A Statistical Parameter Correction Technique for WRF Medium-Range Prediction of Near-Surface Temperature and Wind Speed Using Generalized Linear Model

A statistical post-processing method was developed to increase the accuracy of numerical weather prediction (NWP) and simulation by matching the daily distribution of predicted temperatures and wind speeds using the generalized linear model (GLM) and parameter correction, considering an increase in model bias when the range of the prediction time lengthens. The Land Atmosphere Modeling Package Weather Research and Forecasting model, which provides 12-day agrometeorological predictions for East Asia, was employed from May 2017 to April 2018. Training periods occurred one month prior to and after the test period (12 days). A probabilistic consideration accounts for the relatively short training period. Based on the total and monthly root mean square error values for each test site, the results show an improvement in the NWP accuracy after bias correction. The spatial distributions in July and January were compared in detail. It was also shown that the physical consistency between temperature and wind speed was retained in the correction procedure, and that the GLM exhibited better performance than the quantile matching method based on monthly Pearson correlation comparison. The characteristics of coastal and mountainous sites are different from inland automatic weather stations, indicating that supplements to cover these distinctive topographic locations are necessary.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.