Content uploaded by Frank Decatur Marks
Author content
All content in this area was uploaded by Frank Decatur Marks
Content may be subject to copyright.
Content uploaded by Frank Decatur Marks
Author content
All content in this area was uploaded by Frank Decatur Marks
Content may be subject to copyright.
7D.2 DEVELOPMENT OF A TROPICAL CYCLONE RAINFALL CLIMATOLOGY
AND PERSISTENCE (R-CLIPER) MODEL
Frank Marks
1
and Gretchen Kappler
NOAA/AOML, Hurricane Research Division, Miami, FL
Mark DeMaria
NOAA/NESDIS/ORA, Fort Collins, CO
1. INTRODUCTION
A tropical cyclone (TC) drops copious amounts of
precipitation daily. The rainfall over the oceans is not
a serious concern to society, but once a TC makes
landfall, and even several hundred miles inland after
the hurricane force winds have ceased, it can produce
a serious threat to society from urban and inland
flooding. Indeed, in the last 30 years the majority of
deaths related to TCs have been attributed to flooding
(Rappaport 2000).
Improved quantitative precipitation forecasts (QPF)
in TCs are one of the primary goals of the U. S.
Weather Research Program (USWRP) effort on TC
landfall (Marks and Shay, 1999). The varied nature of
precipitation makes the QPF topic very complex.
Although much of the significant precipitation occurs
in conjunction with convective clouds, stratiform
clouds also account for significant precipitation
accumulations over extended intervals. While all of
these mechanisms are active in TCs, the vortex
structure acts to dynamically constrain the smaller
scale circulations that often confound QPF.
Estimates of rainfall based on radar and other
remote sensors offer promising avenues for
improvement. At the same time numerical models,
both operational and research, have been improved,
running on faster computers at higher resolution, and
with improved model physics. However, there is little
previous work validating the model QPF performance
in TCs. Partly it is because of the lack of accurate
observations of the rain distribution and evolution in
TCs.
A major obstacle to improving QPF in TCs is a lack
of a comprehensive climatology of TC precipitation,
i.e., a description of the distribution of rain in space
and time. Few precipitation climatologies exist for TCs
in the United States, and other TC basins have
similarly limited climatologies. However, remote
sensors such as those on the NASA Tropical Rain
Measurement Mission (TRMM) satellite; in particular
the microwave imager (TMI) and the active
precipitation radar (PR), are providing a first cut at a
credible TC rain climatology (Lonfat et al 2000). This
climatology is precisely what is needed to develop a
simple rainfall climatology and persistence (R-
CLIPER) model, which can be used to validate
numerical models and other QPF methods.
1
Corresponding Author
: Frank Marks, NOAA/ AOML,
Hurricane Research Division, 4301 Rickenbacker
Causeway, Miami, FL 33149, email: Frank.Marks@noaa.gov
2. DATA AND METHOD OF ANALYSIS
2.1 Rain Gauge Data
A TC R-CLIPER was developed originally based
upon 53 years of U.S. rain gauge data (DeMaria and
Tuleya, 2001). The gauge data set includes 125 U.S.
landfalling storms from 1948-2000. There were about
10
6
hourly rain gauge reports within 500 km of the
storm center. Only storms that were hurricanes at
landfall are included (46 storms). There were about
5.6X10
5
hourly gauge reports within 500 km of the
storm center for these cases. The gauge data was
stratified into 50 10-km wide annuli surrounding the
storm center and mean rainfall rates were computed
for each annuli (Fig. 1). In order to handle storms after
they made landfall an inland-decay model was
developed using the rain gauge data set by stratifying
the results by time after landfall. The rainfall
climatology was reduced to a linear fit of the mean
rainfall rates (R) by radius (r) and time (t) after landfall
defined as:
R(r,t) = (ae
-at
+b)e
-(r-rm)/re
(1)
where parameters a and a are defined from the fit to
the gauge data in time, and b by the fit to the gauge
data by radius, where r
m
is the radius of maximum
rainfall (=0) and r
e
=500 km. This approach results in a
circularly symmetric rain distribution that can be
combined with the forecast track to produce a swath
of rain along the forecast track before and after
landfall. The climatology was combined with the
operational track forecasts through the automated
Fig. 1. Gauge-based TC rainfall climatology (in day
-1
)
from 1948-2000 and TRMM TMI-based TC rain
climatology (in day
-1
) from December 1997-2000.
Comparisons are made for tropical storms and
hurricanes.
tropical cyclone forecast (ATCF) system used at
TPC/NHC to compute an integrated rain distribution
for each forecast interval to produce the R-CLIPER.
The gauge-based R-CLIPER model was
implemented at NHC in September 2001. However,
because the hourly gage data is sparse, particularly
within 100 km of the storm center, it is difficult to
obtain a large enough sample to stratify the data by
storm intensity.
2.2 TRMM Data
To overcome this limitation the R-CLIPER was
expanded to include a global satellite-based TC
rainfall climatology based on rain estimates from the
NASA Tropical Rain Measurement Mission (TRMM)
satellite; in particular the microwave imager (TMI)
(Lonfat et al 2000). To date, this climatology includes
global TMI rain estimates in 245 storms from
December 1997 to December 2000, yielding 2121
events, where 64% of the events were tropical
storms, 26% were category 1-2 hurricanes, and 10%
were category 3 or higher.
The climatology provides a mean rain rate and the
rain rate probability distribution in a storm-centered
coordinate system composed of 50 10-km wide annuli
in four quadrants. The results are stratified as a
function of storm intensity (Fig. 2). The results show
that the mean rain rate increases by a factor of four (3
in day
-1
for tropical storms versus 12 in day
-1
for
category 3 and higher TCs) within 50 km of the storm
center with increasing intensity. Figure 2 also
indicates that the radius of maximum rainfall (r
m
) also
decreased with increasing intensity (i.e., from 55 km
for tropical storms, to 45 km for category 1-2
hurricanes, and to 28 km for category 3-5 hurricanes).
A comparison of the satellite-based to the gauge-
based climatology for all hurricanes and tropical
storms depicted in Fig. 1 denotes a surprising
similarity between the two mean rainfall rate curves
with radius and intensity. The major difference is the
high variability of the mean rainfall rate at radii<100
km in the gauge climatology, particularly for tropical
storms, which is caused by the low number of points
in those annuli. There is also a lack of a minimum in
the mean rainfall rate at small radii for the gauge-
based climatology. Despite these slight differences,
this comparison gives a good indication of the veracity
of both climatologies.
The satellite-based R-CLIPER uses the TMI rain
climatology partitioned by storm intensity developed
by Lonfat et al (2000) to provide the storm-centered
mean rain rate distribution out to 500 km radius (r
e
)
from the storm center as a function of radius:
R(r) = (R
0
) + (R
m
-R
0
)(r/r
m
) r<r
m
= R
m
exp(-(r-r
m
)/r
e
) r>r
m
(2)
where parameters R
0
, and R
m
, are the mean rainfall
rates at r
e
and r
m
, respectively. The climatology was
Fig. 2. TMI-based rainfall climatology (in day
-1
) for
tropical storms, Category 1-2, and Category 3-5
hurricanes.
combined with the operational track forecasts through
the ATCF in the same manner as the gauge-based
climatology to compute an integrated rain distribution
for each forecast interval to produce the R-CLIPER.
3. SUMMARY
An important use of the R-CLIPER is to provide a
benchmark for the evaluation of other more-general
QPF techniques. To evaluate the R-CLIPER forecasts
it will be run on a number of past storms to provide
some statistics on model performance and to develop
different data products useful to the hurricane
specialists. With the help of representatives of the
NWS Hydrometeorological Prediction Center (HPC)
and TPC/NHC we selected five cases to test R-
CLIPER: Andrew (1992), Fran (1996), Danny (1997),
Floyd (1999), and Allison (2001). The R-CLIPER
forecasts will be compared with 6-h areal average
rainfall amounts on a 1°X1° grid, used by HPC. This
type of information is available back 10-12 years.
Comparisons with storm total rain gauge data will also
be performed.
Acknowledgement
: The authors would like to thank
Michelle Mainelli of TPC/NHC for her efforts to
implement the R-CLIPER.
4. REFERENCES
DeMaria, M. D., and R. Tuleya, 2001:Evaluation of
quantitative precipitation forecasts from the GFDL
hurricane model. Reprints
Symposium on
Precipitation Extremes: Predictions, Impacts, and
Responses
, AMS, Albuquerque, NM, 340-343.
Lonfat, M., F. D. Marks, S. Chen, 2000: Study of the
rain distribution in tropical cyclones using
TRMM/TMI. Preprints
24th Conference on
Hurricanes and Tropical Meteorology
, AMS, Ft.
Lauderdale, FL, 480-481.
Marks, F. D., L. K. Shay, and PDT-5, 1998:
Landfalling Tropical Cyclones: Forecast Problems
and Associated Research Opportunities.
Bull.
Amer. Meteor. Soc.
, 79, 305-323.
Rappaport, E.N, 2000: Loss of Life in the United
States Associated with Recent Atlantic Tropical
Cyclones.
Bull. Amer. Meteor. Soc.
, 81,
2065–2074.
