Colombian soil texture: building a spatial ensemble model

Abstract

Abstract. Texture is a fundamental soil property for multiple applications in environmental and earth sciences. Knowing its spatial distribution allows a better understanding of the response of soil conditions to changes in the environment, such as land use. This paper describes the technical development of Colombia's first texture maps, obtained via a spatial ensemble of national and global digital soil mapping products. This work compiles a new database with 4203 soil profiles, which were harmonized at five standard depths (0–5, 5–15, 15–30, 30–60, and 60–100 cm) and standardized with additive log ratio (ALR) transformation. A compilation of 83 covariates was developed and harmonized at 1 km2 of spatial resolution. Ensemble machine learning (EML) algorithms (MACHISPLIN and landmap) were trained to predict the distribution of soil particle size fractions (PSFs) (clay, sand, and silt), and a comparison with SoilGrids (SG) products was performed. Finally, a spatial ensemble function was created to identify the smallest prediction errors between EML and SG. Our results are the first effort to build a national texture map (clay, sand, and silt fractions) based on digital soil mapping in Colombia. The results of EML algorithms showed that their accuracies were very similar at each standard depth, and were more accurate than SG. The largest improvement with the spatial ensemble was found at the first layer (0–5 cm). EML predictions were frequently selected for each PSF and depth in the total area; however, SG predictions were better when increasing soil depth in some specific regions. The final error distribution in the study area showed that sand presented higher absolute error values than clay and silt fractions, specifically in eastern Colombia. The spatial distribution of soil texture in Colombia is a potential tool to provide information for water-related applications, ecosystem services, and agricultural and crop modeling. However, future efforts need to improve aspects such as treating abrupt changes in the texture between depths and unbalanced data. Our results and the compiled database (https://doi.org/10.6073/pasta/3f91778c2f6ad46c3cc70b61f02532db, Varón-Ramírez and Araujo-Carrillo, 2022, https://doi.org/10.6073/pasta/d6c0bf5847aa40836b42dcc3e0ea874e, Varón-Ramírez et al., 2022) provide new insights to solve some of the aforementioned issues.