A Study on the Prediction of Daily Snowmelt Depth using Multiple Linear Regression

Abstract

In this study, daily snowmelt was predicted using observed meteorological data and multiple regression analysis. Five observation stations (located in Daegwallyeong, Gwangju, Seosan, Mokpo, and Jeonju) were selected to analyze fresh snow depth from 2000 to 2010. The dependent variable used in the multiple regression analysis was daily snowmelt depth, and the independent variables were fresh snow depth, diurnal temperature range, temperature interception, diurnal humidity range, humidity intercept, and solar radiation. Seventy percent of the total observed data was used to develop a multiple regression model and the regression model was verified using the 30% of remaining data. The adjusted R-squared and Root Mean Square Deviation (RMSE) were used to examine the developed regression model. As a result, the adjusted R-squared was higher than 0.769 (except Daegwallyeong); thus the developed model represented well the daily snowmelt depth. Even Jeonju had an adjusted R-squared of 0.869. Also, the RMSE in all of the five stations was lower than 2.5 cm. The lowest value in Seosan was 1.7 cm. From the two types of verification, the developed multiple regression model was judged to be suitable to predict the daily snowmelt depth. However, multicollinearity should be explained, as rapid increases in temperature and sustained high temperature could not be reflected in the model. Therefore, if the limitations were resolved in further research, the model could be used to predict the amount of daily snowmelt depth more reliably.