Role of Land Surface Processes on land falling Tropical Cyclones and Monsoon Depressions over the Indian Region

The Indian region is suffered by oceanic intense vortices such as tropical cyclones (TCs) and monsoon depressions (MDs) which are periodic in nature and have potential to produce heavy rainfall in the coastal regions during landfall. Generally, the tropical cyclones form in pre- (April – May) and post- (October – December) monsoon season and the MDs form during monsoon season (June – September) in North Indian Ocean (NIO, including both Bay of Bengal and Arabian Sea). The land surface processes are very important in modulating the characteristics of these systems during landfall or when they are close to land or when they are over the land. Further, the Indian region is recognized as one of the major ‘hot spots' for soil moisture variations that significantly influence rainfall. In this study, we hypothesize that the mesoscale model performance can be improved by incorporating more realistic soil moisture/soil temperature (SM/ST) profiles (with depth) as surface initial conditions. However, as of now, long-term high resolution SM/ST observations do not exist at regional scales, particularly over the Indian domain. Further, incorporating the point observations of SM/ST in a high resolution mesoscale model is complicated because of spatiotemporal variations in land surface conditions such as topography, soil texture, and vegetation characteristics. Therefore, in the present study, SM/ST gridded profiles with depth are modelled at 4 km horizontal resolution using High Resolution Land Data Assimilation System (HRLDAS) and investigated its role/credibility on simulation of monsoon depressions (MDs). The LDAS typically use observed/analyzed initial data (SM, ST, canopy water content, skin temperature and water equivalent of accumulated snow depth) and atmospheric surface forcing data such as 2m temperature and mixing ratio, 10m winds, surface pressure, model elevation, surface total rainfall rate, downward short wave radiation at surface and, downward long wave radiation etc. The atmospheric forcing data mentioned above are derived from the Moern-Era Retrospective analysis for Research and Applications (MERRA) reanalysis except rainfall rate which is obtained from (TRMM) 3hourly rain rate. The 3-hourly rainfall is interpolated to 1 hourly rain rate and to the 4 km horizontal grid. The initial land surface fields are derived from global data assimilation system (GDAS) analyses of NCEP. The land surface characteristics are based on the 24-category land use and 16-category soil texture, green vegetation fraction available from USGS. The LDAS output includes SM, ST and surface energy balance components such as latent, sensible and ground heat fluxes etc. For this study, the LDAS was initialized on 1 January 2006 and run up to 30 June 2010. The MERRA-based 2m temperature and TRMM rainfall are in acceptable agreement with the observed data over the IMR. The time series of HRLDAS-based top layer SM was in good agreement with observed data. Two severe cyclones, Aila during pre monsoon season (23-27 May 2009) and Nilam during post-monsoon (28 October – 1 November 2012) were selected for this study as they spent almost 2 days over the land after making landfall and produced heavy to very heavy rainfall. Two numerical experiments are conducted for each cyclone using ARW modeling system at 4 km resolution. Model runs have been started ~12 hours before the landfall time. First experiment uses climatological SM/ST profiles as surface initial conditions (known as CNTL) while, in second experiment (known as LDAS), the LDAS-based SM/ST profiles has been utilized for model surface conditions. Model results showed clear benefit of using realistic SM/ST as surface initial conditions to mesoscale models at high resolution in simulating inland track, intensity changes and therefore the rainfall distribution. The CNTL-simulated tracks are more diverse with respect to observed track and it could not show the intensity evolution. Particularly, Aila cyclone sustained its severe cyclonic storm intensity for about 24 hours even after the landfall. This feature is not captured by the CNTL run and hence the rainfall prediction is poor. However, the LDAS could predict the intensity in more realistic sense and succeeded in predicting the inland rainfall prediction, particularly, the heavy rainfall epochs. In case of Nilam cyclone, after landfall most parts of Andhra Pradesh received heavy to very heavy rainfall which was well reproduced by the LDAS run. It is also noticed that the LDAS could predict the intensity of the system during landfall will be cyclonic stage as observed. These features are not seen in CNTL experiments. A total of six monsoon depressions are studied in the period 2007 – 2011 and all the cases spent at least more than 60 hours over the land. The overall simulation of MDs can be improved if the land surface forcing is improved. Results indicate that the HRLDAS provides more realistic heterogeneity in SM/ST fields as compared to those of climatological values. The SM maps obtained from LDAS are compared with the previous day rainfall as no SM observation is available improved and noted that they are in good coherence as compared with that of CNTL. In case of CNTL run, the movement of MDs is deviated from observed, yielding larger track errors. The track errors are significantly reduced with the incorporation of HRLDAS-based SM/ST profiles at different depths yielding an improvement of approximately 26%, 25%, and 24% at 24, 48, and 72 hour forecast length, respectively. LDAS run improved the portioning of the surface energy fluxes and the boundary layer instability with high equivalent potential temperature and strong updrafts at the correct time as compared to CNTL run. The surface moisture flux in CNTL run is comparatively poorer than LDAS. The vertical structure of mass flux reveals that the positive mass flux (associated with updrafts) contributed to increase the convergence at the right place in LDAS experiments. The warmer land surface in CNTL runs does not support convection as the mass flux is negative and associated with downdrafts. LDAS run could predict reflectivity echoes and associated rainfall bands efficiently. Results highlight the positive impact of incorporation of improved land surface fields on numerical prediction of convection associated with monsoon depression over the Indian region.