Correction of mesoscale model daily precipitation data over Northwestern Himalaya

Maximum numerical weather prediction models have their own inherent biases and these biases have high impact on accuracy of weather forecast. Hence, bias correction is an essential part of any study for any model output datasets. The current study uses a weather research and forecasting (WRF) model, simulated daily precipitation of winter season (December to February: DJF) for the period of 2010–2011 to 2016–2017 (7 years) for the bias correction and validated against observed precipitation of Snow and Avalanche Study Establishment (SASE), India. For the first time, three different methods, i.e., empirical quantile mapping (QM), linear scaling (LS), and regression (REG) have been studied for the bias correction over the Northwest Himalaya region. In order to identify the best method out of these three, four statistical measurements, i.e., skill score (SS) and its decompositions, bias in percentage, root mean square errors (RMSE), and percentile values have been examined. Based on the analysis of SS and RMSE, it is worth to note that the QM method is found to be most suitable method for the December and February forecast of WRF model, whereas the LS approach is most suitable for the January forecast. Comparison based on Taylor’s diagram and percentiles via boxplot shows that the quantile mapping approach is most advisable for bias correction to the model simulated precipitation dataset over Northwest Himalaya region.