Related Experiment Video
Updated: Nov 4, 2025
09:44
Use of Principal Components for Scaling Up Topographic Models to Map Soil Redistribution and Soil Organic Carbon
Published on: October 16, 2018
10.4K
Unveiling spatial variability in herbicide soil sorption using Bayesian digital mapping.
Franca Giannini-Kurina1,2, Susana Hang2, Ariel E Rampoldi2
1CONICET, UFYMA Unidad de Fitopatología y Modelización Agrícola, Córdoba, 5000, Argentina.
Journal of Environmental Quality
|May 29, 2021
Summary
Spatial Bayesian regression effectively predicts herbicide sorption (Kd) across landscapes. This method offers a cost-efficient alternative to traditional analytical quantification for soil risk assessments.
Area of Science:
- Environmental Science
- Soil Science
- Agricultural Science
Background:
- Regional mapping of herbicide sorption to soil is crucial for environmental risk assessment.
- Analytical quantification of adsorption coefficients (Kd) is costly for large-scale studies, necessitating improved spatial prediction methods.
- Spatial Bayesian regression (BR) is an emerging technique for creating continuous spatial maps from discrete sample data.
Purpose of the Study:
- To develop a predictive spatial Bayesian regression (BR) model for herbicide sorption to soil at a landscape scale.
- To assess the performance of BR in predicting herbicide sorption coefficients (Kd) compared to regression kriging.
- To identify key soil and climate variables influencing herbicide sorption.
Main Methods:
- Integrated ancillary soil and climate data with Kd measurements from 80 (glyphosate) and 120 (atrazine) sites in central Argentina.
- Developed spatial mixed-effects models incorporating site random effects.
- Assessed model performance using cross-validation and bootstrap for residual variability estimation, comparing BR with regression kriging.
Main Results:
- Spatial BR predictions significantly outperformed regression kriging.
- The glyphosate Kd model, with a root mean square prediction error of 13%, identified aluminum oxides, pH, and clay content as key predictors.
- The atrazine Kd model, with a root mean square prediction error of 27%, highlighted soil organic carbon, clay content, and climate variables related to water availability.
Conclusions:
- Spatial modeling using BR provides an efficient and modern approach for mapping herbicide sorption to soil at landscape scales.
- The developed models enhance environmental interpretations of complex edaphic processes.
- BR offers a cost-effective strategy for large-scale soil risk assessments related to herbicide fate.
