Simulation of Indian Monsoon Droughts from State-of-Art Climate Models
Corresponding author: Dr. Vinay Kumar, Department of Earth, Ocean & Atmospheric Science Florida State University, Tallahassee, FL-32306, USA, Tel : 850-644-1494, Fax: 850-644-9642, Email: email@example.com
In the wake of the severe socio-economic impacts associated with monsoon droughts in past, there has always been a com- pelling necessity for accurate predictions of seasonal mon- soon rainfall over India [1-5]. Despite advances in seasonal forecasting techniques, the skill in the seasonal predictions of the Indian monsoon rainfall both by statistical models and dynamical models has been poor [6-8]. These studies indicate that the statistical/empirical based predictions generally fail to capture the monsoon rainfall extremes (i.e., droughts and excess rainfall seasons); while dynamically-based GCM predic- tions have large errors in simulating the year-to-year monsoon rainfall variations. All the acronyms used in this study are ex- plained in Table-1.
The low seasonal predictability of the state-of-art GCMs over the Indian monsoon region is partly due to the fact that a ma- jority of the models face a major challenge even in simulating the mean rainfall distribution over the Indian and East Asian monsoon regions . Furthermore, the effects of atmospheric
internal dynamics on the precipitation variability are non-neg- ligible over the Indian monsoon region and compare with the level of variability from the slowly varying external boundary forcing [10-13]. It is now recognized that the strong sensitivity of the simulated Indian monsoon rainfall variations to atmo- spheric initial conditions tends to limit the seasonal predict- ability over the region.
Recent studies have suggested that ocean-atmosphere coupled models performance better than the atmosphere-alone GCMs in forecasting the Indian monsoon rainfall [7,14]. The predic- tions systems designed with atmospheric models (AGCM) are driven with a specified ocean surface temperature. Such a con- figuration pre-supposes the Indian monsoon variability to be a consequence of solely the atmosphere reacting to the ocean. However since the Indian monsoon involves a fully coupled ocean-land-atmosphere system, it is expected that the sea sur- face temperature (SST) field evolves through interactions with the atmospheric circulation. Wang et al.,  have suggested that the lack of skill of AGCMs in simulating the interannual variability over Southeast Asia and west North Pacific results
|AISMR||All India Summer Monsoon Rainfall|
|AGCM/GCM||NCEP Aviation Model|
|AMIP||Atmospheric Model Intercomparison Project|
|CDAC||Centre for Development of Advanced Computing|
|CERFACS||European CEnter for Research and Advanced Training in Scientific
|CMAP||CPC Merged Analysis of Precipitation|
|CNRM||Centre National de Recherches Me´te´orologiques, France|
|COLA-GCMs||Center for Ocean Land Atmosphere – Atmospheric General Circulation
|ECMWF||European Center for Medium range Weather Forecasting, UK|
|EEIO||East Equatorial Indian Ocean|
|ENSO||El-Nino Southern Oscillation|
|EQUINOO||Equatorial Indian Ocean Oscillation|
|INGV||Istituto Nazionale de Geofisica e Vulcanologia, Italy|
|IOD||Indian Ocean Dipole|
|IITM||Indian Institute of Tropical Meteorology, India|
|ITCZ||Inter Tropical Convergence Zone|
|JJAS||June, July, August, September|
|LODYC||Laboratoire d’Océanographie DYnamique et de Climatologie, France|
|MPI||MAX-Planck Institut, Germany|
|NCEP||National Center for Environmental Prediction, USA|
|NOAAOISST||National Oceanic and Atmospheric Administration Optimum Interpolated
Sea Surface Temperature
|SPIM||Seasonal Prediction of Indian Monsoon|
Table 1. Acronym for model’s name and its affiliation with an institute. Additional acronyms used in this study are also included.
from the failure of the models to simulate correctly the rela- tionship between the local summer rainfall and the SST anom- alies in the Philippine Sea, the South China Sea and the Bay of Bengal. In contrast to the design of AGCM experiments, the coupled model simulations take into account SST variations that result from atmospheric forcing. It has been suggested that this interactive ocean-atmosphere coupling improves the skill of the Indian monsoon predictions [7,8,14]. The variabil- ity of rainfall over tropical regions is well demonstrated by Feng et al. .
The analyses of Krishnakumar et al.,  were based on the theoretical assumption that the coupled models have perfect skill in forecasting the Indian summer monsoon rainfall. Since coupled model predictions also involve growth of model fore- cast errors, the perfect forecast skill may not be achievable in actual practice. Therefore, it is not clear if coupled models can actually perform substantially better than the AGCMs in cap- turing the interannual variability of the Indian summer mon- soon rainfall. In order to understand this problem, we have examined an ensemble of hindcast monsoon simulations con- ducted using an AGCM; as well as hindcasts generated by
the DEMETER coupled modeling system (Development of a European Multi-model Ensemble System for Seasonal to Inter- annual Prediction). This paper provides an assessment of the state-of-art climate models in simulating monsoon droughts over India.
Datasets from model and reanalysis
Description of SPIM, DEMETER and other datasets
We shall first examine the interannual variability of the In- dian summer monsoon rainfall simulated by the Center for Ocean Land Atmosphere Studies (COLA) AGCM for 20 summer monsoon seasons of 1985-2004 period. These runs were con- ducted as part of the Seasonal Prediction of Indian Monsoon (SPIM) project, which was launched to assess the skills of the AGCMs, currently used in India, for generating monthly/sea- sonal predictions. All the AGCMs were ported and run on a single computational platform at the Centre for Development of Advanced Computing (CDAC), Bangalore. An AGCM-T30L18 from COLA was one of the six models that included in the SPIM project. The hindcast methodology consists of 5-member en- semble runs starting from the initial conditions of 26-30 April of each year; wherein observed SSTs are specified as lower boundary condition.
The DEMETER system comprises seven global coupled ocean-atmosphere models which are run on a single super- computer to produce a series of six-month multi-model en- semble hindcasts with common archiving and common diag- nostic software [e.g. 16,17]. Details about the atmospheric and oceanic components of the 7 coupled models used in the DE- METER project are given in Table 2.
Mean monsoon rainfall distribution
Figure 1a shows the climatological mean summer monsoon rainfall for the June-September (JJAS) season simulated by the AGCM-COLA; and the corresponding observed rainfall from CMAP is shown in Figure 1b. Superposed on the rainfall maps are the mean winds at 850 hPa in the AGCM-COLA model sim- ulation (Figure 1a) and NCEP reanalysis (Figure 1b). Although the simulated wind-field essentially captures the salient large- scale features of the low-level monsoon cross-equatorial flow;
Table 2. Details of seven coupled models used in DEMETER project (Development of a European Multi-model Ensemble System for Seasonal to Interannual Prediction). The modeling partners are: CERFACS (European Centre for Research and Advanced Training in Scientific Compu- tation, France), ECMWF (European Centre for Medium-Range Weather Forecasts, International Organization), INGV (Istituto Nazionale de Geofisica e Vulcanologia, Italy), LODYC (Laboratoire d’Océanographie Dynamique et de Climatologie, France), Météo-France (Centre National de Recherches Météorologique s, Météo-France, France), MPI (Max-Planck Institut für Meteorologie, Germany) and Met Office (The Met Office, UK) [adapted from Palmer et al., 2004].
In this analysis, we examined the hindcasts from the coupled models datasets over a 22 year (1980-2001) period. During this period, all seven coupled models that participated in the DEMETER project generated hindcasts. Each hindcast is 6-month integration and comprises an ensemble of 9-mem- bers. Since, our primary interest is to investigate the IAV of the Indian summer monsoon; we have examined the DEMETER hindcasts that were initiated from the 1st May initial condition.
Furthermore CMAP rainfall, all Indian summer monsoon rain- fall from IITM, NCEP winds and NOAAOISST were used as the observational datasets and compared against model satasets.
AGCM simulation of Inter-Annual Variability (IAV) of mon- soonal rainfall
it can be seen that the simulated monsoon rainfall over India differs considerably from the CMAP rainfall distribution. It can be seen that the simulated rainfall maxima over Bay of Bengal is located more southward as compared to the CMAP rainfall; while the west coast rainfall maximum is located more to the west over the Arabian Sea in the AGCM simulation. There are also differences between the simulated and observed rainfall distribution over the equatorial Indian and Pacific Oceans.
Gadgil and Sajani  examined the simulation of summer monsoon precipitation from 30 different AGCMs that partic- ipated in the Atmospheric Model Intercomparison Project (AMIP). The AMIP runs were carried out by forcing models with observed monthly SST and sea-ice distribution over a 10-year (1979-1988) period, starting from the atmospheric
initial state of 1 January 1979 . However, the version of COLA model (T30L18) used in SPIM is different from the COLA (R40L18) model used in AMIP. Although the SPIM-version has a coarser resolution than the AMIP-version, the monsoon rain- fall to the west of India is better simulated by the former (Fig- ure 1a) as compared to the latter [see their Figure 9, . This appears to be largely related to the different moist convection schemes used in the two versions.
Figure 1. Climatological rainfall (mm day-1) and 850 hPa winds (m s-1) for June-September (JJAS) monsoon season (a) AGCM-COLA (b) CMAP rainfall and NCEP reanalysis winds. The climatological fields in the GCM are computed from the 20-year simulation. For the CMAP/ NCEP data, the climatology is based on the period (1979-2004).
The convection scheme used in the R40L18 version of AMIP was based on Kuo ; while the T30L18 uses the Relaxed Arakawa Schubert (RAS) scheme . In fact, in the 2nd phase of the AMIP project (AMIP-II), the R40L18 model employed the
RAS convection scheme which led to a marked improvement in the simulation of the mean monsoon precipitation over India and adjoining region . In noting the above, the main point is that accurate simulation of the mean rainfall distribution over the Indian and East Asian monsoon regions still remains a challenging issue for many state-of-art AGCMs [9,18, 22]. Ku- mar and Krishnamurti  analyzed the rainfall climatology from coupled models and found it is far away from observa- tions.
Observed and simulated monsoon droughts during (1985- 2004)
Here, we discuss the simulated interannual variability of monsoon rainfall in the COLA AGCM and also examine spatial maps of precipitation anomalies associated with the 5 mon- soon droughts (1985, 1986, 1987, 2002 and 2004) over India that occurred during the 20-year (1985-2004) period. Figure 2 shows the interannual variations of the summer monsoon rainfall over the Indian region for the period (1985-2004) both from the AGCM simulation and observations. The rainfall anomalies for the AGCM simulation, averaged over the Indian region (70oE – 90oE; 12oN – 32oN), are shown by blue bars. Note that the AGCM runs are 5-member ensemble realizations for each of the 20 years.
The red bars in Figure 2 indicate the CMAP rainfall anomalies averaged over the same region. The black bars in Figure 2 cor- respond to the observed All India Summer Monsoon Rainfall (AISMR) index.
Figure 2. Time-series of the inter-annual variability of all India sum- mer monsoon rainfall (AISMR, black bars) and area averaged summer monsoon rainfall over the Indian region (70oE – 90oE; 12oN – 32oN) for CMAP (red bars) and GCM (blue bars). Rainfall from GCM is based on 5-member ensemble simulations for a 20 year period (1985-2004). The mean and standard deviation (SD) of the 3 rainfall indices are shown in top-right corner. The correlation coefficient (CC) among the 3 rainfall indices is given in topleft corner.
The details of the CMAP and AISMR rainfall data are given in the figure caption. Note the high correlation (0.87) between the CMAP and AISMR indices. On the other hand, the correla- tion between the observed and simulated monsoon rainfall variations is low.
It can be seen that from Figure 2 that the negative rainfall departures associated with the monsoon droughts of (1985, 1986 and 1987) are not captured by AGCM. On the other hand, it is interesting to note that the monsoon droughts of 2002 and 2004 are brought out in a majority of the ensemble members. Further, it is also seen that the 4 consecutive deficient mon- soons (if the AISMR is ≤-10% of climatological normal; ) during (1999-2002) are consistently captured in the AGCM simulation; although the magnitudes of the simulated rainfall anomalies are larger than the observed anomalies. It is not clear as to why the simulation performance of the AGCM is better during certain periods (1999-2002) than during other periods (1985-1987). Likewise, it is intriguing to note that the excess monsoon rainfall during 1988 is consistently captured in the AGCM by all the ensemble members; while on the other hand all the 5-members consistently failed to capture the ex- cess monsoon of 1994. Grimm et al.,  have suggested that the possibility of secular variations in the model forecast skills. It is not obvious if the SPIM simulations of the Indian monsoon are subject to such variations introduced through the specifi- cation of SST forcing over the 20- year period. Furthermore, the AGCM simulated monsoon precipitation exhibits strong variations from member-to-member (model simulations us- ing different initial conditions) in the presence of anomalous SST conditions. For example, the 1987 monsoon drought was accompanied by El Nino conditions in the Pacific [26-31]. De- spite the presence of the anomalous SST boundary forcing, it is surprising to note that the AGCM simulation for 1987 shows significant member-to-member variations (note that 3 mem- bers have positive anomalies; while one member has negative departure and the other is close to zero).
The spatial maps of the observed and simulated rainfall anom- alies for the 5 monsoon droughts (1985, 1986, 1987, 2002 and 2004) are shown in Figure 3. Also included in Figure 3 are maps of rainfall anomaly, over the Indian landmass, from the India Meteorological Department. It can be noticed from the CMAP rainfall that there are significant variations in the rainfall anomalies over the Indo-Pacific region from one drought year to another. Even under El Nino conditions, there are significant differences in the precipitation patterns from one drought case to another. For example, both 1987 and 2002 show enhanced precipitation over the equatorial central Pacific; however the rainfall anomalies over the tropical Indian Ocean are quite different in the two years. The AGCM simulated precipitation anomalies during (1985, 1986, 1987) exhibit significant mis- match with the CMAP anomalies particularly over the Indian subcontinent. Figure 3p shows the rainfall anomaly composite, based on the 5 drought years, for the CMAP data. The corre-
sponding anomaly composites for the GCM and IMD rainfall are shown in Figure 3q and Figure 3r respectively. It can be seen that the GCM broadly captures the enhanced precipitation over the equatorial central Pacific; while the rainfall reduction over the Indian region is not clearly brought out. This essentially indicates that the AGCM might be able to reproduce the ENSO related precipitation anomalies over the tropical Pacific; while the teleconnections over the Indian monsoon region are not realistic in the AGCM simulation. Recent studies have shown that in addition to the ENSO-induced impacts on the monsoon, it will be necessary to take into account the effects of convec- tion changes over the tropical Indian Ocean on the monsoon rainfall variation [32-34].
Figure 3. Spatial distribution of JJAS summer monsoon rainfall anomalies (mm day-1) for the droughts of (1985, 1986, 1987, 2002 and 2004). The panels (a, d, g, j, m, p) are based on the CMAP data; the panels (b, e, h, k, n, q) are from the GCM; and the panels (c, f, i, l, o, r) are for the high resolution (1ox1o) gridded daily data from IMD – version 2 [Rajeevan et al., 2006].
DEMETER coupled model simulations of the IAV of mon- soon rainfall and drought events
In the previous section, it was seen that the AGCM simulated IAV of the Indian monsoon rainfall showed poor correlation with the observed IAV. Although, we had described the results for one particular AGCM (i.e., COLA model) from the SPIM proj- ect, it is known that most of the AGCMs, that participated in the AMIP, had comparable hindcast skills in reproducing the observed IAV [6,9,13]. Several studies have suggested that this problem arises primarily from the AMIP-type of experimental design in which the atmosphere is forced to respond passively to the specified SSTs [7,8,14]. In reality, the SST and rainfall in the monsoon ocean region interact with each other. Therefore, their anomalies tend to be negatively correlated, because the SST anomalies are, to a large extent, a response to monsoon forcing. In a region of an enhanced monsoon, the increased rainfall and cloudiness will tend to reduce the downward solar radiation into the ocean mixed layer, meanwhile, the increased rainfall enhances the monsoon westerly winds, which further enhance the surface evaporative cooling and the entrainment cooling of the mixed layer. Furthermore, it was suggested that coupling of the ocean and atmospheric GCMs can lead to im- proved simulation of the IAV of monsoon rainfall [7, 14]. Here, we have examined the simulation of the monsoon IAV by a suite of coupled ocean-atmosphere models from the DEME- TER ensemble prediction system.
Coupled model simulations of mean and IAV of the Indian summer monsoon
Figure 4a shows the CMAP rainfall climatology for the June-Sep- tember (JJAS) season and the corresponding maps from the 7 coupled models are shown in Figure 4 (b-h). The rainfall cli- matology for each model is based on the 22-year (1980-2001) simulation and is averaged over the 9-ensemble members. The distribution of the mean rainfall in most of the models shows the maximum precipitation over the Indian region; the second- ary rainfall maximum over the equatorial Indian Ocean and the tropical west Pacific. The rainfall band over the equatorial Pacific is associated with the Inter Tropical Convergence Zone (ITCZ). Notice that the rainfall minima over the sub-tropical east Pacific in both the hemispheres, the sub-tropical Indian Ocean and the sub-tropical regions of Arabia and west-Asia are captured well in almost all of the models. While there are dif- ferences in the spatial distribution of the mean rainfall across the models; the large-scale structure of the tropical rainfall distribution is quite consistent among the 7-AGCMs. Further, it can be noticed that the simulated mean low-level winds at 850 hPa by the DEMETER models (Figure 5) captures the salient large-scale circulation features (i.e., monsoon south-westerlies and the cross-equatorial flow; easterly trades over the tropical Pacific, the sub-tropical anti-cyclones, etc.).
The simulated interannual variability of the summer monsoon rainfall over the Indian region (70oE-90oE; 12oN-32oN) by
Figure 4. Climatological rainfall (mm day-1) for JJAS from (a) CMAP
(b) CERFACS (c) ECMWF (d) INGV (e) LODYC (f) Meteo-France (g) MPI (h) Met Office. The DEMETER (Development of a European Multi-model Ensemble System for Seasonal to Interannual Predic- tion) is a project on multimodel ensemble prediction from global cou- pled models.
the DEMETER models for the 22-year period (1980-2001) is shown by blue bars in Figure 6. The corresponding rainfall in- dex based on the CMAP dataset (red) and the observed AISMR index (black) are also shown in Figure 6. It is seen that the DEMETER models show a modest improvement in capturing the observed IAV as compared the COLA-AGCM simulations discussed earlier. Out of the 7 coupled models, the best 5 sim- ulations had correlations in the range of 0.2 to 0.38 with the CMAP rainfall variations over the Indian region. Except for the marginal improvement, it is not evident from Figure 6 that the coupled models significantly out-perform the AGCMs in cap- turing the observed IAV of the Indian monsoon rainfall.
Observed and simulated large-scale anomalies during monsoon droughts
Figure 7a shows the CMAP rainfall anomaly composite based on the 4 monsoon droughts (1982, 1985, 1986, and 1987)
Figure 5. Climatological winds (m s-1) at 850 hPa from (a) NCEP (b-
h) DEMETER models. The shaded regions have wind-speed exceeding 10 m s-1.
during the 22-year period. The corresponding anomaly com- posites from the 7 coupled models are shown in Figure 7(b-h). The striking feature in the CMAP anomaly composite is the in- creased rainfall over the equatorial central-eastern Pacific and decreased rainfall over the Indian sub-continent; Southeast Asia, Indonesia and equatorial west Pacific. A careful examina- tion of Figure 7a shows increased rainfall over the equatorial eastern Indian Ocean; but decreased rainfall over the western Indian Ocean. In the coupled model simulations, a qualitative increase in rainfall over the equatorial central Pacific can be seen in most of the models; although there are differences in the location and intensity of the Pacific rainfall anomaly. How
Figure 6. Time-series of the inter-annual variability of AISMR (black bars) and area averaged summer monsoon rainfall over the Indian region (70oE – 90oE; 12oN – 32oN) from CMAP (red bars) and DEME- TER models (blue bars) (a) CERFACS (b) ECMWF (c) INGV (d) LODYC
(e) Meteo-France (f) MPI (g) Met Office. The rainfall variations are
expressed as percentage departures from normal.
ever, the simulation of the monsoon rainfall deficiency over the Indian region is not adequately captured in majority of models. It can be seen that the simulated regional rainfall anomalies over the Indian subcontinent vary considerably from one mod- el to another.
Figure 7. Composite of seasonal (JJAS) rainfall anomaly (mm day-
1) (a) MAP (b) CERFACS (c) ECMWF (d) INGV (e) LODYC (f) Me- teo-France (g) MPI (h) Met Office. The anomaly composites are based on the 4 monsoon droughts during (1982, 1985, 1986 and 1987) – when the rainfall departures were less than -10% of the normal.
Composite maps of low-level winds and SST anomalies from the DEMETER simulations, based on the 4 monsoon droughts, are shown in Figure 8 and Figure 9 respectively. The corre- sponding anomaly composites of winds from NCEP (Figure 8a) and observed SST (Figure 9a) are shown for comparison.
Figure 8. Circulation anomaly (m s-1) at 850 hPa from (a) NCEP (b-h)
Notice that a majority of the models broadly capture the weak- ening of the tropical Pacific trade winds (i.e., westerly anoma- lies) and the warm SST anomalies in the central-eastern equa- torial Pacific; although there are differences in the location and amplitude of the anomalies across different models. On the other hand, notice that the simulation of SST and wind anoma- lies varies considerably among the different coupled
Figure 9. SST anomaly (oC) from (a) OISST (b-h) DEMETER models.
models over the Indian Ocean region. In particular, it can be seen that the CERFACS, ECMWF, LODYC, Meteo-France, and Met-Office models show larger SST cooling, anomalous east- erly winds and decreased rainfall over the equatorial eastern Indian Ocean (EEIO) as compared to the INGV and MPI models. The bias in the simulation of the ocean-atmosphere anomalies in the EEIO (Figures 7-9) by the CERFACS, ECMWF, LODYC, Meteo-France, and Met-Office models is associated with an en- hanced easterly outflow from the EEIO which tends to increase the monsoon precipitation to the north of the equator . This may be a possible explanation why the CERFACS, ECMWF, LODYC, Meteo-France, and Met-Office models fail to produce the strong rainfall deficiency over the Indian subcontinent; as compared to the INGV and MPI models. Therefore, the above discussion suggests that biases in accurately representing the ocean-atmosphere coupling in the Indian Ocean environment can significantly limit the simulation of IAV of monsoon rain- fall. Therefore, in addition to capturing the ENSO variability, it is also essential for coupled models to realistically depict the Indian Ocean variability in order to be able to improve the simulation of the IAV of monsoon rainfall. For the future pro- jections of the Indian-Pacific SST and ENSO variability is well examined by Soloman and Newman .
Regional aspects associated with the IAV of Indian mon- soon rainfall
Significant advances in our current understanding of air-sea in- teractions in the tropical Indian Ocean environment have taken place during the last decade [37,38]. In particular, the discov- ery of the Indian Ocean Dipole (IOD) phenomenon and its ef- fect on the regional climate variability has drawn considerable attention [35, 39-41]. An IOD episode is characterized by cold SST anomalies in the southeastern part of the tropical Indian Ocean off the Coast of Sumatra; and warm SST anomalies in the west-central Indian Ocean. The zonal SST gradient during IOD events drives anomalous easterlies over the equatorial Indian Ocean; which increases upwelling and cooling in the southeast- ern tropical Indian Ocean and leads to rainfall reduction over the region. Generally, the ocean-atmosphere anomalies during IOD events evolve through the boreal summer and attain max- imum amplitude during the autumn months. During the last 50 years, there have been four major IOD events during 1961, 1994, 1997  and more recently during 2006 . Studies have shown that positive IOD events favor stronger-than-nor- mal monsoonal rains over the Indian sub-continent through enhanced supply of moisture from the south-eastern tropical Indian Ocean into the plains of north-central India [34,35]. Observed rainfall records over India (http://www.tropmet. res.in) provide corroborative evidence for anomalously wet summer monsoons during 1961 and 1994; while the monsoon precipitation was above-normal in 1997 and 2006 – despite 1997 being a very strong El Nino year. Based on model simu- lation experiments, Ashok et al.,  noted that the strength- ening of monsoonal winds during an IOD episode can counter- act and compensate the impact of an ongoing El Nino on the
Figure 10. Seasonal (JJAS) rainfall anomaly (mm day-1) for 1997 (a case of intense El Nino plus a positive IOD) (a) CMAP (b) CERFACS (c) ECMWF (d) INGV (e) LODYC (f) Meteo-France (g) MPI (h) Met Office.
Indian monsoon. Slingo and Annamalai  pointed out that the strong suppression of convection over the southeastern In- dian Ocean and the Maritime Continent by the intense El Nino of 1997 in fact altered the local monsoon Hadley circulation in a manner as to favor above-normal precipitation over the In- dian landmass. Here, we examine the simulation of the rainfall anomalies during the summer monsoon of 1997 by the DEME
Cite this article: Vinay Kumar. Simulation of Indian Monsoon Droughts from State-of-Art Climate Models. J J Earth Science. 2016, 1(1): 001.
Figure 11. Seasonal (JJAS) SST anomaly (oC) for 1997 (a case of intense El Nino plus a positive IOD) (a) OISST (b) CERFACS (c) ECMWF (d) INGV (e) LODYC (f) Meteo-France (g) MPI (h) Met Office.
Figure 12. Seasonal (JJAS) climatology (1981-2010) of rainfall (mm/ day) (a) CMAP (b) CNRM; seasonal (JJAS) climatology (1982-2010) of SST (oC) (c) NOAAOISST (d) CNRM.
TER models; and contrast them with the CMAP rainfall anom- alies (Figure 10). The increased rainfall over the central-east- ern equatorial Pacific in the CMAP data is associated with the strong El Nino conditions which prevailed in 1997. In fact, the
1997 event was the strongest El Nino in the last century. A strong suppression of rainfall is seen over the equatorial west Pacific and eastern Indian Ocean; while above-normal precip- itation is seen over northern India, Indo-China and Eastern China (Figure 10a). Increased precipitation is also seen over the western tropical Indian Ocean in the CMAP data (Figure 10a). While most of the coupled models, capture the increased precipitation over the equatorial eastern-central Pacific and the decreased rainfall over west Pacific; they fail to capture the sign of the rainfall anomaly over the Indian subcontinent. The model simulations mostly show large decrease of rainfall over the Indian landmass. In other words, the DEMETER coupled models basically depict the ENSO-related precipitation chang- es; while the competing regional effects associated with the IOD are not properly depicted in the GCM simulations.
The biases in the rainfall simulations are consistently reflected in the simulated SST during the 1997 IOD event by the DEME- TER coupled models (Figure 11). Although the 1997 El Nino conditions in the Pacific (i.e., positive SST anomalies in the equatorial eastern Pacific and negative anomalies in the west- ern Pacific) are reasonably captured by a majority of models; the regional Indian Ocean SST anomalies associated with the IOD are not so robust in the simulation. For example, the CER- FACS and Meteo-France coupled models show cooling in the south-eastern tropical Indian Ocean, the anomalous warm- ing in the western Indian Ocean is not very clear. Gadgil et al.,
 note that the IAV of the Indian summer monsoon rainfall is strongly determined by the pattern of rainfall/convection anomalies over the equatorial Indian Ocean – which they refer to as the Equatorial Indian Ocean Oscillation (EQUINOO) – the atmospheric component of the IOD. During years of normal or above-normal monsoon rainfall over India, the EQUINOO is as- sociated with suppressed rainfall over the eastern equatorial Indian Ocean and increased rainfall over the western Indian Ocean. The EQUINOO anomalies in the eastern and western Indian Ocean are reversed for the years of monsoon droughts over India. Since the EQUINOO precipitation anomalies are closely linked to the IOD and the regional SST anomalies, it is important that reliable model simulations of the monsoon rainfall critically depend on the treatment of the coupled inter- actions in the Indian Ocean environment.
Furthermore, Krishnan et al.,  have demonstrated the im- portance of the Indian Ocean and monsoon coupled interac- tions in causing droughts over the subcontinent. They point- ed out that prolonged breaks in the monsoon rainfall, which occur on sub-seasonal time-scales, are associated with an ocean-atmosphere feedback in the equatorial Indian Ocean. In this feedback, anomalous equatorial westerly winds push water eastward and cause warm SST anomalies by deepen- ing the oceanic mixed-layer and thermocline in the equato- rial eastern Indian Ocean (EEIO). The anomalous warming in the EEIO, sets-up an east-west SST gradient which in turn
sustains the equatorial westerly anomalies, through a Bjern- kes-type feedback . While this coupled feedback results in increased moisture convergence and enhanced convective/ rainfall activity over the EEIO; the equatorial anomalies induce strong subsidence (sinking of warm and dry air) over the Indi- an landmass and sustain the monsoon-break condition there- by leading to occurrence of droughts over the sub-continent. These points clearly bring out the importance of the regional ocean-atmosphere coupled interactions in modeling the Indi- an monsoon rainfall variability.
Existing models do not carry the correct climatology of rainfall and temperature for the 30 years of time period (1981-2010). An example of climatological rainfall and SST from CNRM cou- pled model (CMIP experiments of extended historical run) are compared against reanalysis datasets in Figure 12 a, b, c, d. The models are not able to simulate maximum rainfall and maxi- mum temperature.
The aim of this paper is to understand the skill of the current state-of-art climate models in simulating the observed IAV of monsoon rainfall and drought occurrences over the subconti- nent. For this purpose, hindcast simulations of both AGCMs and ocean-atmospheric coupled models were examined. It is seen from the analysis that the observed IAV of the Indian monsoon rainfall is not adequately captured both by the AGCMs as well as the ocean-atmosphere coupled models. While the coupled models reasonably simulate the large-scale ENSO response in the Pacific; the IAV of the Indian monsoon rainfall is not ade- quately reproduced in the simulations. The AGCM simulations failed to capture the rainfall deficiencies associated with the monsoon droughts of (1985, 1986 and 1987) over the Indian region. A majority of the DEMETER coupled models, with the exception of the Met Office and INGV model, did not depict the large seasonal monsoon rainfall deficits over subcontinent during the 1987 monsoon drought. In general, the correlations between the observed and simulated monsoon rainfall anoma- lies for the period (1980-2001) by the DEMETER models were not very robust.
It is seen from the present analysis that model deficiencies in accurately representing the regional ocean-atmosphere coupling in the tropical Indian Ocean can pose a major prob- lem for seasonal monsoon prediction in coupled models, as was noticed in the case of the 1997 monsoon simulation by the DEMETER models. In addition, there is considerable vari- ability in the simulation of monsoon rainfall anomalies over the Indian region among the different models and ensemble members. Numerous studies have shown that the strong at- mospheric internal dynamics over the Indian monsoon region renders dynamical forecasting of the seasonal monsoon rains very sensitive to initial conditions [6,9-13,27,46]. It is hoped that with the increasing observational network like moorings, data buoys, ARGO and satellite observations over oceanic and
land regions; together with developments in coupled modeling and data-assimilation will foster significant improvements in extended and seasonal range monsoon rainfall predictions in the coming years.
Authors thank SPIM and the DEMETER prediction systems to
provide various variable datasets to the research community.
- Walker GT. Correlations in seasonal variations of weather,IX. A further study of world weather. Mem Ind Meteor Dept.1924, 24(9): 275-333.
- Gowariker V, Thapliyal V, Kulshrestha SM, Mandal GS, Sen Roy N et al. A power regression model for long range forecast of southwest monsoon rainfall over India. Mausam. 1991, 42: 125–130.
- Krishna Kumar K, Soman MK, Rupa Kumar K. Seasonal fore- casting of south-west monsoon rainfall. Weather. 1995, 50: 449-467.
- Gadgil S, Srinivasan J, Nanjundiah RS, Kumar KK, Munot AA et al. On forecasting the Indian summer monsoon: The intrigu- ing season of 2002. Current Sci. 2002, 83(4): 394-403.
- Rajeevan M, Pai DS, Dikshit, Kelkar RR. IMD’s new opera- tional models for long range forecast of southwest monsoon rainfall over India and their verification for 2003. Current Sci. 2004, 86(3): 422-431.
- Gadgil S, Rajeevan M, Nanjundiah R. Monsoon prediction – Why yet another failure? Current Sci. 2005, 88(9): 1389-1400.
- Krishna Kumar K, Hoerling M, Rajagopalan B. Advancing dy- namical prediction of Indian monsoon rainfall. Geophys Res Lett. 2005, 32(8): L08704.
- Kang IS, Shukla J. Dynamical seasonal prediction and pre- dictability of monsoon. In The Asian Monsoon (B. Wang, ed.), Praxis Pub. Ltd., Chichester, UK. 2005, 585-612585-612.
- Kang IS, Jin K, Wang B, Lau KM, Shukla J et al. Intercompari- son of the climatological variations of Asian summer monsson precipitation simulated by 10 GCMs. Climate Dyn. 2002, 19: 383-395.
- Palmer TN, Anderson DLT. The prospects for seasonal forecasting – A review paper. Quart J Roy Meteor Soc.1994, 120(518): 755–793.
- Goswami BN. Interannual variation of Indian summer mon- soon in a GCM: External conditions versus internal feedbacks. J Climate. 1998, 11: 501-522.
- Brankovic C, Palmer TN. Seasonal skill and predictability of ECMWF PROVOST ensembles. Quart J Roy Meteor Soc. 2000, 126(567): 2035-2068.
- Wang B, Kang IS, Lee JY. Ensemble simulation of Asian-Aus- tralian monsoon variability by 11 AGCMs. J Climate. 2004, 17: 803–818.
- Wang B, Ding Q, Fu X, Kang IS, Jin K et al. Fundamental chal- lenge in simulation and prediction of summer monsoon rain- fall. Geophys Res Lett. 2005, 32(15): L15711.
- Feng X, Porporato A, Rodriguez-Iturbe I. Changes in rainfall seasonality in the tropics. Nat Clim Change. 2013, 3: 811-815.
- Palmer TN, Alessandri A, Andersen U, Cantelaube P, Dav- ey M et al. Development of a European Multi-Model Ensemble System for Seasonal to Inter-Annual Prediction (DEMETER). Bull Am Meteor Soc. 2004, 85: 853-872.
- Hagedorn R, Doblas-Reyes FJ, Palmer TN. The rationale be- hind the success of multi-model ensembles in seasonal fore- casting. Part I: Basic concept. Tellus A. 2005, 57(3): 219-233.
- Gadgil S, Sajini S. Monsoon precipitation in the AMIP runs. Climate Dyn. 1998, 14(9): 659–689.
- Gates WL. AMIP: The Atmospheric Model Intercomparison Project. Bull Ame Met Soc. 1992, 73(12): 1962-1970.
- Kuo HL. On Formation And Intensification Of Tropical Cy- clones Through Latent Heat Release By Cumulus Convection. J Atmos Sci. 1965, 22: 40-63.
- Moorthi S, Suarez MJ. Relaxed Arakawa-Schubert. A Param- eterization of Moist Convection for General Circulation Models. Mon Wea Rev. 1992, 120: 978-1002.
- Mujumdar M, Kumar V, Krishnan R. The Indian summer monsoon drought of 2002 and its linkage with tropical convec- tive activity over northwest Pacific. Clim Dyn. 2007, 28(7-8): 743-758.
- Kumar V, Krishnamurti TN. Improved Seasonal Precipitation Forecasts for the Asian Monsoon Using 16 Atmosphere-Ocean Coupled Models. Part I: Climatology. J Clim. 2012, 25(1): 39-64.
- Sikka DR. Monsoon droughts in India. Joint COLA/CARE Report, 2, Center for Ocean-Land-Atmosphere Studies (COLA), Calverton, Maryland, USA. 1999.
- Grimm AM, Sahai AK, Ropelewski CF. Interdecadal varia- tions in AGCM simulation skills. J Clim. 2006, 19(14): 3406- 3419.
- Krishnamurti TN, Bedi HS, Subramaniam M. The summer monsoon of 1987. J Climate. 1989, 2: 321–340.
- Palmer TN, Brankovic C, Vitervo P, Miller MJ. Modelling in- terannual variations of summer monsoon. J Climate. 1992, 5: 399-417.
- Sperber KR, Palmer TN. Interannual tropical rainfall vari- ability in general circulation model simulations associated with the atmospheric model inter-comparison project. J Cli- mate. 1996, 9: 2727-2750.
- Fennessy MJ, Shukla J. GCM Simulations of active and break monsoon periods. Proc. MONEG International Conference on Moonsoon Variability and Prediction, Trieste, Italy, 1994 9-13 May.
- Krishnan R, Fenneessy MJ. GCM Simulations of Intrasea- sonal Variability in the Indian Summer Monsoon. COLA Tech- nical Reports – 40, 54 pp. 1997.
- Mujumdar M, Krishnan R. GCM simulation of the contrast- ing Indian summer monsoon of 1987 and 1988. IITM Res. Rep., 89, Indian Institute of Tropical Meteorology, Pune, 33 pp. 2001.
- Gadgil S, Rajeevan M, Nanjundiah R. Monsoon prediction – Why yet another failure? Current Sci. 2005, 88(9): 1389-1400.
- Krishnan R, Ramesh KV, Samala BK, Meyers G, Slingo J et al. Indian Ocean-Monsoon coupled interactions and impending monsoon-droughts. Geophys Res Lett. 2006, 33(8): L08711.
- Ashok K, Guan Z, Saji NH, Yamagata T. On the individual and combined influences of the ENSO and the Indian Ocean Dipole on the Indian Summer Monsoon. J Climate. 2004, 17: 3141-3154.
- Behera SK, Krishnan R, Yamagata T. Unusual ocean-atmo- sphere conditions in the tropical Indian Ocean during 1994. Geophys Res Lett. 1999, 26(19): 3001-3004.
- Solomon A, Newman M. Reconciling disparate twenti- eth-century Indo-Pacific ocean temperature trends in the in- strumental record. Nat Clim Change. 2012, 2: 691-699.
- Yamagata T, Behera SK, Luo JJ, Masson S, Jury M et al. Cou- pled ocean atmosphere variability in the tropical Indian Ocean. Earth Climate: The Ocean-Atmosphere Interaction, Wang C, Xie SP, Carton JA (eds.), Geophys. Monogr., 147, AGU, Washington D.C. 2004, 189-212.
- Annamalai H, Murtugudde R. Role of the Indian Ocean in regional climate variability. In Earth Climate: The Ocean-Atmo- sphere Interaction, Wang C, Xie SP, Carton JA (eds.), AGU Geo- physical Monograph. 2004, 147: 213-246.
- Saji NH, Goswami BN, Vinayachandran PN, Yamagata T. A dipole mode in the tropical Indian Ocean. Nature. 1999, 401: 360-363.
- Webster PJ, Moore A, Loschnigg J, Leban M. Coupled ocean-atmosphere dynamics in the Indian Ocean during 1997- 98. Nature. 1999, 40, 356-360.
- Murtugudde RG, McCreary JP, Busalacchi AJ. Oceanic pro- cesses associated with anomalous events in the Indian Ocean with relevance to 1997-1998. J Geopys Res. 2000, 105(C2): 3295-3306.
- Meyers G, Mcintosh P, Linda P, Pook M. The years of El-Nino, La-Nina and interactions with the tropical Indian ocean. J Cli- mate. 2007, 20: 2872-2880.
- Luo JJ, Masson S, Behera SK, Sakuma H, Yamagata T. Suc- cessful experimental forecast of 2006 Indian Ocean Dipole event. APCC Newsletter. 2006, 1(3): 6-7.
- Slingo JM, Annamalai H. 1997: The El Niño of the century and the response of the Indian summer monsoon. Mon Wea Rev. 2000, 128(6): 1778–1797.
- Bjerknes J. Atmospheric teleconnections from the equato- rial Pacific. Mon Wea Rev. 1969, 97(3):163-172.
- Goswami BN, Xavier PK. Dynamics of ‘internal’ interannu- al variability of Indian summer monsoon in a GCM. J Geophys Res. 2005, 110: D24104.