Sun et al. (2026) Translating snowmelt and freeze–thaw legacies into irrigation–drainage zoning of water–salt constraints in seasonally frozen oasis croplands
Identification
- Journal: Journal of Hydrology Regional Studies
- Year: 2026
- Date: 2026-11-07
- Authors: Mingyue Sun, Ping Gong, Pengfei Li, Yingsheng Dang, Hongguang Liu
- DOI: 10.1016/j.ejrh.2026.104178
Research Groups
- College of Water Conservancy & Architectural Engineering, Shihezi University, Shihezi, PR China
- Key Laboratory of Modern Water-Saving Irrigation of Xinjiang Production & Construction Group, Shihezi, PR China
Short Summary
This study developed a management-oriented Water–Salt Constraint Index (WSCI) to integrate multiple soil constraints in snowmelt-fed oasis croplands. It found that antecedent cold-season hydrological legacies significantly improve the prediction of late-season topsoil water–salt constraints, enabling differentiated irrigation–drainage zoning.
Objective
- To construct weighted-average (WSCIWA) and dominant-constraint-sensitive (WSCIDE) formulations integrating water status, salinity, alkalinity, sodicity, carbonate background, and texture-mediated transport constraints.
- To predict the spatial patterns of WSCI across croplands using multisource remote-sensing and environmental covariates.
- To quantify the incremental predictive contribution of Moderate Resolution Imaging Spectroradiometer (MODIS) snow products and ERA5-Land reanalysis covariates within spatial block cross-validation.
- To identify dominant and compound constraints to delineate irrigation–drainage management zones.
- Principal hypothesis: The 2024 late-season topsoil WSCI is statistically associated not only with sampling-period surface, terrain, and groundwater conditions, but also with antecedent cold-season covariates related to snow accumulation, snowmelt, and freeze–thaw transitions.
Study Configuration
- Spatial Scale: Manas River Basin, northwestern China. Cropland mapping at 10 meter resolution. Soil samples collected from the 0–20 centimeter topsoil layer.
- Temporal Scale:
- Soil sampling: Early October to late November 2024.
- Cold-season hydrological legacy (MODIS snow, ERA5-Land): 1 November 2023 – 30 April 2024.
- Sampling-period surface state (Sentinel-2, Sentinel-1): 1 October – 30 November 2024.
- Annual water-balance background (TerraClimate, MOD16A2, MOD11A2): 1 November 2023 – 31 October 2024.
- Groundwater depth: Monthly observations for January–December 2024 (annual), October–November 2024 (autumn), and April–October 2024 (irrigation season).
Methodology and Data
- Models used:
- Water–Salt Constraint Index (WSCI): Weighted-average (WSCIWA) and dominant-constraint-sensitive (WSCIDE) formulations.
- Machine Learning: Extreme Gradient Boosting (XGBoost) (primary), Random Forest, Light Gradient Boosting Machine (LightGBM), Cubist, Partial Least Squares Regression (PLSR).
- Model Interpretation: Shapley Additive Explanations (SHAP), Accumulated Local Effects (ALE).
- Data sources:
- In-situ: 1250 composite topsoil samples (0–20 cm) for surface soil moisture (SSM), total salts (TS), pH, exchangeable sodium percentage (ESP), calcium carbonate equivalent (CCE), and sand, silt, and clay fractions. 66 groundwater monitoring wells (monthly observations).
- Satellite:
- Sentinel-2 MSI Level-2A: Optical indices (bare-soil spectra, salinity, brightness, moisture, residual vegetation).
- Sentinel-1 SAR GRD IW VV/VH: Radar variables (surface moisture, roughness, residual cover, post-irrigation surface conditions).
- MODIS Terra/Aqua MOD10A1/MYD10A1 V6.1: Snow cover, snow duration, snowmelt timing.
- MODIS Terra MOD16A2 V6.1 ET/PET: Actual evapotranspiration, growing-season water consumption.
- MODIS Terra MOD11A2 V6.1 LST: Land surface temperature, freeze–thaw auxiliary information.
- Reanalysis/Model-derived:
- ERA5-Land Hourly land reanalysis: Snowfall, snowmelt, soil moisture, freeze–thaw activity, runoff, evaporation background.
- TerraClimate Monthly climate and water-balance data: Annual mean temperature, extreme temperature, precipitation, potential evapotranspiration, aridity index.
- Geospatial:
- Copernicus DEM GLO-30: Terrain variables (elevation, slope, curvature, topographic wetness index (TWI), flow accumulation, geomorphic position).
- HydroRIVERS River vector data: River connectivity (distance to rivers, river-network density).
- Esri 10 m Annual Land Cover: Cropland mask.
Main Results
- XGBoost achieved the best prediction performance for WSCI_DE under 10-fold spatial block cross-validation, with a coefficient of determination (R²) of 0.761, root mean square error (RMSE) of 0.108, mean absolute error (MAE) of 0.086, and residual prediction deviation (RPD) of 2.05.
- Removing antecedent snow, snowmelt, and freeze–thaw predictors reduced R² to 0.707 and RPD to 1.85, indicating that these cold-season hydrological legacies provided non-redundant statistical information for late-season topsoil WSCI prediction.
- Spatially, WSCI_DE classified 38.7% of cropland as moderate, 19.6% as high, and 3.2% as very high relative constraints.
- Salinity was the dominant constraint, covering 65.0% of cropland, followed by sodicity (12.5%), carbonate content (8.2%), water-status constraints (7.2%), texture influence (5.2%), and alkalinity (1.9%). Compound constraints were present in 24.0% of mapped cropland.
- SHAP analysis identified normalized difference salinity index (NDSI), autumn mean groundwater depth, TWI, bare soil index (BSI), and total snowmelt as the top predictors for WSCI_DE variation.
- Accumulated Local Effects (ALE) analysis showed nonlinear (U-shaped) responses for total snowmelt and cold-season soil moisture, with moderate values associated with lower WSCIDE. WSCIDE decreased with increasing annual mean groundwater depth.
Contributions
- Developed a novel management-oriented Water–Salt Constraint Index (WSCI) that integrates multiple soil properties (water status, salinity, alkalinity, sodicity, carbonate background, texture) to diagnose dominant and compound water–salt limitations, moving beyond single-property mapping.
- Quantified the significant and non-redundant predictive contribution of antecedent cold-season hydrological covariates (snow accumulation, snowmelt, freeze–thaw processes) to late-season topsoil water–salt constraints in snowmelt-fed dryland croplands.
- Provided a robust framework for spatially differentiated irrigation–drainage zoning based on relative constraint intensity and dominant limiting factors, supported by multisource remote sensing, reanalysis data, and interpretable machine learning.
- Demonstrated the ecological and hydrological consistency of the WSCI_DE with vegetation indices and groundwater depth, enhancing its utility for regional water–salt management.
Funding
- Key Research and Development Project of the Xinjiang Production and Construction Corps (XPCC) (Grant No. 2024AB079)
- “Tianshan Talents” Training Program—Young Top Talents in Scientific and Technological Innovation (Grant No. 2023TSYCCX01115)
- Divisional and Municipal Science and Technology Tackling Program Projects (Grant No. 2024GG2301)
Citation
@article{Sun2026Translating,
author = {Sun, Mingyue and Gong, Ping and Li, Pengfei and Dang, Yingsheng and Liu, Hongguang},
title = {Translating snowmelt and freeze–thaw legacies into irrigation–drainage zoning of water–salt constraints in seasonally frozen oasis croplands},
journal = {Journal of Hydrology Regional Studies},
year = {2026},
doi = {10.1016/j.ejrh.2026.104178},
url = {https://doi.org/10.1016/j.ejrh.2026.104178}
}
Original Source: https://doi.org/10.1016/j.ejrh.2026.104178