Verification and Bias Correction of Rainfall and Temperature Forecasts over the Babai River Basin of Nepal

Abstract

Weather Research & Forecasting model is known to exhibit systematic biases for weather variables. These biases need to be post-processed to get an optimal result before applying the weather forecast data in hydrological modeling or similar applications. In this paper, we examined the performance of Weather Research & Forecasting Model forecasts for rainfall and temperature over the Babai River basin of Nepal considering various performance indicators using statistical approach. The model was able to capture the rainfall event forecast (Rain/No Rain) sufficiently. However, the model showed poor skill in forecasting the amount and over-forecasted the rainfall. The multi-category (No Rain, Light Rain, Moderate Rain and Heavy Rain) verifications showed over-forecast in light rain and moderate rain categories and under-forecast in no rain and heavy rain categories. We examined various bias correction schemes such as distribution-derived transformations, parametric transformations and nonparametric transformations to get de-biased results. The empirical quantile mapping is the best scheme for bias correction of both rainfall and temperature. In case of temperature, the linear transformation, robust quantile mapping and smoothing spline schemes also performed well.