A study on high resolution mesoscale modeling systems for simulation of tropical cyclones over the Bay of Bengal

Abstract

Land falling tropical cyclone (TC) is one of the natural disasters producing extremely strong winds, torrential rains, floods influencing many kilometers from the point of landfall and storm surges that overwhelm miles of shores resulting loss of lives, and damages to properties. This disaster is higher in the regions covering Bay of Bengal (BoB). Therefore, the India Meteorological Department (IMD) initiated a field project, "Forecast Demonstration Project (FDP) of land falling cyclones" over the BoB to acquire detailed understanding of genesis, intensity, and structure evolution of TCs so as for better TC forecasting. A comprehensive performance of state-of-the-art mesoscale modeling systems such as Advanced Research Weather Research and Forecasting (ARW), non-hydrostatic mesoscale model of WRF (NMM) and Hurricane Weather Research and Forecasting (HWRF) etc for the simulation of land falling TCs during pilot phase of FDP (2008-2011) is presented. The study is not meant for the inter-comparison of different modeling systems. In the present study, six TCs namely Rashmi (2008), KhaiMuk (2008), Nisha (2008), Giri (2010), Jal (2010) and Thane (2011) are considered. Though different aspects of the TC such as track, intensity, structure and rainfall are studied in detail, this paper is mainly emphasized on the track and intensity prediction and associated errors. Results indicates that the high resolution mesoscale modeling systems provide better guidance for TC forecast up to 72 hours. However, the track and intensity error is relatively more when these models are initialized with coarser resolution global analyses and forecast fields. This error can be significantly reduced with the assimilation of additional regional observations into model initial conditions. The track forecast errors are calculated with respect to IMD best track observations. In case of ARW system, the forecast errors are 138, 135 and 182 km from no-assimilation experiment. The assimilation of all available observations during FDP period into model initial condition decreases the errors 72, 99 and 126 km at 24, 48 and 72 hour, respectively with an improvement of about 47%. In case of NMM model, the mean (based on 30 sub-cases) track errors are improved by about 32%, 22%, 23%, 28%, 24% and 16% at 00, 24, 48, 72, 96 and 120 hrs, respectively with data assimilation experiments compared to no-assimilation experiment. The HWRF model improved the initial position and structure significantly because of its improved vortex-relocation and initialization procedures and hence captures the rapid intensification of the TC Giri in the subsequent forecast hour.