Influence of Water Body Change Drivers Using Boosted Regression Trees: Songhua Water Resource Region

This study analyzed the spatiotemporal evolution of water areas based on land use data interpreted from remote sensing imagery of the Songhua Water Resource Region from 1980 to 2020. The Boosted Regression Tree model was employed to evaluate the relative influence (RI) and partial dependence (PD) of four categories of factors: hydrometeorological fluctuations, hydraulic engineering construction, agricultural development intensity, and socioeconomic changes. The results showed that: (1) The water area exhibited an overall shrinking trend, decreasing by 26,573 km2 over 40 years, with a dynamic rate of −31.10%. All sub-regions experienced shrinkage, with the Sanjiang Plain suffering the largest loss and the Heilongjiang Mainstream Basin showing the greatest relative decline. (2) Agricultural development intensity had the highest comprehensive RI (41.7%), and sown area had the highest single-factor RI (26.6%). Among sub-regions, the highest RI values were observed for hydraulic engineering in the Erguna River Basin region, agricultural development in the Changbai Mountain Basin region, and socioeconomic factors in the Heilongjiang Mainstream Basin region. (3) Cultivated land area, temperature, grain yield, the proportion of primary industry GDP, and levee length all exhibited inflection points, beyond which they positively drove water area shrinkage; reference evapotranspiration and urbanization rate showed opposite trends. This study can provide a reference for aquatic ecological protection and rational water resource utilization in the Songhua Water Resource Region.