Coupling regional climate model and machine learning to model high-resolution precipitation

Coupling regional climate model and machine learning to model high-resolution precipitation

Phenomenological Aspects of Civil Engineering (PACE) - an International Congress
Volume 1 - Issue 1 - PACE-2021

Trinh Quang Toan Tran Duc Huy

Abstract

Accurate areal rainfall is result not only from atmospheric boundary conditions, but also from the quality and availability of soil, topography, and vegetation data. As a result, rainfall model errors are exacerbated by uncertainties in both atmospheric and land surface conditions. Hybrid technique combining dynamical and statistical downscaling was investigated in this research. The proposed downscaling method incorporates information from three global reanalysis data sets: ERA-Interim, ERA20C, and CFSR. The Weather Research and Forecasting (WRF) model is used to hybrid downscale this reanalysis of atmospheric data, which is then followed by the use of an artificial neural network (ANN) model to further downscale the WRF performance to a finer resolution over the studied area. The findings of this study indicate that the proposed method, which combines model simulations with observations over the modeled area, will improve the accuracy of simulated data. Another advantage of this method is the low cost of computation, both in terms of computation time, and performance storage.

Keywords

Artificial neural network (ANN), weather research and forecasting (WRF), ERA-Interim, ERA20C, and CFSR.
https://www.acapublishing.com/dosyalar/baski/PACE_2021_216.pdf