Abstract
The study presents the results of analysis and digital mapping of exchangeable calcium (Caex) and magnesium (Mgex) content in the upper soil layer (0–30 cm) of the Cis-Salair drained plain (Toguchinskii district, Novosibirsk oblast, Russian Federation). Mapping by the Random Forest (RF) machine learning algorithm on the Google Earth Engine platform was based on raster maps of 92 predictors characterizing soil formation factors (climate, relief, vegetation, spatial position, soil properties) and archival soil data (ZapsibNIIgiprozem, 1983–1994). The training dataset contained information on the content of Caex and Mgex for 524 soil profiles, and the validation dataset, for 100 soil profiles. For the Сaex model, the following indicators of effectiveness of RF modeling were obtained: coefficient of determination for the training dataset \(R_{{{\text{TD}}}}^{2}\) was 0.86, coefficient of determination for the validation dataset \(R_{{{\text{VD}}}}^{2}\) was 0.21, the root mean square error RMSEVD was 8.4 cmol(+)/kg, and the mean absolute error MAEVD was 6.3 cmol(+)/kg. For the Mgex model, simulation efficiency indicators were as follows: \(R_{{{\text{TD}}}}^{2}\) was 0.81, \(R_{{{\text{VD}}}}^{2}\) was 0.32, RMSEVD was 4 cmol(+)/kg, and the MAEVD was 2.9 cmol(+)/kg. The Caex content in the top 0–30 cm of the soil ranged from 3.1 to 78.2 cmol(+)/kg, and the Mgex content, from 0.2 to 46.9 cmol(+)/kg. The highest content of Caex and Mgex was detected in the soils of the eastern and northeastern parts of the Toguchinskii district: more than two times higher than in soils of its western and southern parts. Soils of the studied area can be classified into different groups based on the level of exchangeable calcium they contain. The levels range from elevated (10–15 cmol(+)/kg) to high (15–20) and very high (>20). Similarly, the exchangeable magnesium levels can be categorized as elevated (2.1–3.0 cmol(+)/kg), high (3.1–4.0), or very high (>4). Areas of soil distribution with critical ratios of these exchangeable elements have been identified based on the Caex : Mgex ratio map, including thresholds as follows: <2 : 1 suggests risk of colloids peptization under high sodium (Na+) content in soils, coupled with calcium deficiency for plants, leading to soil structure degradation and impaired nutrient availability; >8 : 1 points to possible magnesium deficiency for plants. When the Caex : Mgex ratio was less than 2 : 1, the predominant soils (with the physical clay content of 30–65%) were meadow chernozems (leached, meadow-chernozemic solonetzic, and solonchakous); leached chernozems; gray and dark gray forest soils, and meadow-chernozemic solonetz. When the ratio exceeded 8 : 1, the predominant soils (with the physical clay content of 20–61%) included meadow soils (leached, podzolized, typical, and carbonate); meadow-chernozemic and chernozemic-meadow soils (leached, typical, and podzolized), chernozems (leached and podzolized), and light gray, gray, and dark gray forest soils.
