Liu et al. (2025) A combined spatial interpolation method of co-Kriging with inverse distance weighting and random forest for soil water and salt in arid oasis

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-11-08
• Authors: Shuiqing Liu, Songhao Shang
• DOI: 10.1016/j.jhydrol.2025.134569

Research Groups

• State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, China
• Department of Hydraulic Engineering, Tsinghua University, Beijing 100084, China
• Key Laboratory of Hydrosphere Sciences of the Ministry of Water Resources, Tsinghua University, Beijing 100084, China

Short Summary

This study developed a combined spatial interpolation method integrating Co-Kriging (CK), Inverse Distance Weighting (IDW), and Random Forest (RF) to accurately characterize soil water and salt distribution in arid oasis regions. The novel combined method significantly improved prediction accuracy for both soil water content and total salt content compared to traditional methods, establishing a transferable framework for multi-method integration.

Objective

• To systematically evaluate and develop an improved spatial interpolation method for accurate characterization of soil water and salt distribution in arid oasis regions.

Study Configuration

• Spatial Scale: Yarkand River Basin, Northwest China; 100 topsoil samples (0–20 cm depth).
• Temporal Scale: Snapshot of soil conditions based on collected samples; specific sampling period not detailed.

Methodology and Data

• Models used: Ordinary Kriging (OK), Co-Kriging (CK), Inverse Distance Weighting (IDW), Random Forest (RF). A combined approach integrating IDW-derived covariates into CK, further optimized by Random Forest.
• Data sources: 100 topsoil (0–20 cm) samples collected from the Yarkand River Basin.

Main Results

• OK and CK showed comparable accuracy (correlation coefficients (R) of 0.48 and 0.49 for soil salt; 0.17 and 0.16 for soil water content) and were superior to IDW (R of 0.19 for soil salt; 0.10 for soil water content) in hold-out validation.
• CK outperformed OK in cross-validation, with R improving from approximately 0.35 to over 0.80, demonstrating its sensitivity to localized variability.
• The developed combined method significantly increased the correlation coefficient (R) of test datasets by 319 % for soil water content and 49 % for total salt content.
• The method is applicable for soil properties with varying spatial heterogeneity (coefficients of variation: 38.7 % for soil moisture content, 93.9 % for total salt content).
• Interpolation results showed soil water content ranging from 0.04 to 0.28 g/g with no obvious overall trend, and total soil salt content from 0.87 to 12.1 g/kg, exhibiting clear spatial heterogeneity (non-salinized upstream, slightly salinized midstream, moderately salinized downstream).

Contributions

• Development of a novel combined spatial interpolation method integrating Co-Kriging with Inverse Distance Weighting-derived covariates and Random Forest optimization.
• Significant improvement in the accuracy of soil water and salt content prediction in arid regions compared to traditional methods.
• Establishment of a transferable framework for multi-method integration in spatial interpolation of soil properties, balancing algorithmic strengths with environmental heterogeneity.

Funding

• Not specified in the provided text.

Citation

@article{Liu2025combined,
  author = {Liu, Shuiqing and Shang, Songhao},
  title = {A combined spatial interpolation method of co-Kriging with inverse distance weighting and random forest for soil water and salt in arid oasis},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134569},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134569}
}