Theoretical and Applied Climatology
https://doi.org/10.1007/s00704-018-2432-6
ORIGINAL PAPER
Aspect of ECMWF downscaled Regional Climate Modeling in simulating
Indian summer monsoon rainfall and dependencies on lateral
boundary conditions
Soumik Ghosh 1 & R. Bhatla 1,2 & R. K. Mall 2,3 & Prashant K. Srivastava 2,3 & A. K. Sahai 4,5
Received: 28 February 2017 / Accepted: 19 February 2018
# Springer-Verlag GmbH Austria, part of Springer Nature 2018
Abstract
Climate model faces considerable difficulties in simulating the rainfall characteristics of southwest summer monsoon. In this
study, the dynamical downscaling of European Centre for Medium-Range Weather Forecast’s (ECMWF’s) ERA-Interim
(EIN15) has been utilized for the simulation of Indian summer monsoon (ISM) through the Regional Climate Model version
4.3 (RegCM-4.3) over the South Asia Co-Ordinated Regional Climate Downscaling EXperiment (CORDEX) domain. The
complexities of model simulation over a particular terrain are generally influenced by factors such as complex topography,
coastal boundary, and lack of unbiased initial and lateral boundary conditions. In order to overcome some of these limitations, the
RegCM-4.3 is employed for simulating the rainfall characteristics over the complex topographical conditions. For reliable rainfall
simulation, implementations of numerous lower boundary conditions are forced in the RegCM-4.3 with specific horizontal grid
resolution of 50 km over South Asia CORDEX domain. The analysis is considered for 30 years of climatological simulation of
rainfall, outgoing longwave radiation (OLR), mean sea level pressure (MSLP), and wind with different vertical levels over the
specified region. The dependency of model simulation with the forcing of EIN15 initial and lateral boundary conditions is used to
understand the impact of simulated rainfall characteristics during different phases of summer monsoon. The results obtained from
this study are used to evaluate the activity of initial conditions of zonal wind circulation speed, which causes an increase in the
uncertainty of regional model output over the region under investigation. Further, the results showed that the EIN15 zonal wind
circulation lacks sufficient speed over the specified region in a particular time, which was carried forward by the RegCM output
and leads to a disrupted regional simulation in the climate model.
Keywords ECMWF . Regional climate model . Dynamical downscaling . Summer monsoon . Boundary condition . CORDEX
Key Points
1. Assessment of sensitivity of RegCM’s Mix99 (Grell -> Land &
Emanuel -> Ocean) Convective Parameterization Scheme (CPS) in
simulating Intraseasonal variability of Indian summer monsoon (ISM).
2. Evaluation of mixed CPS and its dependencies on boundary conditions.
3. Low efficiency of in situ lateral boundary conditions over the particular
region causes a disturbed rainfall characteristic in RegCM simulation.
4. Low efficiency in EIN15 zonal wind circulation over Indian region
during the phases of monsoon.
* R. Bhatla
rbhatla@bhu.ac.in
3
Institute of Environment and Sustainable Development, Banaras
Hindu University, Varanasi, India
1
Department of Geophysics, Institute of Science, Banaras Hindu
University, Varanasi, India
4
Indian Institute of Tropical Meteorology, Pune, Maharashtra, India
2
DST-Mahamana Centre of Excellence for Climate Change Research,
Institute of Environment & Sustainable Development, Banaras
Hindu University, Varanasi, India
5
India Meteorological Department, Pune, India
S. Ghosh et al.
Abbreviations
BATS
Biosphere-Atmosphere Transfer Scheme
BoB
Bay of Bengal
CCSM3
Community Climate System Model version 3
CLM
Climate Local Model
CMO
Conditional model output
CORDEX Co-Ordinated Regional Climate Downscaling
EXperiment
COSMO
Consortium for Small-Scale Modeling
CPS
Convective Parameterization Scheme
ECDF
Empirical cumulative distribution function
ECHAM5 Fifth-generation atmospheric GCM developed at
the Max Planck Institute for Meteorology
ECMWF
European Centre for Medium-Range Weather
Forecasts
EIN15
ERA-Interim
ERSST
Extended Reconstructed Sea Surface
Temperature
GCM
Global climate model
hPa
Hectopascal
ICBC
Initial condition and boundary condition
ICTP
International Center for Theoretical Physics
IMD
India Meteorological Department
ISM
Indian summer monsoon
ISMR
Indian summer monsoon rainfall
LLJ
Low-level jet
MM5
Mesoscale model version 5
MPIOM
Max Planck Institute Ocean Model
MSLP
Mean sea level pressure
NCAR
National Center for Atmospheric Research
NCDC
National Climate Data Center
NOAA
National Oceanic and Atmospheric
Administration
OI_WK
OISST in weekly pattern
OISST
Optimum Interpolation Sea Surface Temperature
OLR
Outgoing Longwave Radiation
Q-Q
Quantile–quantile
RCM
Regional Climate Model
RegCM
RCM by ICTP
SD
Standard deviation
SST
Sea Surface Temperature
1 Introduction
One of the most important features of global atmospheric
circulation is the Indian summer monsoon (ISM)
(Webster et al. 1998). Any alteration in this circulation
causes a direct impact on the developing country like
India, where 60% of agriculture depends only on the seasonal monsoonal rainfall (Central Statistical Organization
1998). For understanding this, many studies have been
conducted in the past and well documented in the
technical literature domain by using the state of the art
Regional Climate Model (RCM) (Bhatla et al. 2016;
Singh et al. 2017) and Global Climate Model (GCM)
(ParthSarthi et al. 2016, 2015).
The RCM and GCM are the two widely used tools, which
are generally used to simulate climate circulation processes
and understanding the system. In recent decade, a number of
studies have been conducted to simulate the ISM variability
by using the RegCM (Bhatla et al. 2016; Maharana and Dimri
2016; Dash et al. 2015; Raju et al. 2015; Sinha et al. 2013;
Maurya et al. 2017; Dobler and Ahrens 2010). However, the
use of high-resolution RegCM in simulation of intraseasonal
monsoon variability and its different epochs is still limited in
the literature. In the last decade, several studies have been
conducted with downscaled global reanalysis datasets from
RegCM over particular boundaries (Bhatla et al. 2016;
Dobler and Ahrens 2010; Saeed et al. 2009; Dash et al.
2015; VenkataRatnam and Kumar 2005; Park and Hong
2004). In the sensitivity analysis by Dash et al. (2006) and
Bhatla and Ghosh (2015), the authors have found that the
Grell CPS is performing better than the other CPSs in simulations of rainfall distribution by RegCM. Bhatla et al. (2016)
have shown the performance of CPSs in the RegCM-4.3 over
Indian land-sea continental margin and found that the Tiedtke
CPS is highly efficient when compared with the in situ
observations. Dobler and Ahrens (2010) have considered the
RegCM simulations with the Consortium for Small-Scale
Modeling (COSMO) CLM RegCM over South Asia by using
the input data from ECHAM5–MPIOM and a 45-year re-analysis of European Centre for Medium-Range Weather
Forecasts (ECMWF). The study showed that the mean variability of monsoon indices improves when simulated using
the COSMO RegCM and compared against the driving field
of ECHAM5–MPIOM. With the use of GCM downscaled
output at regional level (Almazroui 2016, 2012; Saeed et al.
2009), Tugba et al. (2016) simulated the seasonal rainfall with
the RegCM-4.3.5 over Central Asia. All the above studies
showed a better performance of the RCM by capturing the
regional phenomena, although it varies depending on the seasons, parameterization schemes along with some biases, and
the role of initial condition. Singh et al. (2017) suggested a
proper evaluation of RCM model output and its utilization in
further climate studies by highlighting the poorer simulation
of GCM downscaled output than the respective GCM simulations. According to them, most of RCMs have not added any
significant output for the past, present, and future time scales
in simulating Indian summer monsoon rainfall (ISMR) behavior as compared to the GCMs. Most of the studies have raised
questions in RCM’s performance by considering various initial condition and boundary condition (ICBC) for climate
modeling study (Mishra et al. 2014; Singh et al. 2017; Dash
et al. 2015) rather than cross verifying the host GCM simulations and finding out the authenticity of the RCM’s ICBC.
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
In purview of the above, this study aims to improve the
simulation of the dynamical mechanism of ISM by using the
time-dependent ECMWF ICBC, through the RegCM-4.3
model downscaling along with the various sea surface temperature (SST) specified over the ocean. Further, through this
study, an attempt has been made to understand the dependencies of model simulation with ERA-Interim (EIN15) ICBC
and deducing its impact on the model simulated rainfall over
a particular region during the phases of ISM rainfall.
2 Data and methodology
2.1 RegCM-4.3 outline
The conceptual innovation of the RegCM was originally developed in the late 1980s from the National Center for
Atmospheric Research (NCAR), USA. With the core of
MM4 (Grell et al. 1994), Giorgi and Bates (1989) and
Dickinson et al. (1989) introduced the first version of
RegCM model in 1989. After several upgrades in model physics, RegCM-3 was introduced in which model grid spacing
was extended between 10 and 100 km with simultaneous
ranging from seasonal to centennial periods covering all land
regions over the world (except Polar region). This version of
RegCM was more portable and developed with an aim to
simulate the tropical climates (Giorgi and Anyah 2012). The
RegCM-4 is then upgraded with integration of new land surface physics, planetary boundary layer conditions and air–sea
flux scheme. With the modification in pre-existing radiative
transfer and boundary layer schemes, the mixed convection
scheme was developed and the model code was upgraded to
improve its flexibility and applicability (Giorgi et al. 2012).
The RegCM-4.3 is a hydrostatic model with sigma-p vertical
coordinate system and having capability to run over a large
range of Regional Climate Modeling system (Giorgi et al.
2012). It is the first limited area model with the mesoscale
model version 5 (MM5) dynamical core, developed by the
International Center for Theoretical Physics (ICTP) for a
long-range climate simulation running on Arakawa B-grid. In
this version, two types of land-use (BATS) have been added for
the better representation of the urban and suburban environments. For urban development, the surface albedo has been
modified and the surface energy balance is modified for alter.
The Climate Local Model (CLM) version 3.5 is coupled with
the RegCM-4.3, which deals with the bio-geophysical-based
parameterization and to describe the exchanges of energy, momentum, water, and carbon (Tawfik and Steiner 2011) along
with the physical process given by Holtslag et al. (1990) for the
planetary boundary layer (PBL). The RegCM-4.3 has four core
CPSs: Grell (Grell 1993), Emanuel (Emanuel 1991; Emanuel
and Živković-Rothman 1999), Tiedtke, and Kuo. The
Arakawa–Schubert or the Fritsch and Chappell (1980a,
1980b) type closures are available to use with the Grell CPS.
It has the capacity to perform combinations of different CPSs
over the land and ocean. The RegCM-4.3 has been forced to
simulate the mixed convection scheme mode: Mix98 (Grell
over the ocean, Emanuel over the land) and Mix99 (Grell over
the land, Emanuel over the ocean). The lower boundary conditions are derived from six hourly forcing of EIN15 re-analysis.
These ICBCs are obtained with the 1.5o horizontal grids and 37
vertical levels, which incorporate the CCSM3 radiation scheme
of NCAR Community (Kiehl et al. 1996; Collins et al. 2006).
The CPSs are forced with the Optimum Interpolation Sea
Surface Temperature (OISST) in weekly pattern (OI_WK)
and Extended Reconstructed Sea Surface Temperature
(ERSST) wherever applicable. The dataset of OI_WK at
weekly and ERSST at six hourly timescales are obtained from
the National Oceanic and Atmospheric Administration
(NOAA) and National Climate Data Center (NCDC), respectively. The details of the RegCM-4.3 configuration are listed
in Table 1.
2.2 Experiment design
The complicated characteristic of South Asian summer monsoon rainfall faces considerable difficulty in rainfall simulation over the region (Webster et al. 1998). Dealing with the
complex topography and coastline terrain, the RegCM is
equipped with advance numerical and physical schemes
(Giorgi et al. 2001). With the flexibility in choosing the appropriate CPS over a particular domain, this study is continued with Mix99 CPS. Several studies have showed the better
performance of Grell CPS over the land region (Bhatla and
Ghosh 2015; Elguindi et al. 2013) and produced good simulation of Emanuel CPS over the ocean. The combination of
core CPSs (Grell and Emanuel) with their better side is termed
as Mix99 scheme, where Grell CPS is used over the Land and
Emanuel over the ocean. The variation in mean seasonal rainfall has a strong association with the global phenomena
through the influence of SST (Krishnamurthy and Kinter
2003). Therefore, for sensitivity analysis, the conditional
model outputs (CMOs) are continued with two types SST
(ERSST and OI_WK) as lower boundary conditions. For better representation in response to climate dynamics, associations with atmospheric convection and topographical complexity have been conducted at 0.5o × 0.5o fine resolution
through the model simulation. For this synoptic study, the
model simulated daily rainfall, Outgoing Longwave
Radiation (OLR), mean sea level pressure (MSLP) and wind
at 925 hPa and 850 hPa vertical pressure intervals over the
South Asia CORDEX region (22oS-50oN and 10°E-130°E).
The actual onset dates (Table 2) are considered as per the India
Meteorological Department (IMD) guidelines, while the
active/break spells are obtained from the NCC Research
Report (2013) provided by the IMD (Table 3). 30 years
S. Ghosh et al.
Table 1 Model configuration of
the RegCM4.3
Dynamics
Hydrostatics
Model domain
South Asia CORDEX domain
Domain cartographic projection
Resolution
Vertical level
(15oS–45oN; 10°E–130oE)
ROTMER––rotated Mercator
50 km horizontal
18 sigma vertical levels
Initial and boundary conditions
ERA15
SST
1. OI_WK – OISST weekly optimal interpolation dataset
Land surface parameterization Radiation
Parameterization PBL
2. ERSST – ERA-Interim 6 hourly 1.5°×1.5° SST
Modified CCM3
Modified Holtslag
Convective parameterization
Mix99 (Emanuel over the ocean and Grell over the land)
Grell closure scheme
Arakawa and Schubert (1974)
(1981–2010) climatological composites of the considered parameters are analyzed to find out the predictive skills.
Maharana and Dimri (2016) have illustrated the simulation
process and the physics of the RegCM, which plays a dominant role in defining the intraseasonal variability. But the model simulation is not well correlated with the in situ observations obtained during the intraseasonal/interannual rainfall
simulations, and the correlation coefficient is found close to
zero (Maharana and Dimri 2014, 2016). Therefore, the correlation coefficient has not been considered as a verification tool
for different CMOs with the in situ observations. In this study,
the performance verification between in situ and the model
derived outputs is verified using quantile–quantile (Q-Q) distribution, empirical cumulative distribution function (ECDF),
SD, and absolute bias (bias) for different phases of summer
monsoon. The bias can be measured as follows:
h i
Bias ¼ y−x
where the optimal value is 0 and the low magnitude value
indicates high accuracy in model simulation. x represents
the mean of the observed/reanalyzed data and y is the mean
of the model output.
3 Results and discussion
3.1 Variability of summer monsoon with different
lower boundary conditions
3.1.1 Onset phase
In order to simulate the summer monsoon rainfall variability,
the ICBC ERSST and OI_WK SST forcing are considered.
Spatial distributions of CMOs and in situ datasets for different
phases of summer monsoon are provided in Figs. 1 to 3. Due
Table 2 Actual and model simulation onset dates along with their
deviation from actual onset of ISM with different initial condition
Onset dates
Year
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Actual
30 M
28 M
13 J
30 M
28 M
04 J
02 J
26 M
03 J
19 M
02 J
05 J
28 M
28 M
05 J
03 J
09 J
02 J
25 M
01 J
23 M
29 M
08 J
18 M
05 J
26 M
28 M
31 M
23 M
31 M
Where M May and J June
OI_WK
–
04 J (+ 7)
21 J (+ 8)
31 M (+ 1)
29 M (+ 1)
25 M (− 10)
08 J (+ 6)
23 M (− 3)
25 M (− 9)
19 M (0)
07 J (+ 5)
06 J (+ 1)
30 M (+ 2)
06 J (+ 9)
11 J (+ 6)
24 M (− 10)
–
30 M (− 3)
02 J (+ 8)
25 M (− 7)
26 M (+ 3)
17 M (− 12)
06 J (− 2)
21 M (+ 3)
07 J (+ 2)
25 M (− 1)
–
31 M (0)
–
19 M (− 12)
ERSST
24 M (− 6)
02 J (+ 5)
21 J (+ 8)
31 M (+ 1)
06 J (+ 9)
24 M (− 11)
07 J (+ 5)
27 M (+ 1)
29 M (− 5)
18 M (− 1)
02 J (0)
25 M (− 11)
31 M (+ 3)
04 J (+ 7)
11 J (+ 6)
03 J (0)
29 M (− 11)
30 M (− 3)
21 M (− 4)
–
22 M (− 1)
17 M (− 12)
–
12 M (− 6)
09 J (+ 4)
28 M (+ 2)
01 J (+ 4)
26 M (− 5)
–
21 M (− 10)
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
Table 3
Active/break periods and days of summer monsoon during 1981–2010
Year
Active spells
Break spells
No. of active/break spells No. of active/break days
1981
1982
7–10 J
24–27 A
Act = 1, brk = 1
Act = 5, brk = 4
21–23 A, 12–14 A
1–8 J
Act = 2, brk = 1
Act = 6, brk = 8
1983
1984
1985
18–20 A
1–3 A, 9–11 A, 16–19 A
15–17 J, 30 J–2 A, 6–8 A
14–16 J
28–30 J
23–25 A
Act = 1, brk = 1
Act = 3, brk = 1
Act = 3, brk = 1
Act = 3, brk = 3
Act = 10, brk = 3
Act = 10, brk = 3
1986
21–24 J, 13–15 A
23–31 A
Act = 2, brk = 1
Act = 7, brk = 9
1987
1988
24–29 A
26–28 J
23–25 J, 30 J–4 A, 9–13 A
14–17 A
Act = 1, brk = 3
Act = 1, brk = 1
Act = 6, brk = 14
Act = 3, brk = 4
1989
1990
–
2–4 J, 21–24 A, 29–31 A
18–20 J, 30 J–3 A
–
Act = 0, brk = 2
Act = 3, brk = 0
Act = 0, brk = 8
Act = 10, brk = 0
1991
1992
21–24 J, 29–31 J, 22–24 A
27–29 J, 16–21 A
1-3 J
4–10 J
Act = 3, brk = 1
Act = 2, brk = 1
Act = 10, brk = 3
Act = 9, brk = 7
1993
7–11 J, 15–17 J
2–4 J, 9–16 J, 30 J–2 A, 18–20 A, 25–27
A, 29–31 A
18–20 J, 22–25 J
24–28 J, 19–21 A
30 J–1 A, 21–25 A
30 J–1 A, 21–25 A
7–9 A
12–14 J, 17–20 J
9–12 J
23–25 A
23–28 J, 27–29 A
8–13 A, 21–23 A
1-5 J, 25–29 J, 31 J–2 A
3–6 J, 27 J–7 A, 13–15 A, 17–20 A
1–9 J, 6–9 A
10–12 A
20–23 J, 8–13 A, 22–28 A
–
Act = 2, brk = 3
Act = 6, brk = 0
Act = 8, brk = 17
Act = 24, brk = 0
3–7 J, 11–17 A
11–13 A
11–15 J, 9–17 A
21–26 J, 16–21 A
1–5 J, 12–18 A, 22–24 A
29 J–8 A
31 J–2 A, 27–29 A
4–16 J, 22–31 J
–
10–13 J, 19–22 J, 26–31 A
7–14 A, 24–31 A
–
18–23 J, 15–17 A
16–21 J, 21–24 A, 28–30 A
Act = 2, brk = 2
Act = 2, brk = 1
Act = 2, brk = 2
Act = 2, brk = 2
Act = 1, brk = 3
Act = 2, brk = 1
Act = 1, brk = 2
Act = 1, brk = 2
Act = 2, brk = 0
Act = 2, brk = 3
Act = 3, brk = 2
Act = 4, brk = 0
Act = 2, brk = 2
Act = 1, brk = 3
Act = 7, brk = 12
Act = 8, brk = 3
Act = 8, brk = 14
Act = 8, brk = 12
Act = 3, brk = 15
Act = 7, brk = 11
Act = 4, brk = 6
Act = 3,brk = 23
Act = 9, brk = 0
Act = 9, brk = 14
Act = 13, brk = 16
Act = 23, brk = 0
Act = 13, brk = 9
Act = 3, brk = 13
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
13–16 J, 20–23 J
2010
–
Total (1981–2010) No. of spells: Act = 59, brk = 44
29 J–9 A, 17–19 A
Act = 2, brk = 2
–
Act = 0, brk = 0
No. of days: Act = 237, brk = 246
Act = 8, brk = 15
Act = 0, brk = 0
J July, A August, act active, brk break
to unavailability of regular time period of SST, the data length
for in situ and CMO ERSST are considered only for the period
1981–2010, while for CMO OI_WK runs are provided for the
time period 1982–2010. The rainfall and OLR distribution
during the onset phase of summer monsoon rainfall are represented through Fig. 1a–f. For simulating the onset phase, IMD
onset simulation criteria (Pai and Rajeevan 2007) are considered. The onset criterion is emphasized by using the peak of
rainfall over Kerala with four (4) newly developed conditions,
which is dependent upon the three climatic parameters. Sixty
percent of rainfall stations is found in the Kerala (a total of 14
stations namely Minicoy, Amini, Thiruvananthapuram,
Punalur, Kollam, Allapuzha, Kottayam, Kochi, Trissur,
Kozhikode, Talassery, Cannur, Kasargode, and Mangalore)
with a minimum rainfall of 2.5 mm/day for the two
consecutive days (after 10th of May) which is considered as
the onset of monsoon. For declaration of onset, another three
criteria also have to be satisfied along with the station wise
rainfall criteria. The depth of westerly will have to be maintained at 600 hPa over 0oN–10oN and 55°E–80°E regions.
Zonal wind will have to be blown at a speed of 15–20 knots
at 925 hPa over the Lat 5oN–10oN and Lon 70°E–80°E, and
INSAT OLR should be less than 200 Wm−2 over Lat 5oN–
10oN and Lon 70°E–75°E.
High-resolution (0.25 × 0.25o) IMD rainfall data is considered for spatial and temporal rainfall analysis which is distributed only over the Indian subcontinent. Two stations (Minicoy
and Amini) among the 14 stations do not have the rainfall
distributed area in IMD rainfall data. Therefore, only 12 stations are considered to carry out this present study. Table 2
S. Ghosh et al.
Fig. 1 a–f Spatial pattern of in
situ and CMO-simulated rainfall
and OLR distribution during
1981–2010 by following IMD
criteria for declaring monsoon
onset
illustrates the actual onset along with CMO onset simulation
followed by IMD criteria. Some years are left blank that do not
satisfy the said criteria. IMD mean rainfall distribution
(Fig. 1a) during the two consecutive days (previous day of
actual onset and the actual onset) showed a densely rainfall
distribution over the adjoining part of the Kerala and Western
Ghat region and represents 2.5 mm/day of rainfall criteria over
the Kerala region. The ERSST CMO (Fig. 1b) and OI_WK
Fig. 2 a–d Distribution of in situ
and CMOs simulation of OLR
over the region Lat 5°N–10°N
and Lon 70°E–75°E and zonal
wind at 925 hPa over the region
Lat 5°N–10°N and Lon 70°E–
80°E for the respective onset date
CMO (Fig. 1c) are following the pattern of IMD rainfall distribution during two consecutive days of model-simulated onset. Because of unavailability of the long-range climatological
INSAT OLR, in this study, the CMO OLR distribution is
cross-verified by using the NOAA in situ data instead of
INSAT OLR. Figure 1d represents the NOAA in situ OLR
distribution where CMO ERSST and CMO OI_WK are considered and shown in Fig. 1e, f, respectively. Distributions of
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
Fig. 3 a–i Composite spatial
pattern of in situ and CMOsimulated rainfall, OLR, and
MSLP distribution during active
phases for duration 1981–2010.
EIN15 and CMO-derived wind
circulation is superimposed at
850 hPa for respective cases
CMO OLR pattern are in close match with the in situ OLR
distribution and correctly following the in situ OLR distribution over Lat 5oN–10oN and Lon 70°E–75°E and satisfying
the OLR distribution of less than 200 Wm−2 criteria. The
temporal distributions of OLR over the region are also considered in Fig. 2a, which represent warm bias for some cases
of NOAA OLR on actual onset date. But declaring onset over
Indian region, NOAA as well as CMOs should maintain the
limit of less than 200 Wm−2 over the respective region (IMD
criteria) on the onset day. In this figure, the dotted line is
indicating the limit band at 200 Wm−2. But with a large variation, in situ OLR distribution (Fig. 2b) is found in between ~
160–190 Wm−2, while the 25th and 75th percentiles and its
median are found towards the lower head of the inter quartile
range box. Therefore, the spatial pattern of in situ OLR
(Fig. 1d) represents the best fit with the IMD onset criteria.
CMO ERSST and OI_WK are showing some deviations in the
simulated onset dates, as shown in the box plots (Fig. 2b).
Zonal wind speed at 925 hPa over Lat 5oN–10oN and Lon
70°E–80°E is considered in Fig. 2c in which red dotted line
represents the upper limit band (20 knots) and blue dotted line
indicates the lower limit band (15 knots). In Fig. 2c, EIN15 in
situ zonal wind speed is showing large variations in the actual
onset date during some of the years. More than 36% in situ
zonal wind dataset is not satisfying the IMD onset zonal wind
speed criteria. Therefore, over a specified region, a large variation is being observed in EIN15 dataset from its normal
wind speed (Srivastava et al. 2013) on the actual onset date.
The simulations using CMO ERSST and OI_WK (Fig. 2c, d)
are kept within the limit band during the analysis, which
causes a mean deviation of ± 5 day from the actual onset for
a 30-year time period (from Table 2).
S. Ghosh et al.
A relation between OLR and zonal wind speed is
depicted in Fig. 2, which represents a negative correlation
between wind speed and radiation (Bett and Thornton
2016) where the OLR value (Fig. 2a) is reduced with
the increase in zonal wind speed (Fig. 2c). This feature
is also well simulated in the CMO ERSST as well as
CMO OI_WK simulation. The depth of westerly of in
situ, CMO ERSST, and CMO OI_WK are maintained at
600 hPa over 0oN–10oN and 55°E–80°E region during the
last few days of actual onset (figures not shown). Using
the said criteria and by considering the above analysis, we
have tried to simulate the onset date for Indian region for
a 30-year time period (Table 2), which shows a mean
deviation of ± 5 day and a maximum of ± 12 day deviation in CMOs (ERSST and OI_WK) from the actual date
of onset. With a robust skill in capturing the intraseasonal
monsoon variability, the RegCM has the limitation to simulate the dates in correspondence to in situ observations
(Maharana and Dimri 2014, 2016). With the MM5 dynamical core, the RegCM-4.3 has performed moderately
in simulation of onset dates and it might be due to the
large variation in ICBC over the particular region.
Therefore, in this study, we particularly have focused on
the physical processes of the data distribution instead to
find out the physics behind the uncertainty of the RegCM4.3 model output.
3.1.2 Active/break phase
To evaluate the model performance in intraseasonal time
scale and to analyze the RegCM-4.3 sensitivity with EIN15
Fig. 4 a–d Distribution of rainfall
and OLR during active spells of
1981–2010
downscaled model output with two types of SST, this study
has been carried out for another two major phases (active
and break) of summer monsoon (Krishnamurti 1985;
Krishnamurti and Ardunay 1980). The core region is considered between 71°E–83°E and 21oN–28oN which is closer to the monsoon region of Rajeevan et al. (2010). Active
and break periods are considered from the NCC Research
Report (2013) as given in (Table 3), which has been identified by standardized rainfall anomaly of greater than + 1
and less than − 1 for minimum three consecutive days respectively. The above criteria were obtained by averaging
the daily rainfall over the core monsoon region and by
standardizing the daily rainfall data by subtracting from
its long-term normal and dividing by its daily SD
(Rajeevan et al. 2010). A total of 59 active spells, 44 break
spells along with 237 active days, and 246 break days are
considered for 30 years during the period 1981–2010
(Table 3). The simulation of intraseasonal monsoon variability during active phases is considered in Figs. 3 and 4.
MSLP has an important role for active and break phases
(Krishnamurti and Bhalme 1976) and hence considered for
further study. The composite analysis of active days for
synoptic patterns is spatially distributed in Fig. 3a–i, in
which wind circulation at 850 hPa is superimposed on
rainfall (Fig. 3a–c), OLR (Fig. 3d–f), and MSLP
(Fig. 3 g–i) distributions. Strong monsoon rainfall distribution (Fig. 3a–c) along with the low-level jet (LLJ) over the
core region during the active phase is depicted with the
CMO ERSST (Fig. 3b) and CMO OI_WK (Fig. 3c), which
are closely following the IMD spatial rainfall distribution
(Fig. 3a) and temporal pattern (Fig. 4a) during the phases.
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
Fig. 5 a–i Composite spatial
pattern of in situ and CMOsimulated rainfall, OLR, and
MSLP distribution during break
phases for duration 1981–2010.
EIN15 and CMO-derived wind
circulation is superimposed at
850 hPa for respective cases
Temporal pattern of CMO rainfall distribution (Fig. 4a–b)
is showing a dry bias during this phase. The IMD rainfall
(Fig. 4a; blue line) and its distribution (Fig. 4b) are
spreaded between 12 to 19 mm of the rainfall amount
where CMO ERSST (Fig. 4a; red line) and OI_WK
(Fig. 4a; green line) are showing under estimation with 4
to 7 mm (Fig. 4b) of rainfall during the active days. During
summer monsoon period, convection plays an important
role in forming the cloud, which is inversely correlated
with the OLR. Positive/negative anomalies in OLR distribution are the causes of negative/positive rainfall anomaly
(Raju et al. 2009). The shrinking/rising distribution of in
situ OLR (Fig. 3d) over core region is showing less than
200 Wm−2 during the active phase, which is poorly simulated by CMO ERSST (Fig. 3e) and CMO OI_WK
(Fig. 3f) OLR distribution. The temporal pattern of CMO
OLR (Fig. 4c) is showing warm bias, and its distributions
(Fig. 4d) are varying up to 260 Wm−2, whereas NOAA
represents 200 Wm−2 over the region. ISMR and OLR
relationship has been already established in previous
study by Raju et al. (2009) and indicated an opposite relationship between ISMR and OLR. In the current study, the
CMO OLR distribution are showing warm biases and at the
same time, the model-simulated rainfall distribution is
representing dry biases. Simultaneously, a pressure belt
(Fig. 3) is showing an average pressure of 995–1005 hPa
over the Monsoon Convergence Zone (MCZ), and the
monsoon trough covers the Gangetic region as a proxy of
rainfall distribution and spreads up to foot of the Bay of
Bengal (BoB) through the Gangetic plain. Similar results
are also provided in the IMD Monsoon Report 2013, which
is in agreement with the model-simulated MSLP (Fig. 3 h–
S. Ghosh et al.
Fig. 6 a–d Distribution of rainfall
and OLR during break spells of
1981–2010
Fig. 7 a–d Vertical level-days (a,
b) and days-longitude at 850 hPa
(c, d) of EIN15 zonal wind
distribution (Knots/s; shaded)
during active and break phase
over the core region (71oE-83oE
and 21oN-28oN)
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
Fig. 8 Distribution of EIN15 zonal wind during active and break phases
of Intraseasonal summer monsoon period
i) as compared to NCEP (Fig. 3 g). In in situ MSLP distribution (Fig. 3 g), a low pressure belt of 995–1000 hPa is
formed over the MCR and this gradient is not well simulated by the CMOs (Fig. 3 h–i). Due to topographical terrain over the Himalayan region, a deep low-pressure (<
990 hPa) belt is formed in the CMO MSLP (Fig. 3 h–i)
and the model is failed to simulate the MSLP over the
respective zone. As a proxy parameter in the RegCM rainfall simulation, MSLP does not have a big role in CMO
over the MCR. Therefore, non-simulating OLR distribution could be a possible cause behind the agitated rainfall
simulation during the active phase.
Composite distribution of synoptic parameters during
break phases is spatially distributed in Fig. 5a–i. These
figures illustrate a weak monsoon rainfall over the core
region, which is depicted well in IMD rainfall distribution
(Fig. 5a). CMO ERSST and OI_WK are able to simulate
this distribution over the core region (Fig. 5b, c). The temporal rainfall distributions during break phase are considered in Fig. 6a, which indicates that the CMOs are following the rainfall distribution pattern as IMD. Although, the
rainfall deviation from the mean is slide high for CMO
ERSST and OI_WK rainfall as compared to IMD rainfall
(Fig. 6b). The less OLR distributed area is shifted towards
the eastern part from the core region of India (Fig. 5d–f)
Table 4
3.2 Dependencies of RegCM-4.3 output on lateral
boundary conditions
The above study shows that the sensitivity of
model-simulated rainfall, OLR, MSLP, and zonal wind distributions over Indian subcontinent is showing an over or
under estimation in simulating different phases of ISMR.
Lateral boundary conditions of u and v wind components,
surface pressure, temperature, and water vapor along with
SSTs specified over the ocean are the regulators and are
used to run the RegCM-4.3 over the specified region. The
study of Bett and Thornton (2016) has shown the relationship of these parameters and responses to rainfall simulation. Zonal wind disturbances have been identified as the
main mechanism in organizing rainfall pattern over a particular region (Diedhiou et al. 1999). Therefore, to find out
possible causes behind the uncertain behavior of
Statistical scores of the RegCM-4.3 simulation with in situ during the phases of monsoon
Onset phase
In situ
ERSST
OI_WK
and found more than 220 Wm−2 over the Indian core region during this phase (Raju et al. 2009). Temporal distributions of the CMO analyses are also capturing the peak as
NOAA OLR (Fig. 6c) with a slight over estimation in respect to in situ (Fig. 6d). At the same time, the monsoon
trough in Fig. 5 g–i has been shifted towards the
Himalayan foothills from the Gangetic plain and has extinct from BoB foot (Bhatla and Ghosh 2015). During the
break phase, the pressure belt (995–1005 hPa) generally
remains active and spreads over the Gangetic plain during
the active period and shifted to the Himalayan foothill in in
situ (Fig. 5 g) as well as CMOs (Fig. 5 h–i) and monsoon,
while become weaker over the core region. These are the
prominent feature during the break phases of ISMR and
well simulated by different CMOs of Mix99 CPSs. The
spatial gradient is also well simulated by CMOs over all
the remaining part including BoB, Arabian Sea, and Indian
Ocean (except the Himalayan foothill region) during
active/break phases. Another major dependency of model
performance is the ICBC. Therefore, to find out the possible causes behind disturbed rainfall and non-simulating the
OLR distribution, the role of the RegCM’s ICBC for longterm simulation is carried out in further section.
OLR
SD
29.24
12.26
11.68
BIAS
− 2.77
− 3.2
Active phase
U-wind
SD
4.46
1.32
1.87
BIAS
− 1.54
− 1.08
Rainfall
SD
BIAS
5.06
2.84
− 9.8
2.71
− 9.5
Break phase
OLR
SD
20.02
8.56
9.22
BIAS
55.68
55.82
Rainfall
SD
BIAS
1.65
2.25
− 1.60
2.76
− 2.04
OLR
SD
21.50
13.16
13.18
BIAS
− 22.28
− 21.8
S. Ghosh et al.
Mix99 CPS, we focused on the ICBC (specially zonal
wind) provided for model run. Figure 7a–d represents
Hovmöller diagram of vertical level wind structure-days
Fig. 9 a–f Q-Q distribution of
different synoptic parameters
during onset, active, and break
phase in respect to in situ dataset
with 5% significant level
(Fig. 7a–b) and days-longitude-wise distribution (Fig. 7c–
d) of EIN15 zonal wind at 850 hPa during active and break
phases over the Indian core region. The shaded area
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
represents the zonal wind speed in knots/s. This distribution illustrates the low efficiency of zonal wind distribution with the color shaded of 0–15 knots/s during active
phase (Fig. 7a) and 10–20 knots/s on vertical level 900–
800 hPa during break phase (Fig. 7b). Generally, LLJ
usually blows over the Indian subcontinent with a higher
speed during the active condition than the weak and normal monsoon conditions (Ruchith et al. 2014). But
EIN15 zonal speed circulation speed never achieved a
value over 25 knots/s at the lower level. At 850 hPa,
most of the active days attained wind speed of 0–10
knots/s and very few cases are there that achieved more
than 15 knots/s during the active days (Fig. 7c). During
the break conditions, the wind speed has shown an average of 10–25 knots/s (Fig. 7d) over the region. The
study by Ruchith et al. (2014) and Varikoden (2006)
showed that during active phase, the zonal wind blows
with a higher speed over the core region rather than the
normal and weak monsoon condition. Zonal wind blows
from the Arabian Sea with the core speed of ~ 34 knots/s
and crosses through the central India (core region) with
average speed of ~ 28 knots/s at 850 hPa during active
phase. During break phase, the direction of LLJ turns
towards south and becomes weak and blows with a
speed of less than ~ 15 knots/s (Varikoden 2006). In
Fig. 8, box plot has been considered for clear representation of in situ zonal wind distribution which represents
the EIN15 zonal wind distribution during active and
break days with 25 and 75 percentile over the region.
With a mean of 4 knots/s, in situ EIN15 zonal wind is
blowing over the core region during active phase and in
break phase, the mean speed is elevated up to 14 knots/s.
Therefore, a sharp conclusion of the above figure might
be drawn that the in situ zonal wind speed is showing a
large under estimation in speed during active phase,
whereas a higher speed is expected during the active
phase. The dynamics of zonal wind circulation pattern
during active and break phases of summer monsoon
showed a large bias (~ 24 knots/s) in in situ zonal wind
speed at 850 hPa during the active phase. During the
break phase, the EIN15 zonal wind speed follows its
regular condition (speed of less than ~ 15 knots/s) and
blows with the speed ~ 14 knots/s.
Rainfall is the most chaotic factor depends upon a
combination of five environmental factors, and zonal
wind is one of them. The model-simulated rainfall distribution purely depends on the ICBC in situ components.
Therefore, a large bias in in situ distribution will have a
direct impact in the model-simulated rainfall distribution.
In this study, the lack/excess of zonal wind circulation
speed during the ISM season is carried forward during
model simulation process. Due to the negative correlation
between the zonal wind speed and OLR distribution, the
in situ zonal wind distribution forces the CMO OLR simulation. At the same time, the disturbance in OLR distribution affects the excess/lack of rainfall distribution over
the rain belt area (Diedhiou et al. 1999) in the RegCM-4.3
simulation during the intraseasonal monsoon season
(Sylla et al. 2010, 2011).
3.3 Statistical score and model verification
3.3.1 Statistical scores and biases
Table 4 represents the statistical scores of modelsimulated output and in situ data during different phases
of monsoon. Statistical representation of CMOs have
showed less than half of SD in ERSST, OI_WK CMO
OLR, and U-wind distribution with respect to the in situ
during onset phase and the bias over the respective region, which represents dry for the parameters of OLR
and zonal wind. Previous analyses (Section 4.1) showed
a high fluctuation in in situ over the respective region,
and the mean for a 30-year period has induced the analyses to disfavor the CMO performance. In this study,
onset simulations by CMOs are considered following
the IMD criteria; therefore, the biases are unexpected
in model simulation, which lay the overestimation in in
situ datasets. During active/break phases, the SD in
CMO rainfall and OLR is depicted roughly a half of
in situ, while biases represent the opposite relationship
among rainfall and OLR distribution with high positive
biases in OLR as stated in the above section. The bias
in OLR distribution over the core region is higher during active period than break period, which attains about
double in respect to break phase. Analyses from
Section 4.1 and Section 4.2 showed the mechanism of
ICBC in the RegCM and impact of wind circulation on
OLR as well as rainfall distribution, and a relationship
between zonal wind, OLR, and rainfall is also evident in
the observed datasets. Those sections particularly dealing
with the uncertainties in the RegCM rainfall simulation
due to disturbed ICBC over some specified region during different phases of ISM, where this section
supporting the analyses and highlighting the nature of
CMOs by the statistical score rather than verifying the
model with the in situ itself.
3.3.2 Model verification
Verification statistics of CMOs are considered using Q-Q distribution and ECDF for the considered parameters for the synoptic analysis during the phases of ISMR.
Q-Q plot is a probability plot to compare the shapes of
d i s t r ib u t i o n b e t w e e n t w o d a t a s e r i e s ( Wi l k a n d
Gnanadesikan 1968) or for its theoretical distribution itself
S. Ghosh et al.
(a) Onset: Q-Q Plot of EIN15 wind
30
(b) Onset: Detrended Q-Q Plot of EIN15-wind
3
2
Deviation from Normal
Expected Normal Value
25
20
15
1
0
-1
10
-2
5
-3
5
10
15
20
25
30
5
10
15
Observed Value
(c) Onset: Q-Q Plot of Mix99 wind (ERSST)
20
20
25
30
Observed Value
1.5
(d) Onset: Detrended Q-Q Plot of Mix99 wind (ERSST)
1.0
18
Deviation from Normal
Expected Normal Value
19
17
16
0.5
15
0.0
14
13
-0.5
13
14
15
16
17
18
19
20
14
15
16
(e) Onset: Q-Q Plot of Mix99 wind (OI_WK)
22
18
19
20
(f) Onset: Detrended Q-Q Plot of Mix99 wind (OI_WK)
1.5
20
1.0
Deviation from Normal
Expected Normal Value
17
Observed Value
Observed Value
18
16
14
0.5
0.0
-0.5
12
-1.0
12
14
16
18
Observed Value
20
22
14
15
16
17
18
19
20
21
Observed Value
Fig. 10 a–l Q-Q distribution of the individual datasets with 1% significant band along with the detrended Q-Q distribution of the respective dataset
considered for onset simulation
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
240
(g) Onset: Q-Q Plot of NOAA-olr
(h) Onset: Detrended Q-Q Plot of NOAA-olr
30
220
200
Deviation from Normal
Expected Normal Value
20
180
160
10
0
140
-10
120
100
100
150
200
-20
140
250
160
200
220
240
260
Observed Value
Observed Value
210
180
(i) Onset: Q-Q Plot of Mix99 olr (ERSST)
10
(j) Onset: Detrended Q-Q Plot of Mix99 olr (ERSST)
Deviation from Normal
Expected Normal Value
200
190
180
170
5
0
160
150
150
160
170
180
190
200
-5
160
210
170
210
180
190
200
210
Observed Value
Observed Value
(k) Onset: Q-Q Plot of Mix99 olr (OI_WK)
5.0
(l) Onset: Detrended Q-Q Plot of Mix99 olr (OI_WK)
200
Deviation from Normal
Expected Normal Value
2.5
190
180
170
0.0
-2.5
-5.0
160
150
150
160
170
180
Observed Value
Fig. 10 (continued)
190
200
210
-7.5
150
160
170
180
Observed Value
190
200
S. Ghosh et al.
(a) Act: Q-Q Plot of IMD rain
30
(b) Act: Detrended Q-Q Plot of IMD rain
4
25
Deviation from Normal
Expected Normal Value
3
20
15
10
2
1
0
5
0
-1
0
10
20
30
40
0
10
20
Observed Value
(c) Act: Q-Q Plot of Mix99 rain (ERSST)
15
2
Deviation from Normal
Expected Normal Value
40
(d) Act: Detrended Q-Q Plot of Mix99 rain (ERSST)
3
10
5
0
1
0
-5
-1
-5
0
5
10
15
0
2
4
Observed Value
15
30
Observed Value
6
8
10
12
14
Observed Value
(e) Act: Q-Q Plot of Mix99 rain (OI_WK)
(f) Act: Detrended Q-Q Plot of Mix99 rain (OI_WK)
2
Deviation from Normal
Expected Normal Value
10
5
1
0
0
-1
-5
0
5
Observed Value
10
15
0
5
10
15
20
Observed Value
Fig. 11 a–l Q-Q distribution of the individual datasets with 1% significant band along with the detrended Q-Q distribution of the respective dataset
considered for active phase simulation
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
(h) Act: Detrended Q-Q Plot of NOAA olr
(g) Act: Q-Q Plot of NOAA olr
15
260
240
Deviation from Normal
Expected Normal Value
10
220
200
180
5
0
160
140
140
160
180
200
220
240
-5
140
260
160
180
290
200
220
240
260
Observed Value
Observed Value
(j) Act: Detrended Q-Q Plot of Mix99 olr (ERSST)
(i) Act: Q-Q Plot of Mix99 olr (ERSST)
6
280
Deviation from Normal
Expected Normal Value
4
270
260
250
2
0
-2
-4
240
-6
230
220
230
240
250
260
270
280
220
290
230
240
Observed Value
280
10.0
270
260
250
240
270
280
290
7.5
5.0
2.5
0.0
240
250
260
270
Observed Value
Fig. 11 (continued)
260
(l) Act: Detrended Q-Q Plot of Mix99 olr (OI_WK)
12.5
Deviation from Normal
Expected Normal Value
(k) Act: Q-Q Plot of Mix99 olr (OI_WK)
290
230
230
250
Observed Value
280
290
300
-2.5
230
240
250
260
270
Observed Value
280
290
300
S. Ghosh et al.
(a) Brk: Q-Q Plot of IMD rain
8
(b) Brk: Detrended Q-Q Plot of IMD rain
3
6
Deviation from Normal
Expected Normal Value
2
4
2
0
1
0
-2
-4
-1
-4
-2
0
2
4
6
8
0
2
4
Observed Value
6
8
Observed Value
(c) Brk: Q-Q Plot of Mix99 rain (ERSST)
(d) Brk: Detrended Q-Q Plot of Mix99 rain (ERSST)
3
10.0
7.5
Deviation from Normal
Expected Normal Value
2
5.0
2.5
1
0.0
0
-2.5
-1
-5
0
5
10
15
0
2
4
Observed Value
6
8
10
12
Observed Value
(e) Brk: Q-Q Plot of Mix99 rain (OI_WK)
(f) Brk: Detrended Q-Q Plot of Mix99 rain (OI_WK)
4
15
3
Deviation from Normal
Expected Normal Value
10
5
2
1
0
0
-5
-1
-5
0
5
Observed Value
10
15
0
5
10
15
Observed Value
Fig. 12 a–l Q-Q distribution of the individual datasets with 1% significant band along with the detrended Q-Q distribution of the respective dataset
considered for break phase simulation
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
(g) Brk: Q-Q Plot of NOAA olr
(h) Brk: Detrended Q-Q Plot of NOAA olr
5
300
Deviation from Normal
Expected Normal Value
0
275
250
225
-5
-10
-15
-20
200
200
225
250
275
-25
180
300
200
220
240
260
280
300
Observed Value
Observed Value
(j) Brk: Detrended Q-Q Plot of Mix99 olr (ERSST)
(i) Brk: Q-Q Plot of Mix99 olr (ERSST)
15
320
300
Deviation from Normal
Expected Normal Value
10
280
260
5
0
240
220
220
240
260
280
300
-5
240
320
260
Observed Value
300
5.0
280
260
240
320
2.5
0.0
-2.5
240
260
280
Observed Value
Fig. 12 (continued)
300
(l) Brk: Detrended Q-Q Plot of Mix99 olr (OI_WK)
7.5
Deviation from Normal
Expected Normal Value
(k) Brk: Q-Q Plot of Mix99 olr (OI_WK)
320
220
220
280
Observed Value
300
320
-5.0
230
240
250
260
270
Observed Value
280
290
300
S. Ghosh et al.
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
Fig. 13
a–f Empirical cumulative distribution function (ECDF) of different parameter during onset, active and break phase of ISM
(Singh et al. 2014). This distribution verifies the probability
by plotting quantile graphic against each other or with its
theoretical distributions to understand the properties of
data. The graphical representation of Q-Q plot (Fig. 9a–
f) generally used to compare the CMO performance with
a nonparametric approach along with their distributions
with respect to in situ datasets at 5% level of significance.
The scatters in x- and y-axes correspond to one of the
quantile of second distribution (y-coordinate) and plotted
against same quantile of first distribution (x-coordinate).
If CMOs and in situ data distributions have similarity,
then the quantiles will be laid approximately on the
dashed line with 45o angle. Otherwise, if distributions
are linearly related, then the scatters will be laid approximately on a line. But it is not necessarily to be laid on the
line y = x. The red scatters indicate the distribution of
ERSST CMO, while blue is for OI_WK CMO. For
checking the goodness of fit of CMOs with in situ
quantiles, most of the datasets during onset are near to
linear line. A slight under estimation is depicted in
CMO’s zonal wind speed during onset phase with respect
to in situ quantiles (Fig. 9a). At the same time, an overestimation is also observed in OLR distribution among the
CMOs and in situ quantiles (Fig. 9b). The wind and OLR
distributions are showing a clear opposite relationship
during the ISM onset. Figure 9c–f represents their goodness of fit with in situ by showing their sequentially
congested distribution of data. Although, during active
phase, the model-simulated output suffers from lack of
rainfall distribution in comparison to the in situ and their
quantiles are showing under estimation with in situ observation (Fig. 9c). On other side, due to surplus in OLR
distribution with in situ, the quantiles (Fig. 9d) are showing an overestimation during active phase. During break
phases, the scatters (Fig. 9e–f) are very close to in situ
and the quantiles for rainfall distribution are showing
goodness for CMO (Fig. 9e). The model-simulated OLR
quantiles of different CMOs are found very close to the
linear line (Fig. 9f). It is also observed that the ERSST
CMO is showing better performance with in situ than the
CMO OI_WK. On the other hand, the CMO ERSST is
showing its quantiles a bit nearer to the in situ or it is
overlapping over CMO OI_WK.
For further verification of CMOs and in situ datasets with
their respective normal quantiles, Figs. 10 and 12 have been
considered for the respective parameters during the phases of
monsoon. In these figures, the normal Q-Q plots represent the
validation of the data with the distributional assumption of the
respective dataset by providing visual comparison of the
sample quantiles to the corresponding theoretical quantiles.
Detrended normal Q-Q plots have also been considered for
monitoring the deviation of the actual data from its normal.
All the figures represent their significance with 1% significant
band. The normal Q-Q plot for all the parameters of in situ as
well as CMOs are plotted in Fig. 10 for onset, Fig. 11 for
active phase, and Fig. 12 for break phase simulation. These
figures are showing the goodness of fit with their theoretical
distributions itself by illustrating their fitness over the linear
line with 1% significant band (Fig. 10a, c, e, g, i, and k). The
detrended normal Q-Q plots show a goodness of fit for every
monsoon phase by representing the crowds of scatters from
the normal. It has been also observed that the deviation of
CMO data from its normal is much less than in situ
(Fig. 10b: IMD rain, Fig. 10h: NOAA OLR) distribution either in the case of rainfall (Fig. 10d, f) or OLR (Fig. 10j, l).
With a slight deviation from the normal, the quantiles are
gathered together over the line during active and break phases
and their detrended normal are also showing a high significance with less deviation (Fig. 11a–l and Fig. 12a–l). ECDF
(Fig. 13a–f) is another nonparametric statistical estimator
tools to interpret similar groups of random datasets by their
own distributions (Gibbons and Chakraborti 2003, 1992). In
this ECDF estimator, it is possible to visualize the data
series distribution by the slope of the line and with the
increase of ECDF from 0 to 1. Because of the limited
number of data series, the curve in Fig. 13a, b looks
rough rather than smooth. The ECDF fitted results of
three distributions (in situ, ERSST, and OI_WK) for every single parameter is presented in one box. These boxes
of different parameters and different phases showed that
the CMO ERSST and CMO OI_WK are very close to the
in situ distribution and following the pattern of in situ for
every ECDF key quantile. The overall statistical study
shows a clear distribution pattern for all the synoptic
parameters in CMO ERSST and OI_WK considered for
the analysis of onset, active, and break phase with respect
to the in situ. Their distribution patterns are showing
goodness of fit with in situ observation and their theoretical distributions itself. This section deals with the validation of the RegCM output with in situ during different
phases with statistical inference which performs well in
comparison to in situ. In other words, the RegCM follows the ICBC with some over/under estimation in the
model simulation.
4 Summary and conclusions
The study indicates the sensitivity of the RegCM simulation
for ISM with different lateral boundary conditions. In order to
evaluate the causative factors behind the uncertainties in the
S. Ghosh et al.
RegCM simulation, various factors are considered and subsequently analyzed. The conclusions of this study are as follows:
Sensitivity of RegCM’s Mix99 (Grell->Land &
Emanuel->Ocean) CPS has a minor dependency with the
ERSST and OI_WK SST lateral boundary conditions in simulating the intraseasonal variability of ISM.
The zonal wind of EIN15 reanalysis ICBC lacks enough
speed over the core Indian region and around Kerala during
definite rainy days (onset and active phases) of ISMR.
Model-simulated rainfall distribution entirely depends on
the distribution of in situ ICBC and the disturbed zonal wind
distribution in EIN15, which is being carried forward in the
RegCM simulation process.
Statistical distribution of all the parameters are in agreement with their respective theoretical quantiles and in situ
distribution, and CMOs are following the pattern of in situ
datasets with some under/over estimation, which illustrates
the RegCM capabilities towards ISM simulation.
The RegCM is able to capture the prominent data distribution very well during the phases of monsoon, if an unbiased
ICBC is provided.
Therefore, it is advised to cross verify the ICBC over the
specified study region before using it directly for the RegCM
simulations.
Acknowledgements This work is a part of a R&D project, funded by the
Department of Science and Technology (DST), Ministry of Earth Science
(MoES), Govt. of India. The authors wish to thank to The India
Meteorology Department (IMD), NOAA/OAR/ESRL (Boulder,
Colorado, USA; http://www.esrl.noaa.gov/psd/), and European Centre
for Medium-Range Weather Forecasts (ECMWF) for providing gridded
datasets. The authors seem their sincere gratitude to Prof. T.N.
Krishnamurti, Florida State University, USA for his valuable comments
on the manuscript to improve publication quality. Special thanks to the
International Center for Theoretical Physics (ICTP), Italy, for providing
the RegCM. The authors wish to extend their sincere gratitude to the
Journal Editor and the Reviewers for their insightful comments on the
paper.
References
Almazroui M (2016) RegCM4 in climate simulation over CORDEXMENA/Arab domain: selection of suitable domain, convection and
land-surface schemes. Int J Climatol 36:236–251
Almazroui M (2012) Dynamical downscaling of rainfall and temperature
over the Arabian Peninsula using RegCM4. Clim Res 52:49–62
Arakawa A, Schubert WH (1974) Interaction of a Cumulus Cloud
Ensemble with the Large-Scale Environment, Part I. J. Atmos. Sci.
31: 674-701. https://doi.org/10.1175/1520-0469(1974)031<0674:
IOACCE>2.0.CO;2
Bett PE, Thornton HE (2016) The climatological relationships between
wind and solar energy supply in Britain. Renew Energy 87:96–110
Bhatla R, Ghosh S, Mandal B, Mall RK, Sharma K (2016) Simulation of
Indian summer monsoon onset with different parameterization convection schemes of RegCM-4.3. Atmos Res 176–177:10–18.
https://doi.org/10.1016/j.atmosres.2016.02.010
Bhatla R, Ghosh S (2015) Study of break phase of Indian summer monsoon using different parameterization schemes of RegCM4.3. Int. J.
Eart. Atmos Sci 2(3):109–115
Central Statistical Organization (1998) Compendium of environment statistics. Central Statistical Organization, Department of Statistics,
Ministry of Planning and Program Implementation, Government
of India: New Delhi
Collins WD, Bitz CM, Blackmon ML, Bonan GB, Bretherton CS, Carton
JA, Chang P, Doney SC, Hack JJ, Henderson TB, Kiehl JT, Large
WG, McKenna DS, Santer BD, Smith RD (2006) The community
climate system model version 3 (CCSM3). J Clim 19:2122–2143
Dash SK, Pattnayak KC, Panda SK, Vaddi D, Mamgain A (2015) Impact
of domain size on the simulation of Indian summer monsoon in
RegCM4 using mixed convection scheme and driven by
HadGEM2. Clim Dyn 44:961–975
Dash SK, Shekhar MS, Singh GP (2006) Simulation of Indian summer
monsoon circulation and rainfall using RegCM3. Theor Appl
Climatol 86:161–172
Dickinson RE, Errico RM, Giorgi F, Bates GT (1989) A regional climate
model for the Western United States. Clim Chang 15:383–422
Diedhiou A, Janicot S, Viltard A, de Felice P, Laurent H (1999) Easterly
wave regimes and associated convection over West Africa and tropical Atlantic: results from NCEP/NCAR and ECMWF reanalyses.
Clim Dynam 15:795–822
Dobler A, Ahrens B (2010) Analysis of the Indian summer monsoon
system in the regional climate model COSMO-CLM. J Geophys
Res 115:1–12
Elguindi N, Bi X, Giorgi F, Nagarajan B, Pal J, Solmon F, Giuliani G
(2013) Regional climate model RegCM user manual version 4. 4.
The Abdus Salam International Centre for Theoretical Physics,
StradaCostiera, Trieste, Italy, 54 pp.
Emanuel KA (1991) A scheme for representing cumulus convection in
large-scale models. J Atmos Sci 48(21):2313–2335
Emanuel KA, Živković-Rothman M (1999) Development and evaluation
of a convection scheme for use in climate models. J Atmos Sci 56:
1766–1782
Fritsch JM, Chappell CF (1980a) Numerical prediction of convectively
driven mesoscale pressure systems. Part I: convective parameterization. J Atmos Sci 37:1722–1733
Fritsch JM, Chappell CF (1980b) Numerical prediction of convectively
driven mesoscale pressure systems. Part II: mesoscalemodel. J
Atmos Sci 37:1734–1762
Gibbons JD, Chakraborti S (1992) Nonparametric statistical inference.
3rd edition: Marcel Dekker
Gibbons JD, Chakraborti S (2003) Nonparametric statistical inference.
Marcel Dekker, New York
Giorgi F, Anyah RO (2012) The road towards RegCM4. Clim Res 52:3–
6. https://doi.org/10.3354/cr01089
Giorgi F, B Hewitson, J Christensen, M Hulme, H Von Storch, P
Whetton, R Jones, L Mearns, C Fu, (2001). Regional climate information: evaluation and projections (chapter 10). In climate change
2001: the scientific basis, contribution of working 32 group I to the
third assessment report of the IPCC [Houghton, J. T. Y. Ding, D. J.
Griggs, M. Noguer, P. J. Van der Linden, X. Dai, K. Maskell, and C.
A. Johnson (eds.)]. Cambridge U. Press: Cambridge, pp. 739–768
Giorgi F, Bates GT (1989) The climatological skill of a regional model
over complex terrain. Mon Weather Rev 117:2325–2347
Giorgi F, Coppola E, Solmon F, Mariotti L, Sylla MB, Bi X, Elguindi N,
Diro GT, Nair V, Giuliani G, Turuncoglu UU, Cozzini S, Güttler I,
O’Brien TA, Tawfik AB, Shalaby A, Zakey AS, Steiner AL, Stordal
F, Sloan LC, Brankovi’c C (2012) RegCM4: model description and
preliminary tests over multiple CORDEX domains. Clim. Res 52:7–
29. https://doi.org/10.3354/cr01018
Grell GA (1993) Prognostic evaluation of assumptions used by cumulus
parameterizations. Mon Wea Rev 121(3):764–787
Aspect of ECMWF downscaled Regional Climate Modeling in simulating Indian summer monsoon rainfall and...
Grell GA, Dudhia J, Stauffer DR (1994) Description of the fifthgeneration Penn State/NCAR Mesoscale Model (MM5). National
Center for Atmospheric Research (NCAR) Technical Note
NCAR/TN-398+STR, NCAR, Boulder, CO. doi: https://doi.org/
10.5065/D60Z716B
Holtslag AAM, de Bruijn EIF, Pan HL (1990) A high resolution air mass
transformation model for short-range weather forecasting.
MonWeather Rev 118:1561–1575
IMD Monsoon report (2013) Monsoon 2013: a report. (edited by D. S.
Pai and S.C. Bhan), India Meteorological Department (IMD),
Government of India. IMD Met Monograph: ESSO/IMD/
SYNOPTIC MET/01-2014/15
Kiehl JT, Hack JJ, Bonan GB, Boville BA, Breigleb BP, Williamson D,
Rasch P (1996) Description of the NCAR community climate model
(CCM3). Technical Report NCAR/TN-420+STR, National Center
for Atmospheric Research, Boulder, CO. doi: https://doi.org/10.
5065/D6FF3Q99
Krishnamurthy V, Kinter III JL (2003) The Indian monsoon and its relation to global climate variability. Global Climate (X. Rodó and F. A.
Comín, Eds., Springer-Verlag) 186–236
Krishnamurti TN (1985) Summer monsoon experiment: a review. Mon.
Wea. Rev. 113:1590–1626
Krishnamurti TN, Ardunay P (1980) The 10–20 day westward propagating model and ‘break’ in the monsoon. Tellus 32:15–26
Krishnamurti TN, Bhalme HN (1976) Oscillations of monsoon system.
Part I: observational aspects J Atmos Sci 45:1937–1954
Maharana P, Dimri AP (2014) Study of seasonal climatology and interannual variability over India and its sub-regions using a regional
climate model (RegCM3). J. Earth. Sys. Sci. 123(5):1147–1169
Maharana P, Dimri AP (2016) Study of intraseasonal variability of Indian
summer monsoon using a regional climate model. Clim Dyn 46(3):
1043–1064
Maurya RKS, Sinha P, Mohanty MR, Mohanty UC (2017) Coupling of
community land model with RegCM4 for Indian summer monsoon
simulation. Pure Appl Geophys 174:4251–4270. https://doi.org/10.
1007/s00024-017-1641-8
Mishra V, Kumar D, Ganguly AR, Sanjay J, Mujumdar M, Krishnan R,
Shah RD (2014) Reliability of regional and global climate models to
simulate precipitation extremes over India. J Geophys Res Atmos
119:9301–9323
NCC Research Report (2013) Development and analysis of a new high
spatial resolution (0.25o x 0.25o) long period (1901-2010) daily
gridded rainfall data set over India. [Ed/author: D. S. Pai, Latha
Sridhar, M. Rajeevan, O. P. Sreejith, N.S. Satbhai and B.
Mukhopadhyay]. NCC Research Report No 1/2013. 63
Pai DS, Rajeevan M (2007). Indian summer monsoon onset: variability
and prediction. National Climate Centre, India Meteorological
Department
Park S, Hong SY (2004) The role of surface boundary forcing over South
Asia in the Indian summer monsoon circulation: a regional climate
model sensitivity study. Geophys Res Lett 31:L12112. https://doi.
org/10.1029/2004GL019729
ParthSarthi P, Ghosh S, Kumar P (2015) Possible future projection of
Indian summer monsoon rainfall (ISMR) with the evaluation of
model performance in coupled model inter-comparison project
phase 5 (CMIP5). Glob Planet Chan 129:92–106
ParthSarthi P, Kumar P, Ghosh S (2016) Possible future rainfall over the
Gangetic Plains (GP), India, in multi model simulations of CMIP3
and CMIP5. Theor Appl Clim 124(3–4):691–701
Rajeevan M, Gadgil S, Bhate J (2010) Active and break spells of the
Indian summer monsoon. J Earth Sys Sci 119(3):229–247
Raju PVS, Bhatla R, Almazroui M, Assiri M (2015) Performance of
convection schemes on the simulation of summer monsoon features
over the South Asia CORDEX domain using RegCM-4.3. Int J
Climatol 35:4695–4706. https://doi.org/10.1002/joc.4317
Raju PVS, Bhatla R, Mohanty UC (2009) The evolution of mean conditions of surface meteorological fields during active/break phases of
the Indian summer monsoon. Theor Appl Climatol 95:135–149
Ruchith RDP, Raj E, Kalapureddy MCR, Deshpande SM, Dani KK
(2014) Time evolution of monsoon low-level jet observed over an
Indian tropical station during the peak monsoon period from highresolution Doppler wind lidar measurements. J Geophys Res Atmos
119:1786–1795. https://doi.org/10.1002/2013JD020752
Saeed F, Hagemann S, Jacob D (2009) Impact of irrigation on the South
Asian summer monsoon. Geophys Res Lett 36:L20711. https://doi.
org/10.1029/2009GL040625
Singh D, Tsiang M, Rajaratnam B, Diffenbaugh NS (2014) Observed
changes in extreme wet and dry spells during the South Asian summer monsoon season. Nat Clim Chan 4:456–46.1. https://doi.org/
10.1038/nclimate2208
Singh S, Ghosh S, Sahana AS, Vittal H, karmakar S (2017) Do dynamic
regional models add value to the global model projections of Indian
monsoon? Clim Dyn 48:1375–1397
Sinha P, Mohanty UC, Kar SC, Dash SK, Kumari S (2013) Sensitivity of
the GCM driven summer monsoon simulations to cumulus parameterization schemes in nested RegCM3. Theor Appl Climatol 112:
285–306
Srivastava PK, Han D, Miguel A, Ramirez R, Islam T (2013)
Comparative assessment of evapotranspiration derived from NCEP
and ECMWF global datasets through Weather Research and
Forecasting model. Atmos Sci Let 14:118–125
Sylla MB, Dell’Aquila A, Ruti PM, Giorgi F (2010) Simulation of the
intraseasonal and the interannual variability of rainfall over West
Africa with a Regional Climate Model (RegCM3) during the monsoon period. Int J Climatol 30:1865–1883
Sylla MB, Giorgi F, Ruti PM, Calmanti S, Dell’Aquila A (2011) The
impact of deep convection on the West African summer monsoon
climate: a regional climate model sensitivity study. Q J R Meteorol
Soc 137:1417–1430
Tawfik AB, Steiner AL (2011) The role of soil ice in land–atmosphere
coupling over the United States: a soil moisture precipitation winter
feedback mechanism. J Geophys Res 116:D02113
Tugba O, Tufan TT, Turke M, Kurnaz M, Levent M (2016) Projected
changes in temperature and precipitation climatology of central
asiacordex region 8 by using regcm4.3.5. Atmospheric Research.
doi:https://doi.org/10.1016/j.atmosres.2016.09.008
Varikoden H (2006) Dynamic characteristics of atmospheric boundary
layer during different phases of monsoon. Department of
Atmospheric Sciences Cochin University of Science and
Technology, Lakeside Campus, Cochin, India, Doctoral Thesis
VenkataRatnam JK, Kumar K (2005) Sensitivity of the simulated monsoons of 1987 and 1988 to convective parameterization schemes in
MM5. J Clim 18:2724–2743
Webster PJ, Magana VO, Palmer TN, Shukla J, Tomas RA, Yanai M,
Yasunari T (1998) Monsoons: processses, predictability, and the
prospects for prediction. J Geophys Res 103:14451–14510
Wilk MB, Gnanadesikan R (1968) Probability plotting methods for the
analysis of data. Biometrika 55(1):1–17