future research.
111
When resolving the diurnal cycle of penetrating deep convection do observations
show that overshooting deep convection is most prevalent over the western tropical
Pacific or over Africa? This question is particularly important since several hypothesis
have been used to determine where deep convective clouds penetrate into the lower
stratosphere. The first hypothesis proposed by Newell and Gould─Stewart [1981]
suggests that troposphere-stratosphere transport occurs over the Maritime Continent and
western Pacific described as the ―Stratospheric Fountain‖ via rather strong deep
convection. With the advent of space-borne radars, Liu and Zipser [2005; 2007] argue
that upper tropospheric and lower stratospheric water vapor exchange is primarily driven
by deep convection over Africa. Although the original Stratospheric Fountain hypothesis
suggests that air only enters the tropical tropopause in the western tropical Pacific,
northern Australia, Indonesia and Malaysia between November-March, we presently
know that water vapor may enter the lower stratosphere during any time of the year, and
while it predominantly enters within the tropics, it also has a tendency to enter the lower
stratosphere from extratropical regions via penetrating deep convection (Chapter 4) and
tropopause folding4 [Holton et al., 1995]. However, in support of the stratospheric
fountain hypothesis the results shown in Figure 5.5 suggests that the western tropical
Pacific has the highest occurrence frequencies of penetrating deep convective clouds
supporting the results of the CloudSat distribution in Figures 3.3 and 3.4 and the results
4 Tropopause folding is a source of midlatitude stratospheric-tropospheric exchange. It
occurs in areas with large vertical shear and strong meridional thermal gradients.
Downward transport of stratospheric air into the troposphere occurs along the sloping
lines of constant potential temperature.
112
of Russo et al., [2010] who modeled the impact of deep convection on the tropical
tropopause layer.
Given that +BTD signatures and cloud brightness temperatures 210 K
sufficiently represent penetrating deep convection, do discernable trends or patterns of
variability exists? If so, how do these trends correspond to changes in lower
stratospheric water vapor? According to studies investigating the temporal variability of
lower stratospheric water vapor, a positive trend totaling ~ 1 ppmv occurred between
1980─2000 (e.g., Oltmans et al., 2000; Solomon et al., 2001; Rosenlof et al., 2010). Post
2000 trends show a decrease totaling ~ 0.4 ppmv (e.g., Solomon et al., 2010). Although
much interannual variability is shown throughout the tropics, negative trends in lower
stratospheric water vapor at 82 mb are consistent with trends in the monthly frequency
anomalies evaluated for the entire tropics and the western-central Pacific and Indian
Oceans which dominates the tropical record. The explained variance provided in Table
5.3, shows that at a lag of 3 months, BT11 ≤ 210 K time series of the western-central
Pacific explains ~8.25% of the variance of anomalies in the 82 mb water vapor time
series. Since anticorrelations suggest that increases in the frequency of penetrating deep
convection decreases water vapor at 82 mb it is possible that penetrating deep convective
clouds are dehydrating the lower stratosphere. However, few observations support this
conclusion and both observational records show decreasing trends. Thus it is actually
more justifiable to conclude that the overall impact of penetrating deep convective clouds
is to hydrate the lower stratosphere; whereby fewer amounts of penetrating deep
convection tend to decreases water vapor at 82 mb and larger amounts of penetrating
deep convection tends to increase water vapor at 82 mb.
113
For these results, we acknowledge that several levels of uncertainty. For example,
while we have concluded that the BT11 210 K signal is dominated by penetrating tops,
it is not composed solely of penetrating tops. Furthermore, the 82 mb water vapor data is
provided as monthly averages over an entire latitudinal band (15°N-15°S). It is possible
that monthly means for such a wide latitudinal range may obscure some relevant
information related to the influence of penetrating deep convective tops. Finally, while
the QBO signal is more dominant away from the tropical tropopause rather than at the
tropical tropopause, there is still some modulation of tropical tropopause temperatures
and water vapor mixing ratio due to the QBO and its variability. Minimizing this
influence is important to better resolve the relationship between penetrating deep
convection and lower stratospheric water vapor.
Although Chapter 5 has investigated the climatology of the BT11 210K
distribution from 1998-2008, other results for years prior to 1998 should also be analyzed
to give a better indication of the relationship between penetrating deep convective active
within the tropics and lower stratospheric water vapor variability. The comparison of this
data with upper tropospheric/lower stratospheric water vapor variability should also be
performed at ~ 15.5 km (~110 mb) and up to ~ 17.5 km (82 mb). A longer time series
analysis at multiple levels may provide more evidence regarding the connection between
lower stratospheric water vapor and penetrating deep convective clouds. It is possible that
penetrating deep convective clouds have a stronger impact on lower stratospheric water
vapor than these observations lead us to conclude. However, the key to better unlocking
this relationship may not be obtained according to the shear frequency but some other
conditions of penetrating tops such as areal extent and duration.
114
Table 5.1: Linear regression statistics for tropical and regional standardized frequency
anomalies in monthly IRWBT 210 K where the slope/trend is equivalent to the
correlation coefficient, p-value and t-stat are both standards for identifying the
significance of the tropical and regional relationships and are compared to =0.05 and a
critical t-value of 1.962.
Tropics
w/
%
Tropics
y-int (10-6)
Slope/trend
p-value
t-stat for
significance of
slope
Variance
Explained
(%)
W. Pacific
24.15
-4.110
0.425
<.001
5.348
18.032
C. Pacific
18.28
-3.305
0.213
0.014
2.480
4.516
E. Pacific
8.89
-3.752
-0.047
0.594
-0.534
0.219
Indian
17.52
-2.840
0.313
<.001
3.757
9.794
Atlantic
5.31
-3.126
0.291
0.001
3.469
8.471
S. America
10.98
-3.700
0.121
0.167
1.389
1.462
Africa
14.86
-4.029
0.159
0.069
1.834
2.523
Table 5.2: Time series regression statistics for standardized frequency anomalies in
monthly IRWBT 210 K with number of years (
ˆ
n *
) of monthly data needed to detect
the trend provided for each region at the 95% confidence level as a function of the
autocorrelation (
ˆ
N
) and standard deviation (
ˆ N
) of the noise (Weatherhead et al., 1998).
y-int
slope/trend
ˆ
N
ˆ N
ˆ
n *
Tropics
0.390
-0.072
0.203
0.031
14.020
W. Pacific Ocean
0.317
-0.058
0.373
0.037
18.149
C. Pacific Ocean
-0.234
0.043
0.573
0.050
26.928
E. Pacific Ocean
0.437
-0.080
0.430
0.039
15.035
Indian Ocean
0.100
-0.018
-0.105
0.023
28.694
Atlantic Ocean
-0.262
0.048
0.272
0.034
19.511
S. America
0.047
-0.009
0.082
0.028
53.715
Africa
-0.109
0.020
0.176
0.031
32.945
115
Table 5.3: Linear regression statistics for anomalies in monthly frequency of IRWBT
210 K and monthly anomalies in 82 mb water vapor mixing ratio at lags of 0 to 6 months
for the tropics, western Pacific, central Pacific and Indian Oceans. Again, the p-value and
t-stat is respectively compared to =0.05 and a critical t-value of 1.962.
Lag in
82 mb
SWV
y-int
slope/trend
p-value
t-stat for
significance of
slope
Variance
Explained (%)
Tropics
0
0.000
-0.074
0.400
-0.844
0.548
1
-0.003
-0.133
0.132
-1.516
1.742
2
-0.020
-0.143
0.100
-1.656
2.103
3
-0.031
-0.072
0.410
-0.826
0.533
4
-0.045
-0.107
0.211
-1.256
1.232
5
-0.057
-0.052
0.544
-0.608
0.292
6
-0.072
-0.048
0.572
-0.567
0.260
western Pacific
0
0.000
0.035
0.690
0.400
0.123
1
-0.011
-0.111
0.206
-1.270
1.232
2
-0.021
-0.129
0.138
-1.492
1.716
3
-0.032
-0.192
0.026
-2.251
3.842
4
-0.049
-0.188
0.029
-2.214
3.725
5
-0.059
-0.144
0.093
-1.690
2.220
6
-0.073
-0.163
0.055
-1.933
2.924
central Pacific
0
0.000
-0.244
0.005
-2.864
5.954
1
-0.008
-0.240
0.006
-2.815
5.905
2
-0.016
-0.233
0.007
-2.721
5.476
3
-0.025
-0.209
0.017
-2.426
4.410
4
-0.040
-0.175
0.044
-2.033
3.168
5
-0.054
-0.083
0.340
-0.958
0.723
6
-0.068
-0.091
0.292
-1.058
0.903
Indian Ocean
0
0.000
0.044
0.617
0.501
0.193
1
-0.001
0.054
0.536
0.621
0.298
2
-0.020
0.016
0.857
0.180
0.020
3
-0.029
0.079
0.372
0.896
0.628
4
-0.046
-0.013
0.887
-0.143
0.016
5
-0.058
-0.124
0.153
-1.439
1.632
6
-0.074
-0.098
0.258
-1.138
1.032
116
Figure 5.1: GridSat 2.5° x 2.5° annual
frequency distributions of annual +BTD for
years 1998-2002 and over 35°N - 35°S.
a) 1998
b) 1999
a) 1998
b) 1999
c) 2000
c) 2000
d) 2001
d) 2001
e) 2002
e) 2002
Figure 5.2: Same as Figure 5.2 except for cloud
brightness temperatures 210 K for years
1998-2002 and over 35°N - 35°S.
a) 1998
b) 1999
117
Figure 5.1: (cont.) for years 2003-2007.
Figure 5.2: (cont.) for years 2003-2007.
118
g) 2008
g) 2008
Figure 5.1: (cont.) for 2008.
Figure 5.2: (cont.) for 2008.
119
Figure 5.3: November 2005 observations for the Maritime continent (15N-15S, 90E-
150E) of cloud top height from a) ISSCP (D1), b) MODIS-Terra c) MODIS-Aqua
Comparison (cf., Russo et al., 2010) and cloud brightness temperatures 210 K from d)
GridSat.
0.018 0.074 0.111 0.185 0.259 0.333 0.407 0.481
Frequency
120
Figure 5.4: Time averaged diurnal cycle per year for tropical regions of Africa, the Indian
Ocean, South America, and the western Pacific Ocean.
Africa
Indian Ocean
western Pacific Ocean
South America
55
Figure 5.5 Normalized monthly frequency of penetrating tops given for all seven regions
within the tropics from January 1998 through December 2008. The Western Pacific
Ocean clearly has higher frequencies of penetrating deep convection among all other
regions considered.
122
Figure 5.6: Standardized frequency anomalies for a) the Tropics (15°N-15°S ) b) Africa
c) the Indian Ocean d) the western Pacific Ocean e) central Pacific Ocean f) eastern
Pacific Ocean g) South America and h) the Atlantic Ocean.
123
Figure 5.8: 1998-2008 anomalies in 82 mb lower stratospheric water vapor mixing
ratio from 15°N-15°S shown with a slightly negative trend.
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Year
Figure 5.7: 1998-2008 monthly zonal averages of water vapor volume mixing ratio at
82mb and from 15°N-15°S.
y=-0.118x+0.643
124
CHAPTER 6
CONCLUSIONS
This thesis evaluates deep convection penetrating the TTL using radar and IR
observations to explore the hypothesis that penetrating and overshooting deep convection
has a strong influence on lower stratospheric water vapor variability. More specifically,
CloudSat/Calipso and Aqua MODIS observations from 2007 were used to 1) obtain a
statistically robust sample of penetrating deep convection and evaluate their areal extent
to determine how well penetrating deep convection may be resolved from IR sensors; 2)
quantitatively compare IR and radar distributions of penetrating deep convection using
traditional IR techniques to determine the extent to which traditional IR techniques
capture penetrating deep convection; and 3) given the uncertainty of traditional IR
techniques the variability of penetrating and overshooting deep convection is captured
from 11 years (1998─2008) of GridSat observations to compare monthly frequency
anomalies in penetrating deep convection with anomalies in lower stratospheric water
vapor. The main findings are summarized below to specifically address science questions
posed in previous chapters.
What new insights on the characteristics of penetrating and overshooting deep
convection does CloudSat provide? The normalized frequency distribution of penetrating
deep convective clouds from CloudSat is consistent with IR studies. However, the
CloudSat distribution does not compare as well with the distribution of penetrating deep
convective clouds provided from TRMM, which is based on precipitation size particles.
From CloudSat estimates of the areal extent of penetrating deep convection, it is
concluded that at a horizontal resolution of 10 km, about two thirds of penetrating deep
125
convective clouds are large enough to be detected from conventional IR sensors.
Examination of the CloudSat data also suggests that penetrating deep convection occurs
in the extratropics. CloudSat observations provided important insights into the
interpretation IR signatures of penetrating convection. It was shown that penetrating
deep convection reaching 14 km shows a range of cloud brightness temperatures > 210 K
and all do not exhibit +BTD signatures. However, higher penetrating deep convective
clouds consistently have cloud brightness temperatures 210 K and do exhibit +BTD
signatures.
To what extent is penetrating deep convection exclusively captured using traditional
IR methods and does this information change the perception of the frequency distribution
of penetrating deep convection as presented in other IR studies? Cold cloud features
210 K and +BTD signatures were sampled along the CloudSat orbital track and compared
with independent CloudSat/Calipso observations of penetrating deep convective clouds.
This was done to better understand the information content provided from the IR
distributions, since IR observations do not directly resolve cloud vertical structure and
there has been some debate on their usefulness to characterize observations of penetrating
tops. By comparing the IR and radar distributions, it is concluded that neither of the IR
schemes completely separates between penetrating deep convection and other types of
high clouds. However, the predominant fraction of +BTD signatures and cold cloud
features/pixels 210 K are associated with penetrating tops. This result is in contrast to
studies that suggest the majority of cold cloud features/pixels 210 K are cirrus/anvil
cloud fractions that coexist with deep convective clouds.
126
What are the microphysical and optical properties of penetrating deep convection in
comparison with other high level clouds? How does the incorporation of these
parameters aide in the evaluation of penetrating deep convective clouds using traditional
IR techniques? Since the fraction of cold cloud features 210 K that are also penetrating
tops is so much greater in this analysis compared with previous studies, the optical and
microphysical properties of other high level clouds and BT 210 K observations were
analyzed to interpret the conclusions that have been made regarding the IR distributions.
Comparison of deep convection, cirrus, anvil clouds and penetrating deep convective
clouds showed that while cirrus clouds reside at altitudes similar to penetrating deep
convection, their IR brightness temperatures are generally much warmer because they are
not as optically thick. Anvil clouds occasionally have cloud brightness temperatures
similar to penetrating deep convection. However, single layered anvil clouds rarely have
optical depths > 34. The evaluation of all optical depth values corresponding to
observations of cold cloud features/pixels 210 K shows that cloud brightness
temperatures 210 K are predominantly associated with deep convective clouds of rather
high optical thickness. This result further substantiates the interpretation that the 210 K
cloud field is predominantly associated with deep convective cores and more importantly
it does not require the use of radar data.
How well do +BTD signatures and cloud brightness temperatures 210 K from
GridSat compare when diagnosing their ability to observe annual, interannual and
diurnal variability of penetrating deep convective clouds? Based on differences in the
interannual variations in normalized frequency distributions of +BTD signatures and
cloud brightness temperatures 210 K, it was concluded that cloud brightness
127
temperatures 210 K are at present more suitable dataset from which to evaluate the
spatiotemporal properties of penetrating deep convective clouds from the GridSat
observations due to issues related to the 6.7 m brightness temperatures.
Do significant trends or patterns of penetrating deep convection exits? If so, how do
these trends correspond to changes in lower stratospheric water vapor? The annual and
interannual variability of penetrating deep convection was inferred from standardized
frequency anomalies in cloud brightness temperatures 210 K. It has been concluded
that the western-central Pacific and Indian Oceans provide the largest contributions to the
BT11 210 K signal for the entire tropics, followed by Africa, S. America, the Eastern
Pacific and Atlantic Ocean. Slightly negative trends occur in the tropics and most other
regions between 1998─2008, with the exception of Africa, the central Pacific, and the
Atlantic Ocean. However, these trends are not significant at the 95% confidence level and
would require much more data to meet this criteria.
The cross-correlation of anomalies in water vapor mixing ratio at 82 mb with monthly
anomalies in the frequency of the coldest penetrating tops supports the inference that the
strongest and most significant relationships were more consistently observed for lags
between 2 – 3 months. At a lag of 3 months, the combined western-central Pacific had a
small but significant anticorrelation, where the largest amount of variance explained by
the combined western-central Pacific monthly frequency anomalies was 8.25%. This
result indicates that events within the western-central Pacific variation dominate the
anomalies in the tropical 82 mb water vapor record. In comparison to Tselioudis et al.
[2010], it is unclear if similar, stronger or weaker correlations exist when the 82 mb
monthly anomalies are analyzed against all deep convection, which Tselioudis et al.
128
suggests is a more likely mechanism driving stratospheric water vapor variability
patterns.
The results of this study are however limited by the characteristics of the
evaluated data. For example, CloudSat observations have high vertical and horizontal
resolutions. In addition, it is a nadir scanning radar. By scanning such a narrow slice of
the atmosphere, CloudSat is unable to observe entire large-scale cloud systems. In
addition, the return period or length of its repeat cycle presents difficulties in identifying
short-lived deep convective clouds and fully resolving their diurnal cycle. CloudSat‘s
polar orbit minimizes the degree to which it observes the tropics, where penetrating deep
convective clouds are most often present. Such limitations may warrant some caution
with regard to the interpretation of CloudSat results that have been presented.
For the evaluation of IR properties, limitations arise due to explicit assumptions
that have been made with regard to infrared window brightness temperatures at ~11 m
from Aqua MODIS and GridSat. Unfortunately, there are differences in the 11 m central
wavelengths, bandwidths, and bit resolutions for Aqua MODIS and GridSat, which
present challenges for the intercalibration of multiple geostationary satellites. Another
layer of complication arises since observations of low cloud brightness temperatures have
fewer matchups of polar and geostationary observations that are used for intercalibration
purposes. This means that while CloudSat and Aqua MODIS observations have been
explicitly linked via analysis of collocated observations, no such procedure has been
presented here to evaluate the consistency between observations from Aqua MODIS and
GridSat, nor has this relationship been provided in the referenced literature.
Finally, with regard to observations of water vapor mixing ratio at 82 mb, several
129
other studies that have analyzed lower stratospheric water vapor anomalies have shown
that there is a clear presence of the QBO that is more obvious at higher levels extending
away from the tropopause rather than at the tropopause. However, the QBO signal is
likely present to some extent at the 82 mb level. Without quantifying the degree to which
the QBO signal is present in anomalies of water vapor mixing ratio at 82 mb and
modulates these values, the relationship between the frequency in penetrating deep
convective clouds and water vapor variability is highly subject to misinterpretation
Comparison of the optical and microphysical properties of penetrating deep
convection compared with other types of high clouds in Chapter 3 was only performed
for January 2007. The short time period of this part of the analysis leaves some question
regarding the representativeness of these results. As shown by Hong and Yang [2007],
there is seasonal variability in cirrus and deep convection that is missed when we limit
our evaluation to 1 month only. What would be the outcome of a similar analysis if it
were performed over a longer time period or at least for January and July? Since no
other studies have coupled CloudSat observations of vertical cloud structure and followed
this process with the retrieval of optical and microphysical properties from Aqua
MODIS, a more rigorous analysis of high-level clouds could be performed especially to
evaluate the degree to which we may disentangle deep convective cores and surrounding
anvil cloud fractions.
With additional regard to the time extent of the evaluations, it is also noted that
the atmosphere is highly variable on annual and interannual time scales that are
considerably modulated by the El Niño Southern Oscillation. Since the 2007 observations
occurred during a moderate La Niña phase, it is also important to ask, what is the impact
130
of weak and strong El Niño and La Niña phases of the Southern Oscillation with regard
to the spatiotemporal variability of penetrating deep convective clouds? A multi-year
analysis that evaluates El Niño /La Niña impacts on the spatiotemporal variability of
penetrating deep convective clouds may also advance our understanding of these
phenomena and their impacts.
Other areas where the analysis of lower stratospheric water vapor and/or the
impact of penetrating deep convection on lower stratospheric water vapor requires further
evaluation includes the following:
Representation of vertical convective transport in models: Modeling is
used to evaluate our conceptual view of physical processes. Vertical
transport in most models is highly parameterized and is evaluated at
resolutions that are too coarse for deep convection to be explicitly
represented. The evaluation of the representation of tropical deep
convection in atmospheric models by Russo et al. [2011] who compare the
representation of tropical deep convection to various observational data
sets considered to be proxies for deep convection is an excellent basis with
which to determine the schemes that are most useful for characterizing
vertical transport. Then, modeling of the mass flux into the lowermost
stratosphere should be improved.
Cloud thermodynamics and microphysics: A better understanding of
cloud thermodynamics and microphysics is a very important as it relates to
the regulation of observed quantities such as cloud top height and cloud
top temperature. Thermodynamic parameters associated with overshooting
131
deep convection events and their implications for UT/LS exchange should
also be analyzed to statistically characterize and evaluate environmental
properties that determine the intensity and depth of penetration into the
tropical tropopause layer and lower stratosphere. In addition, ice formation
in atmospheric clouds influences the cloud life-cycle, precipitation
processes and cloud radiative properties (e.g., Khvorostyanov and Curry,
2009). For deep overshooting convection, ice is specifically important as it
participates in hydration/dehydration of the lower stratosphere. The impact
of ice formation on cloud buoyancy has been addressed (Brahams, 1952;
Riehl, 1979; Zipser, 2003). Yet for deep convection clouds it is rarely
included in the evaluation of undilute adiabatic ascent or the analysis of
thermodynamic indices designed to diagnose convective available
potential energy (e.g., Luo, 2008; Liu and Zipser, 2005). Such analysis
may be achieved via a simple one dimensional cloud parcel model. The
details of this model are provided in Appendix B.
Improved instruments that can better measure changes in the quantity
of water vapor in the lower stratosphere: There is much uncertainty in the
observations of lower stratospheric water vapor since it is present in such
small concentrations. To reduce the uncertainty of these measurements
more accurate instruments should be developed. For example, in the
evaluation of 82 mb water vapor mixing ratio we have used a combination
of UARS-HALOE and Aura MLS data. For version 1.51 of the MLS
water vapor data, which has an estimated precision of 20% and are reliable
132
at altitudes above 316 mb. The vertical resolution in the upper troposphere
is roughly 3 km and the horizontal resolution is roughly 15 longitude and
1.5 latitude. To better pinpoint the impact of various processes on lower
stratospheric water vapor, higher resolution measurements with less
uncertainty are needed.
Dehydration processes: According to Brewer [1949], all air entering the
stratosphere must be freeze-dried near the tropical tropopause, where
temperatures are regularly below 195 K. However, according to
observations, the stratosphere contains less water than it would if the air
entered at the average minimum tropical temperature. Understanding why
the stratosphere contains less water than it would if the air entered at the
average minimum tropical temperature is still eluding. While several
explanations, have evolved there is still little observational evidence to
strongly support any of the recently proposed processes. Determining the
details of dehydration mechanisms, in conjunction with competing
hydration mechanisms, is also an important area of research that remains
challenging.
Future work focused on each of these areas, could make considerable contributions to our
understanding of penetrating deep convection, the controls of stratospheric water vapor,
and our capability to project how stratospheric water vapor may change in a warming
climate.
133
Appendix A
More Details of the 2B-Geoprof Lidar Product
Although the lidar and radar have different vertical and spatial resolutions, the
spatial domain of the output products in the 2B-Lidar GEOPROF algorithm is defined in
terms of the spatial grid of the CPR, which is represented as some spatial field of
probability. Figure A.1 shows a conceptual downward looking view of the spatial field
of probability in blue with some number of coincident lidar footprints in red. The solid
and dashed circles surrounding the lidar footprints and the solid and dashed ellipses
surrounding the radar footprint represent the 1 and 2 standard deviation pointing
uncertainty in the CPR and Lidar.
As also shown in Figure A.1, the probability field forms an ellipse whereby the
highest probability values within the spatial field are found at the center of the profile and
decrease outward from that point. Lidar observations that overlap the spatial region
enclosing the 2-sigma boundary of the radar observational domain are considered to
contribute to the spatial description of the overlap region. Below 8.2 km, as many as 9-
10 separate lidar profiles will be included in the spatial description of the overlap region
and above 8.2 km, 3-4 profiles will potentially contribute to the hydrometeor description.
The degree to which a lidar observation contributes to a given radar resolution volume is
calculated in terms of the degree to which that particular observation contributed to the
spatial overlap in the radar observational domain. This calculation is computed using a
weighting scheme that is based on the spatial probability of overlap according to:
134
where i counts the lidar profile in a particular radar observational domain, x and y
represent spatial dimensions to form an area enclosing the radar domain, subscripts r and
l represent the radar and lidar, respectively and P is the spatial probability that a
particular element of area defined by x and y contributes to the observation. The lidar
cloud fraction (Cl) within a radar footprint is expressed as a weighted combination of the
lidar observations within the radar probability field according to:
where δi, the lidar hydrometeor occurrence, may have a value of 1, which indicates that
hydrometeors exist or it may have a value of 0 which indicates the nonoccurrence of
hydrometeor. In calculating Cl the number, i, of lidar observations within a particular
radar resolution volume is determined by the total number of lidar profiles that could
potentially contribute to the radar probability field. This quantity, Cl, effectively
quantifies the partial filling of the radar volume by hydrometeors and is one of the output
quantities of the radar-lidar combined product.
Here vertically connected CloudSat bins with cloud mask values (≥ 30) define a
cloud layer. A layer boundary is defined as the first encounter of a cloudy range level
(either radar or lidar) following the occurrence of a cloud-free range level by radar or a
cloud range level with a lidar hydrometeor fraction < 0.5.
A.1
wiPrPl
i
y
dxdy
x
Cl
wi i
i l
#lidarobs
wi
#lidarobs
A.2
135
To illustrate the concept of the lidar hydrometeor fraction see Figure A.2 which
denotes hydrometeor fractions by Cnh where n denotes a lidar range number and five lidar
profiles are included in the radar observational domain. A lidar range is defined as cloudy
if the value Cnh≥ 0.5. This indicates that at least 50% of the lidar profiles within 1 radar
range level are cloudy In Figure A.2, the first cloudy layer encountered moving from
bottom to top would be the second layer with C2h≥ 0.6.
136
Figure A.1 (cf., Mace et al., 2007) Illustration of lidar hydrometeor fractions, Cnh, (in red)
that occur within a CPR range resolution volume (in blue). Lidar hydrometeor fractions
reported for each horizontal level are reported in percentages on the right.
Plan View of Radar-Lidar Footprint
Figure A.2 (cf.,Mace et al., 2007) Conceptual view of CPR-Lidar overlap with radar
footprint in blue and lidar footprints in red. The black (red) solid and dashed ellipses
(circles) represent the 1 and 2 standard deviation pointing uncertainty of the radar
(lidar).
137
Appendix B
Description of Parcel Model
An unsaturated parcel is assumed to begin its ascent absent of liquid or ice
condensate. The parcel is raised from the surface in 1mb increments. For temperatures
above freezing, the parcel maintains at an RH=100% and is not allowed to become
supersaturated with respect to liquid (or ice). Liquid condensate is formed by converting
any excess moisture above saturation with respect to liquid to liquid water according to,
rvrvsl rl
w
where rv is the water vapor mixing ratio, rvsl is the saturation mixing ratio with respect to
liquid water, and rl is the liquid water mixing ratio. At temperatures between 273.15 and
238.15 K, condensation is calculated using a modified version of the theoretical ice
nucleation scheme proposed by ECMWF [2007]. The ECMWF formulation and its
modifications characterize the phase state in clouds in effort to parameterize the liquid, fl,
and ice fractions, fi, of the total condensate. The result of this formulation shows that in
pure liquid clouds with warm temperatures slightly above 0 C, fl is observed to be close
to 100 %. In the case of cold clouds where temperatures decline below T < -35 to – 40 C
clouds become purely crystalline [Borovikov et al., 1963; Pruppacher and Klett, 1997].
For the ECMWF [2007] scheme, the fraction of liquid condensate, fl is calculated
fl[(T Tice )/(T0Tice )]2
B.1
B.2
138
and the fraction of ice condensate, fi, is calculated,
fi1fl
where, fl = 0 at T < Tice,, T0 = 273.16 and Tice has been modified from the ECMWF (2007)
value of 250.16 K to 238.15 K. This modification better corresponds with the climatology
of ice and liquid fractions presented by Borovikov et al. [1963] and Pruppacher and Klett
[1997], ensuring that the cloud parcel is not fully glaciated until temperatures decline
below -35 C. The fractions of liquid and ice phase states are shown in Figure B.1. Note,
the conditions 0 ≤ fl ≤ 1 and 0 ≤ fi ≤ 1 are both strictly enforced.
The mixing ratios of water vapor, ice, and liquid water are then diagnosed by the
following
rvFlrvsl Firvsi
riFi(rtrv)
rlrtrvri
where saturations with respect to liquid water and ice, rvsl and rvsi are calculated according
to,
rvsl
esl
p esl
and
rvsi
esi
p esi
and esl was fit to Wexler‘s results, extrapolated for T<
0 C, to an accuracy of 0.1% for -30 C T 35 C and esi is calculated using the
formulation [Bolton, 1980],
es(T) 6.112exp 17.67
T243.5
and esi is calculated using [Marti and Mauersberger, 1993]
B.3
B.4
B.5
B.6
B.7
139
esi(T) (10^( 2663.5/T)12.537) /100.
The model is further constrained according to the ice liquid water potential temperature
defined by Tripolli and Cotton [1981] as
il 1Lv(T0)rl
cpmax(T,253)
Ls(T0)ri
cpmax(T,253)
1
il, represents reversible adiabatic ascent and is conserved under vapor to liquid, vapor to
ice, and liquid to ice phase changes. For pseudoadiabatic ascent, we use Bolton [1980],
substituting the latent heat of sublimation for that of vaporization when necessary and
also in fractions consistent with fl and fi. At temperatures below 238.15 K, supersaturation
with respect to ice is not allowed. Ice condensate is formed by converting any excess
moisture above saturation with respect to ice to ice according to
rvrvsi ri
As the parcel is lifted, the vertical temperature profile of the cloud is determined by
solving the model equations (B1-B10) using the Broyden method of solving simultaneous
nonlinear systems of equations.
B.8
B.9
B.10
140
Figure B.1. Frequency of Liquid versus Mixed Phase states using modified version of ECMWF
[2007]. At temperatures above 0 C the cloud condensate is all liquid water. Between 0 C and -
35 C condensate is a mixture of ice and liquid water. At temperatures below -35 C the cloud is
fully glaciated.
141
References
Adler, R. F., and R. A. Mack (1986), Thunderstorm cloud top dynamics as inferred from
satellite observations and a cloud top parcel model, J. Atmos. Sci., 43, 1945–1960.
Alcala, C. M., and A. E. Dessler (2002), Observations of deep convection in the tropics
using the Tropical Rainfall Measuring Mission (TRMM) precipitation radar, J.
Geophys. Res., 107, 4792, doi:10.1029/2002JD002457.
Aumann, H. H., S. G., DeSouza–Machado, and A. Behrangi (2011), Deep convective
clouds at the tropopause, Atmos. Chem. Phys., 11, 1167–1176.
Bain, C. L., G. Magnusdottir, P. Smyth, and H. Stern, 2010: Diurnal cycle of the
Intertropical Convergence Zone in the east Pacific. Journal of Geophysical
Research, 115, D23116.
Barrett, E. W., P. M. Kuhn, and A. Shlanta, (1973), Recent measurements of the injection
of water vapor and ozone into the stratosphere by thunderstorms, paper presented
at Second Conference on the Climatic Impact Assessment Program, Department
of Transportation, Washington D. C.
Battaglia, A., and C. Simmer (2008), How does multiple scattering affect the spaceborne
W-band radar measurements at ranges close to and crossing the surface-range?,
IEEE Trans. Geosci. Remote Sens., 46, 1644–1651.
Bedka, K. M., and P. Minnis (2010), GOES 12 observations of convective storm
variability and evolution during the Tropical Composition, Clouds and Climate
Coupling Experiment field program, J. Geophys. Res., 115, D00J13,
doi:10.1029/2009JD013227.
——, J. Brunner, R. Dworak, W. Feltz, J. Otkin, and T. Greenwald (2010), Objective
satellite-based detection of overshooting tops using infrared window channel
brightness temperature gradients, J. Appl. Meteorol. Climatol., 49, 181–202.
Berendes, T. A., J. R. Mecikalski, W. M. MacKenzie, Jr., K. M. Bedka, and U. S. Nair
(2008), Convective cloud identification and classification in daytime satellite
imagery using standard deviation limited adaptive clustering, J. Geophys. Res.,
113, D20207, doi:10.1029/2008JD010287.
Boering, K. A., E. J. Hintsa, S. C. Wofsy, J. G. Anderson, B. C. Daube, A. E. Dessler, M.
Loewenstein, M. P. McCormick et al. (1995), Measurements of stratospheric
carbon dioxide and water vapor at northern latitudes: Implications for
stratosphere-to-troposphere transport. Geophys. Res. Lett., 22, 2737–2740.
Bolton, D. (1980), The computation of equivalent potential temperature, Mon. Weather
Rev., 108, 1046–1053.
142
Borovikov, A. M., I. I. Gaivoronsky, E. G. Zak, V. V. Kostarev, I. P. Mazin, V. E.
Minervin, A. Kh. Khrigian, and S. M. Schmeter (1963), Cloud Physics,
Leningrad, Hydrometeoizdat, (1961), Transl. by Israel Program Scientif.
Translation, U. S. Dept. Commerce, Washington, DC, 392 pp.
Borsche, M., G. Kirchengast, and U. Foelsche (2007), Tropical tropopause climatology as
observed from radio occultation measurements from CHAMP compared with
ECMWF and NCEP, Geophys. Res. Lett., 34, L03702,
doi:10.1029/2006GL027918.
Brewer, A. W. (1949), Evidence for a world circulation provided by the measurements of
helium and water vapour distribution in the stratosphere, Q. J. R. Meteorol. Soc.
75, 351–363.
Brunner, J. C., S. A. Ackerman, A. S. Bachmeier, and R. M. Rabin (2007), A quantitative
analysis of the enhanced-V feature in relation to severe weather, Weather
Forecast., 22, 853–872.
Chaboureau, J. P., J. P Cammas, J. Duron, P. J. Mascart, N. M. Sitnikov, and H. J.
Voessing (2007), A numerical study of tropical cross-tropopause transport by
convective overshoots, Atmos. Chem. Phys, 7, 1731–1740.
Chen, T., W. B. Rossow, and Y. Zhang (2000), Radiative effects of cloud-type variations,
J. Climate, 13, 264–286.
Chou, C., and J. D. Neelin (1999), Cirrus detrainment-temperature feedback. Geophys.
Res. Lett., 26, 1295–1298.
Chung, E. S., B. J. Sohn, and J. Schmetz (2008), CloudSat shedding new light on high
reaching tropical deep convection observed with Meteosat, Geophys. Res. Lett.
35, L02814, doi:10.1029/2007GL032516.
Corti, T., B. P. Luo, Q. Fu, H. Vömel, and T. Peter (2006), The impact of cirrus clouds on
tropical troposphere-to-stratosphere transport, Atmos. Chem. Phys, 6, 2539–2547.
——, B. P. Luo, M. de Reus, D. Brunner, F. Cairo, M. J. Mahoney, G. Martucci, R.
Matthey et al. (2008), Unprecedented evidence for deep convection hydrating the
tropical stratosphere, Geophys. Res. Lett. 35, L10810,
doi:10.1029/2008GL033641.
Danielsen, E. F. (1982), A dehydration mechanism for the stratosphere, Geophys. Res.
Lett., 9, 605–608.
—— (1993), In situ evidence of rapid, vertical, irreversible transport of lower
tropospheric air into the lower tropical stratosphere by convective cloud turrets
143
and by larger-scale upwelling in tropical cyclones, J. Geophys. Res., 98, 8665–
8681.
Dessler, A. E. (2002), The effect of deep, tropical convection on the tropical tropopause
layer, J. Geophys. Res., 107(D3), doi:10.1029/2001JD000511.
—— and S. C. Sherwood (2003), A model of HDO in the tropical tropopause layer.
Geophys. Res. Lett., 31, 2173–2181.
Donner, L. J., C. J. Seman, R.S. Hemler, and S. Fan (2001), A cumulus parameterization
including mass fluxes, convective vertical velocities, and mesoscale effects:
Thermodynamic and hydrological aspects in a general circulation model, J.
Climate, 14, 3444–3463.
——, L. W. Horowitz, A. M. Fiore, C. J. Seman, D. R. Blake, and N. J. Blake (2007),
Transport of radon-222 and methyl iodide by deep convection in the GFDL
Global Atmospheric Model AM2, J. Geophys. Res., 112, D17303,
doi:10.1029/2006JD007548.
Folkins, I., and R. V. Martin (2004), The vertical structure of tropical convection and its
impact on the budgets of water vapor and ozone, J. Atmos. Sci., 62,1560–1573.
——, M. Loewenstein, J. Podolske, S. J. Oltmans, and M. Proffitt (1999), A barrier to
vertical mixing at 14 km in the tropics: Evidence from ozonesondes and aircraft
measurements, J. Geophys. Res., 104, 22,095–22,102.
Forster, P. M. F., and K. P. Shine (1999), Stratospheric water vapour changes as a
possible contributor to observed stratospheric cooling, Geophys. Res. Lett., 26,
3309–3312.
——, and K. P. Shine (2002), Assessing the climate impact of trends in stratospheric
water vapor, Geophys. Res. Lett, 29, 6, doi:10.1029/2001GL013909.
Fritz., S., and I. Laszlo (1993), Detection of water vapor in the stratosphere over very
high clouds in the tropics, J. Geophys. Res., 98, 22,959–22,967.
Fu, R., A. D. Del Genio, and W. Rossow (1990), Behavior of deep convective clouds in
the tropical Pacific deduced from ISCCP radiances, J. Climate, 3, 1129–1152.
Fueglistaler, S., A.E. Dessler, T. Dunkerton, I. Folkins, Q. Fu, P.W. Mote (2009),
Tropical Tropopause Layer, Rev. Geophys., 47, RG1004,
doi:10.1029/2008RG000267.
——, and P. H. Haynes (2005), Control of interannual and longer-term variability of
stratospheric water vapor, J. Geophys. Res., 110, D24108,
doi:10.1029/2005JD006019.
144
——, H. Wernli, and T. Peter (2004), Tropical troposphere-to-stratosphere transport
inferred from trajectory calculations, J. Geophys. Res., 109, D03108,
doi:10.1029/2003JD004069.
Fujita, T. T. (1982), Principle of stereoscopic height computations and their applications
to stratospheric cirrus over severe thunderstorms, J. Meteor. Soc. Japan, 60, 355–
368.
Gettelman, A., M. L. Salby, and F. Sassi (2002), Distribution and influence of convection
in the tropical tropopause region, J. Geophys. Res., 107, 4080,
doi:10.1029/2001jd001048.
——, E. J. Fetzer, A. Eldering, and F. W. Irion (2006), The global distribution of
supersaturation in the upper troposphere from the atmospheric infrared sounder, J.
Climate, 19, 6089–6103.
——, T. Birner, V. Eyring, H. Akiyoshi, S. Bekki, C. Brühl, M. Dameris, D. E. Kinnison
et al. (2009), The tropical tropopause 1960-2100, Atmos. Chem. Phys., 9, 1621–
1637.
——, M. I. Hegglin, S.-W. Son, J. Kim, M. Fujiwara, T. Birner, S. Kremser, M. Rex et
al. (2010), Multi-model assessment of the upper troposphere and lower
stratosphere: Tropics and global trends, J. Geophys. Res., 115, D00M08,
doi:10.1029/2009JD013638.
Gierens, K., R. Kohlhepp, P. Spichtinger, and M. Schroedter-Homscheidt (2004), Ice
supersaturation as seen from TOVS. Atmos.Chem. Phys., 4, 539–547.
Gould-Stewart, R. E. N. a. S. (1981), A stratospheric rountain?, J. Atmos. Sci., 38, 2789–
2796.
Grosvenor, D. P., T. W. Choularton, H. Coe, and G. Held (2007), A study of the effect of
overshooting deep convection on the water content of the TTL and the lower
stratosphere from cloud resolving model simulations, Atmos. Chem. Phys.
Discuss., 7, 7277–7346.
Hansen, J., M. Sato, R. Ruedy, L. Nazarenko, A. Lacis, G. A. Schmidt, G. Russell, I.
Aleinov et al. (2005), Efficacy of climate forcings, J. Geophys. Res., 110,
D18104, doi:10.1029/2005JD005776.
Hartmann, D.L., M. E. Ockert-Bell, and M. L. Michelsen (1992), The effect of cloud type
on Earth‘s energy balance: global analysis. J. Climate, 5, 1281–1304.
Heymsfield, A. J. (1993), Microphysical structures of stratiform and cirrus clouds, in
Aerosol-Cloud-Climate Interactions, edited P. V. Hobbs, pp. 97–122, Academic
145
Press, New York.
Heymsfield, G. M., R. Fulton, and J. D. Spinhirne (1991), Aircraft overflight
measurements of midwest severe storms: Implications on geosynchronous
satellite interpretations, Mon. Weather Rev., 119, 436–456.
Highwood, E. J., and B. B. Hoskins (1998), The tropical tropopause, Q. J. Roy. Meteorol.
Soc., 124, 1579–1604.
Holton, J. R., and A. Gettelman (2001), Horizontal transport and the dehydration of the
stratosphere, Geophys. Res. Lett., 28, 2799–2802.
——, P.H. Haynes, M. E. McIntyre, A. R. Douglass, R. B. Rood, and L. Pfister (1995),
Stratosphere-troposphere exchange, Rev. Geophys., 33, 403–439.
Hong, G., G. Heygster, J. Miao, and K. Kunzi (2005), Detection of tropical deep
convective clouds from AMSU-B water vapor channels measurements, J.
Geophys. Res., 110, D05205, doi:10.1029/2004JD004949.
——, G. Hyegster, and C. A. M. Rodriguez (2006), Effect of cirrus clouds on the diurnal
cycle of tropical deep convective clouds, J. Geophys. Res., 111, D06209,
doi:10.1029/2005JD006208.
——, P. Yang, B. C. Gao, B. A., Baum, Y. X. Hu, M. D. King, and S. Platnick (2007),
High cloud properties from three years of MODIS terra and aqua collection-4 data
over the tropics, J. Appl. Meteorol. Climatol., 46, 1840–1856.
Intergovernmental Panel on Climate Change (IPCC) (2007), Climate Change 2007: The
Physical Science Basis, edited by S. Solomon et al., Cambridge Univ. Press, New
York.
Jensen, E., and L. Pfister (2004), Transport and freeze-drying in the tropical tropopause
layer, J. Geophys. Res., 109, D02207, doi:10.1029/2003JD004022.
Jiang, J. H., B. Wang, K. Goya, K. Hocke, S. G. Eckermann, J. Ma, D. L. Wu, and W. G.
Read (2004), Geographical distribution and interseasonal variability of tropical
deep convection: UARS MLS observations and analyses, J. Geophys. Res., 110,
D05205, doi:10.1029/2004JD03756.
Johnston, H. S., and S. Solomon (1979), Thunderstorms as possible micrometeorological
sink for stratospheric water, J. Geophys. Res., 84, 3155–3158.
Kahn, B. H., M. T. Chahine, G. L. Stephens, G. G. Mace, R. T. Marchand, Z. Wang,
C. D. Barnet, A. Eldering et al. (2008), Cloud type comparisons of AIRS,
CloudSat, and CALIPSO cloud height and amount, Atmos. Chem. Phys., 8, 1231–
1248.
146
Khaykin, S., J.-P. Pommereau, L. Korshunov, V. Yushkov, J. Nielsen, N. Larsen, T.
Christensen, A. Garnier et al. (2009), Hydration of the lower stratosphere by ice
crystal geysers over land convective systems, Atmos. Chem. Phys., 9, 2275–2287.
Khvorstyanov, V. I., and J. A. Curry (2009), Critical humidities of homogeneous and
heterogeneous ice nucleation: Inferences from extended classical nucleation
theory, J. Geophys. Res., 114, D04207, doi: 10.1029/2008JD011197.
Kirk-Davidoff, D. B., E. J. Hintsa, J. G. Anderson, and D. W. Keith (1999), The effect of
climate change on ozone depletion through changes in stratospheric water vapour,
Nature, 402, 399–401.
Kley, D., J. M. Russell, and C. Phillips (2000), SPARC assessment of upper tropospheric
and stratospheric water vapour, WCRP Report No. 113, WMO/TD Report No.
1043, World Climate Research Programme, Geneva.
Knapp, K. R. (2008a), Calibration of long-term geostationary infrared observations using
HIRS, J. Atmos. Oceanic Technol., 25, 183–195.
—— (2008b), Scientific data stewardship of international satellite cloud climatology
project B1 global geostationary observations, J. Appl. Remote Sensing, 2(023458),
doi: 10.1117/1.3043461.
Knollenberg, R. G., K. Kelly, and J. C. Wilson (1993), Measurements of high number
densities of ice crystals in the tops of tropical cumulonimbus, J. Geophys. Res.,
98(D5), 8639–8664.
Kossin, J. P., K. R. Knapp, D. J. Vimont, R. J. Murnane, and B. A. Harper, 2007a: A
globally consistent reanalysis of hurricane variability and trends. Geophysical
Research Letters, 34, L04815.
Kubar, T., D. L. Hartmann, and R. Wood (2007), Radiative and convective driving of
tropical high clouds, J. Climate, 20, 5510–5517.
Li, W., and C. Schumaker (2010), Thick anvils as viewed by the TRMM precipitation
radar, J. Climate, 24, 1718–1735.
Liu, C., and E. J. Zipser (2005), Global distribution of convection penetrating the tropical
tropopause, J. Geophys. Res., 110, D23104, doi:10.1029/2005JD006063.
——, E. J. Zipser, and S. W. Nesbitt (2007), Global distribution of tropical deep
convection: Different perspectives from TRMM infrared and radar data, J.
Climate, 20, 489–503.
Liu, X. M., E. D. Rivière, V. Marécal, G. Durry, A. Hamdouni, J. Arteta, and S. Khaykin
147
(2010), Stratospheric water vapour budget and convection overshooting the
tropopause: modeling study from SCOUT-AMMA, Atmos. Chem. Phys., 10,
8267–8286.
Luo, Z., G. Y. Liu, and G. L. Stephens (2008), CloudSat adding new insight into tropical
penetrating convection, Geophys. Res. Lett., 35, L19819,
doi:10.1029/2008GL035330.
——, G. Y. Liu, and G. L. Stephens (2010), Use of A‐Train data to estimate convective
buoyancy and entrainment rate, Geophys. Res. Lett., 37, L09804,
doi:10.1029/2010GL042904.
Mapes, B. E., and R. A. J. Houze (1993), Cloud clusters and superclusters over the
oceanic warm pool, Mon. Weather Rev., 121, 1398–1415.
Marti, J., and K. Mauersberger (1993), A survey and new measurements of ice vapor
pressure at temperatures between 170 and 250 K, Geophys. Res. Lett., 20, 363–
366.
Murphy, D. M., and T. Koop (2005), Review of the vapor pressures of ice and
supercooled water for atmospheric applications, Q. J. R. Meteorol. Soc., 131,
1539–1565.
Nesbitt, S., E. Zipser, and D. Cecil (2000), A census of precipitation features in the
Tropics using TRMM: Radar, ice scattering, and lightning observations, J.
Climate, 13, 4087–4106.
Oltmans, S., H. Vömel, D. Hoffman, K. Rosenlof, and D. Kley (2000), The increase in
lower-stratospheric water vapor from balloon-borne frostpoint hygrometer
measurements at Washington, D. C. and Boulder, Colorado, Geophys. Res. Lett.,
27, 3453–3456.
Pruppacher, H. R., and J. D. Klett (1997), Microphysics of Clouds and Precipitation, 2nd
ed., Kluwer, Norwell, Mass.
Randel, W. J., F. Wu, S. J. Oltmans, K. Rosenlof, and G. E. Nedoluha (2004), Interannual
variability of stratospheric water vapor and correlations with tropical tropopause
temperatures, J. Atmos. Sci., 61, 2133–2148.
——, F. Wu, H., Vömel, G. E. Nedoluha, and P. Forster (2006), Decreases in
stratospheric water vpaor after 2001: Links to changes in the tropical tropopause
and the Brewer Dobson circulation, J. Geophys. Res., 111, D12312,
doi:10.1029/2005JD006744.
148
Read, W. G., D. L. Wu, J. W. Waters, and H. C. Pumphrey (2004), Dehydration in the
tropical tropopause layer: Implications from the UARS Microwave Limb
Sounder, J. Geophys. Res, 109, (D06110), doi:10.1029/2003JD004056.
Riehl, H. (1979), Climate and Weather in the Tropics, Academic Press, London.
Roca, R., M. Viollier, L. Picon, and M. Desbois (2002), A multisatellite analysis of deep
convection and its moist environment over the Indian Ocean during the winter
monsoon, J. Geophys. Res., 107, 8012, doi:10.1029/2000JD000040.
Röckmann, T., J. Grooss, and R. Muller (2004), The impact of anthropogenic chlorine
emissions, stratospheric ozone change and chemical feedbacks on stratospheric
water, Atmos. Chem. Phys., 4, 693–699.
Rohs, S., C. Schiller, M. Riese, A. Engel, U. Schmidt, U. Wetter, I. Levin, T. Nakazawa,
and S. Aoki (2006), Long-term changes of methane and hydrogen in the
stratosphere in the period 1978-2003 and their impact on the abundance of
stratospheric water vapor, J. Geophys. Res., 111, D14315,
doi:10.1029/2005JD00687.
Rosenlof, K. H. (2003), How water enters the stratosphere, Science, 302, 1691–1692.
——, and G. C. Reid (2008), Trends in the temperature and water vapor content of the
tropical lower stratosphere: Sea surface connection, J. Geophys. Res., 113,
D06107, doi:10.1029/2007JD009109.
——, S. J. Oltmans, D. Kley, J. M. Russell, E. W. Chiou, W. P. Chu, D. G. Johnson, K.
K. Kelly et al. (2001), Stratospheric water vapor increases over the past half-
century, Geophys. Res. Lett., 28, 1195–1198.
Rossow, W. B., and C. Pearl (2007), 22-Year survey of tropical convection penetrating
into the lower stratosphere, Geophys. Res. Lett., 34, L04803,
doi:10.1029/2006GL028635.
——, and R. A. Schiffer (1999), Advances in understanding clouds from ISCCP, Bull.
Amer. Meteorol. Soc., 80, 2261–2287.
——, G. Tselioudis, A. Polak, and C. Jakob (2005), Tropical climate described as a
distribution of weather states indicated by distinct mesoscale cloud property
mixtures, Geophys. Res. Lett., 32, L21812, doi:10.1029/2005GL024584.
Russell, J. M, L. Gordley, J. H. Park, S. R. Drayson, W. D. Hesketh, R. J. Cicerone, A. F.
Tuck, J. E. Frederick et al. (1998), The halogen occultation experiment, J.
Geophys. Res., 98 (D6), 10,777–10,797.
Sassen, K., and Z. Wang, (2008), Classifying clouds around the globe with the CloudSat
149
radar: 1-year of results. Geophys. Res. Lett., 35, L04805,
doi:10.1029/2007GL032591.
Schiffer, R. A., and W. B. Rossow (1999), ISCCP global radiance data set: A new
resource for climate research, Bull. Amer. Meteorol. Soc., 66, 1498–1505.
Schmetz, J., S. A. Tjemkes, M. Gube, and L. van de Berg (1997), Monitoring deep
convection and convective overshooting with METEOSAT, Adv. Space Res., 19,
433–441.
Setvák, M., and R. M. Rabin (2003), MODIS observations of deep convective cloud tops,
paper presented at 2003 EUMETSAT Meteorological Satellite Conference,
Weimar, Germany, <http://www.chmi.cz/meteo/sat/setvak/pub_ms.html>
——, R. M. Rabin, and P. K. Wang (2007), Contribution of the MODIS instrument to
observations of deep convection storms and stratospheric moisture detection in
GOES and MSG imagery, Atmos. Res. 83, 505–518.
Sherwood, S. C., and A. E. Dessler (2000), On the control of stratospheric humidity,
Geophys. Res. Lett., 27, 2513–2516.
——, J.-H. Chae, P. Minnis, and M. McGill (2004), Underestimation of deep convective
cloud tops by thermal imagery, Geophys. Res. Lett., 31, L11102,
doi:10.1029/2004GL019699.
Shindell, D. T. (2001), Climate and ozone response to increased stratospheric water
vapour, Geophys. Res. Lett., 28, 179–182.
——, G. Faluvegi, N. Bell, and G. Schmidt (2005), An emissions–based view of climate
forcing by methane and tropospheric ozone, Geophys. Res. Lett., 32, L04803,
doi:10.1029/2004GL021900
Smith, C. A., J. D. Haigh, and R. Toumi (2001), Radiative forcing due to trends in
stratospheric water vapour, Geophys. Res. Lett., 28, 179–182.
Soden, B. J. (2000), Diurnal cycle of convection, clouds, and water vapor in the tropical
upper troposphere, Geophys. Res. Lett., 27, 2173–2176.
Sohn, B. J., and J. Schmetz (2004), Water vapor-induced OLR variations associated with
high cloud changes over the Tropics: A study from Meteosat-5 observations. J.
Climate, 17, 1987–1996.
Solomon, S., K. Rosenlof, R. Portmann, J. Daniel, S. Davis, T. Sanford, and J. Plattner
(2010), Contributions of stratospheric water vapor to decadal changes in the rate
of global warming, Science, 327, 1219–1223.
150
Spichtinger, P., K. Gierens, and W. Read (2003), The global distribution of ice
supersaturated regions as seen by the Microwave Limb Sounder. Q. J. R. Meteor.
Soc., 129, 3391–3410.
SPARC (2000), Assessment of upper tropospheric and stratospheric water vapour,
WMO/TD No. 1043, (SPARC Report #2), Geneva.
Stephens, G. L., D. G. Vane, S. Tanelli, E. Im, S. Durden, M. Rokey, D. Reinke, P.
Partain et al. (2008), The CloudSat Mission: Performance and early science after
the first year of operation, J. Geophys. Res., 113, D00A18,
doi:10.1029/2008JD009982.
Stuber, N., M. Ponater, and R. Sausen (2001), Is the climate sensitivity to ozone
perturbations enhanced by stratospheric water vapor feedback?, Geophys. Res.
Lett., 28, 2887–2890.
Tripoli, G. J., and W. R. Cotton (1981), The use of ice-liquid water potential temperature
as a thermodynamic variable in deep atmospheric models. Mon. Weather Rev.,
109, 1094–1102.
Tselioudis, G., E. Tromeur, W. B. Rossow, and C. S. Zerefos (2010), Decadal changes in
tropical convection suggest effects on stratospheric water vapor, Geophys. Res.
Lett., 37, L14806, doi: 0.1029/2010GL044092.
Wang, P. K. (2003), Moisture plumes above thunderstorm anvils and their contributions
to cross tropopause transport of water vapor in midlatitudes, J. Geophys. Res.,
108, 4194, doi:10.1029/2003JD002581.
—— (2004), A cloud model interpretation of jumping cirrus above storm top, Geophys.
Res. Lett., 31, L18106. doi:10.1029/2004GL020787.
—— (2007), The thermodynamic structure atop a penetrating convective thunderstorm.
Atmos. Res., 83, 254–262.
——, M. Setvák, W. Lyons, W. Schmid, and H. Lin (2009), Further evidences of deep
convective vertical transport of water vapor through the tropopause, Atmos. Res.,
94, 400–408.
Wang, Z., and K. Sassen (2007), Level 2 cloud scenario classification product process
description and interface control document, Version 5.0, 50 pp.,
<http://www.cloudsat.cira.colostate.edu/ dataICDlist.php?go=list&path=/2B-
CLDCLASS>
Waters, J. W., L. Froidevaux, R. S Harwood, R. F. Jarnot, H. M. Pickett, W. G. Read, P.
H. Siegel, R. E. Cofield et al. (2006), The Earth Observing System Microwave
151
Limb Sounder (EOS MLS) on the Aura satellite, IEEE Trans. Remote Sens., 44,
1075–1092.
Webster, P. J., and H. Chang (1988), Equatorial energy accumulation and emanation
regions: Impacts of a zonally varying basic state, J. Atmos. Sci., 45, 803–829.
——, G. J. Holland, J. A. Curry, and H.-R. Chang (2005), Changes in tropical cyclone
number, duration, and intensity in a warming environment, Science, 309, 1844–
1846.
Weisz, E., J. Li, W. P. Menzel, A. K. Heidinger, B. H. Kahn, and C.-Y. Liu (2007),
Comparison of AIRS, MODIS, CloudSat and CALIPSO clout top height
retrievals, Geophys. Res. Lett., 34, L17811, doi:10.1029/2007GL030676.
Yuan, J., and R. A. Houze, Jr., and A. Heymsfield (2011), Vertical structures of anvil
clouds of tropical mesoscale convective systems observed by CloudSat, J. Atmos.
Sci., doi:10.1175/2011JAS3687.1, in press.
——, and R. A. Houze, Jr. (2010), Global variability of mesoscale convective system
anvil structure from A-Train satellite data, J. Climate, 23, 5864–5888.
Yuan, T., and Z. Li (2010), General macro- and microphysical properties of deep
convective clouds as observed by MODIS, J. Climate, 23, 3457–3473.
Zipser, E. J. (2003), Some view on ―hot towers‖ after 50 years of tropical field programs
and two years of TRMM data, in Cloud Systems Hurricanes, and the Tropical
Rainfall Measuring Mission (TRMM), edited by W.-K. Tao and R. Adler,
Meteorol. Monogr., 29, 49–58.
——, D. J. Cecil, C. Liu, S. W. Nesbitt, and D. P. Yorty (2006), Where are the most
intense hurricanes on Earth?, Bull. Am. Meteorol. Soc., 87, 1057–1071.
