Revealing the Driving Factors of Water Balance in Lake Balkhash Through Integrated Attribution Modeling

Abstract. Understanding the impacts of climate change and human activities on large endorheic lakes is crucial for sustainable water management, yet quantitative attribution remains a significant challenge. This study introduces the Hydrological Attribution and Analysis Framework (HAAF), a novel three-stage methodology, to provide a comprehensive explanation for the nearly-centennial (1931–2024) water balance dynamics of Lake Balkhash. The HAAF first establishes a high-fidelity hydrological reconstruction using a Physics-Informed Machine Learning (PIML) model, then employs the Budyko framework to attribute runoff changes, and finally links these catchment-scale drivers to the lake's terminal water balance. Our results confirm the robustness of the PIML model in simulating historical runoff (KGE > 0.75). The attribution analysis then reveals a complex interplay of competing forces. During the intensive intervention period (1970–1990), a substantial human-induced runoff reduction of -9.21 km³ completely masked a significant climate-driven wetting potential (+6.13 km³), triggering the lake's sharp decline. In the recent period (1991–2024), the basin's hydrology has been governed by a fragile stalemate in which a massive, climate-driven potential for increased runoff (+10.80 km³) was almost entirely neutralized by the persistent negative impact of human water use (-11.36 km³). At the lake level, this translated into an apparent stability sustained only by a favorable climatic subsidy. Future projections under various climate scenarios indicate that this climatic buffer is transient and unlikely to persist, exposing the lake to a high risk of rapid decline. We conclude that the recent stability of Lake Balkhash is not a sign of systemic recovery but a "masked vulnerability." This highlights the urgent need for sustainable and forward-looking water management strategies that account for these underlying, competing drivers.