Spatial inference of soil mechanical properties based on mechanical feature similarity

Soil mechanical properties play a critical role in geotechnical engineering, agricultural production, and military applications. Their spatial characterization at regional scales is vital for engineering planning, land use management, and surface mobility assessments. Due to the high cost and difficulty of field sampling for mechanical parameters, sample sizes are often sparse. Traditional spatial interpolation methods frequently fail to provide reliable estimates when dealing with extremely limited data. This paper applies a spatial inference approach grounded in the third law of geography to the estimation of soil mechanical properties, by introducing the concept of mechanical feature similarity. Its core principle posits that regions with similar mechanical features should exhibit analogous mechanical properties. Unlike traditional interpolation, this method fully integrates multisource environmental and geographic data—including soil classification, bulk density, and digital elevation models—to construct a multidimensional mechanical feature space within a GIS framework. Spatial inference is achieved through similarity calculations, with uncertainty distributions output to characterize prediction reliability. We evaluated this method using a bearing capacity dataset from Wasit Province, Iraq, comparing its performance against inverse distance weighting (IDW) and ordinary kriging (OK) methods. Results demonstrate that the approach significantly improves inference accuracy (RMSE=1.53, MAE=1.29), outperforming IDW (RMSE=2.49) and OK (RMSE=2.22).