Reproducibility Evaluation of High-Resolution Downscaled Climate Data Using Automated Weather Station and Precise Temperature Monitoring Data.

Abstract

Forest-related disasters are recently increasing in terms of frequency and severity due to climate change. This study aims to produce and evaluate high-resolution gridded climate data for a better understanding and prediction of forest-related disasters in the context of climate change. Thus, ERA5 re-analysis data was collected, and high-resolution observation gridded data (100m interval) were produced using the Improved GIS-based Regression Model (IGISRM). The produced daily data is bias-corrected using the Automated Surface Observing System (ASOS) data, and then a series of reproducibility evaluations have been performed using the Automated Weather Station (AWS) data and nine on-site monitored Precise Temperature (PT) data. The reproducibility evaluation results using AWS data show that the precipitation has a comparatively low reproducibility as it tends to underestimate the observed temporal fluctuation and rainfall amount. However, the produced high-resolution gridded climate data are found to have excellent estimation performance for the minimum and maximum temperatures. The evaluation using the monitored temperature data showed that the statistical efficiency of minimum temperature was somewhat lower than that of maximum temperature, but overall reproducibility was still high (NRMSE<0.5). In the measurement period (May 2021 ~ April 2022), the maximum temperature was slightly underestimated overall, while the minimum temperature showed overestimation, particularly in winter (after November) when the temperature drops sharply. Excluding one station, the statistical efficiency remained high, indicating that the ERA5 reanalysis data can be used along with the Improved GIS-based Regression Model (IGISRM) statistical downscaling technique for producing high-resolution downscaled observation data in mountainous/forest areas with insufficient observation data.