Dynamic change and driving response of shallow groundwater level based on random forest in southwest Songnen Plain

Abstract

Study regionThe southwest Songnen Plain in China.Study focus:This study investigated the multi-time groundwater dynamic types and the trends of groundwater level. Random forest model was applied to analyse on driving factors of groundwater level.New hydrological insights:This study selected the following data from four periods in 2005, 2010, 2015, and 2020: groundwater depth, precipitation, evaporation, groundwater extraction, dry field area, and paddy field area. Pearson correlation analysis, Kriging interpolation, Mann-Kendall test, and other methods were employed to determine the groundwater dynamic types and changes in level within the four periods. Pearson correlation analysis shows that the dynamic types of groundwater table change greatly in the four periods, and the correlation between groundwater depth and precipitation decreases first and then increases. Kriging interpolation showed that the groundwater depth was lower in the west, and the depth around irrigation areas was higher. The Mann-Kendall test showed that the groundwater depth changed abruptly over time during the irrigation period and rainy season; the average groundwater depth increased by 1.36 m from 2005 to 2010, but only by 0.09 m from 2010 to 2020. The random forest model of groundwater level dynamics and its driving factors was established, and it was found that the size of the paddy field had the greatest influence on groundwater dynamics, affecting 32.15%.