Appraisal of data assimilation techniques for dynamical down scaling of the structure and intensity of tropical cyclones

The dynamical down scaling is one of the most useful techniques for the understanding of physical mechanisms associated with atmospheric phenomena. The advanced data assimilation (DA) techniques and complex numerical weather prediction (NWP) models are being used for dynamical downscaling. In this work, we have generated high resolution reanalysis (6 km) for three tropical cyclones (TCs) Phailin (2013), Nilofar (2014), and Chapala (2015) which underwent Eyewall replacement cycle and rapid intensification over North Indian Ocean (NIO). The present work aims to identify appropriate methodology for generating analysis useful for understanding the environmental controls and internal physical mechanisms associated with intensification processes and structural changes of TCs. Weather Research and Forecasting (WRF) model and its four-dimensional variation (4DVAR), hybrid three-dimensional ensemble-variational (3DEnVAR) and hybrid four-dimensional ensemble-variational (4DEnVAR) DA techniques are compared to assess their impact on the accuracy of the generated analysis. The impact of data assimilation is quantified by calculating errors in position, mean sea level pressure (MSLP) and maximum surface wind speed (V_max) with respect to Indian meteorological Department (IMD) best track dataset of TCs. The intensity evolution simulated by the model is validated by comparing changes in MSLP, V_max, and boundary layer and mid tropospheric relative humidity. The concentric double eyewall structure of TC Phailin and Chapala is validated using Meteosat-7 image developed by United States Naval Research Laboratory. The model forecast skill scores viz. equitable treat score (ETS), false alarm ratio (FAR), the probability of detection (POD) and bias (BIAS) are calculated to identify best DA technique. The intensification and structural features of these TCs are compared for different DA techniques. It is found that the hybrid DA techniques improve the quality of reanalysis over the non-hybrid variational DA techqine. Overall, the simulation using hybrid 4DEnVAR DA technique is found to be the better simulation of track, intensity changes and structural characteristics of TCs over NIO.