The months of June, July, August and September comprise of the summer monsoon season which is the most important period for the Indian Sub-continent from a view point of agriculture and water management. The monsoon gets generated due to the differential heating of the Indian sub-continent and the North Indian Ocean. The model simulations for the Indian sub-continent would be termed as good if they can reproduce the primary features of the Indian summer monsoon namely, the heat low over the north western parts of India, the axis of the monsoon trough, the low level jet over the south peninsula and the Tropical Easterly Jet near to 100 N.
4.1. Spatial distributions:
In the following, the seasonal (JJAS) mean spatial distributions of 2m air temperature, mean sea level pressure (MSLP), wind at 850 hPa and 200 hPa and surface rainfall are presented and the salient features only are portrayed.
4.1.1.2m air temperatures: The spatial distributions of mean 2m air temperature (0K) simulated by the WRF model and the CCSM4 model and ERA5 reanalysis are presented (Fig.2 a,b,c). The difference between WRF and ERA5, which indicate model bias are presented in (Fig.2d.) and the difference of the mean 2m air temperatures between the two periods of (2007-2021)and(1976-2005) for the WRF model and ERA5 reanalysis are presented in (Fig.2e and 2f). The temperature minima over the Himalayan and the Tibetan plateau region are well reproduced. The WRF model simulated temperatures show higher values over the northwest part of the Indian sub-continent in comparison to the temperatures of CCSM4 and ERA5. The model also shows higher values in the northern parts of India and slightly lower values over Kerala and parts of Karnataka compared to the CCSM4 but with higher values over the Western Ghats compared to the ERA5 reanalysis. This slight variation may be attributed to the differences in the representation of orography of the Western Ghats in the WRF and CCSM4 models. The meridional gradient of the temperature is more in the WRF model as compared to the CCSM4 and ERA5. The model simulated surface temperatures are similar to those in the CCSM4 while in the ERA5 analysis they show slightly lesser temperatures over the Western Ghats. The bias for the surface temperature (WRF-ERA) shows positive bias over Rajasthan and parts of Uttar Pradesh while it shows negative bias over Tamil Nadu, Kerala and western part of the Himalayan region. From the figures (2e and 2f) a strong warm tendency of 0.50 K is observed over Indian landmass region in ERA5 analysis whereas in the WRF model a warm tendency of 0.250K is observed in most parts of India except Central India which indicates underestimation of temperatures over that region by the WRF Model. Further the deviation is higher over central India which is generally hotter than the southern parts. This underestimation of the temperatures may be attributed to the adopted physical parameterization schemes in the WRF model.
4.1.2. Mean Sea Level Pressure (MSLP): The spatial distributions of MSLP (hPa) simulated by the WRF model and the CCSM4 model and ERA5 reanalysis are presented (Fig.3 a,b,c). The difference between WRF and ERA5, which indicate model bias are presented in Fig.3d. and the difference of the MSLP between the two periods of (2007-2021) and (1976-2005) with the WRF model and ERA reanalysis are presented in Fig.3e and 3f. The model simulated MSLP is in good agreement with CCSM4 simulation and ERA5 analysis but the WRF model shows strong pressure gradients between northern parts of India and southern parts of India, which correspond to higher temperature gradients over the same region. The observed lowest pressure over the Northwest India is around 1000hPa in WRF, CCSM4 and ERA5 analyses whereas this area of lowest pressure extends up to the eastern parts of India in the WRF Model. The extent of the monsoon trough is better simulated in the WRF model where the negative bias could be seen right from the Northwest parts up to the Northeast parts of India (Fig.3d). The figure (3e, f) depicts the MSLP anomalies for the WRF and ERA5. The strong positive pressure tendency is observed over North-India, Northeast India and weak positive pressure tendency is observed over Tamil Nadu in ERA5 analysis whereas in the WRF model the weak positive pressure tendency is observed over South Peninsular India only.
4.1.3. Wind flow at 850 hPa and 200 hPa:
The wind flow at 850hPa level (Fig.4 a,b,c) shows similar patterns between CCSM4 and ERA5, while the WRF model simulated winds show a slightly weaker magnitude winds over the Arabian Sea and higher over the Bay of Bengal. The monsoon westerlies and the monsoon trough are very well simulated by the WRF model. The WRF model simulated stronger westerlies over the Bay of Bengal as compared to the CCSM4 and ERA5. A stronger cyclonic flow in the vicinity of the monsoon trough region can be clearly seen as compared to that in the CCSM4 and ERA5 analysis. The figure 4 d and 4e depicts the wind anomalies for the ERA5 and WRF model. A strong positive tendency of magnitude (0.75 to 1 m/s) is observed over the Arabian Sea extending up to the Bay of Bengal in WRF model whereas a weak positive tendency is limited to the southwest part of the Arabian Sea as observed in the ERA5 analysis. A strong negative tendency is observed over the Northwest (Gujarat and Rajasthan States) and Central India regions in ERA5 analysis whereas a weak negative tendency is observed in the WRF model with a lesser magnitude of 0.5 (m/s).The WRF simulated 200hPa winds (not shown) are in good agreement with CCSM and ERA analysis, while the WRF simulated higher winds around 0 to 50 N. The anti-cyclonic flow around 270 N in WRF model is slightly less with a magnitude of 6 m/s when compared to CCSM4 and ERA5. The easterlies around 100 N are well simulated, in correspondence with CCSM4 and ERA5.
4.1.4. Rainfall: The spatial distribution of rainfall of the CCSM, WRF model and IMD gridded rainfall are depicted in figure 5 (a, b, c). The model simulated rainfall over the Northern part of India shows good correspondence with CCSM and IMD rainfall. The WRF model is able to capture the spatial patterns of the precipitation over Central India and Northeast regions with higher magnitudes and West coast region with lesser magnitudes as compared to the IMD rainfall. The maximum amount of rainfall during the summer monsoon is due to the convection generated in the region of Monsoon trough and its adjoining low pressure areas. The deviation of rainfall between IMD and WRF over the west coast region may be due to the hilly terrain in that region. From the figure 5d which shows the bias of mean rainfall of the period 2007-2021 for the monsoon season that a positive bias (higher rainfall) over the central India and northeast parts of India and a negative bias (lower rainfall) over the west coast, northwest and northern parts of India are noted. The figure 5(e, f) represents the rainfall anomalies of WRF and IMD for the periods (2007-2021) minus (1976-2005). A positive tendency is observed over the Western Ghats and central parts of India in the WRF model whereas in the IMD rainfall the positive tendency is observed in some parts of Western Ghats and in Gujarat. This may be attributed to a changing climate in the global warming era.
4.1.5. Vertical Integrated Moisture Flux Convergence: The VIMFC along with its transport is calculated in Fig.6(a-c) to investigate the transport of moisture towards the Indian landmass region of CCSM4, WRF and ERA5. The VIMFC is calculated by integrating vertically the horizontal moisture flux convergence/divergence between the surface and 300 hPa. The negative values indicate the convergence and the positive values indicate the divergence. It is evident that there is a moisture convergence along the coast of Somalia and the northern part of the Arabian Sea and that the moisture is transported from the Arabian Sea to the Indian landmass region in CCSM4, WRF and ERA5. There is a strong convergence over the west coast and northeast parts of India as observed in the WRF model compared to the CCSM and ERA which is in good correspondence to the rainfall over those regions. Also the WRF model had simulated the cyclonic circulation very well over the Head Bay of Bengal region which is in good agreement with CCSM4 and ERA5 analysis.
4.2. Rainfall in homogenous regions:
The evaluation of rainfall in the different homogenous regions has been made from a view point of statistical metrics (i.e.) Mean, Standard deviation, Root mean square error and correlation coefficients (CC). As a part of the evaluation, six homogenous regions (Fig.7) are considered as mentioned earlier Parthasarathy et al. 1995. Among the 6 homogeneous regions, the various statistical metrics have been calculated for 5 homogenous regions after excluding the hilly regions and for All India are presented in Table 2. The CC values are higher in all the homogenous regions 0.8 and above. The CC for All India is 0.6. The correlation is highest in the Central North East region which is the area of convergence of the Monsoon Trough. The RMSE shows minimum value in the South Peninsular region where there is usually substantial rainfall especially during the Monsoon onset. The maximum Mean and standard deviation values are observed over the West-Central region in both the WRF and IMD. The various statistical metrics show that the model could simulate the characteristics of the monsoon rainfall.
4.3.1 Tropospheric temperature gradient
Fig.8 shows TTG from the WRF model and ERA analysis. The temperature gradient between Southern and Northern regions averaged over 600-200hPa is defined as the TTG. The TTG characterizes the strength, onset and withdrawal of the Indian Summer Monsoon. For accessing the intensity of the model simulated land-sea gradient, the temperature difference between South (150S-50 N, 400E-1000E) and North (50 N-350 N, 400E-1000E) regions of the Indian sub-continent is analysed and presented. The blue line indicates the ERA analysis and red line indicates the WRF model simulation. The WRF model very well simulated the structure of the Tropospheric temperature gradient throughout its annual cycle. This can be attributed to the fact that the southern region contains more ocean than the northern region and that in the pre-monsoon season the temperature between land and sea would be more which gradually decreases once the monsoon sets in. The model TTG is in perfect agreement with the ERA5 analysis which indicates the good simulation of temperature distribution over the Indian sub-continent.
4.3.2. Meridional distribution of temperature: Fig.9 (a-c) depicts the time-latitude sections of the monthly mean temperatures at 850 hPa level along 780 E. From the figures, it is clearly observed that there is a gradual increase of temperatures from lower latitudes to higher latitudes with a maximum during April-June. The maximum temperature is observed between 200N to 260 N in CCSM whereas the maximum is observed between 180N to 300 N in WRF model. The WRF model is slightly overestimated the temperatures compared to ERA analysis.
4.3.3.Zonal wind cross-section: Fig.10(a-d) shows the latitude-height cross section of the zonal winds along 780 E for pre-monsoon and monsoon Season. The WRF model simulated the patterns both for pre-monsoon and monsoon season well which are in good correspondence with ERA5 analysis. The strong westerlies in the upper troposphere are observed at 220N to 300 N and slightly strong easterlies are observed at lower latitudes in WRF model compared to ERA during pre-monsoon season. During the monsoon season, the strong westerlies are observed in the lower troposphere and stronger easterlies in upper troposphere at a level of nearly 200hPa during monsoon season which are in good agreement with ERA5 but with slight stronger magnitudes. The strong westerlies in the upper troposphere are observed over middle latitudes in the WRF which are consistent with ERA5.
4.3.4. Vertical velocity distribution: The latitude-pressure profile of vertical velocity (ω) has been analysed to observe the vertical motion of atmosphere and is presented in Fig.11 (a, b). The vertical motion in the atmosphere plays a key role in the transfer of mass and energy leading to the formation of clouds which affects the atmospheric stability. From the figure, the positive values signify sinking and negative values signify rising motion. A strong rising motion in the WRF is observed at around 270 N and a weak rising motion has been seen over 100 N-200 N which is in agreement with ERA. A rising branch indicates an active convection over that region.
4.4. 1 Sub-divisional evaluation of temperatures for the pre-monsoon
In the pre-monsoon season, weather over the Indian sub-continent largely defines the performance of the ensuing monsoon since monsoon is a magnificent heat engine that evolves due to the differential heating between land and sea. An evaluation of temperature simulation is performed in terms of Mean and Standard deviation for the subdivisions (fig 12) during the pre-monsoon season. The statistical metrics are shown for 30 sub divisions in table 3. From the results, it can be seen that the WRF model simulates the better mean values of temperatures over 20 sub divisions whereas standard deviation is over 22 sub divisions better than CCSM4. The pre-monsoon temperatures are very important since much of the moisture inflow will be there from the adjoining ocean to the main land. The gradient between the land and the ocean facilitates the advection of moisture from the ocean to the land. The pre-monsoon heating of the land helps the building of vertical velocities and thereby leading to rainfall during the monsoon season. A good simulation in terms of the statistical metrics implies a better prediction of sub regional scale temperatures by the model.
4.4.2 Sub-divisional evaluation of surface temperature and rainfall during monsoon season
The evaluation of model simulated surface temperature and rainfall is performed by calculating the Mean and Standard deviation for the monsoon period in the sub divisions (Fig.12) of Indian sub-continent with IMD gridded data and CCSM4 data. Tables 4 and 5 represent the statistical metrics for temperature at 2m and rainfall respectively.
The WRF model had simulated the mean temperatures better than CCSM4 for 22 subdivisions and a positive bias is observed over Northern and Central parts of India compared to IMD and CCSM4. The WRF model had shown the higher standard deviation values than CCSM4 in most of the subdivisions. The magnitudes of standard deviation range from 0.5 to 2.1 for IMD whereas the values range from 0.5 to 3.5 in WRF model.
A good amount of rainfall is observed during the monsoon season with 24 sub divisions yielding rainfall higher than 5mm/day. The highest amount of rainfall is received over Konkan& Goa (25.7 mm/day) and Coastal Karnataka (22.9 mm/day) and the lowest rainfall is over West Rajasthan (2.6 mm/day) and Tamil Nadu (3.3 mm/day. The model had produced better mean values of rainfall over 20 sub divisions and better standard deviation values of rainfall over 14 sub divisions compared to the rainfall simulated by CCSM4.
The above presentation of the results simulated by the WRF model run on a climate mode clearly demonstrates the advantage of using a WRF model on a finer resolution to simulate future climate.