Regional Data Assimilation with the NCMRWF Unified Model (NCUM): Impact of Doppler Weather Radar Radial Wind

• DOI: 10.1007/s00024-019-02159-7
• Publication Date: 2019-03-26T16:03:35Z
• Authors: Devajyoti Dutta; Ashish Routray; D. Preveen Kumar; John P. George
• Keywords: 13. Climate action; 01 natural sciences; 0105 earth and related environmental sciences

The Indian sub-continent often receives heavy rainfall events, especially during active phases of the south-west monsoon (SWM). The accurate prediction of such events requires a high-resolution model with data assimilation techniques which can assimilate high-density spatial and temporal observed data, like radar observations. This study is undertaken to evaluate the impact of assimilation of radial velocity from multiple Doppler weather radars (DWRs) for simulation of three convective rainfall events during the SWM using the high-resolution (4 km) National Centre for Medium Range Weather and Forecasting (NCMRWF) Unified Model (NCUM) with a three-dimensional variational (3DVAR) data assimilation system. Two numerical experiments are carried out viz. control simulation (CNTL) and a second experiment (RAD, assimilation of all observations used in CNTL plus radial winds from Indian DWRs). The study clearly shows that the assimilation of DWR data has a positive impact in the simulation of heavy rainfall events by the model. The root mean square error (RMSE) of the analyzed (simulated) u- and v-components of wind is reduced by 46% (25%) and 37% (19%) in the RAD experiment as compared to the CNTL experiment. The simulated wind fields at different pressure levels are stronger after assimilation of DWR observations, and it is also successfully produced the cyclonic circulation as compared to the CNTL simulation. The skew-T plots suggested that the convection is more properly represented by the RAD experiment than the CNTL. The value of stability parameters is properly simulated in the RAD experiment, while compared with the CNTL, the values are reasonably well matched with the observed values in all cases. The intensity of the rainfall is reasonably well simulated by the RAD as compared to the CNTL experiment. The statistical skill scores of rainfall with different thresholds are significantly improved in the RAD experiment for all the cases.