Improvement in Near Surface NWP Model Output using Kalman Filtering Technique: A Case Study for Trombay Site

Abstract

Numerical Weather Prediction (NWP) models exhibit systematic errors in the forecast of near surface atmospheric parameters due to various factors like grid resolution, parameterization schemes, treatment of sub-grid scale phenomena, data for initial and boundary conditions and interpolation techniques. One of the methods for reduction in model errors is the use of Kalman filter algorithm which recursively combines model output and observations such that the systematic errors are minimized. In the present study, the Kalman filter algorithm is utilized for correction of model output from The Air Pollution Model (TAPM) for the year 2013. The variables corrected are 2-m air temperature, 2-m relative humidity and zonal and meridional wind components at 10-m. Hourly observations of the same variables available at Trombay site are used in the study. In the present study, it is seen that, both wind speed and wind direction are better reproduced after Kalman filtering, in addition to near surface air temperature and relative humidity. Also, on an annual basis, biases in all the variables are eliminated. The standard statistical indices of model performance computed after Kalman filtering are superior to those computed using only model output. Time series plots of bias and RMSE in model after Kalman filtering indicate the advantage of Kalman filtering.