High-resolution Reconstruction and Causal Framing of Multi-sector Water Withdrawals in China: Emerging Domestic Hotspots and Shifts in Coupled Human-water Regimes

China faces chronic water scarcity and strong spatial-temporal mismatches between water availability and demand, with particularly severe stress in North and Northwest China. Rapid urbanisation, industrial restructuring, and expanding irrigated agriculture have intensified competition among domestic, irrigation, manufacturing, and thermal-cooling water uses. These dynamics reflect coupled human-water feedbacks: socio-economic development reshapes withdrawals, while evolving water constraints and hydroclimatic extremes influence exposure, management responses, and future demand trajectories. A key gap is to causally attribute multi-sector water-use changes to socio-economic and hydroclimatic drivers and to anticipate how their co-evolution may reshape water-use hotspots.

We analyse a new 0.1° gridded dataset of monthly sectoral water withdrawals for China (1965-2022), focusing on emerging domestic-use hotspots and their interaction with other sectors as a first step towards diagnosing cross-sector trade-offs and human-water feedback pathways. National annual domestic withdrawals increased from 1.9×10¹⁰ to 9.3×10¹⁰ m³ (1965-2022). A piecewise linear fit indicates three growth phases and a recent slowdown: moderate growth before 1975, faster growth during 1976-1992, rapid acceleration in 1993-2010 (slope = 2.3×10⁹ m³yr^-1), and a weaker, statistically noisy trend in 2011-2022. Despite the volume increase, domestic seasonality remains stable (amplitude ratio = 0.19; JJA share = 27%).

At the grid-cell level, we compute (i) the long-term trend in annual domestic withdrawals (1965-2022), (ii) relative seasonal amplitude, and (iii) mean annual domestic use in 2000-2022. Hotspots are cells exceeding the 75th percentile in all three metrics. They occupy 17.5% of valid land cells yet account for 24.3% of recent domestic withdrawals and 10.9% of the national domestic-use increase over 2000-2022. The correlation between local trends and recent mean use is extremely high (r = 0.99), indicating growth is concentrated where domestic withdrawals are already substantial, typically along rapidly urbanising corridors.

A complementary multi-sector analysis shows total withdrawals rise from 3.7×10¹¹ to 5.4×10¹¹ m³yr^-1 across 1965-1989, 1990-2009, and 2010-2022. Irrigation remains dominant (80%, 68%, 64% of mean withdrawals), but its contribution to growth turns negative in 1990-2009, when domestic and thermal-cooling withdrawals explain 85% and 68% of the net increase. Together, these patterns indicate a transition from an irrigation-dominated regime to a more complex urban- and energy-driven water-use system, with domestic hotspots emerging as critical pressure points for water security.

Ongoing work links these patterns with socio-economic indicators and hydroclimatic variables using Neural Granger Causal and PCMCI+ frameworks, and couples them with deep learning prediction under plausible population, urbanisation, and climate trajectories to assess future hotspot shifts and inform adaptive, resilient water management.