资源环境科技发展态势分析平台

Swelling soils contain high amounts of expansive clay minerals which swell upon wetting and shrink upon drying. Urban growth across swelling soils causes billions of dollars in infrastructure damage annually throughout the United States. Here, continuous spatial information of soil swelling properties is provided across Colorado, a state experiencing extensive urban growth across swelling soils. The objectives were: 1) model the spatial variability and associated uncertainty of soil swelling-related properties across Colorado; 2) generate a continuous statewide map of soil swelling potential and associated uncertainty at 1x1km resolution; and 3) identify urban areas prone to experience swelling conditions. A digital soil mapping (DSM) framework for extracting pedologically-relevant information from legacy maps was used to train machine learning models to analyze multiple sources of information to represent the soil forming environment (e.g., soils, climate, organisms, topography). Best predictions were based on regression trees and explain over 80% of soil swelling spatial variability across the state (10-fold cross validation strategy). Over 20% of urbanized areas were identified as being prone to experience swelling conditions. Special considerations must be undertaken to couple urban development with soil functionality in order to prevent future damages and economic losses.