DETECTING DOMINANT FACTORS OF CHANGE CHARACTERISTICS OF WATER TEMPERATURE IN RIVER BASINS USING RANDOM FOREST AND VARIABLE IMPORTANCE ANALYSIS

Abstract

 To clarify the factors that govern the water temperature characteristics in a basin, in this study, the observed water temperature and the equilibrium water temperature were used to build regression tree models with the modified thermal sensitivity (dM_mT_w/dM_mT_eq) or the thermal equilibrium difference (M_aT_w−M_aT_eq) as the objective variable. Additionally, 33 characteristics related to the basin and channel were used as the explanatory variables. The variable importance analysis was performed using the developed models. The results revealed that the channel characteristics such as the channel length and slope, and basin characteristics such as the catchment area and elevation are the most important variables for the gradient. Additionally, with regard to the deviation, the variables related to the land use or cover in the catchment areas, such as forest, grassland, and agricultural land, and the aspect-slope that receives solar radiation are the main factors that drive an increase in water temperature.