Chaiyana et al. (2025) Evaluating trade-offs among cotton yield, groundwater extraction, and future projections for sustainable water management in the Texas High Plains

Identification

• Journal: Agricultural Water Management
• Year: 2025
• Date: 2025-12-01
• Authors: Akkarapon Chaiyana, Anita Kumari, S. V. Krishna Jagadish
• DOI: 10.1016/j.agwat.2025.110045

Research Groups

• Department of Plant and Soil Science, Texas Tech University, Lubbock, TX, USA

Short Summary

This study developed a novel data-driven framework integrating remote sensing and in-situ observations to quantify groundwater extraction (GWE) and evaluate trade-offs with crop water productivity (WPc) in irrigated cotton fields across the Texas High Plains (THP) from 2008 to 2030. The findings reveal that much of the central and northern THP exhibited unsustainable water use patterns (overuse and inefficiency) from 2008 to 2023, with GWE projected to increase from 2892 billion liters to 3439 billion liters by 2030, posing a significant risk to groundwater sustainability.

Objective

• To develop a novel approach for mapping cotton cultivation using machine learning applied to Tasseled Cap Transformation (TCT) data and identify irrigated areas through OpenET-derived evapotranspiration (ET) information.
• To estimate cotton lint yield by integrating phenology-based remote sensing metrics with climatic variables.
• To evaluate historical trends in GWE and project short-term future extraction based on observed patterns of cotton expansion.
• To assess spatial and temporal trade-offs between WPc and GWE to identify areas where groundwater use may be unsustainable within cotton-producing regions.

Study Configuration

• Spatial Scale: Texas High Plains (THP) region, encompassing 41 counties, approximately 102,300 square kilometers (km²). Remote sensing data was resampled to a 50-meter (m) spatial resolution.
• Temporal Scale:
  • Historical analysis: 2008 to 2023 (excluding 2012).
  • Future projections: 2024 to 2030.
  • In-situ observation data: 2005 to 2024.

Methodology and Data

• Models used:
  • CatBoost (for cotton extent mapping and lint yield estimation)
  • Random Forest (RF)
  • Support Vector Regression (SVR)
  • Extreme Gradient Boosting (XGBoost)
  • Tasseled Cap Transformation (TCT) for dimensionality reduction of Landsat data.
  • Linear regression for groundwater extraction (GWE) projection.
• Data sources:
  • Remote Sensing: Landsat 5, 8, and 9 imagery (Level-2 surface reflectance products) accessed via Google Earth Engine (GEE). OpenET (satellite-based crop evapotranspiration, ETc). USDA Crop Data Layer (CDL) for cotton and non-cotton areas.
  • Observation/In-situ: In-situ lint yield observations from the Texas Alliance for Water Conservation (TAWC).
  • Climatic/Auxiliary: Precipitation, minimum temperature (Tmin), maximum temperature (Tmax), relative humidity (rh), gross primary productivity (GPP), and Palmer Drought Severity Index (PDSI) accessed via GEE.

Main Results

• Cotton Extent Mapping: The TCT-CatBoost model achieved high classification performance with an overall accuracy of 0.97 and an F1-score of 0.89 for cotton classification. Spatial accuracy at the county level showed an R² of 0.97 and an RMSE of 7412 hectares (ha) in 2023.
• Lint Yield Estimation: The CatBoost model, integrating combined remote sensing and climate data, yielded the best performance with an R² of 0.55 and an RMSE of approximately 305 kilograms per hectare (kg/ha). Mid- and late-season phenological stages (flowering and boll ripening) and associated remote sensing indicators (e.g., Wetness Index, Normalized Difference Water Index, Shortwave Infrared 2) were identified as the most influential predictors.
• Crop Water Productivity (WPc): WPc in irrigated cotton systems ranged from approximately 2.00 to 4.00 kilograms per cubic meter (kg/m³), with higher values concentrated in the northern and central regions during productive years.
• Groundwater Extraction (GWE): Historical GWE showed an increasing trend from 2008 to 2023, with notable peaks exceeding 2500 billion liters (BL) in several years.
• GWE Projection: Based on historical trends, annual GWE is projected to increase from 2892 BL in 2024 to 3439 BL by 2030, with an average annual increase of approximately 91 BL.
• Trade-offs (WPc vs. GWE): Spatio-temporal analysis revealed that much of the central and northern THP fell into "overuse" (high WPc, high GWE) and "inefficient" (low WPc, high GWE) categories from 2008 to 2023, indicating unsustainable water use practices.

Contributions

• Developed a novel data-driven framework integrating remote sensing and in-situ observations for quantifying groundwater extraction (GWE) and evaluating trade-offs with crop water productivity (WPc).
• Presents the first study to integrate long-term spatiotemporal cotton classification and pixel-level lint yield estimation in the Ogallala Aquifer region with pixel-level trade-off assessments between WPc and GWE.
• Provides the first projections of future GWE volumes under cotton cultivation in the Texas High Plains (THP).
• Offers a scalable framework for historical and future yield estimation, overcoming limitations of publicly available county-level yield data.
• Provides forward-looking insights into the sustainability of irrigation practices, informing policy and guiding conservation efforts to address aquifer depletion.

Funding

• "Love Tito’s" (philanthropic arm of Tito’s Handmade Vodka)
• Cotton Inc.
• AFRI Sustainable Agricultural Systems (SAS) grant no. 2025–68012–44235 from the USDA National Institute of Food and Agriculture.
• Texas Water Development Board (for the Texas Alliance for Water Conservation (TAWC) program).

Citation

@article{Chaiyana2025Evaluating,
  author = {Chaiyana, Akkarapon and Kumari, Anita and Jagadish, S. V. Krishna},
  title = {Evaluating trade-offs among cotton yield, groundwater extraction, and future projections for sustainable water management in the Texas High Plains},
  journal = {Agricultural Water Management},
  year = {2025},
  doi = {10.1016/j.agwat.2025.110045},
  url = {https://doi.org/10.1016/j.agwat.2025.110045}
}