Du et al. (2026) Retrieving long-term topsoil moisture in Qingtongxia irrigation district using a modified OPTRAM model

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-01-06
• Authors: Shuai Du, Yuanyuan Zha, Yuzhe Ji, Yue Wang, Xiangsen Xu, Liu Yang, Meijun Zheng, Yang Zhang, Shenshen Wu
• DOI: 10.1016/j.ejrh.2025.103074

Research Groups

State Key Laboratory of Water Resources Engineering and Management, Wuhan University, Wuhan 430072, China

Short Summary

This study developed a modified Optical Trapezoid Model (OPTRAM) using quadratic functions for dry and wet edges to improve long-term surface soil moisture (SSM) estimation in the Qingtongxia Irrigation District, demonstrating superior accuracy, especially with Sentinel-2 data, and enabling field-scale irrigation detection and crop mapping across full annual cycles.

Objective

• To develop a modified OPTRAM model that uses quadratic functions to better capture non-linear Shortwave Infrared Transformed Reflectance (STR)-Normalized Difference Vegetation Index (NDVI) edges, thereby improving long-term surface soil moisture (SSM) estimates in large-scale irrigation areas across crop growth and fallow periods.

Study Configuration

• Spatial Scale: Qingtongxia Irrigation District (QID), Ningxia, China, an arid, fully irrigated region. Field-scale (10 meters to 30 meters resolution).
• Temporal Scale: 2022–2024, covering both crop growth and fallow periods.

Methodology and Data

• Models used: Modified OPTRAM (with quadratic dry/wet edges), original OPTRAM (linear dry/wet edges), Thermal-Optical Trapezoid Model (TOTRAM), Random Forest (for crop type mapping).
• Data sources:
  • Satellite: Sentinel-2 Level-2A (multispectral surface reflectance, 10 meters resolution), Landsat 8 Level 2 (multispectral surface reflectance, Land Surface Temperature (LST), 30 meters resolution), SMAP Level-3 (soilmoistuream, 9 kilometers resolution).
  • Reanalysis: ERA5-Land (daily mean precipitation).
  • In-situ observations: Soil moisture (0–10 centimeters depth) and crop types collected at approximately 20–40 sampling locations during five field surveys.
  • Ancillary: Google Earth Engine (GEE) for image processing and visual interpretation, Statistical Bureau of Ningxia Hui Autonomous Region for crop planting areas.

Main Results

• The modified OPTRAM model achieved the highest accuracy for SSM estimation, particularly with Sentinel-2 data (correlation coefficient (R) = 0.86, unbiased root mean square error (ubRMSE) = 0.06, bias = 0.00).
• The original OPTRAM model with linear edges showed significant underestimation (bias = -0.08 for Sentinel-2, -0.07 for Landsat 8).
• The TOTRAM model with universal parameterization exhibited poor estimation accuracy (R = 0.05).
• The modified OPTRAM model, especially with Sentinel-2 imagery, effectively captured field-scale SSM dynamics and detected irrigation events, including pronounced winter irrigation signals.
• A distinct STR-NDVI feature space for QID was revealed by incorporating the entire crop growth and fallow periods, showing expanded STR variability near both maximum and minimum NDVI values.
• Crop type mapping for QID achieved an overall accuracy of 0.8835 and a Kappa coefficient of 0.8466.
• Sentinel-2's higher spatial resolution (10 meters) provided more accurate SSM estimates compared to Landsat 8 (30 meters). SMAP (9 kilometers) failed to capture crop-specific and field-scale SSM differences.
• Sensitivity analysis showed ubRMSE to be negligibly sensitive to the assumed local minimum dry soil moisture content (θd), while mean absolute error (MAE) could deviate by up to 40%. Nonlinear models were more robust to θd variations.
• Sentinel-2 maintained effective SSM monitoring even at a strict cloud threshold of 0.1%, while Landsat 8 required a more relaxed threshold (20%) for comparable performance. Recommended thresholds for QID are 40% for Sentinel-2 and 70% for Landsat 8.

Contributions

• Introduction of a novel modified OPTRAM model incorporating quadratic functions for dry and wet edges, which more accurately represents the non-linear STR-NDVI relationships observed in large-scale irrigation districts across full annual cycles (crop growth and fallow periods).
• First comprehensive comparative assessment of OPTRAM and TOTRAM for SSM estimation in an irrigation district, utilizing high-resolution Sentinel-2 imagery and validated with extensive ground-truth data, including fallow periods.
• Demonstration of the modified model's capability to capture field-scale SSM heterogeneity, detect irrigation events, and support crop type mapping and growth stage analysis, offering new insights into water use patterns in arid irrigated regions.
• Detailed analysis of the dynamic STR-NDVI feature space for QID, highlighting the impact of winter irrigation and crop-specific spectral responses.
• Evaluation of the impact of cloud cover thresholds on satellite data usability for OPTRAM applications, providing practical recommendations for data selection in irrigated dryland regions.

Funding

• National Natural Science Foundation of China (Grant No. 52479045, 52279042)
• Hubei Provincial Natural Science Foundation Youth A-Class Project (2025AFA091)
• Key Research and Development Program in Guangxi (AB23026021)

Citation

@article{Du2026Retrieving,
  author = {Du, Shuai and Zha, Yuanyuan and Ji, Yuzhe and Wang, Yue and Xu, Xiangsen and Yang, Liu and Zheng, Meijun and Zhang, Yang and Wu, Shenshen},
  title = {Retrieving long-term topsoil moisture in Qingtongxia irrigation district using a modified OPTRAM model},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2025.103074},
  url = {https://doi.org/10.1016/j.ejrh.2025.103074}
}