Abstract: For effective regional water resource management and sustainable agricultural development, it is crucial to assess the water footprint of crops and clarify the determinants. We analyzed the spatiotemporal variations of soybean water footprints across Heilongjiang Province from 2004 to 2020. By employing the BP Neural Network and Decision-making Trial and Evaluation Laboratory (DEMATEL) (BP-DEMATEL) model, we identified the driving forces and typological determinants within the causal network that affect the water footprint of soybean, and elucidated the hierarchical interdependencies among them. The results showed that the average annual soybean water footprints ranged from 1.99 to 2.72 m³·kg^-1 across Heilongjiang Province, with compositional dominance being green water (71.4%) > grey water (17.9%) > blue water (10.7%). Qiqihar, Hegang, and Daxing’anling exhibited stronger interannual variability than other areas. Among the factors influencing soybean water footprint, the key driving factor was the average relative humidity, while the moderate driving factors included the sunshine duration, precipitation, the total power of agricultural machinery and fertilizer consumption. The key typological determinants encompassed average temperature, per capita agricultural GDP, and average wind speed. The soybean water footprint in Heilongjiang Province was primarily influenced by local meteorological conditions and  per capita agricultural GDP, with the chemical fertilizer application and agricultural machinery having the secondary impact on soybean water footprint. The primary means to improve water resource utilization efficiency in soybean cultivation are to optimize planting systems through localized climate resource utilization and enhancing agricultural economic capacity. Supplementary measures are implementing precision fertilizer management and mechanization intensity regulation.

Key words: soybean, water footprint, back propagation neural network, Decision-making Trial and Evaluation Laboratory (DEMATEL)