Abstract:The effective use of environmental covariates in characterizing soil–environment relationships is key to successful digital soil mapping. The typical way to use environmental covariates in digital soil mapping is by selecting diverse environmental covariates considering the overall geographical characteristics of the study area and considering these covariates to have consistent applicability across the whole area. However, this practice ignores the fact that the applicability of each environmental covariate in characterizing soil–environment relationships varies over complex environmental conditions, especially in large areas. This study proposed a method to adaptively consider covariate applicability in large-area digital soil mapping using soil–environment relationships. The applicability of each covariate at each location was quantified from the terrain conditions using the newly designed fuzzy functions in the study. Then the covariate applicability was regarded as the importance weight and integrated into an existing representative method, iPSM (individual predictive soil mapping). The integration was separately performed at the similarity calculation and soil estimation stages of iPSM to generate two new methods: iPSM weighting on the applicability of all covariates (iPSM_WCovar_all), and iPSM weighting on the applicability of the limiting covariate (i.e., the covariate with the minimum similarity between two locations that constrains the overall similarity) (iPSM_WCovar_limit). Experiments were carried in Anhui Province, China. The two new methods were used to predict the soil organic matter content of topsoil and outperformed the original iPSM and random forest kriging methods. The root mean square error of the iPSM_WCovar_all, iPSM_WCovar_limit, iPSM and random forest kriging methods were 8.14, 8.00, 8.88 and 9.65 g/kg, respectively, while the mean absolute error of those methods were 6.48, 6.31, 6.61 and 6.82 g/kg. Both proposed methods outperformed the iPSM method and the other commonly used method, i.e., random forest kriging. Moreover, the performance was stable under different parameter settings. Experimental results indicate that the idea of adaptively considering covariate applicability in digital soil mapping is feasible and effective.

A Case-Based Reasoning Strategy of Integrating Case-Level and Covariate-Level Reasoning to Automatically Select Covariates for Spatial Prediction

A GIS-based CBR model for land-use inference

Local Learning for Covariate Selection in Nonparametric Causal Effect Estimation with Latent Variables

Model Selection with Misspecified Spatial Covariance Structure

A Case-Based Reasoning Approach for Task-Driven Remote Sensing Image Discovery under Spatial–Temporal Constraints

Prediction of network public opinion features in urban planning based on geographical case-based reasoning

Variable Selection of Spatial Logistic Autoregressive Model with Linear Constraints

A Hybrid Retrieval Strategy for Case-based Reasoning Using Soft Likelihood Functions

A backward elimination discrete optimization algorithm for model selection in spatio-temporal regression models

A New Covariate Selection Strategy for High Dimensional Data in Causal Effect Estimation with Multivariate Treatments

Importance of spatial predictor variable selection in machine learning applications -- Moving from data reproduction to spatial prediction

Cokrig-and-Regress for Spatially Misaligned Environmental Data

Landslide Susceptibility Evaluation Using Hybrid Integration of Evidential Belief Function and Machine Learning Techniques

Using Automated Machine Learning for Spatial Prediction—The Heshan Soil Subgroups Case Study

Spatial Tools for Case Selection: Using LISA Statistics to Design Mixed-Methods Research

Causal Learning Empowered OD Prediction for Urban Planning

Mineral Prospectivity Mapping Method Integrating Multi-Sources Geology Spatial Data Sets and Case-Based Reasoning

Spatial Prediction of Soil Organic Matter Using Terrain Indices and Categorical Variables As Auxiliary Information

Digital soil mapping with adaptive consideration of the applicability of environmental covariates over large areas

A novel strategy to assimilate category variables in land-use models based on Dirichlet distribution