1. INTRODUCTION
In an earlier study the apparent relation between winter and spring Eurasian snow and subsequent Indian summer monsoon was examined using observational snow cover and snow depth data based on satellite sensors (Bamzai and Shukla, 1998; Bamzai and Shukla, 1999). In contrast to previous studies that use average snow cover over entire Eurasia as a single number, the frequency of occurrence of snow cover at each grid point over Eurasia was correlated with the Indian summer monsoon rainfall. Thus specific geographical regions over Eurasia that are responsible for the well-known inverse relationship between Eurasian snow cover and the Indian monsoon rainfall were delineated. It is found somewhat surprisingly, that western Eurasia is the only geographical region for which a moderately significant inverse correlation is found between winter snow cover and subsequent summer monsoon rainfall. Composites for high and low snow cover years were also studied using temperature data. Winters of high snow cover for Eurasia are found to be associated with colder temperatures over large-scale regions of the Eurasian continent and vice versa. These temperature anomalies persist until the pre-monsoon months giving rise to higher land-ocean meridional temperature contrast for low snow cover years.
The effect of seasonal snow on summer monsoon circulation is a result of complex competing feedback mechanisms. The boundary forcing due to anomalous soil moisture is implicitly included in the snow-monsoon hypothesis since the anomalous snow melt affects soil moisture and subsequent evaporation: excessive soil moisture over the mid-latitude rain belt tends to prolong the moisture anomaly and increases precipitation by recycling water between land surface and atmosphere. These changes have an impact on the temperature of the atmosphere from the surface to the upper troposphere. This evaporation and convection feedback is particularly feasible in summer, over the continents where large-scale convergence in the lower troposphere is dominant. Enhanced snow may provide increased soil moisture for rainfall leading to a positive snow-monsoon relationship. On the other hand, reduced snow increases surface sensible heat flux, increases local convergence, leading to increased precipitation. This would imply a negative snow-monsoon relationship. It has been suggested that if a strong regional monsoon circulation is initiated in early summer, greater precipitation that occurs provides surface moisture that contributes to sustaining rainfall through the monsoon season by means of the soil moisture positive feedback mechanism (Meehl, 1994a).
The basic idea tested here is whether variations in initial snow mass field cause an atmospheric response that affects subsequent monsoon circulation through subsequent melting of anomalous snow, evaporation and influence of soil hydrology. The result expected a priori is that enhanced Eurasian snow mass leads to below normal monsoon rainfall and vice versa. To understand how spring snow mass anomalies in western and central Eurasian region influence subsequent monsoon circulation and rainfall through snow albedo and soil moisture effects, an 11-member ensemble sensitivity experiment is conducted using the Center for Ocean-Land-Atmosphere Studies Atmospheric General Circulation Model (COLA AGCM) forced by observed sea surface temperature (SST). Using the period 1948-92, the climate run was examined for the inverse snow-monsoon hypothesis. Since the relationship was found to hold well for the period 1982-1992 (that uses SSTs that include satellite observations), the AGCM sensitivity to spring snow mass was conducted using the 1982-1992 period as the control.
The AGCM is a unique tool to test under controlled experimentation since one may study the physics entailing surface-atmosphere interactions by examining the spatial and temporal evolution of the model output where observations are not necessarily available. The question of whether the snow-monsoon correlation is a cause-effect one may also be addressed
more effectively by a numerical experiment. Further the AGCM provides an opportunity of several realizations through the use of ensemble runs.
Satellite-derived observations of seasonal snow depth and snow cover variability have been examined and compared to simulated values (Bamzai, 1997; Bamzai and Kinter, 1997). The standard deviation for observed Eurasian winter snow depth (see Section 3.2 for details on the snow depth data used) is in the range 5-10´ 10-2 m, corresponding to ~1´ 10-2 m water equivalent. Snow depth anomalies have a magnitude as large as ~3´ 10-2m only for eastern Eurasia. For spring, the range is of the same order of magnitude as maximum values of standard deviation for the central Eurasian region. Since it is known that the satellite-derived snow depth data underestimates depth in shallow regions, the extent of the anomalies may be larger than that suggested by this observational data set.
A brief review of various modeling studies addressing the Eurasian snow - Indian summer monsoon relation is presented in Section 2. The GCM used in this study is described in Section 3.1. Model simulated seasonal snow cover and depth are compared with satellite-derived global observational data in Section 3.3. Section 3.4 describes the design of the sensitivity study. Results and conclusions are in Section 4, and Section 5, respectively.
2. REVIEW OF SNOW-MONSOON MODELING STUDIES
GCM modeling studies indicate theoretical support for the snow-monsoon hypothesis
(Kutzbach (1981); Yeh et al. (1983); Yeh et al. (1984); Barnett et al. (1988); Barnett et al. (1989); Delworth and Manabe (1989); Cohen and Rind (1991); Yasunari et al. (1991); Zwiers (1993); Kitoh et al. (1994); Meehl (1994b); Douville and Royer (1996); Ose (1996); Kitoh et al. (1997); Dong and Valdes (1998)). Table 1 summarizes various GCM studies that have focused on the role of Eurasian snow on subsequent Asian monsoon. The studies demonstrate that the GCMs used reproduce an inverse snow-monsoon relationship to varying degrees, suggesting Eurasian snow cover and snow mass impacts climate on a seasonal time scale.
Some of the past sensitivity experiments have used snow anomalies that are unrealistically large in spatial extent, sometimes encompassing the entire Eurasian continent. This may cause enhanced signals in models to give an inverse snow-monsoon relation. It is difficult to validate model results for physical processes related to ground and snow hydrology, both due to inadequate field observations and due to the fact that most models still treat processes related to snow and ground hydrology by overly simple parameterizations.
In the experiment of Barnett et al. (1989) the snowfall rate in the model's condensation schemes were doubled or halved, corresponding to snow depth anomalies of around 13´ 10-2 m water equivalent in May. These anomalies are high compared to those evidenced in the satellite-based observations. Yasunari et al. (1991) added 5´ 10-2 m water equivalent uniformly over entire Eurasia between 30oN and 60oN. The magnitude of this anomaly is realistic compared to observations; however observed seasonal snow mass anomalies exhibit a spatial structure. Using climatologically varying SST, Ose (1996) considered anomalies of 5´ 10-2 m over three separate regions of Eurasia (Tibet, eastern Europe and Siberia), thus addressing in part, the question of how the response of the atmosphere and the ground depends on the location of the anomaly. The results of this experiment suggest that the cooling source over Tibet works to significantly delay the seasonal transition from spring to summer in the Northern Hemisphere. Douville and Royer (1996) prescribed 10´ 10-2 m water equivalent over the region 30oN to 80oN in March and demonstrated an effect on subsequent monsoon circulation.
Two anomaly experiments of `heavy' and `light' snow, based on extremes of maximum and minimum values at each grid point over Eurasia (as computed from observations of snow depth) were performed using the COLA AGCM (Zhou, 1993; Vernekar et al., 1995). Initial snow depth for the rest of the globe was prescribed from the nine-year climatology. The results of this experiment suggest heavy snow is unfavorable for the following summer Indian monsoon. There is an associated weak monsoon circulation, a delay in the monsoon onset, a delay in the monsoon withdrawal, higher sea level pressure over India, a weaker Somali jet, weaker lower tropospheric westerlies and weaker upper tropospheric easterlies.
The main differences between our experiment and that of Vernekar et al. (1995) are the following: (i) introduction of the snow anomalies in a specific region rather than for the entire Eurasian continent; (ii) use of the T30 COLA AGCM here as opposed to the R40 version; (iii) use of observed SST in this experiment as opposed to climatological SST; (iv) incorporation of the RAS scheme of convection in the AGCM as opposed to the Kuo scheme. Furthermore, in the experiment of Vernekar et al.(1995), the difference in snow depth anomalies (high snow - low snow) are ~10´ 10-2m water equivalent over the Himalayan region, and ~5´ 10-2 m water equivalent over eastern Eurasia. In contrast, this experiment introduces almost no snow anomalies over both the Himalayas and eastern Eurasia.
3. METHODOLOGY
The importance of a realistic simulation of the mean climate by a dynamical model to be used for sensitivity studies relating to the effect of slowly varying boundary conditions has been discussed (Shukla, 1984). In order to detect the effects of changes in snow forcing, the model should be able to predict the seasonal mean snow accurately. It is thus necessary to first assess the model's ability to reproduce seasonal snow cover and snow depth.
The model, initial diagnostics of the control run, observational data sets used for comparison with the model results and the design of the sensitivity experiment are described here. The COLA AGCM (Kinter et al., 1988; Kinter et al., 1997) is based on the NCEP global spectral model used for operational medium range weather forecasting. The full primitive equations of motion are expressed in spectral form and projection coefficients in the spherical harmonics are used. In this study, spectral coefficients are truncated at a wave number 30, using triangular truncation. Vertical derivatives are expressed in centered difference form using a normalized pressure vertical coordinate in 18 discrete unevenly distributed levels.
The cloud longwave radiative scheme is that of Harshvardhan et al. (1987) and the shortwave scheme is that of Lacis and Hansen (1974). The diurnal cycle is explicitly resolved. The convection scheme is the Relaxed Arakawa-Schubert Scheme, RAS (Moorthi and Suarez, 1992; DeWitt 1996). A biophysical treatment of the terrestrial biosphere is incorporated including its interaction with the atmosphere in terms of radiative flux, heat flux, moisture flux and momentum flux (Xue et al., 1991). This biosphere model is a simplified version of the SiB model of Sellers et al. (1986) and simulates interactions between land and atmosphere by treating vegetation explicitly. The hydrologic cycle in the model agrees more closely with observations than the so-called `bucket hydrology' used previously (Delworth and Manabe, 1988; Delworth and Manabe 1989). The daily variation of observed SST and the seasonal variation of vegetation cover are prescribed boundary conditions. Soil moisture and snow mass are prescribed initially; at later time steps, they predicted by the model. Orography is represented by a modified mean orography (Fennessy et al., 1994).
3.1 Snow Physics in the model
The various effects of snow and ice within SiB have been accounted for by simple modifications to some of the parameters and calculations in the model (Sellers et al., 1986; Xue et al., 1991). The main processes are described below.
(i) Definition of snowfall
The water that accumulates on the ground and the canopy surfaces may be in either liquid or solid form (snow). As precipitation nears the ground in the model, it is assumed to equilibrate with the air at the reference height. If this is below the freezing point of water, precipitation is assumed to be snow.
(ii) Interception and accumulation of snowfall
If the temperature of the vegetation canopy or the ground surface is below freezing, water storage on the ground is assumed to be frozen and accumulates. Snow accumulates entirely on the ground. Once the snow has accumulated, if the reference temperature rises above freezing, snow begins to melt. If temperature at either one of the two subsurface levels in the land surface treatment is above freezing when snow is melting, meltwater infiltrates into the ground. If the surface layer soil moisture becomes saturated, meltwater is assumed to run off. Snow accumulation on the ground thus has a direct effect on the GCM surface hydrologic balance. Snow increases the surface albedo in the model, and also significantly alters the partitioning of the surface latent and sensible heat fluxes. The effects of sublimation, cloud physics and blowing snow are neglected in the model. It is assumed that the depth of snow intercepted by the ground is equivalent to five times the water depth and full coverage of the ground is reached when the snow is 0.005 m deep.
(iii) Radiation
The scattering coefficient of leaf elements in the canopy and reflectance of the ground are adjusted according to the amount of intercepted snow. Snow is assumed to reflect radiation and there is no transmission through the depth of snow layer. The value of snow albedo and
scattering coefficient of snow are assumed to be 0.8 in the visible and 0.4 in the near-infrared wavelength. These values are reduced by 40% when surface temperature is close to, or at the melting point of snow.
(iv) Energy partition
The heat capacity of the canopy or the ground is calculated as a function of the biomass and/or soil material and intercepted water mass. An adjustment is made for the latent heat flux when snow is present. If the surface temperature goes above or below the freezing point of water, melting or freezing of the intercepted precipitation occurs and the surface temperature is modified accordingly.
(v) Runoff
In the model scheme, there is no infiltration or runoff when the surface temperature is 0.1oK below the freezing point. However when the surface temperature is between - 0.1oK and the freezing point, snowmelt may run off. The initial snow depth can be set using an offline program.
3.2 Observational Data Sets
The snow cover data set used is the NOAA NESDIS weekly gridded data set spanning the period January 1973 to September 1994 (Kukla and Robinson, 1981; Ropelewski et al., 1984; Robinson et al., 1991; Robinson et al., 1993). For snow depth, the fine resolution, albeit shorter time series of monthly means of the NIMBUS-7 NASA SMMR data for the period November 1978 to August 1987 has been utilized (Chang et al., 1987; Chang et al., 1990; Chang et al., 1992). The total snow-covered area derived from snow depth data is typically about 10% less than the snow cover data (Foster et al., 1996). The precipitation data used for this study is the Xie-Arkin data based on gauge and satellite productions (Xie-Arkin, 1996). The surface air temperature data set used is the 1979-1993 Climate Anomaly Monitoring System (CAMS) station data based on station data (Ropelewski et al., 1985), interpolated on an R40 Gaussian grid corresponding to an approximate resolution of 1.76o latitude and 2.81o longitude.
Recently, new snow products such as the Historical Soviet Daily Depth, Northern Hemisphere snow cover and sea ice are available from the NSIDC. Efforts are also underway to use the NASA SMMR data through a more sophisticated retrieval algorithm than the one based on a single coefficient (Foster et al., 1996). Parameters that account for forest cover, snow grain size based on field observations could be introduced to give estimates of snow depth that are free from the inaccuracies in the present data. In the Himalayan region, the temperature data set has regions of missing data making it difficult to assess the model in this region.
3.3 Comparison of the control run simulation with observational data
The model run of the COLA AGCM studied is the Climate of the Twentieth Century (C20C) 1948-1994 run. Initial conditions were the November 1 atmospheric conditions from another climate run, and the model was run through November 1994. The SST boundary condition applied to C20C integrations was taken from an objective analysis of SST measurements, based on the UK Hadley Centre product which is referred to as the Global Ice/Sea Surface Temperature, GISST 2.2 (Rayner et al.,1995; Parker et al., 1995). For a comparison of model with observations, the climatology of snow cover frequency, standard deviation, annual cycle and interannual variability of seasonal snow cover extent has been computed using the common time period between the two, 1973-1994.
The GCM does not simulate snow directly. The prognostic variable, water storage on the ground (in meters water equivalent) may be used to compare with observational data. The model output of monthly values of storage to ground, stg(t), was constructed from daily output fields. To obtain snow cover (presence or absence) for comparing with observations, the following assumption was made: a location has a value of snow cover equal to 1 (presence of snow) if there is a minimum critical stg that has a value of 0.005 m, and 0 (absence of snow) otherwise. Simulated snow cover is sensitive to the choice of the threshold value. Since comparisons are being made with satellite observations and the satellite detects snow on the ground only if the thickness is about 2.5 ´ 10-2 m. A value of 0.005 m is appropriate as a threshold since this corresponds to a snow depth of 2.5 ´ 10 -2m since the model assumes a conversion factor of 5 as water equivalent for snow.
The area average of simulated snow depth for Himalayan region (60oE to 120oE and 30oN to 50oN) and the simulated precipitation over the land region (65oE to 95oE and 5oN to 35oN) were examined. No strong correlation's between snow anomalies and precipitation anomalies were found for the 44-year period 1948-1992; however, for the period 1982-1992, the SST data set uses satellite data giving improved spatial coverage. For this period, the inverse snow-monsoon relation held for 10 out of 13 years. Seasonal correlations between model generated snow depth anomalies and model Indian monsoon rainfall anomalies were constructed. However the model was unable to reproduce the structure of the correlation of the snow-monsoon evidenced in the observational data sets.
Figure 1 shows a comparison of observed and simulated winter snow depth climatology and its standard deviation. The results for both model and observations correspond to the same period 1979-1987. Winter snow depth is underestimated in the region of western Eurasia and central Eurasia. In the Himalayas and some regions of Eurasia; winter snow depth is overestimated by a factor of ~100 mm water equivalent. The observations suggest large interannual variability in eastern Eurasia; however the model does not capture this feature. Figure 2 shows the results for spring. The spring snow depth climatology is overestimated by 60-100 mm water equivalent. The model overestimates both spring snow depth climatology and its interannual variability in the Himalayan region.
Figure 3 (top panel) shows the annual cycle of snow cover for Eurasia based on observed snow cover, observed snow depth and model output using the common time period 1979-1987. The simulated annual cycle is much more pronounced compared to both observational data sets with differences of ~3 million square kilometers (corresponding to ~10% of the total snow cover extent) during winter months; simulated spring snow cover is overestimated (~15% for April). There is a lack of snow cover during summer months and fall snow cover is underestimated. Figure 3 (lower panel) shows the observed and simulated annual cycle of area averaged snow depth over Eurasia. Area average snow depth over Eurasia is overestimated (up to 100%) during winter months; furthermore, the observed annual cycle suggests maximum snow depth occurs around mid-February whereas the simulated snow depth peaks around April.
The interannual variability of Eurasian snow cover based on the NOAA NESDIS snow cover and model simulation has been examined for the period 1973-1994. The top and lower panels of Fig. 4 show the results of area average values for the region (zero to 120oE and 35oN to 65oN) for winter and spring, respectively. Also shown is the snow cover based on the snow depth data (assuming a grid box has snow cover if it has a minimum snow depth of 2.5 ´ 10-2 m) for the period 1979-1987. For winter the snow depth data has ~10% less snow cover for the region considered, whereas for spring the reverse holds. Fig. 5 shows the interannual variability of area averaged winter and spring snow depth over Eurasia for the region (zero to 120oE and 35oN to 65oN) respectively. The model captures the interannual variability but overestimates the mean by up to 100 % for both seasons for certain years.
To address the issue of whether this overestimation is mainly one of partitioning of precipitation (into solid and liquid) or whether it is arising due to overestimation of precipitation itself, the seasonal precipitation over Eurasia has been examined and compared to observations (Xie and Arkin, 1996). The time period for both model simulation and observations is 1982-1996. Figure 6 shows the comparison of simulated and observed precipitation for winter, and Fig. 7 show the comparison of simulated and observed precipitation for spring. The simulated precipitation is larger than observed especially for northeastern Eurasia (over the Taklimakan desert). The mean seasonal position of the freezing line is also shown in these plots. The data for this is based on a gridded data set of the Climate Anomaly Monitoring data (Ropelewski et al., 1985) for the period 1979-1992.
In summary, the model reproduces the main qualitative features associated with seasonal mean snow cover and snow depth. The annual cycle of area averaged snow depth over Eurasia gives an overestimation (by a factor of two) due to the tendency of the model to
produce large snow depths in northern high latitudes. The maximum simulated snow depth occurs around April, more than a month after what is indicated in the observations.
3.4 Experimental Design
Based on the observational study (Bamzai and Shukla, 1998, 1999), the Eurasian region (zero to 120oE and 35oN to 65oN) was selected for introducing the snow mass anomalies. Two sets of experiment were made, one for heavy snow and the other for light snow. Both sets of experiments contain eleven runs each (using initial atmospheric conditions for March 1 1982, March 1 1983, 1992) from the C20C run. The control run for this experiment is the period 1982-1992. The final model output is on a grid of resolution of 4o latitude and 5o longitude.
The observational study suggests an area of small extent to have the maximum correlation with spring snow depth and snow cover. Since the model underestimates snow depth in this region, a larger region (zero to 120oE and 35oN to 65oN) was selected so that the anomalies would be substantial in extent. The result of this study should suggest whether a snow anomaly on a continental scale or over a limited region (the Tibetan region, north of the subcontinent) is crucial for establishment of the Indian monsoon in the AGCM. Only the part of the Tibetan plateau (north of 35oN) falls in the region considered for prescribing snow anomalies in our experiment.
For the enhanced runs, the March 1 climatology was added to the snow depth of March 1 for that particular year of the C20C run, whereas for the reduced run snow depth case 50% of the March climatology was subtracted from the snow depth of March 1 of that year (set to zero at all grid points for which values were negative). If Mcl is the climatology for March 1, the Enhanced Snow Anomaly, Esa, introduced on March 1 of any year is given by
Esa = Mcl
Similarly the reduced snow anomaly, Rsa introduced on March 1 of any year is given by
Rsa = - 0.5 Mcl
The snow anomalies introduced are not of constant magnitude but have an associated structure which is model-dependent. The anomalies prescribed here correspond approximately
to doubling or halving snow depth of the mean March 1 climatology of the control run for the period 1982-1992. For the midlatitude region of western Eurasia, the anomalies have a magnitude of only 2 x 10-3 m water equivalent which is smaller than the standard deviation values as evidenced by the snow depth data. In the high latitudes (where the model greatly overestimates snow depth compared to the observed snow depth data), the anomalies are larger than observational values. However, since the region for introducing the anomalies is below 65oN, the greater part of the region of overestimated model anomalies is excluded from this experiment.
These enhanced and reduced snow anomalies are substantial. The reason for choosing the enhanced snow anomaly as the model March 1 climatology is because the snow climatology in the region of interest is low. The reduced snow anomaly is not equal to the enhanced snow anomaly in magnitude: since the effect of snow is being examined, only half the March 1 climatology is subtracted from the initial conditions for March 1. The effect for the entire ensemble is to double or half the mean initial snow. Fig 8 shows the initial snow depth (on March 1) for one of the ensemble runs.
Most previous experiments have assumed constant anomalies for Eurasia; however, snow depth shows considerable regional variability. Consequently using a constant magnitude for the snow anomaly over a widely distributed region such as the Eurasian continent is unrealistic since introducing anomalies over areas where such anomalies have never been observed in the past data could give rise to unrealistic circulation response. Further, most experiments have considered the effect of prescribing positive snow anomalies; the experiment here considers the effect of both enhanced and reduced snow anomalies. The study of Vernekar et al. (1995) considered both enhanced and reduced snow depth initial conditions; a comparison of Fig. 8 with the initial snow depth conditions in this previous experiment using the COLA AGCM (Fig. 4, Vernekar et al.(1995)) indicates the anomalies introduced here are north of 35oN where the winter snow depth values are high, and towards the central and western Eurasian region. In contrast initial snow depth anomalies in the Vernekar et al. (1995) experiments were located in the high latitudes in eastern Eurasia and somewhat lower latitudes in the rest of the continent, including the Himalayan region.
4. RESULTS
The results are presented for ensemble mean differences. They suggest the following sequence of events leading to the enhanced (reduced) snow anomalies giving below (above) normal monsoon rainfall:
enhanced (reduced) spring snow anomalies over Eurasia lead to;
more (less) snowmelt in late spring and early summer (Fig. 9);
more (less) spring/summer soil wetness; (Figs. 10, 11);
less (more) May sensible heat flux as well as latent heat flux (Fig. 12);
smaller (larger) May temperature contrast ( Figs. 13, 14);
less (more) May sea level pressure over India (Fig. 15) ;
weaker (stronger) JJAS monsoon circulation (Fig. 16);
less (more) JJAS precipitation (Fig. 17).
4.1 Seasonal variation of snow melt and snow depth
Figure 9 shows the ensemble average snow melt for the enhanced, control and reduced runs. These plots are centered on the first of the month. Most of the snow melts by the end of July for the entire Eurasian continent (top panel); the region where the snow anomalies were prescribed (middle panel), and the Himalayan region (lower panel). The rate of snow melt is higher in the enhanced case with a maximum in June.
The area average value of storage on ground, stg, for all the runs is shown in Fig. 10. There are 11 curves each for the enhanced snow, reduced snow and control, respectively. Enhanced and reduced runs are clearly separated with the control runs in between. The snow depth observations are shown in the top panel of Fig. 10; the ensemble average values of stg as well as anomalies from the control run are shown in the middle and lower panels, respectively. Irrespective of the initial snow introduced in March, there is a complete meltdown of snow occurring around June (top panel of Fig. 10). There is a clear separation of the three sets of runs, with the enhanced snow curves well above the reduced snow curves until all the snow is melted in all the runs. The area average observational values based on the snow depth data are also shown in this panel; the model melts all the snow completely, while observations suggest non-zero values over summer months. The control run curves are closer to the reduced curves than the enhanced.
The middle panel of Fig. 10 shows the ensemble mean for the enhanced, control and reduced runs. There is a separation in the ensemble means of all three viz., ensemble members of a particular run (control, enhanced and reduced) remain closer and have a low intra-ensemble standard deviation. The bottom panel of Fig. 10 shows the ensemble mean anomaly from the control run for the enhanced and reduced runs. The ensemble mean snow anomaly for the enhanced snow run is positive, and the ensemble mean snow anomaly for the reduced snow run is negative.
4.2 Pre-monsoon circulation
The monsoon circulation is established at the time of the onset. The climatological onset date at the southern tip of the Indian peninsula is 1 June with a standard deviation of eight days (Soman and Kumar, 1993); however, features of the monsoon circulation appear in May (Krishnamurti, 1985). We demonstrate that prescribed initial enhanced and reduced snow mass anomalies give rise to features that are significantly different a season after the anomalies are introduced.
A local significance test is applied to the enhanced and reduced snow anomaly experiments using the two-tail t-test (Chervin and Schneider, 1976). In all the figures to follow, the geographical location for which the difference in ensemble means for the two sets of runs is significant at the 90% and 95% level are the light and darkly shaded region, respectively. If there is no shaded region in a figure, it implies the t-test was not satisfied at either level. The difference of ensemble means is shown with the convention that the value for the enhanced ensemble mean is subtracted from the reduced ensemble mean.
In order to study the effect of enhanced (reduced snow), three land surface variables, storage on ground stg, soil wetness of surface sws and soil wetness of root zone swr, have been examined over the Eurasian continent for both enhanced and reduced model runs. Whereas the variable sws is rapidly changing, swr is the soil wetness of an intermediate zone and is more slowly varying. For the pre-monsoon month, the reduced snow has less soil moisture than the enhanced snow case.
Figure 11 shows these results. The top panel of this figure shows that in May, the reduced ensemble mean has less storage on ground (6 - 9 mm water equivalent) over a large region over Eurasia. The difference is significant at the 95% level. The middle panel shows the soil wetness of the surface: the soil is less wet for the ensemble mean of the reduced case by ~9%. For the same region, soil wetness of the subsurface is also drier in the case of the reduced set. The differences for soil wetness in both these cases are significant at the 95% level.
Sensible heat flux, latent heat flux and surface temperature for May are shown in Fig. 12. From the top panel of Fig. 12, it is seen that sensible heat flux in May is significantly higher for the reduced snow experiment in the region where the snow was reduced. The middle panel indicates the latent heat flux is higher for the reduced snow set. The ensemble mean difference is significant at the 95% level. There are also other regions over the land and ocean where there are significant differences. These have not been examined for field significance.
Although temperatures are higher in the reduced snow case (differences of up to 2o K over the Eurasian region), the t-test is not passed at 90% level for temperature. However the differences are large-scale and are of the same order of magnitude as those obtained in the composite analysis for high and low snow years in western Eurasia. Ensemble mean differences in temperature for the 700 mb, 500 mb and 200 mb levels were computed. Since the average height of the Tibetan Plateau is above 4,000 m, the 500 mb level is within the planetary boundary layer and hence the 500 mb temperature is most influenced by surface conditions in this region. Very large differences at the 500 mb level extending all the way to northeast Asia are apparent in Fig. 13. There is a suggestion of a wave train emanating from the region where snow anomalies were introduced at the 500 mb level, with alternate regions of heating and cooling eastward, resulting in a remote response as far as ~140oE.
Individual ensemble members for both enhanced and reduced snow were examined for the feature of a wavetrain: some of the individual runs showed this while others did not. The temperatures are not statistically significant. Over the broad region surrounding Tibet, the ensemble difference is positive, implying heating over this region. This increases the north-south temperature gradient between the equator and 30oN over the Indian subcontinent for the reduced snow case. The impact of the increased temperature gradient is to increase the intensity of the circulation.
Fig. 14 shows the vertical cross section of May ensemble mean temperature (reduced - enhanced) along 60oE to 120oE area average over the latitudes (40oN to 60oN), (28oN to 40oN) and (equator to 28oN). These three regions correspond to land area south of the Himalayas (equator to 28oN), the Himalayan region (28oN to 40oN) and the Eurasian land region north of the Himalayas (40oN to 60oN). The top panel of Fig. 14 shows that for the latitudes (40oN to 60oN) there is a region of warming (cooling) in Eurasia west (east) of 60o E. For the Himalayas (middle panel of Fig. 14), warming of over 1.2oK is seen at the upper levels (200 mb). The land area of the subcontinent also gets warmed to a maximum of 0.4o K at 500 mb and to a lesser extent at lower and higher levels. However the significance test is not passed implying the response due to internal dynamics is high. It is possible that an ensemble of larger size would give a significant difference but this has not been investigated.
One of the features of the monsoon circulation is the establishment of the monsoon trough from northwest to the southeast of India. The May sea level pressure difference is shown in the Fig. 15; negative values over India indicate that the monsoon trough is more pronounced in the low snow years. The t-test is passed at 90% level.
These results indicate that ensemble mean differences suggest that favorable conditions for above normal monsoons are set up for the reduced snow case. The JJAS low level winds as well as JJAS precipitation is examined to see if the snow-monsoon relationship holds for the AGCM runs.
4.3 Monsoon Winds and Precipitation
The low level ensemble mean JJAS 850 mb level wind differences (reduced - enhanced) have been computed (Fig. 16). Southwesterlies are more pronounced in the reduced snow case compared to the enhanced snow. The lower tropospheric circulation over the Indian Ocean and the Pacific Ocean show differences in the two experiments. The circulation is weaker in the enhanced snow experiment compared to the reduced snow experiment. The JJAS 200 mb difference (reduced - enhanced) winds indicated an easterly flow between the latitudes 30oN to 45oN. There is enhanced anticyclonic flow over the Tibetan region. These features suggest the low snow years have wind patterns conducive to above normal monsoons.
Finally Fig. 17 shows the ensemble mean difference in JJAS precipitation. There is significantly increased precipitation over the Indian subcontinent. Light and dark areas in this figure are regions where the t-test is passed at the 90% and 95% levels of confidence levels, respectively.
5.CONCLUSIONS
The AGCM is able to simulate the seasonal features of snow cover and snow depth for the Eurasian continent. The AGCM simulation of seasonal snow cover as well as snow depth differ from satellite-derived snow cover and snow depth. However, the process of determining which is correct is complicated by the fact that the snow cover and snow depth sensors themselves have regions of underestimation. The regions where the model is underestimating compared to observational data need to be examined more carefully, since it is known that both
the snow cover and snow depth are underestimated in certain regions such as the boreal forests and regions close to the snowline.
Motivated by an earlier observational study of the snow-monsoon relation, an AGCM sensitivity experiment was conducted, choosing a region for reduced (enhanced) snow mass anomalies in spring. The area encompassing the Himalayas was excluded from the region of snow anomalies. It was shown that model anomalies persist for at least two months, leading to less soil wetness for the reduced snow case. The difference in means for soil wetness is significant at the 95% confidence level.
Higher temperature anomalies persisting through the troposphere, and extending well beyond the region where the snow anomalies were introduced, were established by May for the reduced snow case. May sea level pressure showed a significant difference in the two sets of runs over the land region of the subcontinent. Subsequently, there was a significant difference in both the JJAS low-level southwesterlies as well as the JJAS precipitation for the difference between the ensemble means of the enhanced and reduced cases. There was a redistribution in the precipitation pattern, with an increase in precipitation over the land area of the subcontinent and a decrease in the precipitation over the oceans south of the continent.
Large-scale temperature anomalies, persisting through the troposphere, extending well beyond the region where snow anomalies were introduced in spring were established by May. Although these large-scale temperature anomalies did not pass the t-statistic, they are of the same order of magnitude as the anomalies observed in the composite analysis using observational data sets. The model standard deviation for temperature in the mid latitude region was fairly large (~5o K).
We have conducted an experiment for a specific region using model-based snow anomalies. Determining observed snow depth anomalies is greatly restricted by the lack of global data. An experiment using observed anomalies over specific regions would elucidate the role of the location of the snow anomalies and their magnitude on the monsoon. Further model development, accounting for more realistic properties of snow, such as the varying density of snow pack, would make the snow simulation more realistic. The current model uses the surface temperature for snow accumulation and melting process and a constant water equivalent for snow.
Acknowledgements
We acknowledge financial support from NSF ATM-9321354. ASB would like to thank M.J. Fennessy, J.L. Kinter III and J. Shukla for helpful discussions.
REFERENCES:
Bamzai, A. S., 1997: Observational and Modeling Study of Snow and its relation to the Indian summer monsoon. Ph.D Thesis, Institute of Computational Sciences and Informatics, George Mason University, Va, USA. pp.178.
Bamzai, A. S. and Kinter, 1997: Climatology and Interannual Variability of Northern Hemisphere Snow Cover and Snow Depth based on Satellite Observations. COLA Technical Report # 52. Center for Ocean-Land-Atmosphere Studies, Calverton, Md 20705.
Bamzai, A. S. and J. Shukla, 1998: Relation between Eurasian Snow Cover, Snow Depth and the Indian Summer Monsoon: An Observational Study. COLA Technical Report # 53. Center for Ocean-Land-Atmosphere Studies, Calverton, Md 20705.
Bamzai, A. S. and J. Shukla, 1999: Relation between Eurasian Snow Cover, Snow Depth and the Indian Summer Monsoon: An Observational Study. J. Climate (in press)
Barnett, T. P., L. Dumenil, U. Schlese and E. Roeckner, 1988: The Effect of Eurasian Snow Cover on global climate. Science, 239, 504-507.
Barnett, T. P., L. Dumenil, U. Schlese, E. Roeckner and M. Latif, 1989: The Effect of Eurasian Snow Cover on regional and global climate variations. J.Atmos.Sci., 46, 661-685.
Chang, A. T. C., J. L. Foster and D. K. Hall, 1987: Nimbus-7 SMMR Derived Global Snow Cover Parameters. Annals of Glaciology, 9, 39-44.
Chang, A. T. C. , J. L. Foster and D. K. Hall, 1990: Satellite sensor estimates of Northern Hemisphere snow volume. Inter. J. Rem. Sen., 11, 167-171.
Chang, A. T. C., J. L. Foster, D. K. Hall, H. W. Powell and Y. L. Chien, 1992: The Pilot Land Data System: Data Set Documentation, Product Description and User's Guide for the Nimbus-7 SMMR Derived Global Snow Cover and Snow Depth Data Set, NSIDC DAAC, Boulder, Colorado.
Chervin, R. M. and S. H. Schneider, 1976: On determining the statistical significance of climate experiments with general circulation models. J.Atmos.Sci., 33, 405-412.
Cohen, J. and D. Rind, 1991: The Effect of Snow Cover on Climate. J. Climate, 4, 689-706.
Delworth, T. and S. Manabe, 1988: The influence of potential evaporation on the variability of simulated soil wetness and climate. J. Climate, 1, 523-547.
Delworth, T. and S. Manabe, 1989: The influence of soil wetness on near-surface atmospheric variability. J. Climate, 2, 447-1462.
DeWitt, D., 1996: The Effect of the Cumulus Convection on the Climate of the COLA General Circulation Model. COLA Technical Report No. 27, Center for Ocean-Land-Atmosphere Studies, Calverton, Md 20705.
Dong B. and P. J. Valdes, 1998: Modelling the Asian summer monsoon rainfall and Eurasian winter/spring mass. Q.J. R. Meteorol. Soc., 124, 2567-2596.
Douville, H. and J.-F. Royer, 1996: Sensitivity of the Asian summer monsoon to an anomalous Eurasian snow cover within the Meteo France GCM. Clim.Dyn., 12, 449-466.
Fennessy, M. J., J. L. Kinter III, B. Kirtman, L. Marx, S. Nigam, E. Schneider, J. Shukla, D. Straus, A. Vernekar, Y. Xue and J. Zhou, 1994: The Simulated Indian Monsoon: AGCM Sensitivity Study. J. Climate, 7, 33-43.
Foster, J., G. Liston, R. Koster, R. Essery, H. Behr, L. Dumenil, D. Verseghy, S. Thompson, D. Pollard, and J. Cohen, 1996: Snow Cover and Snow Mass Intercomparisons of General Circulation Models and Remotely Sensed Datasets. J. Climate, 9, 409-426.
Harshvardhan, R. Davis, D.A. Randall and T.G. Corsetti, 1987: A fast radiation parameterization for general circulation models. J.Geophys.Res., 92,1009-1016.
Kinter, J. L. III, J. Shukla, L. Marx and E. K. Schneider, 1988: A simulation of the winter and summer circulation with the NMC global spectral model. J. Atmos. Sci., 45, 2486-2522.
Kinter, J. L. III, D. Dewitt, P. A. Dirmeyer, M. J. Fennessy, B. P. Kirtman, L. Marx, E. K. Schneider, J. Shukla and D. M. Straus, 1997: The COLA-Atmosphere-Biosphere General Circulation Model Volume 1 Formulation. COLA Technical Report # 51, Center for Ocean Land Atmosphere Studies, Calverton Md
Kitoh, A., 1994: AGCM experiments on Tibetan Snow and Monsoon. WMO International Conference on Monsoon Variability and Prediction, International Center for Theoretical Physics, Italy, 9-13 May, pp. 661-665.
Kitoh, A., S. Yukimoto, A. Noda, S. Nakagawa and T. Motoi, 1997: Interannual variability and future possible changes of the Asian summer monsoon as simulated in the MRI coupled GCM. 1st WMO International Workshop on Monsoon Studies (IWM-1), Denpasar, Indonesia, 24-28 February.
Kukla, G. and D. A. Robinson, 1981: Accuracy of snow and ice monitoring. Snow Watch 1980. Glaciological Data Report GD-5, pp. 91-97.
Krishnamurti, T. N., 1985: Summer Monsoon Experiment-a review. Mon.Wea.Rev., 113, 1590-1626.
Kutzbach, J., 1981: Monsoon climate of the early Holocene: Climate experiment with the earth's orbital parameters for 9000 years ago. Science, 214, 59-61.
Lacis A.A. and J.E. Hansen, 1974: A parameterization for the absorption of solar radiation in the earth's atmosphere. J. Atmos. Sci., 31, 118-133.
Meehl G.A., 1994a: Influence of land surface in the Asian summer monsoon: External conditions versus internal feedbacks. J. Climate, 7, 1033-1049.
Meehl G.A., 1994b: Coupled land-ocean-atmosphere processes and south Asian monsoon Variability. Science, 266, 263-267.
Mooley, D., and B. Parthasarathy, 1983: Variability of the Indian summer monsoon and tropical circulation features. Mon. Wea. Rev., 111, 967-978.
Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa-Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 978-1002.
Ose T., 1996: The Comparison of the Simulated Response to the Regional Snow Mass Anomalies over Tibet, Eastern Europe, and Siberia. J. Met. Soc. Japan, 74, 845-866.
Parker, D. E., M. Jackson and E. B. Horton, 1995: The GISST2.2 sea surface climatology and sea-ice climatology. Climate Research Technical Note CRTN 63, Hadley Centre Meteorological Office, Bracknell, Berkeshire RG12 2SY, UK.
Rayner, N. A., C. K. Folland, B. Horton and D. E. Parker, 1995: A new Global sea Ice and Sea Surface Temperature (GISST) data set for 1903-1994. Presented by C.K. Folland to the Wadati Conference on Global Change and the Polar Climate, Tsukuba, Japan, 7-10 November. Available as Hadley Centre Internal Note No. 69, Hadley Centre Meteorological Office, Bracknell, Berkeshire UK.
Robinson, D. A., F. T. Keimig and K. F.Dewey, 1991: Recent variations in Northern Hemisphere snow cover. Proc 15th Annual NOAA Climate Diagnostics Workshop, Asheville North Carolina, pp. 219-224.
Robinson, D. A., K. F.Dewey and R. R. Heim, Jr. 1993: Global Snow Cover Monitoring: An Update. Bull. Am. Met. Soc., 74, 1689-1696.
Ropelewski, C. F., A. Robock, and M. Matson, 1984: Comments on "An Apparent relationship between the Eurasian Snow Cover and Indian Monsoon Rainfall". J. Clim. Appl. Meteor., 23, 341-342.
Ropelewski, C. F., J. E. Janowiak and M. S. Halpert, 1985: The analysis and display of real time surface climate data. Mon. Wea. Rev., 113, 1101-1107.
Sellers P.J., Y. Mintz, Y.C. Sud and A. Dalcher, 1986: A simple biosphere model (SiB) for use within general circulation models. J. Atmos Sci., 46, 503-531.
Shukla, J., 1984: Predictability of Time Averages: Part II. The Influence of the Boundary Forcing. Problems and Prospects in Long and Medium Range Weather Forecasting,(eds.) D.M. Burridge and E. Kallen, Springer-Verlag, London, pp. 155-206.
Soman, M. K. and K. Krishna, 1993: Space-time evolution of meteorological features associated with the onset of the Indian summer monsoon. Mon. Wea. Rev., 121, 1177-1194.
Vernekar, A. D., J. Zhou and J. Shukla, 1995: The Effect of Eurasian Snow Cover on the Indian Monsoon. J. Climate, 8, 248-266.
Xie, P. P., and P. A. Arkin, 1996: Analyses of Global Monthly Precipitation Using Gauge Observations, Satellite Estimates, and Numerical Model Predictions. J. Climate, 9, 840-858.
Xue, Y., P. J. Sellers, J. L. Kinter III, and J. Shukla, 1991:A simplified biosphere model for global climate studies. J. Climate, 4, 345-364.
Yasunari, T., A. Kitoh, and T. Tokioka, 1991: Local and remote responses to excessive snow mass over Eurasia appearing in the northern spring and summer climate-A study of the MRI GCM. J. Met. Soc. Japan, 69, 473-487.
Yeh, T. -C., R. Wetherald and S. Manabe, 1983: A model study of the short term climatic and hydrological effects of sub-snow cover removal. Mon. Wea. Rev., 111, 1013-1024.
Yeh, T.-C., R. Wetherald and S. Manabe, 1984: The effect of soil moisture on the short-term climate and hydrology change - A numerical experiment. Mon. Wea. Rev., 112, 474-490.
Zhou, J., 1993: GCM Study of Interannual variability of Indian Summer Monsoon: The impact of Anomalous Eurasian Snow Cover. Ph.D. Thesis, University of Maryland. pp 136.
Zwiers, F. W., 1993: Simulation of the Asian summer monsoon with the CCC GCM-1. J. Climate, 6, 470-486.
FIGURE CAPTIONS
Figure 1: Model simulated and observed winter (DJFM) snow depth climatology and standard deviation. Top and lower panels of left-hand-side show results based on observational data; top and lower panels of right-hand-side show results based on model simulation. Unit is mm water equivalent.
Figure 2: Spring (AM) snow depth climatology and standard deviation. Top and lower panels on left-hand-side show results based on observational data; top and lower panels on right-hand-side show results based on model simulation. Unit is mm water equivalent.
Figure 3: Annual cycle of Eurasian snow. Top panel shows simulated and observed annual cycle of Eurasian snow cover. Unit for snow cover is 106km2. Lower panel shows simulated and observed annual cycle of area-average snow depth for Eurasia. Unit is mm water equivalent.
Figure 4: Interannual variability of Eurasian snow cover. Top and lower panels show results for winter (DJFM) and spring (AM) seasons for region (zero to 120oE and 35oN to 65oN). Unit is 106km2.
Figure 5: Interannual variability of Eurasian snow depth. Top and lower panels show results for winter (DJFM) and spring (AM) seasons for region (zero to 120oE and 35oN to 65oN). Unit is mm water equivalent.
Figure 6: Winter (DJFM) precipitation over Eurasia. Top and lower panels show results based on observations and model simulation. Unit is mm day-1. Thick dashed line shows the T=0 (freezing line) for DJFM.
Figure 7: Spring (AM) precipitation over Eurasia. Top and lower panels show results based on observations and model simulation. Unit is mm day-1. Thick dashed line shows the T=0 (freezing line) for AM.
Figure 8: Initial conditions for spring snow depth for enhanced and reduced sensitivity run for 1982 run. Lower panel shows difference in snow depth between the two. Unit is mm water equivalent.
Figure 9: Seasonal snowmelt: enhanced, reduced and control runs. Top, middle and lower panels show the ensemble average snow melt for regions (zero to 190oE and 20oN to 90oN), (zero to 120oE and 35oN to 65oN) and (70oE to 100oE and 28oN to 40oN), respectively. Unit is mm water equivalent.
Figure 10: Evolution of area average snow depth for enhanced, reduced and control runs. Top panel shows storage to ground, stg. Unit is m. Middle and lower panels show ensemble mean snow depth and ensemble average anomaly, respectively. Region considered is (zero to 120oE and 35oN to 65oN). Unit for middle and lower panel is mm water equivalent.
Figure 11: May storage to ground (stg), soil wetness (sws) and soil wetness of root zone (swr): Ensemble mean difference (reduced - enhanced). Top panel shows result for stg. Unit in mm water equivalent. Middle and bottom panel show results for sws and swr, respectively. Unit for middle and lower panel is % to saturation level. Regions for which difference in ensemble means is significant at 90% and 95% levels are light and dark shaded regions, respectively.
Figure 12: May sensible heat flux, latent heat flux and surface temperature: Ensemble mean difference (reduced - enhanced). Top and middle panels show results for sensible heat flux and latent heat flux, respectively. Unit is Wm-2. Lower panel shows results for surface temperature. Unit is oK. Regions for which difference in ensemble means is significant at 90% and 95% levels are light and dark shaded region, respectively.
Figure 13: May mean temperature: Ensemble mean difference (reduced - enhanced). Top, middle and lower panels show results for 200 mb, 500 mb and 700 mb, respectively. Unit is oK.
Figure 14: May vertical cross section of Temperature: Ensemble mean difference (reduced - enhanced). Top, middle and lower panels show area average over latitude belts (40oN to 60oN), (28oN to 40oN) and (equator to 28oN), respectively. Bold line in middle panel shows profile of mean orography of model over region. Unit is oK.
Figure 15: May Sea Level Pressure: Ensemble mean difference (reduced - enhanced). Regions for which difference in means is significant at 90% and 95% levels are light and dark shaded region, respectively. Unit is departure from 1000 mb.
Figure 16: JJAS Winds: Ensemble mean difference (reduced - enhanced). Top and lower panels show winds at 850 mb and 200 mb levels, respectively. Unit is m s-1.
Figure 17: JJAS precipitation: Ensemble mean difference (reduced - enhanced). Unit is mm day-1. Regions for which difference in means is significant at 90% and 95% levels are light and dark shaded region, respectively.