S. et al. (2025) Soil infiltration variability across diverse soil reference groups, textures, and landuse types

Identification

• Journal: Geoderma
• Year: 2025
• Date: 2025-10-14
• Authors: Farnaz Sharghi S., Sara L. Bauke, Mehdi Rahmati, Dymphie Burger, Harry Vereecken, Wulf Amelung
• DOI: 10.1016/j.geoderma.2025.117550

Research Groups

• Institute of Crop Science and Resource Conservation (INRES), Soil Science and Soil Ecology, University of Bonn, Germany
• Agrosphere Institute IBG-3, Forschungszentrum Jülich GmbH, Germany

Short Summary

This study evaluates the variability of soil infiltration parameters, such as saturated hydraulic conductivity (Ks) and final infiltration rate (ic), across diverse soil reference groups, textures, and land-use types using a global database. It finds that World Reference Base (WRB) soil groups, especially when combined with land-use and texture, are significantly more effective in explaining infiltration parameter variability than soil texture or land-use alone, thereby improving upscaling for hydrological modeling.

Objective

• To evaluate the between- and within-class variabilities of key infiltration parameters (e.g., saturated hydraulic conductivity (Ks) and final infiltration rate (ic)) considering different soil texture classes, World Reference Base (WRB) soil reference groups, and land-use types.
• To hypothesize that including WRB soil reference groups and land-use will significantly improve the representation of infiltration parameters compared to the established use of soil texture alone.
• To evaluate how the representation of infiltration parameters depends on the dimensionality (1D or 3D) of the water flow model.

Study Configuration

• Spatial Scale: Global scale, utilizing a worldwide database of in situ infiltration measurements.
• Temporal Scale: Cross-sectional analysis of existing global infiltration data, without specific temporal dynamics of infiltration measurements within the study itself.

Methodology and Data

• Models used:
  • Horton model (for 1D infiltration data)
  • Haverkamp model (for 3D infiltration data)
  • Random Forest (RF) machine learning algorithm (for predictive modeling)
• Data sources:
  • Soil Water Infiltration Global (SWIG) database (5023 individual infiltration curves initially, 4178 selected for analysis)
  • SoilGrids database (for World Reference Base soil groups)
  • Global Land Data Assimilation System (GLDAS-NOAH) (for land-use types)
  • Statistical measures: Mutual Information (MI), Standard Deviation (STD), median, mean.
  • Model evaluation metrics: Root Mean Square Error (RMSE), Normalized RMSE (NRMSE), Coefficient of Determination (R²), Nash-Sutcliffe Efficiency (NSE).

Main Results

• Soil texture alone is inadequate for upscaling infiltration parameters, showing low mutual information (MI = 0.16 for both ic and Ks) and high standard deviation (STD = 1.08 for ic, 3.65 for Ks).
• Land-use alone also poorly explains observed variation in infiltration parameters (MI = 0.28 for ic, 0.14 for Ks; STD = 1.10 for ic, 4.08 for Ks).
• World Reference Base (WRB) soil groups are superior to texture and land-use in explaining infiltration parameter variability, particularly for ic (MI = 0.52, STD = 1.10). For Ks, WRB had MI = 0.20.
• An integrated classification combining texture, land-use, and WRB reference groups yielded the highest mutual information (MI = 0.66 for ic, 0.54 for Ks) and lowest standard deviation values.
• Random Forest modeling demonstrated that incorporating WRB and land-use as categorical predictors significantly improved prediction accuracy for Ks (23% R² improvement) and A (36% R² improvement) in unseen test data, compared to models using only conventional pedotransfer function predictors.
• Different instrument types used for 3D infiltration measurements introduce additional variability, with tension and mini disc infiltrometers showing less variability than single ring infiltrometers.

Contributions

• Provides the first comprehensive, global evaluation of the impact of World Reference Base (WRB) soil groups on infiltration rates, demonstrating their superior predictive power compared to traditional soil texture or land-use classifications.
• Highlights the significant improvement in explaining infiltration parameter variability through an integrated classification approach that combines soil texture, land-use, and WRB.
• Offers crucial guidance for designing more effective frameworks for land and water resource management by emphasizing the need to account for WRB reference groups and land-use in infiltration modeling applications.
• Demonstrates the value of categorical soil and land-use information for developing robust pedotransfer functions (PTFs) for large, spatially and temporally heterogeneous datasets.
• Calls for future database developments and measurement campaigns to systematically include structural attributes (e.g., macroporosity, biogenic channels) and initial soil moisture conditions to enable more process-based analyses and improve parameterization in hydrological models.

Funding

• German Science Foundation (Deutsche Forschungsgemeinschaft DFG) – SFB 1502/1-2022 (project number 450058266)
• Core funds from the Soil Science and Soil Ecology group, University of Bonn

Citation

@article{S2025Soil,
  author = {S., Farnaz Sharghi and Bauke, Sara L. and Rahmati, Mehdi and Burger, Dymphie and Vereecken, Harry and Amelung, Wulf},
  title = {Soil infiltration variability across diverse soil reference groups, textures, and landuse types},
  journal = {Geoderma},
  year = {2025},
  doi = {10.1016/j.geoderma.2025.117550},
  url = {https://doi.org/10.1016/j.geoderma.2025.117550}
}