Original article
Simulation of Indian summer monsoon rainfall and its intraseasonal variability in the NCAR climate system model M. Lal á G. A. Meehl á J. M. Arblaster Abstract The broad climatological features associated with the Asian monsoon circulation, including its mean state and intraseasonal and interannual variability over the Indian subcontinent as simulated in the National Center for Atmospheric Research (NCAR) global coupled climate system model (CSM) in its control reference experiment, are presented in this paper. The CSM reproduces the seasonal cycle as well as basic observed patterns of key climatic parameters reasonably well in spite of some limitations in simulation of the monsoon rainfall. However, while the seasonality in rainfall over the region is simulated well, the simulated area-averaged monsoon rainfall is underestimated to only about 60% of the observed rainfall. The centers of maxima in simulated monsoon rainfall are slightly displaced southward as compared to the climatological patterns. The cross-equatorial ¯ow in simulated surface wind patterns during summer is also stronger than observed with an easterly bias. The transient experiment with a 1% per year compound increase in CO2 with CSM suggests an annual mean area-averaged surface warming of about 1.73 °C over the region at the time of CO2 doubling. This warming is more pronounced in winter than during the monsoon season. A net increase in areaaveraged monsoon rainfall of about 1.4 mm day±1, largely due to increased moisture convergence and associated convective activity over the land, is obtained. The enhanced intraseasonal variability in the monsoon rainfall in a warmer atmosphere is con®ned to the early part of the monsoon season which suggests the possibility of the date of onset of summer
monsoon over India becoming more variable in future. The enhanced interannual and intraseasonal variabilityin the summer monsoon activity over India could also contribute to more intense rainfall spells over the land regions of the Indian subcontinent, thus increasing the probability of extreme rainfall events in a warmer atmosphere. Key words Global climate model á Indian monsoon á Intraseasonal variability á GHG forcings á Regional climate change
Introduction
Simulation of the Asian monsoon circulation has proven to be a critical test of the ability of global climate models (GCMs) in simulating the tropical climate variability (Dumenil 1998; Webster et al. 1998). Unfortunately, the simulation of seasonal rainfall as well as its spatial and temporal variability over the Indian subcontinent has remained rather poor in most GCMs primarily due to coarse horizontal resolution which restricts the representation of the complexity of topography and coastlines and the limitations in physical parameterization of sub-grid scale processes (Lal et al. 1998). The numerical experiments performed with the state-of-the-art-coupled ocean±atmosphere general circulation models (A±O GCMs) are able to realistically simulate the planetary and regional scale circulations associated with the Indian summer monsoon. These models simulate the large-scale climatological features in response to global forcings such as land±sea distribution, orography and differential heating reasonably well (Cess et al. 1993; Gates et al. 1993 among others). Received: 23 February 1999 / Accepted: 9 August 2000 / Considerable improvement in the simulation of presentPublished online first: 17 October 2000 day climate on regional scales has taken place in recent ã Springer-Verlag years (Intergovernmental Panel on Climate Change 1996). There have also been re®nements made in coupled global M. Lal (&) climate models to make them more self-deterministic, Centre for Atmospheric Sciences, e.g. the use of self-determined cloud optical properties to Indian Institute of Technology, replace the prescribed values (Senior and Mitchell 1993) New Delhi 110016, India or the sensitivities of a coupled system to changes in e-mail:
[email protected] cloud amount and albedo feedbacks have been explored G. A. Meehl á J. M. Arblaster (Stockdale et al. 1994; Meehl and Washington 1995). A±O National Center for Atmospheric Research, Boulder, Colorado, USA GCMs are now being used for studies related to the natural DOI. 10.1007/s101130000017
Reg Environ Change (2000) 1:163-179
163
Original article
variability of the climate system and its response to changes in anthropogenic radiative forcings (Palmer et al. 1992; Latif et al. 1998). Regional peculiarities, e.g., land±sea thermal contrast, orographic features, vegetation characteristics and inland water basins, play an important role in the establishment of summer monsoon over the Indian subcontinent. The thermal structure of the adjoining oceanic areas ± the Arabian Sea, the Bay of Bengal and the South Indian Ocean ± and its temporal variations have a modulating in¯uence on the monsoon. The onset and retreat of summer monsoon over the Indian subcontinent are associated with rather abrupt changes in the atmospheric general circulation. Signi®cant interannual variations in the dates of onset of monsoon rains and intraseasonal variability in the observed monsoon rainfall are also displayed over this region (Parthasarathy et al. 1994). A high degree of natural monsoon variability results in recurrent droughts and ¯oods. In India, drought risk represents potentially the most serious impact of climate change. In this respect, the future projections of changes in monsoon rainfall variability over the Indian subcontinent are very crucial. The summer monsoon season over India is punctuated by intermittent emergence and subsequent decay of wellde®ned synoptic-scale northwestward-moving disturbances. These synoptic-scale disturbances, commonly called monsoon depressions, have a time scale of several days and a length scale of about 500±1000 km and owe their existence to not only the larger and longer lasting components of the monsoon circulation but also the smaller and more transient subsystems. An analysis of seasonal records of daily rainfall at central Indian stations provides many components with time scales ranging from 7 to over 40 days and thus suggests the multiplicity of scales in the monsoon circulation (Bhalme et al. 1987; Nanjundiah et al. 1992). In order to reproduce the seasonal rainfall of right magnitude in A±O GCMs and to understand the intraseasonal variability of monsoon under anthropogenic radiative forcing conditions, one must examine whether the various scales of monsoon circulation are resolved appropriately in the current generation of coupled climate models. We examine here the skill of NCAR's CSM in simulating the annual cycle of observed climatological features associated with the Asian monsoon circulation in the 300year-long control simulation for the current climate forcing conditions. A comparison of the observed and simulated broad scale patterns of mean sea level pressure, surface air temperature, surface winds and precipitation over the Indian subcontinent is made. The onset and advancement of monsoonal activity and temporal rainfall variations as simulated by the model during the summer monsoon season over India are assessed. In this paper we also project the likely changes in climatology of the Indian subcontinent and in the intraseasonal and interannual variability of the summer monsoon as simulated by the CSM at the time of doubling of atmospheric CO2. 164
Reg Environ Change (2000) 1:163-179
The climate system model The NCAR's CSM is a comprehensive model of the physical climate system. The atmospheric component employs a spectral dynamical representation with a horizontal truncation of T42 (approximately 2.8 ´ 2.8 equivalent Gaussian grid) and 18 levels in the vertical. The model includes the deep convection scheme of Zhang and McFarlane (1995) and the shallow convection scheme of Hack (1994). The radiation scheme includes a diurnal cycle wherein the radiation computations are performed at each hour. Cloud properties (e.g., cloud fraction, cloud water and cloud particle size) are diagnostic (Kiehl et al. 1998). The boundary layer scheme included herein is an extension of that developed by Holtslag and Boville (1993). The land surface scheme (Bonan 1998) treats the energy, momentum, water and CO2 exchange between the atmosphere and the land surface. The ocean model has 45 vertical levels (12.5 m thick at the surface) and a horizontal resolution of 2.4° in longitude and between 1.2 and 2.3° in latitude (®nest at the equator and in high latitudes and coarsest in mid-latitudes). The dynamic±thermodynamic sea ice scheme (Weatherly et al. 1998) is based on the cavitating-¯uid rheology of Flato and Hibler (1992) and is formulated on the same grid as the ocean model. The various system components of the model are coupled through a `¯ux coupler' (Bryan et al. 1996) which accommodates heterogeneity of surface types at the grid scale of atmospheric model. For further details of the model, the reader is referred to Boville and Gent (1998). The individual component models are spun up in uncoupled mode in a `phased-in' manner (Boville and Gent 1998) prior to the start of two 300-year-long coupled experiments, the ®rst being a reference control experiment and the other wherein the atmospheric CO2 is allowed to increase at a compound rate of 1% per year into the next century. These integrations have been performed without `¯ux adjustments'. The spin up and initialization procedures (Bryan 1998) are so chosen that the inherent biases in the component models that lead to ¯ux incompatibilities are overcome ef®ciently and there is essentially no drift in global mean surface temperature simulated in the reference control experiment through the 300 years of the model integration.
Study region and data used The geographic region of interest in this study is the summer monsoon area bounded by latitudes 5±35°N and longitude 65±100°E (Indian subcontinent and adjoining seas). The simulated climatology is constructed by averaging over a 10-year period (1981±1990) representative of the present-day climate in the control reference experiment. For all validation purposes, we have analyzed the data for the northern hemisphere winter (December, January and February: DJF) and summer monsoon season (June, July, August and September: JJAS) in addition to the
Original article
annual means. The observed climatology for mean sea level pressure, surface air temperature and surface winds used in this study is based on an average of 23 years of data from National Center for Environmental Prediction (NCEP) re-analyses. The observed precipitation climatology used here is that of Legates and Willmott (1990). The observed precipitation climatology is likely to be less realistic in and around the mountainous regions and over the oceans due to sparsity of observations. Daily data for selected climatic variables have also been analyzed for the period from 1 June until 30 September during each of this 10-year period to explore the intraseasonal and interannual variability in Indian summer monsoon over the study region. Two regions have been selected for obtaining the area-averaged values of key climatic elements for this purpose. These are central India (land points only con®ned to latitudes 18±30°N and longitudes 68±90°E) and all India (land points only con®ned to latitudes 5±35°N and longitudes 65±100°E). For future projections of change in key climatic elements, we have used the data from the transient CO2 experiment over a 10-year period representative of the time of doubling of CO2 in the atmosphere. In the next section, we shall present the skill of CSM in simulating the annual mean and seasonal mean climatology over the Indian subcontinent. Much of our emphasis in model validation is con®ned only to surface parameters. The onset and withdrawal of the summer monsoon over the Indian subcontinent as seen in the area-averaged daily rainfall time series will also be discussed. The upper air features have been given only a cursory look in the model output for the purpose of the study presented here as many of the broad-scale upper air circulation features have already been examined by other workers (e.g., Boville and Fig. 1 Observed (a) and simulated (b) seasonal migration of tropical convergence zone (TCZ) in Africa±Asia monsoon region
Hurrell 1998; Raphael 1998) elsewhere. Relationships between Indian monsoon rainfall and tropical Paci®c sea surface temperatures (SSTs) in the CSM are described by Meehl and Arblaster (1998).
Comparison of model simulation with observed climatology The NCAR CSM has been able to realistically simulate important aspects of the large-scale northern and southern hemispheric climatology as con®rmed by veri®cation with NCEP re-analyses. The monsoon is considered as a manifestation of the seasonal migration of the planetary-scale equatorial trough or tropical convergence zone (TCZ). The simulated seasonal migration of TCZ in the monsoon regions of the Africa±Asia sector is close to the climatological circulation features (Fig. 1). The location and strength of the upper air westerly jet over the northern Indian subcontinent during winter and the tropical easterly jet during the monsoon season are also found to be in fair agreement with observed climatology. In view of this, we shall focus here only on the surface climatology of the Indian subcontinent. Mean sea level pressure and surface winds The annual cycle of the mean sea level (MSL) pressure has the largest amplitude over the Asian continent. Over the Indian subcontinent, the MSL pressure distribution is characterized by high pressure over land and low pressure over the adjoining Indian ocean during winter and the reverse during the summer monsoon season. Figure 2 depicts the spatial distribution of observed and simulated MSL pressure over the Indian subcontinent in the months of January and July. The simulated MSL pressure
Reg Environ Change (2000) 1:163-179
165
Original article
Fig. 2 Spatial distribution of mean sea level pressure (hPa) over Indian subcontinent in January and July a as in observed climatology and b as simulated by the model (10-year mean)
®elds) between the simulated and the observed MSL pressure for January and July. The approach for calculating the SPCC and RMSE is similar to that described in Wigley and Santer (1990) and the calculations have been performed on distribution is fairly realistic in both the representative the grided data at T-42 resolution. The SPCCs between the simulated and the observed MSL pressure in the region months except over the Himalayan region in January. bounded by 5±30°N and 65±100°E (note that the Tibetan The seasonal heat low simulated by the model over northwest India in July is also rather weak as compared Plateau has been excluded in the SPCC and RMSE calcuto observed climatology and leads to a distortion in the lations as the observed climatology for this region is questionable and the model simulations are believed to be normal orientation of the monsoon trough along the poor due to unresolved orographic features) are estimated Indo-Gangetic plains. With a view to quantitatively assessing the skill in model to be 0.76 and 0.83 for January and July months respecsimulation, we have computed spatial pattern correlation tively. The RMSEs are considerably lower in January coef®cients (SPCC ± a measure of similarity of the pattern (2.1 hPa) compared to those in July (3.4 hPa). The simulated surface winds and those inferred from NCEP structure over the region) and root mean square errors (RMSE ± a measure of the absolute error between the two re-analyses for January and July are depicted in Fig. 3. As in
166
Reg Environ Change (2000) 1:163-179
Original article
Fig. 3 somewhat stronger than observed over the region. The Spatial pattern of surface winds (m s±1) over Indian subcontinent in SPCCs between the simulated and the observed surface January and July a as in observed climatology and b as simulated by zonal winds are estimated to be 0.87 and 0.83 for January the model (10-year mean)
observed climatology, the simulated surface winds across the equator over the Indian Ocean are southeasterlies and become southwesterlies over the Arabian Sea, peninsular India and Bay of Bengal during the monsoon season. The model is also able to reproduce the observed features of the monsoon circulation along the Somali coast (Somali jet off the east African coast) and the seasonal reversal in surface winds over the Indian subcontinent. The simulated trade winds are, however, stronger than observed in the equatorial Indian Ocean. The cross-equatorial ¯ow in simulated surface wind patterns during summer is also stronger than observed, with an easterly bias along the eastern sector of the Indian Ocean. The 850-hPa winds over the land are
and July respectively. The RMSEs in January and July range from 1.8 to 3.1 m s±1. In general, the broad-scale features of the monsoon circulation have been captured by the CSM reasonably well. Surface air temperature Figure 4 compares the spatial distributions of observed and simulated surface air temperatures on an annual mean basis and during the winter and monsoon. The simulated surface air temperature distribution is fairly realistic in both the seasons when compared with observed climatology. The seasonal reversal in north to south thermal gradient over the Indian subcontinent is reasonably simulated by the model. The spatial range of temperature over India Reg Environ Change (2000) 1:163-179
167
Original article
(~8.9 °C) simulated by the model during July is fairly close to that recorded (~9.4 °C) in the observed climatology. The hottest region is the western margins of India where the observed mean temperatures of 35 °C in July are reproduced by the model. The land to sea temperature gradient during the monsoon is regarded as the main driving force behind the monsoon circulation over the Indian subcontinent (Meehl 1994). With a view to examining this aspect in the CSM simulation, we have computed seasonal mean surface temperatures (average for eight grid points) over northwest India where the observed heat low dominates during the monsoon season and over the east Arabian Sea area near the east coast of Somalia from where the monsoon winds turn in the northern hemisphere. The difference in simulated surface temperature between the two selected regions is 2.3 °C as against the observed temperature difference of 2.1 °C. Thus, the model is able to produce the observed thermal force at the surface to drive the summer monsoon circulation. The SPCCs between the simulated and the observed surface temperatures are estimated to be 0.91 and 0.89 for winter and monsoon seasons respectively. The RMSEs are signi®cantly low both in winter (1.3 °C) and in the monsoon (1.8 °C). This suggests that CSM has a fairly high skill in reproducing the observed seasonal surface temperature patterns over the Indian subcontinent. Precipitation About 75% of the total annual rainfall over India occurs during the summer monsoon season. During winter, there is some rain over the northern parts of India (associated with western disturbances) and over the southern peninsular India (associated with northeast monsoon). The model is able to realistically simulate the seasonality in rainfall as observed over the Indian subcontinent. The spatial pattern in observed mean monsoon precipitation is, however, fairly complex. The heaviest rains occur over the hilly states in the northeast and along the mountainous west coast (the Western Ghats and the Konkan coast). In addition to the primary monsoon rainbelt over the Indian subcontinent and the secondary one over the equatorial Indian Ocean, signi®cant rainfall occurs over the foothills of the Himalayas. There are major shifts in the simulated locations of rainfall maxima relative to observed patterns during both the winter and monsoon seasons (Fig. 5). This is in spite of the fact that the seasonal cycle in simulated surface-speci®c humidity is consistent with observations, and the moisture convergence over the land regions during the monsoon is well reproduced by the model. The model does not realistically simulate the orographic pattern near the Himalayan foothills. The observed rainfall maxima along the west coast during the monsoon season have a southward shift in the simulated rainfall pattern. The sharp gradient in observed monsoon rainfall from the west to east coast of peninsular India is also not realistically simulated by the model. The area-averaged (land points only) long-term mean climatological monsoon rainfall over India (based on data for the period 1871±1990) is 85.2 cm, with a standard deviation of 8.4 cm (coef®cient of variation is 9.9%). 168
Reg Environ Change (2000) 1:163-179
The observed mean summer monsoon rainfall over India during the recent 10-year period from 1979 to 1988 has been reported to be 81.8 cm, with a standard deviation of 9.8 cm (Parthasarathy et al. 1994). In contrast, the simulated area-averaged (land points only) mean monsoon rainfall over India is only 49.4 cm, with a standard deviation of 3.8 cm (coef®cient of variation is 5.7%). Thus, the CSM is able to simulate only about 60% of the observed rainfall over the land regions of India. A major factor for the de®cit in simulated monsoon rainfall over the land regions seems to be the southward shifts in major rainbelt over the Indian Ocean (see also Meehl and Arblaster 1998). This can be best illustrated in the temporal variation of the observed and simulated monthly mean precipitation over the Asia±Paci®c (70±140°E) region (Fig. 6). While the observed rainfall maxima during July± August is located at about 20°N, it is located close to 10°N in the model simulation. The model also simulates unrealistically excessive precipitation at about 10°S during February±May. Thus, it is obvious that, contrary to the fact that the continental TCZ is dominant over the Indian subcontinent in reality, the rain belt in the equatorial Indian Ocean region dominates the continental monsoon rain belt in the CSM simulation. The transition from such a simulation to one in which continental TCZ (or the monsoon trough) dominates needs to be attained with further improvements in convective parameterization and other physical processes in the model. Understanding the mechanisms of climate variability on a time scale of months to several years is a prerequisite for any prediction of climate. Incorporation of many of the climate feedbacks into CSM has substantially improved its skill in reproducing the broad features of observed natural variability of the climate system in its control simulation. It has been shown to represent most major features of the monsoon system in terms of the interannual variability and connections to the El NinÄo/Southern Oscillation (ENSO) phenomenon in the tropical Paci®c (Meehl and Arblaster 1998). The Indian summer monsoon has linkages to not only global climate but also regional energetics and its own inherent dynamics. To examine the progression of monsoonal currents and associated rainfall over the Indian subcontinent as well as its interannual and intraseasonal variability, we analyzed daily rainfall data for the 10-year period of the control experiment performed with CSM. We shall discuss our ®ndings on this aspect in the next section.
Interannual and intraseasonal variability in monsoon rainfall The Asian summer monsoon is one of the most robust components of the global circulation system. From a global perspective, the Asian monsoon makes the largest contribution to the annual cycle of atmospheric heating, which has a signi®cant impact on the planetary-scale circulation (Nitta 1983; Palmer 1994). It also makes a substantial contribution to the interannual variability of the
Original article
monsoon region (Webster and Yang 1992). The complex nature of the annual cycle of heating in the India±Paci®c region makes the onset of the Indian summer monsoon quite abrupt and there are large intraseasonal variations tropospheric circulation. The centre of the highest inter- between active and weak spells of monsoon rainfall. Sigannual variability of kinetic energy is located in the Asian ni®cant periodicities have been identi®ed in the long time Fig. 4 Spatial distribution of annual mean and seasonal surface air temperature (°C) over Indian subcontinent a as in observed climatology and b as simulated by the model (10-year mean)
Reg Environ Change (2000) 1:163-179
169
Original article
Fig. 5 south to north progressing wave with a recurrence period Spatial distribution of annual mean and seasonal rainfall (mm day±1) of 40 days (Sikka 1980). Its wavelength is estimated to be over Indian subcontinent a as in observed climatology and b as about 20° latitude and its speed is 0.5° latitude day±1. It simulated by the model (10-year mean)
series of observed Indian monsoon rainfall data. An analysis of weekly summer monsoon rainfall shows a 170
Reg Environ Change (2000) 1:163-179
explains about 25% of the variance of the evolution of weekly rainfall as a function of space and time. Active spells of the Indian summer monsoon are associated with
Original article
Fig. 6 Temporal variation of a observed and b model-simulated monthly mean precipitation over Asia±Paci®c (70±140°E) region. Also shown is the temporal variation of simulated monthly mean precipitation at time of doubling of CO2 in the atmosphere (lower panel)
an intense intertropical convergence zone across the monsoon regime over the heated subcontinent. A dominant 30- to 60-day oscillation is also discernible in circulation parameters and cloudiness over the region (TOGA Monsoon Numerical Experimentation Group 1990). The ability of CSM to simulate some of these observed features of intraseasonal and interannual variability in the Indian monsoon circulation is examined here with a view to improving our understanding of the precise and reliable (con®dent) magnitude of regional warming as well as the likely changes in intraseasonal and interannual variability in response to future changes in anthropogenic radiative forcings. We have analyzed the simulated daily data for selected climatic variables from 1 June until 30 September
(122 days) during each of the 10-year periods corresponding to the present-day climate in order to explore the intraseasonal and interannual variability in Indian summer monsoon over central India (land points only con®ned to latitudes 18±30°N and longitudes 68±90°E). Figure 7 depicts the temporal variations in simulated daily values of sea level pressure, surface temperature and total rainfall averaged over central India from 1 June till 30 September for each of the 10 years along with daily mean for the 10-year period (thick line). For comparison purposes, the temporal variations in the daily surface temperature and rainfall as observed over central India during the 10-year period from 1986 to 1995 are depicted in Fig. 8. The area-averaged MSL pressure is lowest at the time when the monsoon circulation has just established itself across India (early July). As simulated by the model, the surface pressure along the monsoon trough is usually at a minimum before the monsoon onset (Rao 1976). Both in observed and simulated data series, the surface temperature starts dropping down due to the presence of cloudiness from mid-June onwards. The rainfall maxima coincides with the peak monsoon activity over the region around mid-July. Although the simulated daily rainfall is lower by an order of magnitude as compared to observed rainfall (Figs. 7 and 8), considerable year-to-year variability is clearly discernible both in simulated and observed areaaveraged daily rainfall. As is observed, both interannual and intraseasonal variabilities in the key climatic elements during the monsoon season simulated by the model are higher over central India which is characterized by intermittent emergence and subsequent decay of synoptic-scale disturbances. The normal onset date of the summer monsoon over the extreme south of the Indian subcontinent is 1 June and it advances over northeast Rajasthan by the ®rst week of July. Within 5 weeks of the onset over the southernmost tip of India, the monsoon normally spreads over the whole of the country. A closer look at the daily rainfall data suggests that CSM simulates the ®rst spell of intense rainfall appearing over the southernmost part of India (5±10°N) during the ®rst week of June. The simulated onset date at 10°N (based on the criteria that rainfall at grid points along 10°N in the region is 2 mm day±1 or more for at least three consecutive days) ranges between 27 May and 6 June during the 10-year period, which is consistent with the observed interannual variability in the onset dates of monsoon rainfall. The maximum monsoon rainfall activity as simulated by the model is, however, con®ned largely at around 15°N latitude throughout the season. The intense rainfall occurs between 10 and 28°N latitudes and to the east of about 75°E. The simulated rainfall is signi®cantly de®cient to the west of 75°E. There is only marginal north and westward advancement of simulated monsoon rains with the progression of the season. Thus, CSM seems to have only limited ability in simulating the south to north progression of observed monsoon rainfall over the Indian subcontinent. A large part of the interannual and intraseasonal variability in monsoon rainfall is associated with the various
Reg Environ Change (2000) 1:163-179
171
Original article
Fig. 8 Temporal variation of area-averaged daily surface air temperature and total rainfall as observed over central India for the period 1 June± 30 September for each of the 10 years during 1986±1995 (thick lines represent trend in 10-year mean)
moist convective clouds and cause intense spells of rainfall over the land regions of central India. The frequency and intensity of these synoptic-scale systems in a particular year largely determine whether the monsoon rainfall has been excess, normal or de®cient. It is, however, not possible to physically resolve these synoptic-scale systems in global climate models with rather coarse horizontal resolution. An alternative approach towards identifying these synoptic-scale disturbances in daily rainfall data is to examine the signi®cant spectral peaks in rainfall through precise mathematical tools such as power spectrum Fig. 7 analysis. Temporal variation of area-averaged daily mean sea level pressure, surface air temperature and total rainfall over central India for the The detrended time series of simulated daily rainfall for period 1 June±30 September for each of the 10 years selected in the the monsoon season (starting from 1 June to 30 Septemcontrol experiment (thick lines represent trend in 10-year mean) ber) averaged over central India has been analyzed in frequency domain to estimate the raw spectral energies. The spectral energy computational procedure is similar to scales of monsoon circulation. It is thus pertinent to that adopted by Krishnamurti and Bhalme (1976). After ascertain whether the dominant scales are resolved apsmoothing with a 3-term weighted average by the Hanning propriately in the CSM control experiment. During the monsoon season, an elongated zone of low pressure called method, the ®nal spectral estimates are calculated from the the monsoon trough is a persistent feature over the Indian raw spectral estimates. The analysis is repeated for each of the ten consecutive years of control simulation and then region with its axis located at about 22°N in the east to about 27°N in the west. The low-pressure systems forming the averaged spectral energies for the selected 10-year over the Bay of Bengal during the monsoon and moving period are calculated. Similar analysis is done for simulated all-India rainfall. Figure 9 shows the dominant inland along this monsoon trough are accompanied by 172
Reg Environ Change (2000) 1:163-179
Original article
spectral peaks obtained in the frequency domain for the observed (based on daily rainfall data of 30 stations for central India and 50 stations for all India, each for the period 1971±1980) as well as simulated central and allIndia rainfall. The signi®cance of the peaks is tested by ®tting a curve at the 90% level of con®dence calculated by chi square distribution. For central India, the simulated daily rainfall exhibits statistically signi®cant spectral peaks at periodicities of about 15 and 11 days in addition to short-term transients. These periodicities are not signi®cantly different from the periodicities of 14 and 10 days obtained in observed rainfall time series. On an all-India basis, the periodicities in oscillations in both observed and simulated daily rainfall are not substantially different to those obtained for central India rainfall. Observations reveal that a large number of depressions within the monsoon trough form over the north Bay of Bengal in the time interval of 3±15 days (Rao 1976). Therefore, the oscillations of time scales of about 7±13 days are likely to be associated with the synoptic-scale disturbances themselves (Lal 1995). The quasi-biweekly cycle with an oscillation of 11 days simulated in CSM suggests that perhaps an average of ten lows/depressions develop in the Bay of Bengal during a monsoon season and move across the central Fig. 9 Average spectral estimates of a observed and b simulated (control simulation) central India (upper panel) and all India (lower panel) daily monsoon rainfall
plains of India, which is close to the number frequency of 9.6 obtained in long-term climatology. From the above, it is clear that the CSM has only limited ability in simulating the south to north progression of monsoon and hence the total seasonal rainfall. The CSM has, however, captured the broad-scale features of the monsoon climatology reasonably well. The model is able to resolve various scales of summer monsoon circulation and simulate a few observed characteristic features of the interannual and intraseasonal variability over the Indian subcontinent. The periodicities in simulated monsoon rainfall are comparable with the observations over the central Indian region. In the next section, we present a brief account of the future projections of likely changes in monsoon climate and its variability at the time of doubling of CO2 in the atmosphere as simulated by the CSM in its perturbed radiative forcing experiment.
Scenarios of future changes in monsoon climate and its variability We shall focus here on the plausible climate change scenario for the Indian subcontinent as inferred from CSM simulation experiment with enhanced greenhouse gas forcings. The precise magnitude of future changes in the mean and/or variance of climatological parameters on
Reg Environ Change (2000) 1:163-179
173
Original article
regional scales due to anthropogenic increases in greenhouse gases is much warranted to evaluate the vulnerability of the region to such changes and to appropriately formulate adaptive response strategies. The key climatic elements examined here are surface air temperature and precipitation. Some additional elements are also considered depending upon their importance in speci®c aspects of variability/change in the selected region. A 2 °C rise in annual global mean surface temperature of the land regions with respect to the present-day atmosphere is simulated by the CSM at the time of doubling of CO2 in the atmosphere near year 70 in a 1% per year compound CO2 increase experiment. Over land regions of the Indian subcontinent, the area-averaged annual mean surface temperature rise at the time of doubling of CO2 in the atmosphere is 1.7 °C, only marginally lower than the global mean temperature increase. The area-averaged surface temperature increase over India during the winter is 2.2 °C, while during the summer monsoon season, it is restricted to only 1.0 °C at the time of doubling of CO2 in the atmosphere. The projected increase in annual average surface air temperature over the Indian subcontinent due to enhanced CO2 concentration in the atmosphere is consistent with those simulated in other A±O GCM experiments (Kitoh et al. 1997). Moreover, the projected warming being more pronounced during the winter than during the summer (monsoon) is also found in most of the current state-of-the-art global climate models. Greenhouse gas-induced surface warming may produce an increase in cloud water content (the water-holding capacity of the air increases with rise in temperature) and could enhance the re¯ectivity of the clouds (a negative feedback), but also contribute to an increase in the longwave emissivity of cloud (a positive feedback especially for the high cloud). Most global climate models seem to disagree about the net cloud radiative effect which depends crucially on the optical properties of clouds at solar and infrared wavelengths. The spatial distribution of surface warming as a consequence of increase in the atmospheric CO2 concentration (change in simulated surface air temperature during the period 2066±2075 when the CO2 concentration in the atmosphere is likely to double with respect to 1981±1990) averaged for the winter and monsoon seasons over the Indian subcontinent as simulated by the CSM is depicted in Fig. 10. It is seen that the spatial pattern of temperature change has a large seasonal dependency. The model simulates peak warming of about 2 °C over north and central India in winter. Over most of the south peninsula, the warming is under 2 °C during the winter season. During the summer monsoon, temperature rise over the south Indian region is less than 1 °C. The surface temperature rise is more pronounced over the northern and eastern regions of India (~1.5 °C) during the monsoon season. As regards the change in precipitation simulated by the CSM over the Indian subcontinent, a marginal decline (±0.04 mm day±1) in the area-averaged winter precipitation is obtained during the period 2066±2075 with respect to 1981±1990. The projected precipitation change over the Indian subcontinent in the winter months is, however, of 174
Reg Environ Change (2000) 1:163-179
relatively less practical importance than that likely during the monsoon. The summer monsoon season is regarded as the peak period for southwest monsoon activity and contributes about 75% of the observed total annual rainfall averaged over the subcontinent. During the monsoon season, the model simulation projects an increase in areaaveraged summer precipitation of about 1.4 mm day±1
Fig. 10 Spatial distribution of simulated changes in surface air temperature (°C) during winter (upper panel) and during summer monsoon season (lower panel) at time of doubling of CO2 in the atmosphere with respect to present-day conditions
Original article
over the land regions. The larger increase in surface temperature over land results in the intensi®cation of heat low over northwest India and increased land±sea pressure gradient which strengthens the summer monsoon ¯ow. The enhanced moisture convergence and associated convective activity over the region in a warmer atmosphere results in an increase in summer monsoon precipitation. The spatial distribution of changes in precipitation during winter and monsoon seasons in a CO2-enriched warmer atmosphere as simulated by the CSM is presented in Fig. 11. The model does not exhibit any appreciable change in spatial pattern of winter precipitation over land regions of the Indian subcontinent. A marginal decrease in precipitation over some parts of northeast India is simulated. During the summer season, the model suggests an increase in precipitation over almost all parts of the Indian subcontinent. The western semi-arid margins and southern regions of India could receive rainfall in excess of 1.0 mm day±1, whereas some parts of central India may receive rain in excess of 2 mm day±1 in a warmer atmosphere. This increase in precipitation is accompanied by enhanced moisture convergence over the central plains of India as evidenced in the spatial distribution pattern of likely changes in surface-speci®c humidity (Fig. 12). In order to examine the likely changes in intraseasonal and interannual variability in summer monsoon over the Indian subcontinent in response to changes in anthropogenic forcings, we have analyzed the simulated daily data for selected climatic variables for the monsoon season during each of the 10-year periods corresponding to the time of doubling of CO2 in the atmosphere. Figure 13 shows the temporal variation in simulated daily values of sea level pressure, surface air temperature and total rainfall averaged over central India from 1 June to 30 September for each of the 10 years, along with daily mean for the 10-year period (thick line). A comparison of Fig. 13 with Fig. 7 reveals many aspects of plausible changes in interannual and intraseasonal variability in Indian summer monsoon activity over the central plains of India. First, the year to year variability in selected climatic parameters seems to enhance considerably in a warmer atmosphere, more prominently in June when the monsoon currents are still progressing northward over the Indian subcontinent. This could imply that the date of monsoon onset over central India is likely to be more variable from one year to another in a warmer atmosphere. A detailed analysis of the daily rainfall data suggests that, under enhanced greenhouse forcing conditions, while the model still simulates the ®rst spell of intense rainfall appearing over the southernmost part of India (5±10°N) during the ®rst week of June on average, the spread of simulated onset date at 10°N (based on the criteria that rainfall at all grid points along 10°N in the region is 3 mm day±1 or more for at least three consecutive days) extends from 24 May to 11 June during the 10-year period corresponding to the time of doubling of CO2 in the atmosphere. This reaf®rms an enhanced variability in the date of onset of summer monsoon over India in a warmer atmosphere. The enhancement in year-to-year variability in daily
Fig. 11 Spatial distribution of simulated changes in rainfall (mm day±1) during winter (upper panel) and during summer monsoon season (lower panel) at time of doubling of CO2 in the atmosphere with respect to present-day conditions
rainfall is clearly discernible both in the early stages of monsoon activity (June) and again from mid-July until the end of September in the later part of the monsoon season (see also Meehl and Washington 1993). The model results also suggest that the enhanced interannual and intraseasonal variability in monsoon rainfall over India is essentially due to increased intraseasonal variability in the strength of convective activity in the
Reg Environ Change (2000) 1:163-179
175
Original article
176
Fig. 12 Spatial distribution of simulated changes in surface-speci®c humidity (g kg±1) during winter (upper panel) and during summer monsoon season (lower panel) at time of doubling of CO2 in the atmosphere with respect to present-day conditions
Fig. 13 Temporal variation in area-averaged mean sea level pressure, surface air temperature and total rainfall over central India for the period 1 June±30 September for each of the 10 years at time of doubling of CO2 in the atmosphere (thick lines represent trend in 10-year mean)
summer season. A signi®cant increase in the variability of convective precipitation (more pronounced over central India) is simulated by the model at the time of doubling of CO2 in the atmosphere relative to present-day conditions (Fig. 14). It is expected that this should lead to more intense rainfall spells over the land regions of the Indian subcontinent, thus increasing the probability of extreme rainfall events in a warmer atmosphere.
In order to ascertain the dominant time scales of monsoon circulation in the model at the time of doubling of CO2 in the atmosphere and to examine the likelihood of changes in frequency of monsoon depressions in a warmer atmosphere which could contribute to enhanced precipitation over the Indo-Gangetic plains, spectral energy computations were performed on the detrended time series of model-simulated daily rainfall for each of the ten consec-
Reg Environ Change (2000) 1:163-179
Original article
utive years of perturbed forcing simulation and averaged over the years. Figure 15 shows the dominant spectral peaks obtained in the frequency domain for the simulated central and all-India rainfall for doubled CO2 atmosphere. For central India, the simulated daily rainfall exhibits statistically signi®cant spectral peaks at periodicities of about 17 and 9 days, not substantially different from those in the control experiment. If we believe that the short period oscillations in the simulated daily rainfall data are a manifestation of synoptic-scale northwestward-moving monsoon depressions, we may infer from this analysis that no appreciable changes are likely in the number of monsoon depressions moving across the central plains of India in a warmer atmosphere.
Conclusions The CSM has captured the seasonal cycle as well as broadscale features of the monsoon climatology over the Indian Fig. 14 Temporal variation in convective precipitation area-averaged over central (left) and all India (right) for the period 1 June±30 September for each of the 10 years for present-day conditions (upper panels) and at time of doubling of CO2 (lower panels) in the atmosphere (thick lines represent trend in 10-year mean)
subcontinent reasonably well. The cross-equatorial ¯ow in simulated surface wind patterns during summer is, however, stronger than observed with an easterly bias. The 850 hPa winds over the land are also somewhat stronger than observed over the region. The location and strength of the upper air westerly jet over the northern Indian subcontinent during winter and the tropical easterly jet during the monsoon season are in fair agreement with observed climatology. However, while the seasonality in rainfall over the region is simulated well, the simulated area-averaged monsoon rainfall is underestimated to only about 60% of the observed rainfall. The centers of maxima in simulated monsoon rainfall are slightly displaced southward as compared to the climatological patterns. The transient experiment with a 1% per year compound increase in CO2 with CSM suggests an annual mean areaaveraged surface warming of about 1.73 °C over the region at the time of CO2 doubling. This warming is more pronounced in winter than during the monsoon season. A net increase in area-averaged monsoon rainfall of about 1.4 mm day±1, largely due to increased moisture convergence and associated convective activity over the land, is obtained. The intraseasonal variability in the monsoon rainfall enhances in a warmer atmosphere. Much of this variability is con®ned to the early part of the monsoon season, which suggests the possibility of the date of onset of summer monsoon over India becoming more variable
Reg Environ Change (2000) 1:163-179
177
Original article
Sulphate aerosols have also been shown to affect the climate system by changing the radiative balance of the atmosphere. As the anthropogenic sulphate aerosol burden in the troposphere would have large spatial and temporal variations in the atmosphere, its future impact on the regional scale would be in striking contrast to the impact from greenhouse gases for which the concentration changes, in most cases, are likely to be uniform throughout the globe. Contrary to simulations that consider only greenhouse gas-induced radiative forcing, a possible decline in summer monsoon rainfall over south-east Asia and more speci®cally over India has been suggested under combined greenhouse gas and sulphate aerosol forcing experiments (Lal et al. 1995; Mitchell et al. 1995). However, only the direct cooling effects of sulphate aerosol produced by industrial activity were considered in these studies. Considerable uncertainty prevails about the indirect effect of sulphate aerosols on tropospheric clouds which could strongly modulate the monsoon climate. We are still unclear about the implications of localized radiative forcing on the deep convection in the tropics and on Hadley circulation. Moreover, recent ®ndings suggest that the greenhouse gas forcings are likely to dominate and override the sulphate aerosol forcings in the future, with several policy measures currently being negotiated under the United Nations Framework on Climate Change Convention. Current studies on climate variability and climate change increasingly rely upon diurnal, seasonal, latitudinal and vertical patterns of temperature trends to provide evidence for anthropogenic signatures. Such approaches require increasingly detailed understanding of the spatial variFig. 15 ability of all forcing mechanisms and their connections to Average spectral estimates of simulated central India (upper panel) and all India (lower panel) daily monsoon rainfall at time of doubling global, hemispheric and regional responses. Precise magof CO2 in the atmosphere nitude as well as the role of these spatially localized potential forcings must be known before a more con®dent prediction of regional changes in climate and its variability in future. The enhanced interannual and intraseasonal can be made. variability in the summer monsoon activity over India could also contribute to more intense rainfall spells, thus Acknowledgments The authors acknowledge the efforts of all increasing the probability of extreme rainfall events in a those who performed the CSM experiments at the National Center for Atmospheric Research (the NCAR is sponsored by the Nawarmer atmosphere. If we believe that the short period oscillations in daily rainfall are manifestation of the syn- tional Science Foundation). The ®rst author (M.L.) acknowledges optic-scale westward-propagating monsoon depressions, the generous support and hospitality extended to him by the no signi®cant changes are likely in the number of mon- NOAA Aeronomy Laboratory and the Atmospheric Chemistry Project of NOAA's Climate and Global Change Program during soon depressions traversing the central plains of India in the time this study was undertaken. M.L. also appreciates the a warmer atmosphere. facilities and scienti®c advice rendered to him by Drs. Susan Changes in the simulated spatial distribution of seasonal Solomon and T.M.L. Wigley during the course of this study. Part rainfall pattern over the Indian subcontinent depend on of the study was supported by the Of®ce of Health and Enviadequate representation of surface orography and hence ronmental Research of the US Department of Energy under its on model resolution. The A±O GCMs are principally de- Carbon Dioxide Research Program. signed to simulate the large-scale features of the global climate. The regional climates simulated by the A±O GCMs lack the ®ner scale details, which could be attributed to References poor representation of mesoscale climate forcings (e.g., orography, complex coastline, vegetation characteristics and inland water bodies). Better representation of surface BHALME HN, RAHALKAR SS, SIKDER AB (1987) Tropical quasibiennial oscillation in the 10-mb wind and Indian rainfall ± topography and implementation of improved parameterimplications for forecasting. J Climatol 7:345±353 ization of sub-grid-scale physical processes in climate BONAN GB (1998) The land surface climatology of the NCAR land simulation experiments should reduce the uncertainty in surface model (LSM 1.0) coupled to the NCAR community climate model (CCM3). J Climate 11:1307±1326 projections of future regional climate change scenarios. 178
Reg Environ Change (2000) 1:163-179
Original article
BOVILLE BA, GENT PR (1998) The NCAR climate system model version 1. J Climate 11:1115±1130 BOVILLE BA, HURRELL JW (1998) A comparison of atmospheric circulations simulated by the CCM3 and CSM1. J Climate 11:1327±1341 BRYAN FO (1998) Climate drift in a multi-century integration of the NCAR climate system model. J Climate 11:1455±1471 BRYAN FO, KAUFFMAN BG, LARGE WG, GENT PR (1996) The NCAR CSM ¯ux coupler. NCAR Tech Note NCAR/TN-424+STR, NCAR, Boulder, Colorado, 48 pp CESS RD, ZHANG M-H, POTTER GL, BARKER HW, COLMAN RA, DAZLICH DA, DEL GENIO AD, ESCH M, FRASER JJ, GALIN V, GATES WL, HACK JJ, INGRAM WJ, KIEHL JT, LACIS AA, LE TREUT H, LI Z-X, LIANG X-Z, MAHFOUF J-F, MCAVANEY BJ, MELESHKO VP, MORCRETTE J-J, RANDALL DA, ROECKNER E, ROYER J-F, SOKOLOV AP, SPORYSHEV PV, TAYLOR KE, WANG W-C, WETHERALD RT (1993) Uncertainties in carbon dioxide radiative forcing in atmosphere general circulation models. Science 262:1252±1255 DUMENIL L (1998) Portrayal of the Indian summer monsoon in the land±ocean±atmosphere system of a coupled GCM. Rep 228, Max Planck Institute for Meteorology, Hamburg, 40 pp FLATO GM, HIBLER III WD (1992) Modelling pack ice as a cavitating ¯uid J Phys Oceanogr 22:626±651 GATES WL, CUBASCH U, MEEHL GA, MITCHELL JFB, STOUFFER J (1993) An intercomparison of selected features of the control climate simulated by coupled ocean atmosphere general circulation models. WCRP-82, WMO/TD no 574, World Meteorological Organization, Geneva HACK JJ (1994) Parameterization of moist convection in the National Center for Atmospheric Research community climate model (CCM2). J Geophys Res 99:5551±5568 HOLTSLAG AAM, BOVILLE BA (1993) Local versus non-local boundary layer diffusion in a global climate model. J Climate 6:1825±1842 Intergovernmental Panel on Climate Change (1996) Second scienti®c assessment of climate change. In: Houghton JT, Meira Filho LG, Callander BA, Harris N, Kattanberg A, Maskell K (eds) WMO-UNEP Report. Cambridge University Press, Cambridge, 572 pp KIEHL JT, HACK JJ, BONAN G, BOVILLE BA, WILLIAMSON D, RASCH PJ (1998) The National Center for Atmospheric Research community climate model: CCM3. J Climate 11:1131±1149 KITOH A, YUKIMOTO S, NODA A, MOTOI T (1997) Simulated changes in the Asian summer monsoon at times of increased atmospheric CO2. J Meteorol Soc Jpn 75(6):1019±1031 KRISHNAMURTI TN, BHALME HN (1976) Oscillations of a monsoon system. Part I ± observational aspects. J Atmos Sci 33:1937±1954 LAL M (1995) Synoptic scale disturbances of Asian summer monsoon as simulated in ECHAM3 T-42 climate model. In: Staniforth A (ed) WMO/CAS-WGNE research activities in atmospheric and oceanic modelling. WMO/TD no 665, Feb, pp 7.41±7.42 LAL M, CUBASCH U, VOSS R, WASZKEWITZ J (1995) The effect of transient increase in greenhouse gases and sulphate aerosols on monsoon climate. Curr Sci 69(9):752±763 LAL M, WHETTON PH, PITTOCK AB, CHAKRABORTY B (1998) Simulation of present-day climate over the Indian subcontinent by general circulation models. Terr Atmos Ocean Sci 9(1):69±96 LATIF M, ANDERSON D, BARNETT T, CANE M, KLEEMAN R, LEETMAA A, O'BRIEN J, ROSATI A, SCHNEIDER E (1998) A review of the predictability and prediction of ENSO. J Geophys Res 103(C7):14375±14394 LEGATES DR, WILLMOTT CJ (1990) Mean seasonal and spatial variability in gauge-corrected global precipitation. Int J Climatol 10:111±127
MEEHL GA (1994) In¯uence of the land surface in the Asian summer monsoon: external conditions versus internal feedbacks. J Climate 7:1033±1049 MEEHL GA, ARBLASTER JM (1998) The Asian±Australian monsoon and El Nino±Southern Oscillation in the NCAR climate system model. J Climate 11:1356±1385 MEEHL GA, WASHINGTON WM (1993) South Asian summer monsoon variability in a model with doubled atmospheric carbon dioxide concentration. Science 260:1101±1104 MEEHL GA, WASHINGTON WM (1995) Cloud albedo feedback and the super greenhouse effect in a global coupled GCM. Climate Dynamics 11:399±411 MITCHELL JFB, JOHN TC, GREGORY JM, TETT SFB (1995) Climate response to increasing levels of greenhouse gases and sulphate aerosols. Nature 376:501±504 NANJUNDIAH RS, SRINIVASAN J, GADGIL S (1992) Intraseasonal variation of the Indian summer monsoon, II. Theoretical aspects. J Meteorol Soc Jpn 70:529±549 NITTA T (1983) Observational study of heat sources over the eastern Tibetan Plateau during the summer monsoon. J Meteorol Soc Jpn 61:590±605 PALMER TN (1994) Chaos and the predictability in forecasting the monsoons. Proc Indian Natl Sci Acad A60:57±66 PALMER TN, BRANKOVIC C, VITERBO P, MILLER MJ (1992) Modeling interannual variations of summer monsoons. J Climate 5:399± 417 PARTHASARATHY B, MUNOT AA, KOTHAWALE DR (1994) All India monthly and seasonal rainfall series; 1871±1993. Theor Appl Climatol 49:217±224 RAO YP (1976) Southwest monsoon. Meteorol Mono 1, India Meteorological Department, New Delhi, 367 pp RAPHAEL MN (1998) Quasi-stationary waves in the southern hemisphere: an examination of their simulation by the NCAR climate system model with and without an interactive ocean. J Climate 11:1405±1418 SENIOR CA, Mitchell JFB (1993) Carbon dioxide and climate: the impact of cloud parameterization. J Climate 6:393±418 SIKKA DR (1980) Some aspects of the large scale ¯uctuations of summer monsoon rainfall over India in relation to ¯uctuations in the planetary and regional scale circulation parameters. Proc Indian Acad Sci (Earth Planet Sci) 89:179±195 STOCKDALE T, BURGERS G, LATIF M, WOLF J (1994) Some sensitivities of a coupled ocean±atmosphere GCM. Tellus 46A:367±380 TOGA Monsoon Numerical Experimentation Group (1990) TOGA Monsoon Climate Research. Report of the second session of the Tropical Ocean and Global Atmosphere Monsoon Numerical Experimentation Group. WCRP-49, WMO-TD no 392, World Meteorological Organisation, Geneva WEATHERLY J, BRIEGLEB B, LARGE WG, BETTGE T (1998) Sea ice and polar climate in the NCAR CSM. J Climate 11:1472±1486 WEBSTER PJ, YANG S (1992) Monsoon and ENSO: selectively interactive systems. Q J R Meteor Soc 118:877±926 WEBSTER PJ, MAGANA VO, PALMER TN, SHUKLA J, THOMAS RA, YANAI M, YASUNARI T (1998) Monsoons: processes, predictability and the prospects for prediction. J Geophys Res 103(C7):14451±14510 WIGLEY TML, SANTER BD (1990) Statistical comparison of spatial ®elds in model validation, perturbation and predictability experiments. J Geophys Res 95:851±865 ZHANG GJ, MCFARLANE NA (1995) Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian general circulation model. Atmos Ocean 33:407±446
Reg Environ Change (2000) 1:163-179
179