Validation of SoilGrids 2.0 in an Arid Region of India using In Situ Measurements

Abstract

As one of the Earth’s most important natural resources, soil plays a prominent role in regulating ecosystem services, human food production systems and in facilitating a region’s sustainable development. Of late, due recognition has been given to soil sciences and soil information systems as they act as a core to achieve the targets of land degradation neutrality and help in fostering soil governance. In this regard, the availability of global soil databases paves the way for implementing successful soil information systems. Currently, harmonized world soil database from the Food and Agriculture Organization and SoilGrids from International Soil References and Information Centre serve various global soil data products in a geospatial-ready format for the scientific fraternity. In this study, SoilGrids 2.0 is validated with in situ measurements in the arid region of the Thar Desert. Soil fractions and pH at the top surface (0–5 cm) and subsurface (5–15 cm) were measured through soil sample analysis collected from the study area and compared with the values retrieved from SoilGrids 2.0 for the same location. This comparison shows that the SoilGrids 2.0 has underestimated the sand fragments up to ~28% and overestimated ~14% for silt and clay fragments. Deviation of pH in SoilGrids 2.0 was also observed with the root mean square error of one unit. However, in the comparison of soil texture classes from the field and the one given by SoilGrids 2.0, a systematic shift was found, indicating the robustness of SoilGrids prediction algorithm that can be fine-tuned by incorporating additional soil profiles (from contributing agencies) as the current snapshot of the soil database lacks dense and well-distributed soil profiles in this arid region.

Assessing the utility of SoilGrids250 for biogeographic inference of plant populations

“…The relationships for soil texture found here were similar to those reported in France (Caubet et al., 2019 ), another area with high density of training data. Results in regions with fewer training points are more variable: no relationships were found between SoilGrids texture predictions and field textures in Norway or Croatia (Huang et al., 2022 ; Radočaj et al., 2023 ), whereas results in arid regions in India were similar to what we observed here (Dandabathula et al., 2022 ). This suggests that biases or noise in SoilGrids predictions of soil texture may be related to regional differences in drivers of soil texture variation rather than the density of training data.…”

“…Results in regions with fewer training points are more variable: no relationships were found between SoilGrids texture predictions and field textures in Norway or Croatia(Huang et al, 2022;Radočaj et al, 2023), whereas results in arid regions in India were similar to what we observed here(Dandabathula et al, 2022). This suggests…”

“…In this study, we explored the utility of SoilGrids for investigating biogeographic patterns within and among species using soil samples from 212 Lobelia population sites representing a broad range of habitats. Most datasets that have been used to evaluate SoilGrids predictions are derived from random or systematic soil sampling distributed across a geographic area of interest (Bodenstein et al., 2022 ; Caubet et al., 2019 ; Dandabathula et al., 2022 ; Dharumarajan et al., 2021 ; Huang et al., 2022 ; Liang et al., 2019 ; Radočaj et al., 2023 ; Tifafi et al., 2018 ). Our test incorporated constraints that are inherent in “presence” datasets for modeling the distributions of individual species (Jeliazkov et al., 2022 ).…”