The increasing water stress projected for China could shift the agriculture and manufacturing industry geographically

The study addresses the growing mismatch between water supply and demand in China under climate change, which threatens sustainable development and may constrain water-intensive sectors such as agriculture and manufacturing. China’s per-capita water resources are low and unevenly distributed, with northern regions facing deficits despite hosting extensive agricultural and industrial activity. Existing water stress indices (WSIs) often rely on limited definitions of water availability and face data uncertainty. The research aims to develop a fuzzy TOPSIS-based water stress prediction index (FTOPWSP) that integrates multiple demand and supply drivers, handles uncertainty, and projects spatiotemporal water stress and associated population impacts. The study also explores implications for potential migration of agriculture, manufacturing, and population within China.

Prior work commonly defines WSI as the ratio of water withdrawals (agriculture, industry, domestic) to availability, often approximated by river discharge. Enhancements have incorporated environmental flow requirements, upstream consumptive withdrawals, and blue/green water components. Studies have projected increasing water scarcity due to climate change globally and in China, with the North China Plain (NCP) and northern regions highlighted as hotspots. Global and national assessments (e.g., Liu et al., Munia et al., He et al.) report rising urban and agricultural water stress and widespread future cropland scarcity. China-specific literature documents severe regional imbalances, reliance on projects like the South-to-North Water Diversion Project (SNWDP), and potential industrial relocation to rebalance resource use. However, computational simplicity and data uncertainty limit traditional WSI applications, and few studies explicitly examine sectoral migration induced by changing water stress. This study fills these gaps by introducing a fuzzy decision-making framework and investigating sectoral/population redistribution.

Study design: Develop the FTOPWSP index to assess spatiotemporal water stress for China (monthly, 0.5° × 0.5° grid) from 2020–2099 under RCP2.6 and RCP6.0, and estimate water-stressed population (FWSPOP) under coupled RCP–SSP2. Inputs (WDSIs): Ten water demand and supply indexes selected for representativeness and data availability: Demand: domestic water withdrawal (DWW), irrigation water withdrawal (IWW), manufacturing water withdrawal (MWW), evapotranspiration (EVAP). Supply: groundwater runoff (GWRO), groundwater recharge resources (GWRC), surface runoff (SRO), subsurface runoff (SSRO), total soil moisture (TSM), snow water equivalent (SWE). Positive property (+) reduces stress; negative (–) increases stress. Weights: Combination weights for each index computed via principal component analysis (PCA) and entropy weight method (EWM) to reduce subjectivity and reflect index importance (Supplementary Table 1 indicates high weights for IWW, MWW, and SWE). Fuzzy TOPSIS construction: For each grid and period, create a fuzzy decision matrix using triplets (aij, bij, cij) from three GCM-driven simulations for each WRPM: GCMs: GFDL-ESM2M, HadGEM2-ES, MIROC5. WRPMs: H08, CWatM, PCR-GLOBWB. Normalize via min–max, apply weights to form the weighted normalized fuzzy matrix. Determine fuzzy positive-ideal solution (FPIS) and fuzzy negative-ideal solution (FNIS) across all grids (2020–2099). Compute distances of each index to FPIS/FNIS, sum to total positive (d+i) and negative (d−i) distances, and define FTOPWSP = d−i / (d+i + d−i). Larger FTOPWSP indicates higher water stress and farther from optimal status. Population impact (FWSPOP): Define water-stressed population as residents in grids with FTOPWSP at or above the 60th percentile threshold (validated against alternative thresholds 50–90%). Population projections from SEDAC SSP2 (one-eighth degree) are used to compute annual FWSPOP under RCP2.6–SSP2 and RCP6.0–SSP2. Data sources and settings: Administrative boundaries from National Earth System Science Data Center. Monthly WDSI data (2020–2099) from ISIMIP2b for the three WRPMs and GCMs, with land use, nitrogen deposition, fertilizer fixed at 2005 levels for consistency. IPSL-CM5A-LR excluded due to cold bias. Grid resolution: 0.5° × 0.5°. Time step: monthly. Hotspot analysis via Getis-Ord (ArcGIS) for both FTOPWSP and FWSPOP. Validation/Comparison: Parallel calculation of conventional WSI shows consistent temporal trends and spatial patterns with FTOPWSP, supporting reliability.

• National trend: FTOPWSP increases from 2020 to 2099 under both scenarios, indicating rising water stress. RCP2.6 Sen’s slope: 5.9×10^-4 (statistically significant), implying ~0.078% increase in 2099 vs 2020. RCP6.0 Sen’s slope: 11.3×10^-4 (positive but not significant), ~0.18% increase. Mean annual FTOPWSP: 0.57 (RCP2.6) vs 0.58 (RCP6.0), indicating higher stress under RCP6.0. Wide spatial variability in both scenarios.
• Spatial patterns: Northern and western China, especially the North China Plain (NCP), show substantially higher stress than southern and eastern regions. High Mountain Asia/Tibetan Plateau shows very low stress, underscoring glacier melt’s current buffering role. Late-century increases are more pronounced than mid-century, with sharp rises in Jiangsu, Anhui, Jiangxi (East), and Hubei, Hunan (Central).
• Hotspots/cold spots: Hotspots (>95% confidence) cluster in the NCP (Beijing, Tianjin, Hebei, Shandong, Henan, Anhui, Jiangsu); cold spots (>95%) in southern provinces (Zhejiang, Fujian, Jiangxi, Hunan, Guangdong, Guangxi, Sichuan) and Tibet.
• Provincial late-century (RCP6.0) highest-stress list (FTOPWSP values): Tianjin (0.600), Beijing (0.596), Shandong (0.591), Jiangsu (0.589), Hebei (0.587), Henan (0.585), Shanghai (0.584), Xinjiang (0.586), Ningxia (0.586), Liaoning (0.585), Inner Mongolia (0.584), Shanxi (0.583). Net annual changes are positive; increases mainly in spring and winter for these provinces.
• Seasonal drivers: Under RCP2.6, annual changes are dominated by spring and autumn contributions; under RCP6.0, winter contributes most in many decades (linked to larger SWE changes under greater warming). Under RCP6.0, larger increases occur notably in 2021–2030 and 2041–2050.
• Process drivers: All supply-side indexes have negative Sen’s slopes (declines), with significant decreases in total soil moisture (TSM), surface runoff (SRO), and snow water equivalent (SWE), driving stress up. Irrigation water withdrawal (IWW) trends differentiate scenarios: significantly positive under RCP6.0 (raising stress) and significantly negative under RCP2.6 (mitigating stress). IWW carries the largest weight among WDSIs, making it pivotal for scenario differences; MWW also has a large weight.
• Water-stressed population (FWSPOP): Using the 60th percentile threshold, on average >20% of China’s population is water-stressed annually in both scenarios. RCP2.6–SSP2: average ~27%; before 2080 values peak mid-century with 49% in 2040; thereafter decline. RCP6.0–SSP2: average ~32%; peak ~37% in 2040. Spatially, late-century FWSPOP concentrates in northern/eastern mid-latitudes; 2099 top-20% FWSPOP grids lie roughly within 111°E–119°E and 123°E–126°E, 26°N–27°N and 33°N–36°N (RCP2.6–SSP2), and 111°E–120°E, ~27°N, 34°N–37°N (RCP6.0–SSP2). Hotspots (99% significance) for FWSPOP include Beijing, Tianjin, Hebei, Shandong, Henan, Jiangsu, Anhui.
• Sectoral/population implications: Increasing, uneven stress likely to catalyze north-to-south migration of agriculture and manufacturing (especially from NCP and Shandong Peninsula), and of population from highly stressed northern provinces to less stressed southern provinces (e.g., to Fujian, Guangdong, Chongqing, Sichuan, PRD).

The FTOPWSP index, by integrating multiple demand and supply factors with fuzzy TOPSIS to manage data uncertainty, robustly captures temporal increases and strong north–south disparities in China’s water stress under climate change. Comparison with the classic WSI shows high consistency in trends and spatial patterns, reinforcing confidence in results. Findings indicate supply-side declines (TSM, SRO, SWE) as primary drivers, with irrigation withdrawals (IWW) critically amplifying stress under higher emissions due to its high weight and positive trend. Seasonal analyses highlight spring/autumn (RCP2.6) and winter (RCP6.0) as key contributors to annual variability, linking climate dynamics (e.g., SWE changes) to stress patterns. The projections imply that persistent, uneven stress will constrain water-intensive sectors and increase the share and concentration of water-stressed populations in northern provinces, potentially inducing geographic shifts of agriculture and manufacturing to less-stressed southern regions and associated labor/population movements. The role of Asia’s water tower in currently alleviating stress is noted but may be unsustainable given ongoing terrestrial water storage deficits. Policy and management implications include strengthening SNWDP, improving agricultural water efficiency, considering strategic reallocation of production, and guiding industrial siting to reduce exposure while accounting for socioeconomic and environmental trade-offs (e.g., migration-related emissions and pollution impacts).

This study introduces the FTOPWSP index, a fuzzy TOPSIS-based water stress predictor that integrates ten demand–supply drivers with objective weighting, enabling robust spatiotemporal assessment under data uncertainty. Projections show increasing water stress across China through 2099 under both low and high emissions, with marked north–south disparities and hotspots in the NCP and northern provinces. Declining water supplies (TSM, SRO, SWE) and rising irrigation withdrawal under higher emissions drive stress increases. A substantial share of the population (>20% annually; averages ~27% under RCP2.6–SSP2 and ~32% under RCP6.0–SSP2) will be water-stressed, concentrated in northern regions. These patterns could prompt north-to-south shifts in agriculture, manufacturing, and population. Policy options include continuing and refining SNWDP, improving agricultural water-use efficiency, enabling southern agricultural expansion (e.g., land consolidation, crop strategies), and incentivizing relocation of high water-consuming industries to lower-stress regions. Future research should incorporate additional dimensions (e.g., water quality, economic development, management efficiency), harmonize detailed supply sources across models (e.g., aqueduct transfers, reservoirs, desalination), test alternative thresholds and socio-economic pathways, and evaluate the socioeconomic–environmental trade-offs of sectoral and population migrations.

• Water transfers and detailed supply sources (e.g., SNWDP aqueduct flows, reservoirs, desalination, renewable/non-renewable groundwater breakdowns) from models like H08 were not fully incorporated to maximize cross-model comparability; this may overestimate northern stress and underestimate southern stress.
• Reliance on three WRPMs (H08, CWatM, PCR-GLOBWB) and three GCMs (GFDL-ESM2M, HadGEM2-ES, MIROC5); IPSL-CM5A-LR excluded due to cold bias. Model structural differences and scenario forcings (fixed 2005 land use, nitrogen deposition, fertilizer) introduce uncertainties.
• Fuzzy TOPSIS reduces but does not eliminate data uncertainty; index weights (PCA+EWM) still reflect methodological choices.
• FWSPOP threshold set at 60th percentile (tested alternatives), which affects absolute counts; results are scenario- and threshold-dependent.
• Not all socioeconomic drivers (e.g., wages, jobs, policy changes) or water quality factors are explicitly modeled in stress estimation, limiting direct causal inference on migration/industrial relocation.