The Once and Future Pulse of Indian Monsoonal Climate

Introduction

The climate of South Asia is dominated by the monsoon, which returns with remarkable regularity each summer and provides the rainfall needed to sustain over 60% of the world’s population. The vastness of the Asian continent and the unique configuration of the Tibetan Plateau means that the Asian summer monsoon is the most vigorous and influential of all the monsoon circulations (Pant and Rupa Kumar, 1997).The subsistence of India's burgeoning population and its rapidly surging economy is intricately tied to the monsoonal climate variability, more so in coming decades. Spreading throughout the Ganges region, Indians pursue a thriving agricultural lifestyle in the fertile alluvial plain that by many accounts is the nation's lifeblood. Variability in agriculture output, driven by the year-to-year strength of the summer (June to September) monsoon rains, tend to have an exaggerated economic and societal consequences. Temperature variability also has societal consequences, especially with a large poor population and infrastructure. Hot and cold spells routinely lead to several deaths every year ( ), in addition to exerting a strong impact on crop health and yield. Thus understanding the variability of the monsoonal climate is of great importance to devise sustainable policies, particularly in the wake of changing climate.

The Indian monsoon has exhibited variability on inter-annual and inter-decadal time scales. The interannual variability is largely driven by the El Niño Southern Oscillation (ENSO) phenomenon in the tropical Pacific (5,6,7) which modulates the strength and location of the Walker circulation and consequently the monsoon winds and the convection over Indian subcontinent. Antecedent land conditions over Eurasia (xxx) and also tropospheric conditions (xxx) have also been shown to affect monsoon variability on a year to year basis. Interdecadal variability of the monsoon has been linked to the decadal variability of the ENSO phenomenon (xxx) and also to that of the land features (Xxx).

Warming climate into the future due to increased green houses gases (GHG) is also shown (IPCC, AR4) to influence the monsoon climate variability – both precipitation and temperature. This could modulate the known monsoon variability mechanisms mentioned above, in unpredictable ways.Climate change will, in all likelihood, predispose India to enhanced threats from natural hazards linked to the atmosphere and oceans, besides stressing the availability of water and health of our key natural and managed ecosystems. There is therefore, an urgent need to undertake systematic research on changes that are happening, and likely to happen, over the Indian region along with their linkages to the global changes and also the impacts on diverse sectors of socio-economic activity.In this study we attempt to provide broad insights into these aspects from a suite of climate model simulations of present and future Indian monsoon climate.

The paper is organized as follows. (i) A detailed description of the various data sets used in this study are first presented. (ii) Results describing the observed variability of the monsoon climate and the ability of the climate model simulations to capture them is next presented. (iii) This is followed by future projections and the impact on socio-economic activities, mainly crop yield and public health, the two key sectors of India’s growth. (iv) Summary and discussion of the findings in a broader policy context concludes the paper.

Data

Monthly data for several variables from a suite of 22 coupled atmosphere ocean general circulation models (CMIP) with a total of 48 ensemble members for the 20th Century Climate (20c3m) and for the future climate simulations based on SRES-A1B green house gas (GHG) emission scenarios (IPCC 2001) are used in this study[1]. Most of the 20c3m experiments were integrated according to observed anthropogenic forcing from the late 19th century and ending in 2000.

A high resolution regional climate model of Hadley Centre for Climate Prediction of UK Meteorological Office, known as PRECIS, is run over the Indian subcontinent domain, with two sets of lateral and surface boundary forcing data generated from the Hadley Centre AOGCM corresponding to the present (1961-90) with observed GHG forcing, referred in the paper as ‘baseline’, and for the future (2071-2100) following SRES-A2 GHG emission scenarios, referred henceforth as A2. Details of the regional model domain, its simulated seasonal climatology and extremes are discussed in Kumar et al (2005).

Observed precipitation CMAP (Xie and Arkin ..) and HadCRU3v land and sea surface temperature data set from University of East Anglia ( ) are used to verify coupled model simulated climatologies. Yearly all-India rice yield data during 1961-2006 is obtained from FAO ( ). All India monthly rainfall based on a network of 306 stations covering the entire country (Parthasarathy et al., 19xx), maximum and minimum temperatures for the period 1871 – 2006 are obtained from Indian Institute of Tropical Meteorology.[2] The all-Indian summer (June-Sep) monsoon rainfall index (AISMR) is computed from the monthly rainfall.This is probably the longest and most extensively used index for understanding the inter-annual fluctuations of the Indian monsoon rains and also as a proxy for several other climate studies (e.g., xxx).

Results

Observed Changes

The percent anomalies of AISMR during 1871-2006 shown in Fig.1a reveal monsoon variability on inter-annual and inter-decadal time scales – but, interestingly there is no apparent long term trend. Occasional excursions of monsoon rainfall from the mean lead to large scale droughts and floods covering large parts of the country. These inter-annual positive(negative) fluctuations generally co-occur with the warm(cold) ENSO events in the Pacific Ocean as can be seen in this figure. This monsoon-ENSO relationship is fairly robust and is a continued source for monsoon forecasting (xxx). In recent years this relationship is seen to have weakened somewhat (Kumar et al., 1999a,b). They argue for the role of warming temperature trend over Eurasian region in this weakening. Large rainfall departures are also observed in the absence of ENSO events – which are attributed to the chaotic internal dynamics (xxx) and other surface boundary forcings (xxxx). .

Low frequency variability of the monsoon is quite clear from this figure – in that, during the middle part of the 20th century the monsoon shows a higher mean relative to the other periods. This epochal nature is believed to be driven by the corresponding epochal variations of ENSO (Kriplani, xxx) and North Atlantic forcings (Goswami xxx). The recent decades of below normal rainfall is interesting as it seems contrary to expectations from substantial warming trends in the Indian Ocean surrounding India (IPCC4).

Though AISMR does not show any perceptible long-term trend during the past century, there is however, an evidence for a change in the rainfall characteristics in recent decades. Goswami et al (2007) show an increasing trend in the magnitude and frequency of extreme rainfall events during post 1950 period in a large region of Central India. They also show a corresponding decrease in the frequency of low rainfall events. This redistribution of rainfall character has serious implications for flood and water resources management.

Surface air temperature is the most important meteorological variable for identifying global warming footprint. Fig. 1b shows the all-India mean annual surface air temperature anomalies with respect to 1961-90 climatology. The steep increasing trend in recent decades is quite dramatic and conforms to rapid trends in global surface temperatures (IPCC AR4).

Future Projections of Monsoonal Climate

Several studies in the past have assessed the performance of different climate models in an uncoupled observed SST forced (generally known as AMIP) mode (xxx Gadgil and Sajani, xxx, Kumar et al., 2005; Wang et al., 2005) and coupled (i.e., CMIP) mode (Annamalai et al., 2007; Kripalani et al., 2007) – in their ability to simulate the climatological features of the Indian monsoon and its ENSO teleconnections. These efforts also attempted to identify a unique subset of models that can be used with confidence for seasonal forecasting and to assess the future changes in the monsoonal climate over India. Unfortunately, these studies did not result in a unique model subset as the metrics used in the selection depends greatly on the choice of spatial and temporal scales of monsoon variability of interest – rendering such a model selection exercise futile. Thus, the multi-model ensemble seems to be the best approach to assess future changes. This is strongly advocated by IPCC4 (Chapters 8 and 10). They mention that – “the reason to focus on the multi-model mean is that averages across structurally different models empirically show better large-scale agreement with observations, because individual model biases tend to cancel (see Chapter 8). The expanded use of multi-model ensembles of projections of future climate change therefore provides higher quality and more quantitative climate change information compared to the TAR.Even though the ability to simulate present-day mean climate and variability, as well as observed trends, differs across models, no weighting of individual models is applied in calculating the mean.” Theuse of multi-model ensembles has been shown in other modelling applications to produce simulated climate features that are improved over single models alone (see Palmer et al., 2004; Krishnamurty et al., 2000; Rajagopalan et al., 2001). Consequently, in this paper we use the multi-model ensemble to assess the future monsoon climate change and refrain from selecting a model subset.

The right panels on Fig. 2 show the simulated and observed annual cycles of rainfall and temperature averaged over India and the two box plots below depict the standard deviation of the simulated rainfall and the correlation with NINO4 SSTs (an ENSO index) - the observed values are shown as stars. It can be seen that almost all the models could simulate the annual cycle in rainfall and temperatures very well. The ensemble mean annual cycle of rainfall is slightly underestimated and the temperature slightly overestimated compared to observations. However, the median values of standard deviation of the monsoon seasonal rainfall and ENSO correlations (horizontal lines in the box plots) are very close to the observed values, albeit, there is a large spread among the models. The spatial patterns of the ensemble mean of simulated monsoon rainfall and annual temperatures are remarkably similar to the observed patterns (Fig.S1) despite substantial differences in the simulated rainfall patterns in the individual models (figures now shown). Thus, the multi-model ensemble mean is a good representation of the observed monsoonal climate features. This provides further credence to the multi-model ensemble approach advocated by IPCC4.

The simulated annual temperature and monsoon rainfall, averaged over the Indian land grids from the ensembles for the 1901-2098 period are shown in Fig 2 a and b, respectively. The ensemble mean (black line) captures the observed variations (blue line) quite well. The ensemble mean suggest an increase of 2C by mid 21st century and over 3.5C by the end of the next century in the average annual temperature. Furthermore, the ensembles are tightly bunched, indicative of close agreements between the models. This substantial increase in average temperature can have a significant manifestation in the variability of extremes that could be detrimental to socio-economic activity. This will be examined in the following section. The ensemble member corresponding to the Hadley Center AOGCM (red line) appears to closely follow the multi-model ensemble mean.

Unlike temperature, the simulated monsoon rainfall has a larger spread between the ensemble members and the ensemble mean rainfall shows a modest increase of about 8-10% by the end of the 21st century. Much of this increase seems to occur during the second half of the century. This increase can be seen to be within the limits of the variability exhibited by the observed monsoon rainfall in the 20th century. We examined the variability and seasonality of the projected rainfall (Fig S2a,b), an important aspect of rainfall. The ensemble mean annual cycle of rainfall for the present (1961-1990) and future (2070-2098) indicate nearly a 20% increase in the rainfall during May and October months in future. This is suggestive of an extended monsoon season in the future and consistent with recent studies (Meehl et al, 2006, Ramanathan et al 2007)) that show increased pre-monsoon rainfall under increased GHG conditions over India.

The standard deviation of monsoon rainfall and the strength of the monsoon-ENSO correlation during the three periods (1961-90, 2041-60 and 2070-2098), shown as box plots in Fig S2b and c – indicate a slight increase (~10%) in the standard deviation in the future compared to the present period. The monsoon-ENSO correlation appears to be quite stable. This assessment, based on the entire suite of IPCC model runs, is in variance with earlier studies that have indicated considerable changes in monsoon rainfall variability and its ENSO teleconnection in future (Meehl and Arblaster, 2003; Kripalani et al , 2007, Annamalai et al, 2007 and Ashrit et al, 2003). This difference is due to the fact that these studies used a single or a small subset of models.

Spatial patterns of expected ensemble mean change in rainfall and surface temperatures at the end of the 21st century relative to present are shown in Figure 3. It is clear that both oceans and land seem to show a substantial warming in the future – with land areas showing more warming than the oceans. Over the oceans, generally the warming is more in the tropical regions compared to extra tropics. In the Pacific ocean, an El Nino like warming pattern – with a warmer central and eastern tropical Pacific is quite conspicuous. There is a debate as to the a nature of SST pattern in the tropical Pacific (Vecchi et al., 2007) in a warmer futre climate, but the consensus from CMIP seems to be clearly towards an El Nino like warming. The associated positive rainfall change (Fig 3b) in the central and eastern Pacific corroborate this El Nino like SST pattern in the future. In addition, there is a projected rainfall increase over the Indian region in conjunction with the central and eastern tropical Pacific ocean warming –contrary to the reduction in Indian monsoon rainfall associated with El Nino (xxxx).

We offer some insights into this paradox from examination of atmospheric circulation features related to the monsoon. The evolution of meridional wind index (Goswami ) over the Indian subcontinent domain (shown in Fig 4a) from CMIP ensembles and the ensemble mean for the 2000-2098 period is shown in Fig 4c. The meridional wind index is defined as the vertical meridional wind shear between 850 and 200 hPa and captures the strength of the monsoon rainfall and in particular, the reverse Hadley circulation – the key monsoon feature (Goswami xx). As can be seen, there is no trend in the strength of ensemble mean meridional wind index during the entire 21st century suggesting no perceptible change in the strength of dynamical monsoon circulation. Mid-tropospheric (500hPa) temperature gradient between the Indian land region and the Indian Ocean, a robust measure of the strength of the monsoon circulation than the surface gradient (Goswami et al., 2007), also indicate no trends in this century (figure not shown). These suggest that the strength of the monsoon circulation is likely to be steady into the future despite El Nino like conditions in the tropical Pacific. We computed the future change in area averaged total water content in the troposphere over the Indian monsoon region (shown in Fig 4d) – which shows a substantial increasing trend. The enhanced moisture availability over the Indian region, likely due to increased temperature and evaporation (IPCC4), seems to be main factor in the projected increase in the monsoon rainfall (Fig 2a).Sensitivity experiments from single climate models, of monsoon rainfall changes in the future under increased GHG forcing also suggest enhanced thermodynamical changes to monsoon rainfall (i.e., more moisture available in the atmosphere) than dynamical (Meehl and Arblaster, 2003, Sugi …, Dariaku and ..,).

Societal Implications

Indian society with modest infrastructure is highly vulnerable to even slight variations in weather extremes. In fact, climate change could have a major tipping in the socio-economic aspects. Therefore, it is important to understand the changes in weather extremes under future climate projections. This is further highlighted by an increasing occurrence of extreme events in recent periods such as - floods across many states during 2005-07, the prominently a 944 mm of rainfall in one day on 26 July, 2005 at Mumbai; high human mortality due to heat waves in one of the southern states in 2003, are to name a few. These events extracted a heavy toll on the economic and social health of India. Some recent studies have examined changes in extremes under future climate projections, using coarse resolution global climate models (e.g., May, , Sun et al 2007).High resolution models are ideal to investigate the variability in extreme weather conditions as they can capture the synoptic scale regional pecularities consistent with global conditions. As mentioned earlier, we used the HadleyCenter’s regional climate model (PRECIS) with the lateral and surface boundary conditions from coarse resolution global model of HadleyCenter (HadCM3). In particular, we focus on the extremes of maximum and minimum temperatures and the character of daily rainfall over India. The simulated rainfall and temperature values from HadCM3 are shown as red lines in Fig. 2. As can be seen this model closely captures the ensemble mean of CMIP simulations and thus, can be considered as its representative. The rainfall and temperature climotologies generated based on 30 year (1961-1990) baseline simulations reproduce the observed features very well (Fig. S1 c,d). Furthermore, the regional model is able to better resolve the orographic rainfall over Indiathan the coarse resolution global coupled models (Fig S1xx). The regional model over estimated the rainfall in June and generally has a cold bias compared to observations (xxx Fig Sxxx) – also see Kumar et al, 2005 for details on the model set up and its performance over India.