Content uploaded by Andrew G. Turner
Author content
All content in this area was uploaded by Andrew G. Turner on Jun 19, 2015
Content may be subject to copyright.
1 3
DOI 10.1007/s00382-015-2614-1
Clim Dyn
The resolution sensitivity of the South Asian monsoon
and Indo‑Pacific in a global 0.35° AGCM
Stephanie J. Johnson1 · Richard C. Levine2 · Andrew G. Turner1 · Gill M. Martin2 ·
Steven J. Woolnough1 · Reinhard Schiemann1 · Matthew S. Mizielinski2 ·
Malcolm J. Roberts2 · Pier Luigi Vidale1 · Marie‑Estelle Demory1 · Jane Strachan2
Received: 11 September 2014 / Accepted: 10 April 2015
© The Author(s) 2015. This article is published with open access at Springerlink.com
and the Maritime Continent results in more precipitation over
the Maritime Continent islands at the expense of reduced pre-
cipitation further north. We also evaluate the resolution sen-
sitivity of monsoon depressions and lows, which contribute
more precipitation over northeast India at higher resolution.
We conclude that while increasing resolution at these scales
does not solve the many monsoon biases that exist in GCMs,
it has a number of small, beneficial impacts.
Keywords Asian monsoon · Maritime Continent ·
Orography · High resolution · African highlands · Monsoon
depressions
1 Introduction
In global circulation model (GCM) assessments such as
the third and fifth Coupled Model Intercomparison Projects
(CMIP3 and CMIP5, Meehl et al. 2007; Taylor et al. 2012)
large biases remain in the simulation of important climate
systems such as the South Asian monsoon (Sperber et al.
2013). These biases span many temporal and spatial scales
and are observed in coupled and atmosphere-only GCMs
(Gadgil and Sajani 1998; Annamalai et al. 2007; Randall
et al. 2007; Sperber et al. 2013). Biases in simulated sea-
sonal mean monsoon precipitation include excess precipi-
tation over the equatorial Indian Ocean, low precipitation
over the Indian subcontinent and excess precipitation over
orography such as the southern slopes of the Himalayas
(Sperber and Palmer 1996; Gadgil and Sajani 1998; Anna-
malai et al. 2007; Bollasina and Nigam 2009; Sperber et al.
2013). GCMs’ simulated monsoon onsets are typically too
late (Sperber et al. 2013) and the boreal summer intrasea-
sonal oscillation (BSISO), which has a particularly large
socio-economic impact in South Asia, is often weak or not
Abstract The South Asian monsoon is one of the most
significant manifestations of the seasonal cycle. It directly
impacts nearly one third of the world’s population and also
has substantial global influence. Using 27-year integrations
of a high-resolution atmospheric general circulation model
(Met Office Unified Model), we study changes in South
Asian monsoon precipitation and circulation when horizon-
tal resolution is increased from approximately 200–40 km at
the equator (N96–N512, 1.9°–0.35°). The high resolution,
integration length and ensemble size of the dataset make this
the most extensive dataset used to evaluate the resolution sen-
sitivity of the South Asian monsoon to date. We find a con-
sistent pattern of JJAS precipitation and circulation changes
as resolution increases, which include a slight increase in
precipitation over peninsular India, changes in Indian and
Indochinese orographic rain bands, increasing wind speeds
in the Somali Jet, increasing precipitation over the Mari-
time Continent islands and decreasing precipitation over
the northern Maritime Continent seas. To diagnose which
resolution-related processes cause these changes, we com-
pare them to published sensitivity experiments that change
regional orography and coastlines. Our analysis indicates that
improved resolution of the East African Highlands results in
the improved representation of the Somali Jet and further sug-
gests that improved resolution of orography over Indochina
Stephanie J. Johnson was previously known as Stephanie J. Bush.
* Stephanie J. Johnson
s.j.bush@reading.ac.uk
1 National Centre for Atmospheric Science - Climate
Directorate, Department of Meteorology, University
of Reading, Earley Gate, Reading RG6 6BB, UK
2 Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB,
UK
S. J. Johnson et al.
1 3
present (e.g. Sabeerali et al. 2013; Sperber et al. 2013).
Monsoon low pressure systems (LPS), including the more
intense monsoon depressions and less intense monsoon
lows, are synoptic systems that generally develop in the
Bay of Bengal and pass up the monsoon trough, increasing
precipitation in central and northern India and generating
many of the most intense rain events in the monsoon (Sikka
1977; Krishnamurthy and Misra 2010). These systems are
often too infrequent and too weak in GCMs (Ashok et al.
2000; Sabre et al. 2000; Stowasser et al. 2009).
Most of these biases improve slightly in CMIP5 GCMs
compared to CMIP3 GCMs. CMIP5 GCMs typically have
higher horizontal and vertical resolution in the atmosphere
and ocean than CMIP3 GCMs and usually include para-
metrisations of more physical processes, such as a more
detailed treatment of aerosols and a more complete repre-
sentation of the carbon cycle (Sperber et al. 2013). Since
the CMIP samples are composed of many diverse GCMs,
it is difficult to attribute the monsoon improvement to any
particular physics or resolution change in the atmosphere
or ocean components.
Here, we evaluate how the simulated Asian monsoon
responds to increasing horizontal atmospheric resolu-
tion from approximately 200 to 40 km in atmosphere-
only climate integrations of the Met Office Unified Model
(MetUM, Sect. 2.2). Previous studies which used different
GCMs, integration lengths, and different initial resolu-
tions have also analysed the South Asian monsoon’s sen-
sitivity to increased resolution in atmosphere-only GCMs
(AGCMs). The studies are summarised in Table 1. We
expect some differences between their findings depending
on the GCM used, for example, the amplitude and sign of
seasonal mean precipitation resolution sensitivity depends
on the choice of convection scheme (Shin et al. 2003). We
also expect the response to depend on how the studies’
the initial resolution, and resolution increase, compare to
the scales of features and processes relevant to the mon-
soon, such as specific orographic features or synoptic scale
variability.
Despite the expected variation, similarities in the reso-
lution sensitivity exist throughout these studies. Improve-
ments in the spatial distribution of orographic precipita-
tion windward of the Western Ghats and the Indochina
peninsula is common (Jha et al. 2000; Kitoh and Kusunoki
2004). The lower tropospheric monsoon circulation, such as
the peak wind speed of the Somali Jet, generally improves
with increasing resolution (Sperber et al. 1994; Sabin et al.
2013). Some studies have also shown increased easterlies
extending further along the monsoon trough just south of
the Himalayas (Sabin et al. 2013). In contrast, the western
equatorial Indian Ocean (WEIO, Sperber et al. 1994; Sabin
et al. 2013) and Indian peninsula (Sperber et al. 1994; Ste-
phenson et al. 1998; Martin 1999; Sabin et al. 2013) are
key monsoon regions in which the resolution sensitivity of
the precipitation is inconsistent across studies in both sign
and magnitude.
In this study, we analyse multiple 27-year MetUM
AMIP-style climate integrations ranging from approxi-
mately 200–40 km horizontal resolution (N96–N512)
generated for the UK on PRACE: weather resolving
Simulations of Climate for globAL Environmental risk
campaign (UPSCALE campaign, Mizielinski et al. 2014).
As many other aspects of the GCM as possible remain
unchanged, which allows us to directly attribute changes
in precipitation and circulation to the resolution increase
and to analyse where and how increasing resolution has
the most impact on the monsoon. Due to the length of
the integrations and number of ensemble members, this
dataset is more extensive than any other individual GCM
dataset used to study the sensitivity of the South Asian
Table 1 Some previous studies of the Asian monsoon’s resolution sensitivity in atmosphere only models
The resolution, integration length, number of ensemble members and type of SSTs (climatological or annually varying, denoted AMIP-style) are
listed. Some of the AMIP-style SSTs are monthly averaged while others are averaged over shorter time scales. The GCM column indicates the
GCM name, but not the version of that GCM. For example, the version of the MetUM used by Martin (1999) differs considerably from that used
here
GCM Resolution Resolution at
equator (km)
Integration
length
Ensemble
members
SSTs References
MetUM N96–N512 208–39 27 years 3–5 AMIP-style This study
LMD Variable resolution mesh ~35 1 year 10 Climatological Sabin et al. (2013)
JMA TL95 (T63)–TL959 317–32 20 years 1–3 AMIP-style Kitoh and Kusunoki
(2008)
JMA T42–T213 476–94 3 years 1 Climatological Kitoh and Kusunoki
(2004)
FSU T42–T120 476–166 1 month 1 AMIP-style Jha et al. (2000)
MetUM N48–N144 417–138 10 years 1 AMIP-style Martin (1999)
ARPEGE T21–T63 953–317 4 years 1 AMIP-style Stephenson et al. (1998)
IFS T21–T106 953–189 1 year 1 Climatological Sperber et al. (1994)
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
monsoon to horizontal resolution. Using this resource, we
analyse the resolution dependence of the JJAS mean state
in the tropical Indo-Pacific, the monsoon seasonal cycle,
monsoon variability and monsoon LPS. We focus on
understanding which resolution-related processes cause
the sensitivity and explore changes in processes across the
Indo-Pacific warm pool that may affect the South Asian
monsoon.
There are many possible drivers of this resolution
sensitivity. Improving the resolution might be expected
to improve the representation of high temporal variabil-
ity events which are responsible for much of the extreme
precipitation in South Asia. For example, numbers of
tropical cyclones and monsoon LPS in the Indian Ocean
increase with resolution in some GCMs (Kitoh and Kusu-
noki 2004; Sabin et al. 2013). Another important source
of resolution sensitivity is the representation of orogra-
phy. Several orographic features are fundamental drivers
of the South Asian monsoon through multiple mecha-
nisms: orographically forced precipitation and the result-
ing diabatic heating (Xie et al. 2006), providing a barrier
between extratropical and tropical air (Boos and Kuang
2010, 2013; Tang et al. 2013), elevated surface heating
(e.g. Li and Yanai 1996) and confinement and modifica-
tion of the cross-equatorial monsoon circulation (Hoskins
and Rodwell 1995; Rodwell and Hoskins 1995; Slingo
et al. 2005). As horizontal resolution increases, more of
the orography is resolved, and the contribution from the
orographic parameterisation should reduce, which may
change the overall impact of orography in the GCM. The
representation of islands and coastlines is also important
in the South Asian monsoon domain. As a collection of
many islands at or below the typical grid-scale of GCMs,
the Maritime Continent is a difficult test for parameteri-
sations and large precipitation biases have existed over
the Maritime Continent in many versions of the MetUM
(Neale and Slingo 2003; Strachan 2007; Schiemann et al.
2014). As a large error in diabatic forcing in the monsoon
domain, any resolution sensitivity of Maritime Conti-
nent biases may affect the precipitation and circulation
over India through alteration of the Walker circulation or
wave activity (Neale and Slingo 2003). We consequently
include the Maritime Continent in the South Asian mon-
soon domain we study here.
In Sect. 2 we describe the configuration of the MetUM
used in this study and our experimental design. In Sect. 3
we present diagnostics of many aspects of the monsoon
and discuss their resolution sensitivity. In Sect. 4 we inves-
tigate the sources of the largest changes found in Sect. 3.
In Sect. 5 we combine our findings into a coherent picture
of the resolution sensitivity in the South Asian monsoon
domain and discuss the implications. In Sect. 6, we sum-
marise our findings.
2 Methodology
2.1 Experimental design
The UPSCALE dataset (Mizielinski et al. 2014) consists of
multiple ensemble members of AMIP-style climate model
integrations at N96, N216, N512 resolution with 85 vertical
levels extending to 85 km. Integrations were conducted for
present and future climate conditions. We analyse the present
climate integrations in this study, which run from February
1985 to December 2011. The five N512 ensemble members
were initialised with conditions taken from consecutive days
of a 5-year N512 spin-up run, which was initialised with
conditions from an N320 (
≈
62 km) integration from the
MetUM model development programme. Three ensemble
members at N216 and five ensemble members at N96 were
initialised from the N512 initial conditions interpolated to
lower resolution. The integrations are summarised in Table 1
and the model configuration is described in Sect. 2.2.
2.2 MetUM configuration
UPSCALE used a version of the Global Atmosphere 3.0
(GA3.0, Walters et al. 2011) configuration of the atmos-
pheric component of the HadGEM3 family of models,
which we will refer to as the Met Office Unified Model
(MetUM). The model configuration is a unique choice of
atmospheric MetUM dynamical core, atmospheric param-
eterisations, settings and forcings. The default GA3 config-
uration is described in Walters et al. (2011). Very few sci-
entific settings in the MetUM are changed with resolution,
but a few parameters must be changed to ensure numerical
stability. An example is the dynamical core’s alternating-
direction implicit (ADI) pseudo time-step (Davies et al.
2005), which is related to the efficiency of the implicit
solver at high latitudes. The parameters changed with reso-
lution, or changed from their default settings to apply at all
resolutions, are listed in Table 2. In addition to these altera-
tions, diffusion is applied to vertical wind velocities in the
upper five levels of the atmosphere to dissipate grid-scale
artifacts in the stratosphere (Mizielinski et al. 2014). Here,
we summarise the aspects of the MetUM that are relevant
to horizontal resolution sensitivity in the South Asian mon-
soon domain. Further details can be found in Walters et al.
(2011).
2.2.1 Prescribed sea surface temperatures
UPSCALE prescribes daily sea surface temperature (SST)
and sea-ice forcings bilinearly interpolated to the MetUM
grids from the OSTIA product, which combines satellite
and in-situ observation into a product with a native reso-
lution of one-twentieth degree (Donlon et al. 2012). The
S. J. Johnson et al.
1 3
JJAS mean prescribed SSTs in the Indo-Pacific warm pool
region are shown in the bottom panels of Fig. 1. While the
SST distribution changes little with resolution, smaller fea-
tures, such as the variation in SST around the islands of the
Maritime Continent, are resolved at higher resolution.
2.2.2 Land surface model and coastal tiling scheme
The land surface fluxes are calculated with the community
land surface model Joint UK Land Environment Simulator
GL 3.0 (JULES, Best et al. 2011; Clark et al. 2011; Walters
et al. 2011). The 2′ Gridded Global Relief Data (ETOPO2,
US Department of Commerce 2001) dataset is degraded to
1° (N96) or 0.25° (N216 and N512) and used to designate
land and sea grid-points. Atmospheric grid-points that have
both land and sea in them (including sub-grid islands) are
designated coastal grid points and assigned a land frac-
tion. The land fraction boundary conditions are shown
in the middle panels of Fig. 1. Red regions are land grid-
points, white regions are sea grid-points and other colours
indicate coastal grid-points. At higher resolution the frac-
tion of coastal points decreases markedly, while the frac-
tion of land and sea points increases, particularly over the
Maritime Continent and around the Arabian Peninsula.
Due to the different course-graining of the N96 and higher
resolution configurations, fewer islands appear in the West
Pacific in the N96 configuration, than in higher resolution
configurations.
Coastal surface fluxes are computed separately for the
land and sea fractions of each grid point, then passed to the
atmosphere such that the total flux is accounted for. This
coastal tiling scheme blurs the coastline over a grid box.
Consequently, the apparent width of coastlines decreases
as resolution increases (Johns et al. 2006). In the MetUM
configuration used here, a “buddy” scheme is added to
the coastal tiling scheme, which splits the lowest level
atmospheric wind speeds over coastal points into separate
land and sea contributions using the average wind speed
over sea grid points adjacent to a coastal grid point. This
increases the wind speed over the sea part of the grid-box,
improving the scalar surface fluxes (Walters et al. 2011).
2.2.3 MetUM response to orography
The MetUM dynamical core uses a semi-implicit, semi-
Lagrangian formulation to solve the non-hydrostatic, fully
compressible deep atmosphere equations of motion (Davies
et al. 2005) using prognostic fields discretised horizontally
on the Arakawa-C grid (Arakawa and Lamb 1977) and
vertically on terrain-following hybrid height levels using
Charney–Phillips staggering (Charney and Phillips 1953).
On the grid-scale, the MetUM dynamical core responds to
the mean height of orography in a grid-box derived from
the GloBal One-km Base Elevation dataset (GLOBE,
GLOBE Task Team 1999), which has a native resolution
of 30″. The grid-scale orographic boundary condition is
shown in Fig. 1. As resolution increases, orographic height
increases in many locations. Examples include the East
African Highlands in Ethiopia and along the horn of Africa,
and the New Guinea Highlands on the Maritime Continent.
The effects of orographic features smaller than the grid-
scale are parametrised.
As described in Walters et al. (2011), there are three
parameterisations that represent different effects of the sub-
grid orography: boundary layer drag, flow blocking and
gravity wave drag. On the smallest scales, the momentum
used by the boundary layer scheme is adjusted through an
increased roughness length over orography (Gregory et al.
1998). On larger scales, where buoyancy effects are impor-
tant, the sub-grid orography is represented by a measure of
amplitude, which is proportional to the standard deviation
Table 2 MetUM configurations
used in this study
The resolution naming convention number, N, defines the resolution of the configuration grid, which has
2N longitude and 1.5N + 1 latitude grid-points
Parameter N96 N216 N512
Nx
×
Ny
192 × 145 432 × 325 1024 × 769
Lon × lat (°) 1.875 × 1.245 0.83 × 0.55 0.35 × 0.23
Lon × lat (at equator, km) 208 × 139 93 × 62 39 × 26
Time step (min) 20 15 10
ADI pseudo time step 8 × 10−43 × 10−410−4
Targeted diffusion W threshold (m s−1) 0.5 1.0 1.0
CAPE closure time scale (s) 3600 3600 3600
Threshold W for reduction of CAPE time scale (m s−1) 0.4 0.4 0.4
Fig. 1 Fixed boundary conditions for the N96, N216 and N512
atmosphere-only configurations of the MetUM. Top Grid-point mean
orographic height (km). Middle Grid-point land fraction. Red indi-
cates a land fraction of one which is a land grid-point, while white
indicates a land fraction of zero which is a sea grid-point. Other col-
ours indicate the land fraction of coastal grid-points. Bottom JJAS
mean SST (°C)
▸
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
20S
10S
0
10N
20N
30N
40N
N96
Orography
N216
0.00
0.25
0.50
1.00
2.00
3.00
4.00
5.00
N512
Height (km)
20S
10S
0
10N
20N
30N
40N
Land-sea mask
0.0
0.2
0.4
0.6
0.8
1.0
Land fraction
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
SST
40E 60E 80E 100E 120E 140E 160E 180E
27.0
27.5
28.0
28.5
29.0
29.5
30.0
40E 60E 80E 100E 120E 140E 160E 180E
SST (oC)
S. J. Johnson et al.
1 3
of the source orography in a model grid-box, and a meas-
ure of anisotropy. The total surface stress from orography
is determined using a simple linear hydrostatic expression,
which is proportional to the bulk lower level winds and
atmospheric stability. The total stress is then partitioned
into a gravity wave component representing flow over the
sub-grid orography and a flow blocking component repre-
senting flow around the sub-grid orography. Flow blocking
drag is exerted over the depth of the sub-grid orography
assuming a linear decrease with height, while the gravity
wave drag is exerted at levels where wave breaking and
wave saturation are diagnosed. The drag is applied as incre-
ments to the MetUM wind fields, and a numerical limiter is
imposed on flow blocking drag to ensure numerical stabil-
ity (Brown and Webster 2004). As horizontal resolution is
increased, more orography is resolved explicitly, and less is
parametrised.
2.2.4 Convective parametrisation
As the South Asian monsoon is dominated by convective
precipitation, it is very sensitive to MetUM parameters
relating to convection. Convective parametrisation in the
MetUM is derived from the bulk mass flux scheme devel-
oped by Gregory and Rowntree (1990). The scheme’s
foundation is single plume parcel theory modified to rep-
resent the average properties of an ensemble of convective
plumes. The scheme triggers convection from the boundary
layer using an undilute parcel calculation, then performs
shallow or deep convection for grid points where convec-
tion is triggered. Mid-level convection can be triggered
from levels in the free troposphere.
Deep convective precipitation dominates in the tropi-
cal Indo-Pacific region. Several adjustments to the deep
convection scheme have been introduced since Gregory
and Rowntree (1990), such as the addition of down-drafts
(Gregory and Allen 1991), convective momentum trans-
port using a flux gradient approach (Stratton et al. 2009)
and a convective available potential energy (CAPE) closure
scheme, based on Fritsch and Chappell (1980a, b), which
uses a dilute CAPE calculation. In the default MetUM GA3
configuration, the time scale for the CAPE closure is set at
90 min unless large-scale vertical velocities over 0.3 m s−1
are detected in the column, in which case the time scale is
reduced to ensure numerical stability. To study the sensi-
tivity to horizontal resolution alone, it is important to keep
these parameters as constant as possible across configura-
tions. To ensure numerical stability at all the resolutions
studied here, the maximum CAPE time scale is decreased
to 60 min and the threshold vertical velocity is increased to
0.4 m s−1 across all configurations.
Convection can also occur at the grid-scale and targeted
horizontal diffusion of moisture is applied above a certain
vertical velocity threshold to dissipate very high vertical
velocities and maintain numerical stability. This thresh-
old is increased with resolution to allow resolution of finer
scale features in the vertical flow (targeted diffusion W
threshold in Table 2).
2.3 Analysis techniques
2.3.1 Tracking synoptic systems
Analysis of synoptic systems in the monsoon trough uses
a tracking algorithm (TRACK-1.4.0, Hodges 1994; Rob-
erts et al. 2015) on 6-hourly 850 hPa relative vorticity. The
data are first filtered to T42 resolution (approximately 2.8°
in latitude and longitude). This coarse resolution filtering is
a typical approach applied in tropical cyclone tracking to
remove sub-synoptic-scale noise from the relative vorticity
fields (e.g. Bengtsson et al. 2006; Strachan et al. 2013). The
common spatial filter also reduces resolution dependence
in the tracking procedure. Intense monsoon depressions
and less intense monsoon lows have a typical length-scale
of 1000–2000 km (e.g. Sikka 1977; Krishnamurthy and
Misra 2010), which is several times the T42 grid-spacing.
Tracking parameters were chosen based on validation
with ERA-Interim reanalysis (1° gridded product regridded
to T42 for tracking). Systems that exceed a (cyclonic) vor-
ticity threshold of 3 × 10−5 s−1 for at least 2 days and that
travel a minimum distance of 5° are diagnosed. The dis-
tance travelled threshold is applied to the greatest distance
found between any two points in the system’s trajectory. To
affect South Asia in a meaningful way, we require that the
system spends at least 60 % of its lifetime in between 70°–
95° E and 10°–30° N. A further criterion is applied to dif-
ferentiate fully recirculating systems from other instances
of vorticity variability such as (diurnal) variations in the
monsoon trough, variations in the heat low development
over India during the early monsoon, and small-scale fea-
tures developing near Himalayan orography. The recircula-
tion criterion requires the presence of flow reversal in both
zonal and meridional directions, with the absolute value of
the flow exceeding 5 m s−1 in both directions within a box
approximately 10° around each system and lasting continu-
ously for 1 day.
Validation of this method applied to ERA-Interim re-
analysis against the Indian Meteorological Department
(IMD) Cyclone and Monsoon Depression eAtlas (IMD
2011) shows reasonable agreement, with most systems
travelling a substantial distance across India diagnosed.
Our method also identifies weaker monsoon lows that trav-
erse a substantial part of India in ERA-Interim re-analysis
and substantially contribute to monsoon rainfall. How-
ever, these systems are absent in the IMD dataset as they
do not qualify as monsoon depressions. Several relatively
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
stationary depressions in the IMD dataset over the northern
Bay of Bengal are not diagnosed in ERA-Interim due to the
minimum distance threshold, which is necessary to avoid
multiple erroneous inclusions. Likewise, stationary mon-
soon lows are not included.
LPS rainfall is estimated in a box approximately 10°
around each system, which estimates the typical length-
scale of monsoon depressions and lows.
2.3.2 Significance testing
To test for significance with respect to ensemble member,
we use a Mann–Whitney rank sum test (Mann and Whitney
1947; Kanji 2006). This test ranks members of two samples
in ascending size and uses the relative positions of the two
samples in the rank to determine the significance of the dif-
ference between the two samples.
2.4 Observational datasets
We compare the MetUM output to many observational
products. To assess seasonal mean precipitation we use
the Global Precipitation Climatology Project (GPCP) Ver-
sion 2.2 Monthly Precipitation Analysis (Adler et al. 2003).
GPCP is a 2.5° gridded merged analysis that incorporates
precipitation estimates from low-orbit satellite microwave
data, geostationary satellite infrared data and surface rain
gauge observations. To estimate observational error, we
also use the monthly CPC Merged Analysis of Precipita-
tion (CMAP) product, a similar gauge and satellite analysis
(Xie and Arkin 1997). We interpolate MetUM output to the
GPCP and CMAP grid for comparison.
To assess daily precipitation, we use the 1° daily gridded
product GPCP Version 1.2 (Huffman et al. 2001), which
is available from 1996 to the present. We also use the 0.5°
daily gridded APHRODITE dataset, version APHRO_MA_
V1003R1 (which has better quality control, incorporates
more data, and has had bugs corrected when compared
with APHRO_MA_V0902, Yatagai et al. 2009). APHRO-
DITE is primarily derived from rain gauges and is available
over land only from 1951 to 2007. Both daily datasets were
interpolated to the N96 grid for comparison with MetUM
data.
To assess latent heat fluxes over the Maritime Continent
we used monthly 1° gridded output from the Objectively
Analysed air–sea Heat Fluxes (OAFlux) project, available
from 1958 to 2012. OAFlux products are constructed over
the ocean from an optimal blending of satellite retrievals
and three atmospheric reanalyses (Yu 2007; Yu et al. 2008).
To assess surface wind speeds, we used the Scatterometer
Climatology of Ocean Winds (SCOW) product, derived
from 122 months (September 1999–October 2009) of
QuikSCAT scatterometer data and available over the ocean
only (Risien and Chelton 2008). SCOW monthly 0.25°
gridded fields were interpolated to N96 for comparison to
MetUM output.
For fields that are not directly observable, including
winds on many pressure levels, we use two reanalysis prod-
ucts: the European Centre for Medium-Range Weather
Forecasts’ (ECMWF) ERA-Interim atmospheric rea-
nalysis product gridded to 0.70 × 0.70° (Dee et al. 2011)
and the NASA’s Modern-Era Retrospective Analysis for
Research and Applications reanalysis product gridded to
0.67 × 0.50° (MERRA, Rienecker et al. 2011). Fields were
interpolated to N96 grid, compared over an equivalent time
period and compared on equivalent pressure levels unless
otherwise stated.
3 Resolution sensitivity of the South Asian
monsoon
3.1 JJAS mean state
In Fig. 2 we show climatological and ensemble averaged
JJAS precipitation in the Indo-Pacific region for GPCP,
CMAP and the N96, N216 and N512 configurations of
the MetUM. We also show the differences between con-
figurations and each configuration’s bias relative to GPCP
and CMAP. In GPCP and CMAP, precipitation maxima
lie over the eastern equatorial Indian Ocean, over and off
the west coast of the Indian peninsula and in the western
Bay of Bengal. The MetUM shows significant biases in
these regions, in particular, excess precipitation over the
equatorial Indian Ocean; a deficit of precipitation over the
eastern Arabian Sea, Indian peninsula and Bay of Bengal;
and excess precipitation over the southern slopes of the
Himalayas.
Increasing resolution produces consistent small changes
in precipitation over large areas including precipitation
increases in the southeast Arabian Sea, over India and
over east Africa; precipitation decreases over the equato-
rial Indian Ocean, South China Sea and western Pacific;
and precipitation increases over the Maritime Continent
islands. The largest grid-point changes in precipitation tend
to be over steep gradients in orography, and often consist
of a large precipitation increase next to a large precipita-
tion decrease where the orographic precipitation has shifted
closer to the orography. Increasing resolution from N96 to
N216 introduces larger changes than increasing resolution
from N216 to N512. This suggests resolution sensitivity
decreases at higher resolution.
These precipitation changes are not necessarily improve-
ments. The precipitation deficit over India and the precipi-
tation excess over the equatorial Indian Ocean are both
slightly reduced, but this comes at the cost of increasing
S. J. Johnson et al.
1 3
precipitation biases over the western equatorial Pacific.
Over the entire domain of Fig. 2, the root mean square error
increases and the pattern correlation decreases, despite
interpolating all configurations to the same resolution, indi-
cating an overall increase in biases.
To examine the changes in South Asian monsoon cir-
culation, in Fig. 3 we show the JJAS ensemble averaged
climatology of MetUM 850 hPa wind at N96, the change
when the resolution is increased to N512 and the bias with
respect to ERA-Interim (N96 and N512 climatological
850 hPa winds are compared later in Fig. 10). These are
overlaid on the comparable JJAS precipitation map for
reference. The largest changes in circulation are also near
steep orography, such as the East African Highlands and
the Hadramawt mountains on the Arabian peninsula. In the
Arabian Sea, there is a strengthening and northward shift
20S
10S
0
10N
20N
30N
40N GPCP
0.0 0.5 1.0 5.0 10.0 15.0
P (mm day-1)
20S
10S
0
10N
20N
30N
40N
40E 60E 80E 100E 120E 140E 160E 180E
CMAP
20S
10S
0
10N
20N
30N
40N
N512
20S
10S
0
10N
20N
30N
40N N216
20S
10S
0
10N
20N
30N
40N N96 N216 - N96 @ N96 N512 - N96 @ N96
N512 - N216 @ N216
N512 - GPCP
RMSE = 3.29 Pattern corr. = 0.67
N216 - GPCP
RMSE = 3.32 Pattern corr. = 0.69
N96 - GPCP
RMSE = 2.97 Pattern corr. = 0.73
-5.0
-4.0
-3.0
-2.0
-1.0
-0.5
0.5
1.0
2.0
3.0
4.0
5.0
∆ P (mm day
-1
)
40E 60E 80E 100E 120E 140E 160E 180E
N512 - CMAP
RMSE = 3.41 Pattern corr. = 0.66
40E 60E 80E 100E 120E 140E 160E 180E
N216 - CMAP
RMSE = 3.39 Pattern corr. = 0.68
40E 60E 80E 100E 120E 140E 160E 180E
N96 - CMAP
RMSE = 2.80 Pattern corr. = 0.76
Fig. 2 Diagonal and left Climatological JJAS precipitation in GPCP
and CMAP and ensemble averaged climatological JJAS precipita-
tion in the N96, N216 and N512 configurations of the MetUM, all on
their native grids, in the tropical Indo-Pacific. Off diagonal, top two
rows Differences between each of the MetUM configurations. The
higher resolution data is interpolated to the lower resolution MetUM
grid for comparison. Shaded grid points in the difference panels indi-
cate precipitation differences are significant at the 95 % level using a
Mann–Whitney rank sum test, less significant changes are outlined.
Off diagonal, bottom two rows Biases between each configuration of
the MetUM with respect to GPCP and CMAP observations. RMSE
and pattern correlations over the entire domain are listed above these
panels
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
of the Findlater/Somali Jet off the coasts of Africa and the
Arabian peninsula, which is an improvement in the weak
Somali Jet bias. As with precipitation, there are also smaller
changes over larger areas. Decreased westerlies extend-
ing from the African coast to the West Pacific are consist-
ent with the reduction in precipitation and diabatic heating
in the equatorial Indian Ocean and West Pacific. Increased
southerly wind components in the southern Bay of Bengal
and South China Sea, and decreased southerly wind com-
ponents just south of the equator are also all consistent with
the precipitation changes in the equatorial Indian Ocean,
West Pacific and over the Maritime Continent.
On average, the GCMs in CMIP5 have higher atmos-
pheric horizontal resolution than the GCMs in CMIP3.
While the slight improvements over the equatorial
Indian Ocean and India are consistent with the differ-
ences between the CMIP3 and CMIP5 multi-model mean
(MMM) bias, the overall increase in bias is inconsistent
with the difference between the CMIP3 and CMIP5 MMM
(Sperber et al. 2013). While the CMIP5 GCMs are usually
coupled and the UPSCALE integrations are atmosphere-
only, this indicates that increasing atmospheric resolution is
one plausible cause for increasing precipitation over India
in the CMIP5 GCMs. Consistency with other individual
GCM resolution sensitivity studies is discussed in Sect. 5.2.
3.2 Seasonal cycle
The South Asian monsoon onset occurs over the Andaman
Sea in the southeast Bay of Bengal in late May and covers
most of the Bay of Bengal and the southern tip of the India
peninsula by the end of May. It progresses northward and
westward to cover the entire Indian peninsula by mid-July.
After the onset, there is competition between the equatorial
Indian Ocean convergence zone and the Indian subcontinent
convergence zone (Gadgil and Sajani 1998), which supplies
monsoon precipitation and drives intraseasonal variability.
The onset and the two convergence zones are visible in the
annual cycle of precipitation averaged from 70° to 90° E,
which is shown in Fig. 4 for GPCP and the MetUM N96,
N216 and N512 configurations. AGCMs (e.g. Gadgil and
Sajani 1998) and coupled GCMs (e.g. Sperber et al. 2013)
typically struggle to correctly simulate the timing of the onset
and the competition between the convergence zones. Figure 4
shows that the MetUM is no exception. There is extremely
low precipitation over India throughout the season, except for
intense precipitation over the southern slopes of the Hima-
layas, and excessive precipitation over the equatorial Indian
ocean, consistent with the JJAS mean precipitation in Fig. 2.
When resolution increases, little improvement is seen in the
competition between the continental and oceanic conver-
gence zones, but there are indications that the monsoon onset
is slightly improved, with increased precipitation in May
from 5° to 15° N, which is highlighted in Fig. 4.
To analyse the detailed seasonal cycle over India, a
time series of climatological N96, N216, N512, GPCP and
2 m s-1
(b) N512-N96
20S
10S
0
10N
20N
30N
40N
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
-5.0 -4.0 -3.0 -2.0 -1.0 -0.5 0.5 1.0 2.0 3.0 4.0 5.0
∆
P (mm day
-1
)
2 m s-1
(c) N96 bias
0.0 0.5 1.0 5.0 10.0 15.0
P (mm day-1)
10 m s-1
(a) N96
20S
10S
0
10N
20N
30N
40N
Fig. 3 a Ensemble averaged climatological JJAS precipitation
(mm day−1) and 850 hPa winds (m s−1) in the N96 configuration of
the MetUM. b Difference in ensemble averaged, climatological JJAS
precipitation and 850 hPa winds between the N512 and N96 MetUM
configurations. c JJAS precipitation and 850 hPa wind bias in the
N96 configuration of the MetUM relative to GPCP and ERA-Interim.
Shaded grid points in (b) indicate precipitation biases are significant
at the 95 % level using a Mann–Whitney rank sum test, less signifi-
cant changes are outlined. Vectors are only shown in (b) if either the
zonal or meridional wind differences are significant at the 95 % level
S. J. Johnson et al.
1 3
APHRODITE daily precipitation averaged over the India
peninsular land south of 25° N and west of 90° E, exclud-
ing the dominant orographic precipitation over the Hima-
layas, are shown in Fig. 5. The large precipitation deficit
over India in all configurations of the MetUM is clear, but
as resolution increases, precipitation increases from May
through August. The increase is at most 1.5 mm day−1, a
small fraction of the precipitation bias over India in the
MetUM. However, it nearly doubles the precipitation at
N96 and as such represents a meaningful improvement for
model development. The increase is largest in June and
July, giving a moderately more accurate distribution of pre-
cipitation through the season.
3.3 Interannual and intraseasonal variability
The largest driver of Indian monsoon interannual variability
is the teleconnection with the El Niño Southern Oscillation
(ENSO, e.g. Webster and Yang 1992; Kumar et al. 2006).
Coupling to a dynamical ocean is required to accurately
simulate the teleconnection between ENSO and the mon-
soon domain (Bracco et al. 2005; Wang et al. 2008), but the
relationship between prescribed Pacific SSTs and the South
Asian monsoon may still change as resolution is increased
in the AGCM. To check the sensitivity of the monsoon-
ENSO teleconnection to resolution, we calculated the lag-
correlation of JJAS all-India rainfall and the JJAS Webster-
Yang dynamical monsoon index (Webster and Yang 1992)
with seasonal Niño 3.4 SST in each ensemble member at
each resolution (not shown). Due to the small number of
years available in our ensemble, we found no statistically
significant change in either relationship with resolution.
To check for changes in intraseasonal variability we also
analysed the northward and eastward propagation of the
BSISO using the lag-correlation of 20–100 days bandpass
filtered precipitation in each ensemble member averaged
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1
-30
-15
0
15
30
45
(a) GPCP
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1
-30
-15
0
15
30
45
(b) N96
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1
-30
-15
0
15
30
45
(c) N216
Jan 1 Feb 1 Mar 1 Apr 1 May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1
-30
-15
0
15
30
45
(d) N512
0.0 0.5 1.0 3.0 5.0 7.0 9.0 11.0 13.0
P (mm day
-1
)
Fig. 4 Seasonal cycle of climatological precipitation averaged from
70° to 90° E in (a) GPCP (1997–2011) and the (b) N96, (c) N216 and
(d) N512 configurations of the MetUM (1985–2011, ensemble aver-
age). All fields are interpolated to the N96 grid for comparison. The
box highlights an increase in precipitation early in the season
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
over 70°–100° E and 5° S–5° N, respectively (not shown).
In each case, very little resolution sensitivity was seen, con-
sistent with other resolution sensitivity studies (Bacmeister
et al. 2014).
3.4 Monsoon low pressure systems
As described in Sect. 2.3.1, we use the tropical cyclone
tracking software TRACK to identify monsoon LPS. The
total number of monsoon LPS in each month of JJAS for
each MetUM ensemble member, ERA-Interim reanalysis
and the IMD e-atlas of observed depressions is shown in
Fig. 6. Consistent with previous studies with the MetUM
and other GCMs (Ashok et al. 2000; Sabre et al. 2000; Sto-
wasser et al. 2009), there are far fewer LPS detected in the
MetUM than in ERA-Interim or the IMD e-atlas, and the
resolution sensitivity is very small compared to the size of
the bias. Variation between ensemble members is also large
compared to the resolution sensitivity. Numbers of LPS
in JJA increase from N96 to N216, consistent with other
GCMS (Kitoh and Kusunoki 2004; Sabin et al. 2013),
while they stay approximately the same from N216 to
N512. In September, the number of LPS tends to decrease
slightly as resolution increases from N96 to N216 and
again to N512. TRACK may be detecting two distinct types
of systems: monsoon LPS, which are mainly present in
JJA, and tropical cyclones in the Bay of Bengal, which are
mainly present in spring and autumn. This would explain
the difference in resolution sensitivity between JJA systems
and September systems. Analysis of tropical cyclones in
the UPSCALE dataset indicates that tropical cyclone num-
bers in the northern Indian ocean decrease in SON when
resolution is increased (Roberts et al. 2015), consistent
with the decrease in September systems seen here.
We associate precipitation with the tracked LPS if it
occurs at the same time as the vorticity anomaly and within
a 10° box centred on the LPS. The derived LPS precipita-
tion in the N96 and N512 configurations of the MetUM and
the difference between them is shown in Fig. 7. As resolu-
tion increases from N96 to N216 (not shown), and further
to N512, changes in monsoon LPS contribute to increased
rainfall in northern and eastern central India and decreased
rainfall in the southern Bay of Bengal. When compared
to the total precipitation change in Fig. 2, an increase in
May 1 Jun 1 Jul 1 Aug 1 Sep 1 Oct 1 Nov 1 Dec 1
0
1
2
3
4
5
6
7
8
9
10
Indian peninsula P (mm da
y-1)
0
1
2
3
4
5
6
7
8
9
10
Indian peninsula JJAS P (mm day
-1)
GPCP
Aphrodite
N96
N216
N512
60 100
0
30
Fig. 5 Five day running median of climatological daily precipita-
tion over the Indian peninsula (region outlined in the inset diagram)
in GPCP (1997–2011, black), APHRODITE (1985–2007, grey), N96
(blue), N216 (green) and N512 (red) configurations of the MetUM
(1985–2011, ensemble average). All fields are interpolated to the N96
grid for the comparison. The first and third quartiles are shown for
N96 and N512 as the dashed lines. The area between the N96 and
N512 time series is shaded if the difference between N96 and N512
is significant at the 95 % level by a Mann–Whitney rank sum test.
The coloured error bars to the right of the figure indicate the ensem-
ble spread of the JJAS mean MetUM precipitation, and the dots indi-
cate GPCP and APHRODITE JJAS mean precipitation
0
50
100
150
Number of system
s
jun
jul
aug
sep
N96 N216 N512
1985-2011
ERAi IMD
Fig. 6 Total number of monsoon LPS in each MetUM integra-
tion, including five ensemble members for N96 and N512 and three
ensemble members for N216. Also shown are numbers of monsoon
LPS in ERA-Interim reanalysis and the IMD e-atlas. Colours indicate
the month in which the system initiated. There are fewer systems in
the IMD e-atlas than ERA-Interim because the IMD e-atlas only con-
tains the more intense monsoon LPS classified as depressions
-1.5 -0.5 0.5 1.5
∆ P (mm day
-1
)
70E 80E 90E 100E
(c) N512-N96
70E 80E 90E 100E
5N
10N
15N
20N
25N
30N
35N
0.0 0.5 1.0
P (mm day
-1
)
(a) N96
0.0 0.5 1.02.03.0 2.03.0
P (mm day
-1
)
70E 80E 90E 100E
(b) N512
Fig. 7 Climatological, ensemble averaged JJAS precipitation attrib-
uted to monsoon LPS in the a N96 configuration of the MetUM on
its native grid and b N512 configuration of the MetUM interpolated
to the N96 grid. c Difference between b and a. Only points sig-
nificant by a Mann–Whitney rank sum test are shown. The colour
scale ranges to ±1.5 mm day−1, while the scale on Fig. 3, which
shows the total JJAS precipitation change with resolution, ranges to
±5 mm day−1
S. J. Johnson et al.
1 3
monsoon LPS may explain much of the increased precipi-
tation over northeast India, but not over southern India.
However, by necessity, the vorticity threshold used to track
systems neglects any smaller, less intense or stationary sys-
tems that still contribute to the rainfall, and the resolution
sensitivity of rainfall, over India.
The version of the MetUM we use here is known to
have low numbers of tropical cyclones compared to obser-
vations (Roberts et al. 2015). It is likely that a GCM with
an improved representation of tropical LPS may also show
greater resolution sensitivity in the number of monsoon
LPS. A detailed study of the intensity and lifetime of mon-
soon LPS is beyond the scope of this article. The resolution
sensitivity of these properties will be discussed in a future
article (Levine et al. in preparation).
4 Analysis of sources of resolution sensitivity
As discussed in Sect. 1, there are several ways in which
increasing the horizontal resolution of the MetUM
could change monsoon precipitation and circulation in
the MetUM. In the following sections, we will attempt
to determine the specific resolution-related changes in
processes that generate the largest changes in precipi-
tation and circulation described in Sect. 3. We analyse
only the largest changes: changes in orographic precipi-
tation (Sect. 4.1), changes in the monsoon circulation in
the Arabian Sea (Sect. 4.1.2) and precipitation changes
over the Maritime Continent region (Sect. 4.2). In
Sect. 5 we speculate about how these processes interact
to form the total monsoon domain response to increas-
ing resolution.
4.1 Orography and the South Asian monsoon
We expect improved resolution of Asian and African orog-
raphy will influence monsoon precipitation and circulation.
In this section, we draw upon sensitivity experiments per-
formed in other studies to understand how the improved
resolution of different orographic features influences the
monsoon in the MetUM.
0
1
2
Height (km)
MetUM Orography
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
(a) N96 Precipitation
0
1
2
Height (km)
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
(b) N216 Precipitation
0
1
2
Height (km)
30 40 50 60 70 80 90 100 110 120 130 140
Longitude
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
(c) N512 Precipitation
W. Ghats
Bilauktang
Cardamom
Annam Cordillera
Philippines
Andaman Isle
Ethiopian Highlands
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
GPCP Precipitation
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
30 40 50 60 70 80 90 100 110 120 130 140
Longitude
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
Fig. 8 Left Zonal transect of climatological ensemble mean JJAS
precipitation averaged from 11° N to 13° N for the a N96, b N216
and c N512 configurations of the MetUM (coloured lines) compared
to transects of the MetUM grid-box mean orographic height (shaded),
all on their native grid. Orographic features are labelled in (c). Right
MetUM precipitation transects (colours) interpolated to the GPCP
grid, compared to a transect of GPCP precipitation (black)
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
4.1.1 South Asian monsoon orographic rainfall
Xie et al. (2006) use tropical rainfall measuring mission
(TRMM) observations to argue that when monsoon pre-
cipitation is examined on a fine scale, it is comprised of a
distinct set of rain bands, each anchored to an orographic
feature: the Western Ghats and Himalayas in India; the Ara-
kan Yoma, Bilauktang, Cardamom, and Annaman Cordil-
lera in Indochina and the Andaman and Philippine Islands
in the surrounding seas. They also show, using sensitivity
tests, that the correct orographically forced diabatic heat-
ing over the monsoon domain can improve the precipita-
tion and large-scale circulation of the monsoon, especially
in the Bay of Bengal.
As resolution is increased in the MetUM, these oro-
graphic features and their forcing of precipitation are
resolved more accurately. In Fig. 8, we show a zonal tran-
sect of the grid-box mean orography and precipitation in
the N96, N216 and N512 configurations of the MetUM at
12.5° N, which crosses all of the orographic features listed
above except the Himalayas. At N96 there are distinct oro-
graphic rain bands windward of well separated orographic
features, such as the Western Ghats. Over closely spaced
features, such as the three mountain ranges over Indochina,
the orographic features, and consequently the rain bands,
blend together in the N96 configuration. As resolution
increases, the orographic height of each mountain range
increases and the valleys between the mountain ranges
appear. This leads to distinct orographically forced rain
bands, with larger maxima windward of the orography,
and lower minima leeward or over the orography. Conse-
quently, while there is a predictable increase in maximum
precipitation windward of orographic features, consistent
with resolution sensitivity seen in the CMIP3 models (Kim
et al. 2008), precipitation integrated over the orographic
feature may not change due to the combination of the
improved resolution of the rain band windward of the orog-
raphy and improved resolution of the rain shadow leeward
of the orography. When the transects are interpolated to
the GPCP resolution and compared to GPCP precipitation
0
1
2
3
4
5
6
Height (km)
MetUM Orography
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
(a) N96 Precipitation
0
1
2
3
4
5
6
Height (km)
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
(b) N216 Precipitation
0
1
2
3
4
5
6
Height (km)
0 10 20 30 40
Latitude
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
(c) N512 Precipitation
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
GPCP Precipitation
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
0 10 20 30 40
Latitude
0
2
4
6
8
10
12
14
16
18
20
P (mm day-1)
Fig. 9 Left Meridional transect of climatological ensemble mean
JJAS precipitation averaged from 83° E to 87° E for the a N96, b
N216 and c N512 configurations of the MetUM (coloured lines),
compared to transects of the MetUM grid-box mean orographic
height (shaded), all on their native grid. The large orographic feature
at approximately 25°–40° N is the Himalayas and Tibetan Plateau.
The lower altitude orography at approximately 20°–25° N is the Dec-
can Plateau and Eastern Ghats on the Indian Peninsula. Right MetUM
precipitation transects (colours) interpolated to the GPCP grid com-
pared to a transect of GPCP precipitation (black)
S. J. Johnson et al.
1 3
(Fig. 8), the improved resolution of orographic rain bands
has no systematic effect on precipitation biases near oro-
graphic features. In some cases, precipitation decreases
when resolution increases, such as off the west coast of
Indochina (95°–100° E). In some cases it increases, such
as over the East African Highlands (40°–50° E). This does
not appear to correlate with the sign of the bias, but is gen-
erally consistent with the broader precipitation changes in
the region. This suggests that while orographic rain bands
are better resolved at high resolution across South Asia,
the integrated change in precipitation over an orographic
feature depends on the interaction with the large-scale
circulation.
In Fig. 9, we show a meridional transect of the orogra-
phy and precipitation at 85° E in the N96, N216 and N512
configurations of the MetUM. This transect crosses the
Himalayas and the Tibetan Plateau, which are the tallest
and broadest orographic feature in the region. Again, the
peak in precipitation windward of the orography increases
and narrows with resolution, and the trough in precipitation
over the orography deepens with resolution. However, the
change in maximum precipitation rate on the native grid,
2 mm day−1, is much more modest than seen in Fig. 8.
This may be because the representation of the Himala-
yas changes little, compared to features like the mountain
ranges over Indochina, when resolution is increased. The
Tibetan Plateau is a particularly large, smooth feature;
degrading it to lower resolution does not change its maxi-
mum orographic height or number of peaks by very much.
This contrasts to the previously shown changes over Indo-
china, where larger changes in maximum precipitation on
the native grid are seen as rain bands and rain shadows
become distinct at higher resolution. The wind direction is
along, rather than perpendicular to, the Himalayas on this
transect, which may also reduce the resolution sensitivity
of the orographically forced precipitation here.
(d) N512
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
(c) N216
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
wind speed (m s
-1
)
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
(b) N96
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
10 m s-1
(a) ERA-Interim
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
Fig. 10 Climatological JJAS 850 hPa circulation in a ERA-Interim
reanalysis and b N96, c N216 and d N512 configurations of the
MetUM (ensemble averaged) all interpolated to the N96 grid. Vectors
are coloured according to wind speed. Grid-points that are below the
orography at 850 hPa are omitted. Grey lines contour the MetUM and
ERA-Interim orography on their native grid at 500 m, 1 km and 3 km
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
When the transect is compared to the GPCP transect
on the GPCP grid, there is a marked decrease in inte-
grated rainfall windward of the Himalayas as resolution is
increased. This again demonstrates that orographic precipi-
tation does not always increase as resolution is increased.
4.1.2 East African highlands and the Somali Jet
The East African Highlands (EAH) are mountain ranges
in Ethiopia, Uganda, Kenya and Tanzania, directly west
of where the Findlater/Somali Jet crosses the equator as
it turns anticyclonically towards South Asia. Sensitivity
tests with the MetUM and other models have shown that
in the absence of the EAH and the Hadramawt moun-
tains (Yemen) on the Arabian peninsula, the cross-equa-
torial flow decreases in velocity and extends in a broad
swath across equatorial Africa (Hoskins and Rodwell
1995; Rodwell and Hoskins 1995; Slingo et al. 2005;
Chakraborty et al. 2006). These highlands also have a pro-
nounced effect on the South Asian monsoon circulation as
it crosses the Arabian Sea and Indian peninsula. Rodwell
and Hoskins (1995) used a global primitive equation model
forced by summer monsoon diabatic heating to show that
the orographic drag of the EAH alters the sign of the jet’s
potential vorticity as it crosses the equator, allowing it to
remain in the northern hemisphere. Slingo et al. (2005) fur-
ther showed in a full GCM that the EAH introduce a sta-
tionary wave in the monsoon circulation that extends from
the Arabian Sea to the South China Sea and beyond.
Ensemble mean 850 hPa wind vectors, coloured by wind
speed, are shown in Fig. 10 for the N96, N216 and N512
configurations of the MetUM and ERA-Interim reanaly-
sis. In all MetUM configurations the Findlater/Somali Jet
is too weak, with maximum wind speeds of approximately
20 25 30 35 40 45 50 55 60 65 70
1000
925
850
700
(b) N96
20 25 30 35 40 45 50 55 60 65 70
Longitude
1000
925
850
700
Pressure (hPa)
(c) N216
5.0 m s-1
-0.05 Pa s-1
20 25 30 35 40 45 50 55 60 65 70
Longitude
1000
925
850
700
(d) N512
20 25 30 35 40 45 50 55 60 65 70
1000
925
850
700
Pressure (hPa)
(a) ERA-Interim
0.0 2.5 5.0 7.5 10.0 12.5
Meridional wind (m s
-1
)
Fig. 11 Climatological JJAS 1000, 925, 850, and 700 hPa meridional
(colours, m s−1), zonal (vectors, m s−1) and vertical wind transects
(vectors, Pa s−1) averaged from 10° N to 15° N for a ERA-Interim
and the b N96, c N216 and d N512 configurations of the MetUM
(ensemble averaged). All fields have been interpolated to the N96
grid. Also shown is a transect of the grid-box mean orography (grey)
in each model configuration on its native grid. If a pressure level is
below the orography at any time that grid-point point is shown in
white
S. J. Johnson et al.
1 3
14 m s−1 in the MetUM and approximately 17 m s−1 in
ERA-Interim. However, the speed increases as resolu-
tion increases, particularly just to the north of Madagascar
and just to the east of the EAH. Jet speeds also increase
across the meridional extent of the Arabian Sea, and have
increased wave-like structure over the tip of the Indian pen-
insula and into the southern Bay of Bengal, more similar
to the structure in ERA-Interim. Northward flow along the
Indus valley (Pakistan) also increases, increasing the south-
westerly bias there. This could be related to the increased
wind speeds in the Somali Jet, or flow being forced south-
westerly by improved resolution of the Western Ghats.
Wind speed also increases through narrow channels such as
the Hindu Kush and the Turkana Channel in east Africa.
To examine the increased speed in the Findlater/Somali
Jet, Fig. 11 shows a 5° wide zonal transect of ensem-
ble mean JJAS wind speed at 12.5° N in each MetUM
configuration and ERA-Interim. A transect of the orog-
raphy, including the EAH, is also shown. The maximum
height of the highlands in this transect increases by over
700 m as resolution increases from N96 to N512. Somali
Jet maximum velocities are shifted east and intensify, likely
as a result of the increased confinement from the orogra-
phy. The increased intensity of the jet is an improvement
relative to ERA-Interim, while the eastward confinement,
and consequent decrease in wind speeds around 50° E, is
detrimental.
Slingo et al. (2005) use sensitivity tests removing the
EAH to show that the integrated wind stress curl in the
Arabian Sea decreases (becomes more negative due to
more clockwise curl) when the EAH are introduced, due
to increased anticyclonic wind stress in the Arabian Sea
from the standing wave introduced into the Somali Jet. In
Fig. 12, we show the JJAS wind stress, JJAS wind stress
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
-4.5
-4.0
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
Integrated wind stress curl (x 10
4 N m-1)
ERA-Interim
N96
N216
N512
(d) Monthly integrated Arabian Sea wind stress curl
0.3 N m-2
(a) ERA-Interim
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
40N
-0.3 -0.2 -0.1 0.0 0.1 0.20.3
wind stress curl (x 10
-7
N m
-3
)
0.3 N m-2
(c) N512
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
0.3 N m-2
(b) N96
40E 60E 80E 100E 120E 140E 160E 180E
20S
10S
0
10N
20N
30N
40N
Fig. 12 Indo-Pacific JJAS wind stress (vectors) and wind stress curl
(contours) in a ERA-Interim reanalysis and the b N96 and c N512
configurations of the MetUM (ensemble averaged). d The seasonal
cycle of the wind stress curl integrated over the Arabian Sea (outlined
in black in a–c) in ERA-Interim (black) and the N96 (blue), N216
(green) and N512 (red) configurations of the MetUM. Each model
is evaluated on its own grid to avoid interpolation effects over coast-
lines. The shaded area represents where the difference between N96
and N512 is significant by a Mann–Whitney rank sum test
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
curl and the annual cycle of the wind stress curl integrated
over the Arabian Sea in the N96 and N512 configurations
of the MetUM, as well as ERA-Interim. As resolution
increases, the summer wind stress curl decreases, improv-
ing relative to ERA-Interim reanalysis. This indicates that,
similar to introducing the EAH in Slingo et al. (2005),
the increased resolution of the East African Highlands is
improving the speed and curl of the Somali Jet.
4.2 Maritime Continent and coastal tiling
The largest area JJAS precipitation sensitivity to resolu-
tion in this study is the precipitation increase over Maritime
Continent islands and the precipitation decrease over the
northern seas of the Maritime Continent. Using sensitivity
experiments, Schiemann et al. (2014) attributes a similar
change in annual precipitation in their resolution sensitiv-
ity study with HadGEM1 (an older version of the MetUM,
Johns et al. 2006), to the decreased influence of the MetUM
coastal tiling scheme with resolution. In recent versions of
the MetUM, grid points that contain sub-grid land and sea,
including points with sub-grid scale islands, are designated
“coastal” grid points, and their properties are an amalgam
of land and sea properties (see Sect. 2.2.2 and Walters
et al. 2011). Strachan (2007) and Schiemann et al. (2014)
found that the coastal tiling scheme that calculated these
combined properties generated unrealistically low latent
heat fluxes on coastal grid points. At higher resolution, the
coastal points with their low latent heat fluxes covered less
area (see Fig. 1), increasing the mean latent heat flux and
precipitation over the Maritime Continent.
The GA3 configuration of HadGEM3 used here includes
a fix to the coastal tiling scheme (“buddy scheme”,
Sect. 2.2.2 and Walters et al. 2011) that raises the latent
heat flux over coastal points (not shown). Building on
past work, we reanalyse the precipitation sensitivity over
the Maritime Continent in light of these alterations to the
MetUM.
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
(a) OAFlux observations
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
(d) N512
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
(c) N216
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
(b) N96
3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5
Latent Heat Flux (mm day
-1
)
Fig. 13 JJAS latent heat flux (mm day−1) over the Maritime Continent region from the a OAFlux observations (1985–2011) over sea only, b
N96, c N216 and d N512 configurations of the MetUM (1985–2011, ensemble average). All fields are shown on their native grid
S. J. Johnson et al.
1 3
To confirm that introducing the buddy scheme has
changed the Maritime Continent latent heat flux resolution
sensitivity, we show the JJAS mean latent heat flux over the
Maritime Continent in OAFlux observations and the N96,
N216 and N512 configurations of the MetUM in Fig. 13.
While the latent heat fluxes surrounding the islands in the
MetUM are lower than over the islands and lower than over
the surrounding seas, low latent heat fluxes do not appear
to be confined to coastal grid points as in Schiemann et al.
(2014). To determine the reasons for the low latent heat
fluxes, we examined surface winds over the Maritime Con-
tinent. Figure 14 shows the JJAS 10 m circulation in the
N96 and N512 configurations of the MetUM as well as the
SCOW climatology of QuickSCAT observations. Overlaid
are contours of Maritime Continent orography. Low 10 m
wind speeds accompany the low latent heat fluxes, and the
lowest wind speeds occur in the lees of the orography on
the Maritime Continent. These lees gain definition at higher
resolution as the orography and coastlines over the Mari-
time Continent contain increased detail.
An example is New Guinea, the large island roughly
centred on 140° E and 5° S, and the small, crescent
shaped, New Britain island to its east. In Fig. 14, in
the N96 configuration, the 10 m wind is southerly over
central New Guinea, whereas in the N512 configura-
tion, the wind is forced southeasterly, around the New
Guinea Highlands. The northern branch of the flow is
also guided by the orography on New Britain, and in
the N512 configuration the wind is channeled tightly
between New Britain and New Guinea. This increases
the wind speed near the coast of New Guinea, and
decreases it in the lee of New Britain. The adjustments
5 m s-1
(c) N512
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
wind speed (m s
-1
)
5 m s-1
(b) N96
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
5 m s-1
(a) SCOW observations
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
2 m s-1
(d) N512-N96
90E 100E 110E 120E 130E 140E 150E 160E
10S
5S
0
5N
10N
15N
20N
Fig. 14 a JJAS surface winds over the ocean from QuickSCAT
(SCOW climatology 2000–2006) and JJAS 10 m winds for the b N96
and c N512 configurations of the MetUM over the Maritime Conti-
nent region (1985–2011, ensemble average). d The difference in 10 m
wind in the N512 and N96 configurations. Note that the reference
vector is 2 m s−1 in (d). All data are interpolated to the N96 grid for
comparison. In all panels, vectors are coloured by wind speed. The
MetUM grid-box mean orography is contoured at 500 m, 1 km, and
3 km on the N96 grid in (b), and on the N512 grid in (c) and (d)
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
to the 10 m wind have a pronounced effect on the latent
heat flux, shifting high latent heat fluxes north of New
Guinea southward and lowering latent heat fluxes in the
lee of New Britain.
Lower latent heat fluxes and adjusted 10 m wind speeds
around the islands in the higher resolution MetUM con-
figurations are overall in better agreement with the obser-
vational products. For example, leeward of New Britain,
the latent heat flux and 10 m wind speed are much lower
in OAFlux (Fig. 13) and QuikSCAT (Fig. 14) observations
than in the MetUM at all resolutions. However, despite
these improvements at higher resolution, latent heat flux
and surface wind biases are still large, and the MetUM pre-
cipitation improves in some places and degrades in others
(Fig. 3).
To explore the effect of improved resolution of Maritime
Continent and Indochina orographic lees on the larger scale
precipitation response, we examined the column-integrated
moisture budgets over the northern Maritime Continent,
where the precipitation decreases with resolution, and
over the southern Maritime Continent, where precipitation
increases with resolution. The budgets are averaged from
90° E to 160° E and from 10° S to 10° N, divided between
northern and southern regions at the equator (shown in
Fig. 16). In Fig. 15, latent heat flux decreases as resolution
increases both north and south of the equator, consistent
with the maps in Figs. 13 and 14. This reduction in evapo-
ration is much less than the reduction in precipitation north
of the equator. The precipitation decrease is instead associ-
ated with a decrease in the moisture flux convergence north
of the equator. South of the equator, the moisture flux con-
vergence increases, which supplies the moisture for the pre-
cipitation increase there despite the evaporation decrease.
As resolution increases, average precipitation north
of the equator worsens, evaporation south of the equator
improves and moisture convergence north of the equator
worsens relative to the reanalyses. However, the large dif-
ference between the reanalysis products prevents a clear
comparison for other terms. Latent heat flux and 10 m
wind speed maps of ERA-Interim and MERRA reanalysis,
comparable to Figs. 13 and 14, show large discrepancies
in regional features as well. Some regions in the analysis
agree better with the MetUM, while others agree better
with OAFlux and SCOW (not shown).
In Table 3 we list the JJAS precipitation and latent heat
flux averaged over land and sea grid-points separately,
defined on the native model grid, in the N96, N216 and
N512 configurations. Precipitation over land increases
both north and south of the equator, while precipitation
over the sea decreases both north and south of the equa-
tor. This indicates a shift in the precipitation from sea to
land. Latent heat flux also increases over land, consistent
with the increased precipitation, and decreases over sea,
both north and south of the equator. These changes indi-
cate that the changes in the mean moisture budgets over the
MFCLHP
ERA
MERRA
N96
N216
N512
ERA
MERRA
N96
N216
N512
ERA
MERRA
N96
N216
N512
0
1
2
3
4
5
6
7
8
9
mm day -1
North
South
Fig. 15 Climatological JJAS column integrated moisture budgets
including precipitation (P), latent heat flux (LH) and moisture flux
convergence (MFC), averaged over grid-points in the Maritime Conti-
nent region (90°–160° E, 10° S–10° N, divided between northern and
southern regions at the equator). ERA-Interim (black) and MERRA
reanalysis (black) and the N96 (blue), N216 (green) and N512 (red)
configurations of the MetUM are shown. For reanalysis, crosses rep-
resent averages north of the equator and dots represent averages south
of the equator. For MetUM configurations, solid lines connect north-
ern averages and dashed lines connect southern averages. The error
bars show the spread in the MetUM ensemble
Table 3 Ensemble averaged
JJAS precipitation and latent
heat flux (mm day−1) in the
MetUM N96, N216 and N512
configurations averaged over
all, land and sea points in the
northern and southern regions
of the Maritime Continent (90°–
160° E, 10° S–10° N, divided
between northern and southern
regions at the equator)
Total Land Ocean
P LH P LH P LH
North
N96 7.81 4.31 6.27 4.10 8.39 4.36
N216 6.48 4.10 6.57 4.26 6.61 4.13
N512 6.03 3.96 7.32 4.40 5.83 3.91
South
N96 5.56 5.02 4.47 3.64 7.11 5.79
N216 6.03 4.96 6.20 3.76 6.36 5.53
N512 6.62 4.84 6.98 3.78 6.65 5.27
S. J. Johnson et al.
1 3
northern and southern Maritime Continent are related to the
higher land fraction south of the equator than north of the
equator. There is a very strong diurnal cycle over the Mari-
time Continent which is often incorrectly timed in GCMs
including the MetUM (e.g. Yang and Slingo 2001). Analy-
sis of the precipitation diurnal cycle at each grid point over
the Maritime Continent (not shown) indicates that there is
no resolution sensitivity of phase at these resolutions, sim-
ply a change in amplitude that is consistent with the mean
precipitation change.
The increase in precipitation over land (Table 3) and the
clear increase in the resolution of the lees of the mountain-
ous Maritime Continent islands in Figs. 13 and 14, suggest
that the increased resolution of orography is playing a role
in the large-scale response. The orography over the Mari-
time Continent and the Indochina peninsula forms an arc
of orographic features that inhibit the southerly and west-
erly flow into the South China Sea and the northern seas
of the Maritime Continent. This analysis suggests that the
increasing influence of these features at higher resolution
causes increased moisture convergence and precipitation
on the windward side of the orography, which leads to
decreased moisture availability on the leeward side, causing
reduced precipitation. Figure 16 shows the moisture flux
into the northern and southern Maritime Continent from
different directions in the N96 and N512 configurations
of the MetUM. While the moisture flux into the southern
Maritime Continent from the south changes very little, the
flux from the south to the north of the Maritime Continent
decreases by nearly 30 %. The westerly moisture flux into
the northern Maritime Continent also decreases by about
30 %. Both of these reductions are likely due to a combi-
nation of increased precipitation windward of orography
decreasing moisture in the flow and decreased circulation
strength due to the decreased latent heating associated
with the reduced precipitation over the northern Maritime
Continent. Similar shifts in precipitation over the Mari-
time Continent are seen in other GCMs (Bacmeister et al.
2014), perhaps suggesting that many GCMs are sensitive
to increased resolution of orography in the Maritime Con-
tinent region.
5 Discussion
5.1 Contributions to the total response
To understand the resolution sensitivity of the South Asian
monsoon, we quantified changes in precipitation and circu-
lation between the N96, N216 and the N512 configurations
of the MetUM on several time scales over the Indo-Pacific
warm pool. We then diagnosed the processes responsi-
ble for the largest changes. Here, we combine the differ-
ent aspects of our analysis to build an overall picture of the
resolution sensitivity of the South Asian monsoon in the
MetUM.
JJAS precipitation increases slightly over India when
resolution is increased. While the precipitation change is
small compared to the bias, and small compared to many
other changes in Indo-Pacific JJAS precipitation, it is still
a large fraction of the JJAS precipitation over India in the
N96 configuration of the MetUM. We showed in Sect. 3.4
that monsoon LPS rainfall contributes to the precipitation
increase over northern central India, but it does not con-
tribute to the precipitation increase over southern India.
No diagnostic we analysed suggested a clear cause for
1.5 mm day-1
N96
N512
80E 90E 100E 110E 120E 130E 140E 150E 160E 170E
15S
10S
5S
0
5N
10N
15N
0.5 mm day-1
80E 90E 100E 110E 120E 130E 140E 150E 160E 170E
15S
10S
5S
0
5N
10N
15N
N512-N96
Fig. 16 Left Moisture flux from different directions into the Mari-
time Continent region (90°–160° E, 10° S–10° N divided between
northern and southern regions at the equator). The northern and
southern regions of the Maritime Continent are marked by the grey
dashed lines. Blue arrows indicate the moisture flux through each
boundary in the N96 configuration of the MetUM, and red arrows
indicate moisture flux in the N512 configuration. Right Difference
between N512 and N96 moisture flux through each boundary. Note
the different scales in the two panels. The total moisture flux conver-
gence increases by 1.03 mm day−1 south of the equator and decreases
by 1.39 mm day−1 north of the equator, with the largest contribution
coming from the change in moisture flux across the equator
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
the improvement over southern India. It is likely due to a
combination of the other changes discussed: changes in
large-scale circulation related to the changes in precipita-
tion over the Maritime Continent, changes in the Somali Jet
and improved resolution of the Western and Eastern Ghats
(mountain ranges on the west and east coasts of peninsu-
lar India). It is also possible that small organised systems
of convection are better able to propagate to southern India
over the Ghats from the west or from the Bay of Bengal to
the east. Sensitivity experiments in a very similar N96 con-
figuration of the MetUM indicate that the Western Ghats
introduce a cyclonic perturbation over southern India and
Sri Lanka, and an anticyclonic perturbation over northern
India, with very little associated change in precipitation
over southern peninsular Indian (Turner et al. 2015). These
responses are not particularly consistent with the response
seen in Fig. 3. This may indicate the increased resolution of
the Western Ghats contributes little to the resolution sensi-
tivity and the remote influences of the Maritime Continent
and the Somali Jet contribute more. Due to the nonlinear
interaction of the resolution-related changes it is difficult to
establish conclusively.
Our analysis in Sect. 4.1.2 indicates the increased wind
speed and wave structure in the Findlater/Somali Jet are
due to the increased resolution of the East African High-
lands and orography over the Arabian peninsula. In con-
trast, the changes over the Maritime Continent are likely
due to a combination of processes. Our analysis in Sect. 4.2
suggests that the increased precipitation over the Maritime
Continent islands and decreased precipitation north of the
Maritime Continent are largely caused by improved rep-
resentation of the orography over the Maritime Continent
and Indochina peninsula, which increases convergence and
precipitation over the orography and decreases the surface
wind speed and latent heat flux in the lees of the orogra-
phy. However, Schiemann et al. (2014) have shown that a
similar pattern of precipitation changes is created due to
improved the resolution of coastlines in an older version
of the MetUM. Schiemann et al. (2014) attribute these
changes to erroneously low latent heat fluxes on coastal
grid-points that, thanks to an improvement in the scheme,
are not present in our integrations. The integrations ana-
lysed here are also at higher resolution than the integra-
tions analysed in Schiemann et al. (2014); coastal grid
points occupy less area in our integrations and should con-
tribute less to the resolution sensitivity. However, changes
in the land fraction of grid-points as resolution increases
may change the surface roughness, which could have a
very similar effect to changing the orography. To diagnose
the roles of the decreasing influence of the coastal tiling
scheme and the increasing influence of the orography in
our integrations, sensitivity experiments similar to those
conducted in Schiemann et al. (2014) are underway with
higher resolution and more recent versions of the MetUM
(Johnson et al. in preparation).
Improved resolution of the atmospheric dynamics may
also contribute to the resolution sensitivity. The JJAS pre-
cipitation decrease in the West Pacific and the equatorial
Indian Ocean corresponds to a precipitation decrease at
the centre of the inter-tropical convergence zone (ITCZ).
Precipitation increases border the precipitation decrease
to the north and south, perhaps indicating a broadening
of the ITCZ as resolution increases. Global changes in
zonal mean cloud and ascent indicate resolution sensitiv-
ity in the MetUM that is consistent with global changes in
the ITCZ (Tsushima personal communication, 2014). In
other GCMs, aqua-planet simulations at different resolu-
tions have demonstrated that changing horizontal resolu-
tion and dynamical time-step can broaden or narrow the
ITCZ, increase or decrease its peak and change its single
or double peaked nature depending on how the dynamical
core interacts with the parametrisation schemes (William-
son 2008; Landu et al. 2014; Zarzycki et al. 2014). Aqua-
planet models of the version of the MetUM used here are
needed to diagnose the contribution of ITCZ changes to
the resolution sensitivity in the monsoon domain. How-
ever, consistent with Schiemann et al. (2014), precipitation
changes in the warm pool region are larger than in other
equatorial regions of the model (not shown). Consequently,
if the ITCZ is changing shape with increasing resolution,
the effect is amplified in the Indo-Pacific, likely through
feedbacks with the processes discussed previously.
5.2 Robustness of resolution sensitivity to GCM
differences
Resolution sensitivity in the South Asian monsoon domain
depends on the resolutions analysed and the GCM used, but
there are some commonalities across studies. The Somali
Jet, due to its reliance on the orography in east Africa,
often shows resolution improvements in speed and posi-
tion (Sperber et al. 1994; Stephenson et al. 1998; Sabin
et al. 2013). Studies in multiple versions of the MetUM
(Schiemann et al. 2014; Johnson et al. in preparation) as
well as other GCMs (Sabin et al. 2013; Bacmeister et al.
2014) show similar precipitation changes over the Mari-
time Continent.
The increase in precipitation over India in the
MetUM is consistent with the resolution sensitivity of the
annual precipitation range over India in the CMIP3 mod-
els (Kim et al. 2008). In other resolution sensitivity stud-
ies using individual GCMs, orographic rain bands over
India tend to become more well defined, but the sign of
precipitation changes over India are highly variable (Sper-
ber et al. 1994; Stephenson et al. 1998; Delworth et al.
2012; Sabin et al. 2013; Bacmeister et al. 2014). This is
S. J. Johnson et al.
1 3
also true within MetUM configurations. We analysed this
MetUM configuration due to the ensemble of high reso-
lution runs available in the UPSCALE dataset. The more
recent GA6 version of the MetUM (Walters et al. 2015)
which contains a new dynamical core and updates to
the gravity wave drag scheme, shows a similar, slightly
larger, resolution sensitivity over most of the South Asian
monsoon domain. The response over India, however,
is much larger (not shown). The reason for this is being
explored in work focusing on monsoon LPS (Levine et al.
in preparation).
5.3 Shift in precipitation over land and sea
Demory et al. (2014) and Schiemann et al. (2014) have
shown in several versions of the MetUM that global precip-
itation shifts from sea to land as resolution increases. In the
UPSCALE integrations, the fraction of global precipitation
occurring over land increases from approximately 21.5 %
in the N96 configuration to 23 % in the N512 configura-
tion of the MetUM. The monsoon domain in JJAS is con-
sistent with that global pattern, particularly over the Mar-
itime Continent. It is not clear whether the local changes
are a cause, an effect, or simply consistent with the global
changes. It is possible that better resolution of orography
in the tropics is causing more tropics-wide convergence,
and consequently precipitation, over land. Future work will
examine how much of the global precipitation shift this
effect can explain.
5.4 Comparison to MetUM biases
We suggest that an improved representation of the orogra-
phy has had a significant impact on the large scale in mul-
tiple regions. While this indicates improved resolution of
a process, it does not necessarily improve the fidelity of
simulated precipitation or circulation with respect to obser-
vations. For example, there is a deficit of precipitation in
the N96 configuration of the MetUM in the northern seas
of the Maritime Continent. While increased orographic
flow blocking from the Maritime Continent and Indo-
china clearly improves the small-scale flows around the
orography, our analysis suggests it further decreases rain-
fall north of the Maritime Continent, increasing the bias.
This indicates that “low” resolution may compensate for
other errors. As GCMs increase in resolution, deficiencies
in parameterisations may become more, rather than less,
apparent because many GCMs parameterisations were
originally designed for low resolution. Scale-aware param-
eterisations may be necessary to realise the full potential
of high resolution models for simulating the South Asian
monsoon.
6 Conclusions
We have studied the sensitivity of South Asian monsoon
domain rainfall to increasing horizontal resolution from
approximately 200–40 km (N96–N512). We find a number
of changes and, where diagnostics indicate a mechanism,
suggest reasons for the changes:
• Small increase in precipitation over the Western Ghats,
southern and central India.
• Improved representation of South Asian monsoon pre-
cipitation bands anchored to individual orographic
features, but without a systematic change in the total
amount of precipitation associated with orography.
• Increase in the speed and curvature of the Findlater/
Somali Jet, due to increased resolution of the East Afri-
can Highlands and orography on the Arabian peninsula.
• Increased precipitation over and around the Maritime
Continent Islands, most likely due to increased con-
vergence over the better resolved Maritime Continent
islands and orography.
• Decreased precipitation in the northern seas of the Mar-
itime Continent due to decreased moisture flux into the
region, likely because of increased precipitation over
the better resolved Maritime Continent and Indochina
peninsula orography. Decreased reliance on the coastal
tiling scheme and sensitivity of the ITCZ shape to
increased resolution may also play a role. Further work
on disentangling these effects is in progress (Johnson
et al. in preparation).
Each of these changes may be partly driven by, or con-
tributing drivers for, global scale changes with resolution
such as changes in the ITCZ (Williamson 2008; Landu
et al. 2014) and a global shift of precipitation from sea to
land (Demory et al. 2014; Schiemann et al. 2014).
At these resolutions, the increase in wind speed in the
Somali Jet and the precipitation changes over the Mari-
time Continent are more similar across GCMs (Sabin et al.
2013; Bacmeister et al. 2014) than the resolution sensitivity
of rainfall over India.
Similar changes occur when resolution is increased from
N216 to N512 as occur when resolution is increased from
N96 to N216. Increasing resolution at lower resolutions
gives the largest benefit, but increasing resolution to 0.35°
continues to significantly redistribute tropical rainfall and
consequently, diabatic heating.
As seen in previous studies, improvements as a result of
increased resolution are small compared to Indian monsoon
biases: the precipitation deficit over India, the excess precipita-
tion over the Indian Ocean, and the weakness of the Somali
Jet. Where changes are larger, such as over the Maritime
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
Continent, some biases improve (increased precipitation over
the islands), while some worsen (weak dry bias north of the
Maritime Continent becomes a large dry bias). This is an indi-
cation that from 200 to 40 km resolution, increasing resolution
is not a substitute for improving parameterisations, which may
have a larger impact on South Asian monsoon biases.
The UPSCALE dataset can be obtained from: http://proj.
badc.rl.ac.uk/upscale/wiki.
Acknowledgments S.J.B. and R.S. were supported by the Joint
Weather and Climate Research Programme, a partnership between
the Natural Environment Research Council (NERC) and the Met
Office, under University of Reading Contract R8/H9/37. AGT held
a NERC Fellowship, No. NE/H015655/1. R.C.L., G.M.M., M.S.M.,
M.J.R. and J.S. are supported by the Joint UK DECC/Defra Met
Office Hadley Centre Climate Programme (GA01101). R.C.L.,
A.G.T. and G.M.M. were additionally supported by the NERC
Changing Water Cycle (CWC) SAPRISE Project (reference NE/
I022841/1 and NE/I022469/1). RCL was further supported by the
European Commissions 7th Framework Programme, under Grant
Agreement No. 282672, EMBRACE Project. S.J.W., R.S., P.L.V.,
M.E.D. were supported by the National Centre for Atmospheric Sci-
ence Climate directorate (NCAS-Climate), a collaborative centre of
NERC. P.L.V. acknowledges the support provided to the Willis Chair
in Climate System Science and Climate Hazards. While working on
the UPSCALE simulations, J.S. was supported by a UK Technol-
ogy Strategy Board Knowledge Transfer Partnership. We also thank
the large team of model developers, infrastructure experts and all the
other essential components required to conduct the UPSCALE cam-
paign, in particular the PRACE infrastructure and the Stuttgart HLRS
supercomputing centre, as well as the STFC CEDA service for data
storage and analysis using the JASMIN platform. We acknowledge
use of the MONSooN system, a collaborative facility supplied under
the Joint Weather and Climate Research Programme, which is a stra-
tegic partnership between the Met Office and the Natural Environ-
ment Research Council. We acknowledge the use of e-Atlas products
developed by Cyclone warning Research Centre, Regional Meteoro-
logical Centre, Chennai, India. GPCP Precipitation data is provided
by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their
web site at http://www.esrl.noaa.gov/psd/.
Open Access This article is distributed under the terms of the Crea-
tive Commons Attribution 4.0 International License (http://creativecom-
mons.org/licenses/by/4.0/), which permits unrestricted use, distribution,
and reproduction in any medium, provided you give appropriate credit
to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made.
References
Adler RF, Huffman GJ, Chang A, Ferraro R, Xie PP, Janowiak J,
Rudolf B, Schneider U, Curtis S, Bolvin D, Gruber A, Susskind
J, Arkin P, Nelkin E (2003) The version-2 global precipitation
climatology project (GPCP) monthly precipitation analysis
(1979–present). J Clim 4:1147–1167
Annamalai H, Hamilton K, Sperber KR (2007) The South Asian sum-
mer monsoon and its relationship with ENSO in the IPCC AR4
simulations. J Clim 20(6):1071–1092. doi:10.1175/JCLI4035.1
Arakawa A, Lamb V (1977) Computational design of the basic
dynamical processes of the UCLA general circulation model, vol
17. Academic Press, New York, pp 173–265
Ashok K, Soman MK, Satyan V (2000) Simulation of monsoon dis-
turbances in a GCM. Pure Appl Geophys 157:1509–1539
Bacmeister JT, Wehner MF, Neale RB, Gettelman A, Hannay C,
Lauritzen PH, Caron JM, Turesdale JE (2014) Exploratory
high-resolution climate simulations using the community
atmosphere model (CAM). J Clim 27:3073–3099. doi:10.1175/
JCLI-D-13-00387.1
Bengtsson L, Hodges KI, Roeckner E (2006) Storm tracks and cli-
mate change. J Clim 19:3518–3543
Best MJ, Pryor M, Clark DB, Rooney GG, Essery RLH, Menard CB,
Edwards JM, Hendry MA, Porson A, Gedney N, Mercado LM,
Sitch S, Blyth E, Boucher O, Cox PM, Grimmond CSB, Harding
RJ (2011) The joint UK land environment simulator (JULES),
model description—part 1: energy and water fluxes. Geosci
Model Dev 4:677–699. doi:10.5194/gmd-4-677-2011
Bollasina M, Nigam S (2009) Indian Ocean SST, evaporation, and
precipitation during the South Asian summer monsoon in
IPCC-AR4 coupled simulations. Clim Dyn 33:1017–1032.
doi:10.1007/s00382-008-0477-4
Boos WR, Kuang Z (2010) Dominant control of the South Asian
monsoon by orographic insulation versus plateau heating. Nature
463(7278):218–222. doi:10.1038/nature08707
Boos WR, Kuang Z (2013) Sensitivity of the South Asian monsoon to
elevated and non-elevated heating. Scientific Reports 3(1192):3–
6. doi:10.1038/srep01192
Bracco A, Kucharski F, Molteni F (2005) Internal and forced modes
of variability in the Indian Ocean. Geophys Res Lett 32:1–4. doi:
10.1029/2005GL023154
Brown AR, Webster S (2004) Orographic flow-blocking scheme char-
acteristics. Q J R Meteorol Soc 130:3015–3028. doi:10.1256/
qj.04.40
Chakraborty A, Nanjundiah RS, Srinivasan J (2006) Theoreti-
cal aspects of the onset of Indian summer monsoon from
perturbed orography simulations in a GCM. Ann Geophys
24(2003):2075–2089
Charney JG, Phillips NA (1953) Numerical Integration of the quasi-
geostrophic equations for barotropic and simple baroclinic flows.
J Meteorol 10(2):71–99
Clark DB, Mercado LM, Sitch S, Jones CD, Gedney N, Best MJ,
Pryor M, Rooney GG, Essery RLH, Blyth E, Boucher O, Hard-
ing RJ, Huntingford C, Cox PM (2011) The joint UK land envi-
ronment simulator (JULES), model description—part 2: carbon
fluxes and vegetation dynamics. Geosci Model Dev 4:701–722.
doi:10.5194/gmd-4-701-2011
Davies T, Cullen MJP, Malcolm AJ, Mawson MH, Staniforth A,
White AA, Wood N (2005) A new dynamical core for the Met
Offices global and regional modelling of the atmosphere. Q J R
Meteorol So 131:1759–1782. doi:10.1256/qj.04.101
Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S,
Andrae U, Balmaseda MA, Balsamo G, Bauer P, Bechtold P, Bel-
jaars ACM, Berg LVD, Bidlot J, Bormann N, Delsol C, Dragani R,
Fuentes M, Geer AJ, Dee DP (2011) The ERA-Interim reanalysis:
configuration and performance of the data assimilation system. Q
J R Meteorol Soc 137:553–597. doi:10.1002/qj.828
Delworth TL, Rosati A, Anderson W, Adcroft AJ, Balaji V, Ben-
son R, Dixon K, Griffies SM, Lee HC, Pacanowski RC, Vecchi
GA, Wittenberg AT, Zeng F, Zhang R (2012) Simulated cli-
mate and climate change in the GFDL CM2.5 high-resolution
coupled climate model. J Clim 25:2755–2781. doi:10.1175/
JCLI-D-11-00316.1
Demory ME, Vidale PL, Berrisford P, Strachan J, Schiemann R, Rob-
erts MJ, Mizielinski MS (2014) The role of horizontal resolution
in simulating drivers of the global hydrological cycle. Clim Dyn
42(7–8):2201–2225. doi:10.1007/s00382-013-1924-4
Donlon CJ, Martin M, Stark J, Roberts-Jones J, Fiedler E, Wim-
mer W (2012) Remote sensing of environment the operational
S. J. Johnson et al.
1 3
sea surface temperature and sea ice analysis (OSTIA) sys-
tem. Remote Sens Environ 116:140–158. doi:10.1016/j.
rse.2010.10.017
Fritsch JM, Chappell CF (1980a) Numerical prediction of convec-
tively driven mesoscale pressure systems—part I: convective
parameterization. J Atmos Sci 37:1722–1733
Fritsch JM, Chappell CF (1980b) Numerical prediction of convec-
tively driven mesoscale pressure systems—part II: mesoscale
model. J Atmos Sci 37:1735–1762
Gadgil S, Sajani S (1998) Monsoon precipitation in the AMIP runs.
Clim Dyn 14:659–689
GLOBE Task Team (1999) The global land one-kilometer base ele-
vation (GLOBE) digital elevation model, version 1.0. Techni-
cal report, National Oceanic and Atmospheric Administration,
National Geophysical Data Center, 325 Broadway, Boulder, Col-
orado 80305-3328, USA. http://www.ngdc.noaa.gov/mgg/topo/
globe.html
Gregory D, Allen S (1991) The effect of convective downdrafts upon
NWP and climate simulations. In: Proceedings of the ninth con-
ference on numerical weather prediction. American Meteorologi-
cal Society, Denver, Colorado, pp 122–123
Gregory D, Rowntree PR (1990) A mass flux convection scheme with
representation of cloud ensemble characteristics and stability
dependant closure. Mon Weather Rev 118:1483–1506
Gregory D, Shutts GJ, Mitchell JR (1998) A new gravity-wave-drag
scheme incorporating anisotropic orography and low-level wave
breaking: impact upon the climate of the UK Meteorological
Office Unified Model. Q J R Meteorol Soc 124:463–493
Hodges KI (1994) A general method for tracking analysis and
its application to meteorological data. Mon Weather Rev
122:2573–2586
Hoskins BJ, Rodwell MJ (1995) A model of the Asian summer mon-
soon. Part I: the global scale. J Atmos Sci 52(9):1329–1340
Huffman GJ, Adler RF, Morrissey MM, Bolvin DT, Curtis S, Joyce
R, McGavok B, Susskind J (2001) Global precipitation at one-
degree daily resolution from multisatellite observations. J
Hydrometeorol 2:36–50
IMD (2011) IMD Cyclone and Monsoon Depression eAtlas. Techni-
cal report, cyclone warning Research Centre, Regional Mete-
orological Centre, Chennai, India. http://www.rmcchennaieatlas.
tn.nic.in
Jha B, Krishnamurti TN, Christides Z (2000) A note on horizontal
resolution dependance for monsoon rainfall simulations. Mete-
orol Atmos Phys 74:11–17
Johns TC, Durman CF, Banks HT, Roberts MJ, McLaren AJ, Ridley
JK, Senior CA, Williams KD, Jones A, Rickard GJ, Cusack S,
Ingram WJ, Crucifix M, Sexton DMH, Joshi MM, Dong BW,
Spencer H, Hill RSR, Gregory JM, Keen AB, Pardaens AK,
Lowe JA, Bodas-Salcedo A, Stark S, Searl Y (2006) The new
hadley centre climate model (HadGEM1): evaluation of coupled
simulations. J Clim 19:1327–1353
Kanji GK (2006) 100 statistical tests, 3rd edn. Sage, Beverly Hills,
CA
Kim HJ, Wang B, Ding Q (2008) The global monsoon variability sim-
ulated by CMIP3 coupled climate models*. J Clim 21(20):5271–
5294. doi:10.1175/2008JCLI2041.1
Kitoh A, Kusunoki S (2008) East Asian summer monsoon simulation
by a 20-km mesh AGCM. Clim Dyn 31:389–401. doi:10.1007/
s00382-007-0285-2
Kobayashi C, Sugi M (2004) Impact of horizontal resolution on the
simulation of the Asian summer monsoon and tropical cyclones
in the JMA global model. Clim Dyn 93:165–176. doi:10.1007/
s00382-004-0427-8
Krishnamurthy V, Misra V (2010) Daily atmospheric variability in the
South American monsoon system. Clim Dyn 37(3–4):803–819.
doi:10.1007/s00382-010-0881-4
Kumar KK, Rajagopalan B, Hoerling M, Bates G, Cane M (2006)
Unraveling the mystery of indian monsoon failure during El
Nino. Science 314:115–119. doi:10.1126/science.1131152
Landu K, Leung R, Hagos S, Vinoj V (2014) The dependence of
ITCZ structure on model resolution and dynamical core in
aquaplanet simulations. J Clim 27:2375–2385. doi:10.1175/
JCLI-D-13-00269.1
Li C, Yanai M (1996) The onset and interannual variability of the
Asian summer monsoon in relation to land–sea thermal contrast.
J Clim 9:358–375
Mann HB, Whitney DR (1947) On a test of whether one of two ran-
dom variables is stochastically larger than the other. Ann Math
Stat 18(1):50–60
Martin GM (1999) The simulation of the Asian summer monsoon, and
its sensitivity to horizontal resolution, in the UK Meteorological
Office Unified Model. Q J R Meteorol Soc 125:1499–1525
Meehl GA, Covey C, Delworth T, Latif M, McAvaney B, Mitch-
ell JFB, Stouffer RJ, Taylor KE (2007) The WCRP CMIP3
multimodel dataset. Bull Am Meteorol Soc 88:1383–1394.
doi:10.1175/BAMS-88-9-1383
Mizielinski MS, Roberts MJ, Vidale PL, Schiemann R, Demory ME,
Strachan J, Edwards T, Stephens A, Lawrence BN, Pritchard M,
Chiu P, Iwi A, Churchill J, del Cano Novales C, Kettleborough
J, Roseblade W, Selwood P, Foster M, Glover M, Malcolm A
(2014) High resolution global climate modelling; the UPSCALE
project, a large simulation campaign. Geosci Model Dev 7:1629–
1640. doi:10.5194/gmdd-7-563-2014
Neale RB, Slingo J (2003) The Maritime Continent and its role in the
global climate: a GCM study. J Clim 16:834–848
Randall et al (2007) Climate models and their evaluation. In: Climate
change 2007: IPCC working group I: the physical science basis.
Cambridge University Press, Cambridge, UK and New York, pp
590–662
Rienecker MM, Suarez MJ, Gelaro R et al (2011) MERRA: NASA s
modern-era retrospective analysis for research and applications.
Society 24:3624–3648. doi:10.1175/JCLI-D-11-00015.1
Risien CM, Chelton DB (2008) A global climatology of surface wind
and wind stress fields from eight years of QuickSCAT scatter-
ometer data. J Phys Oceanogr 38:2379–2413. doi:10.1175/2008
JPO3881.1
Roberts MJ, Vidale PL, Mizielinski MS, Demory ME, Schiemann R,
Strachan J, Hodges K, Camp J, Bell R (2015) Tropical cyclones
in the UPSCALE ensemble of high resolution global climate
models. J Clim 28:574–596
Rodwell MJ, Hoskins BJ (1995) A model of the Asian summer mon-
soon. Part II: cross-equatorial flow and PV behavior. J Atmos Sci
52(9):1341–1356
Sabeerali CT, Dandi AR, Dhakate A, Salunke K, Mahapatra S, Rao
SA (2013) Simulation of boreal summer intraseasonal oscil-
lations in the latest CMIP5 coupled GCMs. J Geophys Res
118:4401–4420. doi:10.1002/jgrd.50403
Sabin TP, Ghattas J, Denvil S, Jl Dufresne, Hourdin F, Pascal T (2013)
High resolution simulation of the South Asian monsoon using a
variable resolution global climate model. Clim Dyn 41:173–194.
doi:10.1007/s00382-012-1658-8
Sabre M, Hodges KI, Laval K, Polcher J, Desalmand F (2000) Simu-
lation of monsoon disturbances in the LMD GCM. Mon Weather
Rev 128:3752–3771
Schiemann R, Demory ME, Mizielinski MS, Roberts MJ, Shaf-
frey LC, Strachan J, Vidale PL (2014) The sensitivity of the
tropical circulation and Maritime Continent precipitation to cli-
mate model resolution. Clim Dyn 42:2455–2468. doi:10.1007/
s00382-013-1997-0
Shin DW, Larow TE, Cocke S (2003) Convective scheme and reso-
lution impacts on seasonal precipitation forecasts. Geophys Res
Lett 30(20):2078. doi:10.1029/2003GL018297
The resolution sensitivity of the South Asian monsoon and Indo-Pacific in a global 0.35° AGCM
1 3
Sikka DR (1977) Some aspects of the life history, structure and
movement of monsoon depressions. Pure Appl Geophys
115:1501–1529
Slingo J, Spencer H, Hoskins B, Berrisford P, Black E (2005) The
meteorology of the Western Indian Ocean, and the influence
of the East African highlands. Philos Trans R Soc 363:25–42.
doi:10.1098/rsta.2004.1473
Sperber KR, Palmer TN (1996) Interannual tropical rainfall variabil-
ity in general circulation model simulations associated with the
atmospheric model intercomparison project. J Clim 9:2727–2750
Sperber KR, Hameed S, Potter GL, Boyle JS (1994) Simulation of the
northern summer monsoon in the ECMWF model: sensitivity to
horizontal resolution. J Clim 122:2461–2841
Sperber KR, Annamalai H, Kang IS, Kitoh A, Moise A, Turner AG,
Wang B, Zhou T (2013) The Asian summer monsoon: an inter-
comparison of CMIP5 vs. CMIP3 simulations of the late 20th cen-
tury. Clim Dyn 41:2711–2744. doi:10.1007/s00382-012-1607-6
Stephenson DB, Chauvin F, Royer JF (1998) Simulation of the Asian
summer monsoon and its dependence on model horizontal reso-
lution. J Meteorol Soc Jpn 76(2):237–265
Stowasser M, Annamalai H, Hafner J (2009) Response of the South
Asian summer monsoon to global warming: mean and synoptic
systems. J Clim 22:1014–1036. doi:10.1175/2008JCLI2218.1
Strachan J (2007) Understanding and modelling the climate of the
Maritime Continent. Ph.D. thesis, University of Reading
Strachan J, Vidale PL, Hodges KI, Roberts MJ, Demory ME (2013)
Investigating global tropical cyclone activity with a hierarchy
of AGCMs: the role of model resolution. J Clim 26:133–152.
doi:10.1175/JCLI-D-12-00012.1
Stratton RA, Stirling A, Derbyshire S (2009) Changes and develop-
ments to convective momentum transport (CMT) parameteriza-
tion based on analysis of CRM and SCM: Technical note 530.
Technical report 530, UK Met Office
Tang H, Micheels A, Eronen JT, Ahrens B, Fortelius M (2013) Asyn-
chronous responses of East Asian and Indian summer monsoons
to mountain uplift shown by regional climate modelling experi-
ments. Clim Dyn 40:1531–1549. doi:10.1007/s00382-012-1603-x
Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5
and the experiment design. Bull Am Meteorol Soc 93:485–498.
doi:10.1175/BAMS-D-11-00094.1
Turner AG, Martin GM, Levine RC (2015) The role of peninsu-
lar India in the South Asian summer monsoon. Clim Dyn
(submitted)
US Department of Commerce (2001) 2-Minute gridded global relief
data (ETOPO2). Technical report, National Oceanic and Atmos-
pheric Administration
Walters DN, Best MJ, Bushell AC, Copsey D, Edwards JM, Falloon
PD, Harris CM, Lock AP, Manners JC, Morcrette CJ, Roberts
MJ, Stratton RA, Webster S, Wilkinson JM, Willett MR, Boutle
IA, Earnshaw PD, Hill PG, Maclachlan C, Martin GM, Palmer
MD, Petch JC, Rooney GG, Scaife AA, Williams KD (2011) The
Met Office Unified Model global atmosphere 3.0/3.1 and JULES
global land 3.0/3.1 configurations. Geosci Model Dev 4:919–
941. doi:10.5194/gmd-4-919-2011
Walters DN, Williams KD, Boutle IA, Bushell AC, Edwards JM, Field
PR, Lock AP, Morcrette CJ, Stratton RA, Wilkinson JM, Willett
MR, Brooks ME, Copsey D, Earnshaw PD, Harris CM, Manners
JC, MacLachlan C, Palmer MD, Roberts MJ, Tennant WJ (2015)
The Met Office Unified Model global atmosphere 6.0/6.1 and
JULES Global Land 6.0/6.1.0 congurations. In preparation
Wang B, Jy Lee, Kang IS, Shukla J, Kug JS, Kumar JS, Schemm J,
Luo KJ, Yamagata T, Park CK (2008) How accurately do cou-
pled climate models predict the leading modes of Asian-Aus-
tralian monsoon interannual variability? Clim Dyn 30:605–619.
doi:10.1007/s00382-007-0310-5
Webster PJ, Yang S (1992) Monsoon and ENSO: selectively interac-
tive systems. Q J R Meteorol So 118:877–926
Williamson DL (2008) Convergence of aqua-planet simu-
lations with increasing resolution in the Community
Atmospheric Model, Version 3. Tellus 60A:848–862.
doi:10.1111/j.1600-0870.2008.00339.x
Xie P, Arkin PA (1997) Global precipitation: a 17-year monthly anal-
ysis based on gauge observations, satellite estimates, and numer-
ical model outputs. Bull Am Meteorol Soc 78(11):2539–2558
Xie SP, Xu H, Saji NH, Wang Y (2006) Role of narrow mountains
in large-scale organization of Asian monsoon convection. J Clim
3420:3420–3429
Yang GY, Slingo J (2001) The diurnal cycle in the tropics. Mon
Weather Rev 129(1994):784–801
Yatagai A, Arakawa O, Kamiguchi K, Kawamoto H, Nodzu MI, Ham-
ada A (2009) A 44-year daily gridded precipitation dataset for
Asia based on a dense network of rain gauges. Sci Online Lett
Atmos 5:137–250. doi:10.2151/sola.2009
Yu L (2007) Global variations in oceanic evaporation (1958–2005):
the role of the changing wind speed. J Clim 20:5376–5390. doi:1
0.1175/2007JCLI1714.1
Yu L, Jin X, Weller RA (2008) Multidecade global flux datasets from
the objectively analyzed air–sea fluxes (OAFlux) project: latent
and sensible heat fluxes, ocean evaporation, and related surface
meteorological variables—OAFlux project technical report (OA-
2008-01). Technical report, January, Woods Hold Oceanographic
Institution, Woods Hole, Massachusetts
Zarzycki CM, Levy MN, Jablonowski C, Overfelt J, Taylor M, Ull-
rich P (2014) Aquaplanet experiments using CAM’s variable-
resolution dynamical core. J Clim 27:5481–5503. doi:10.1175/
JCLI-D-14-00004.1
