Prediction of near-surface air temperature in glacier regions using ERA5 data and the random forest regression method

Abstract

Near-surface air temperature is an essential physical parameter for estimating the glacier ablation model that is significant to understanding the dynamics of the changes and constant monitoring of glaciers. There is a significant difference between the temperatures in glacier and non-glacier areas at the same altitude, and the estimation model that uses low-altitude automatic weather station (AWS) data cannot be applied to glaciers far away from the stations. In this study, a scheme for air temperature prediction using the random forest regression (RFR) model and the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis 5 (ERA5) data was analyzed. The results show that the average mean absolute error (MAE) and root mean square error (RMSE) of the ERA5 raw temperature product for three glaciers were 4.32 °C and 5.72 °C, respectively. After the RFR model revision, the MAE and RMSE significantly decreased to 1.01 °C and 1.41 °C, respectively. In addition, when the Linzhi weather station was used to calculate the temperature in the Parlung No. 4 Glacier area, the MAE of the Greuell and Böhm (GB) model decreased by 0.18 °C compared to that obtained using the lapse rate method. With our method, long time series of glacier surface air temperature data with high accuracy can be obtained by using short term in-situ temperature measurements.