In semi-arid ecoregions of temperate zones, focused snowmelt water infiltration in topographic depressions is a key, but imperfectly understood, groundwater recharge mechanism. Routine monitoring is precluded by the abundance of depressions. We have used remote-sensing data to construct mass balances and estimate volumes of temporary ponds in the Tambov area of Russia. First, small water bodies were automatically recognized in each of a time series of high-resolution Planet Labs images taken in April and May 2021 by object-oriented supervised classification. A training set of water pixels defined in one of the latest images using a small unmanned aerial vehicle enabled high-confidence predictions of water pixels in the earlier images (Cohen’s Κ = 0.99). A digital elevation model was used to estimate the ponds’ water volumes, which decreased with time following a negative exponential equation. The power of the exponent did not systematically depend on the pond size. With adjustment for estimates of daily Penman evaporation, function-based interpolation of the water bodies’ areas and volumes allowed calculation of daily infiltration into the depression beds. The infiltration was maximal (5–40 mm/day) at onset of spring and decreased with time during the study period. Use of the spatially variable infiltration rates improved steady-state shallow groundwater simulations.

This study presents an approach for predicting soil class probabilities by integrating synthetic composite imagery of bare soil with long-term vegetation remote sensing data and soil survey data. The goal is to develop detailed soil maps for the agro-innovation center “Orlovka-AIC” (Samara Region), with a focus on lithological heterogeneity. Satellite data were sourced from a cloud-filtered collection of Landsat 4–5 and 7 images (April–May, 1988–2010) and Landsat 8–9 images (June–August, 2012–2023). Bare soil surfaces were identified using threshold values for NDVI (<0.06), NBR2 (0.10). Synthetic bare soil images were generated by calculating the median reflectance values across available spectral bands. Following the adoption of no-till technology in 2012, long-term average NDVI values were additionally calculated to assess the condition of agricultural lands. Seventy-one soil sampling points within “Orlovka-AIC” were classified using both the Russian and WRB soil classification systems. Logistic regression was applied for pixel-based soil class prediction. The model achieved an overall accuracy of 0.85 and a Cohen’s Kappa coefficient of 0.67, demonstrating its reliability in distinguishing the two main soil classes: agrochernozems and agrozems. The resulting soil map provides a robust foundation for sustainable land management practices, including erosion prevention and land use optimization.

The results of baseline evaluation of soil organic carbon (SOC) contents and stocks, soil bulk density, and their spatial variability and vertical distribution patterns within the upper 40-cm-thick layer of agrosoddy-podzolic soils developed from glacial till (Albic Gleyic Retisols (Episiltic, Anoloamic, Aric, Ochric)) and agrosoddy podzols developed from sands (Albic Podzols (Arenic, Aric, Ochric)) are considered. Thirteen monitoring plots (30 × 30 m) have been established on drained (closed drainage) arable land, hayfield, and unmanaged fallow land of the experimental site of the All-Russia Research Institute of Reclaimed Lands (VNIIMZ) in Tver oblast, European Russia. The spatial variability of the SOC content within the plots exceeds the reproducibility of SOC determination by the dry combustion method by 5–10 times. The standard deviation of the SOC stocks nonlinearly increases with increasing thickness of the assessed layer from 1–5 t/ha for a layer of 0–5 cm to 10–20 t/ha for a layer of 0–40 cm. Differentiated soil sampling from thin layers within the upper 40-cm-thick layer demonstrated a uniform vertical distribution of SOC in the plow layer of the arable land; differences between SOC stocks in the layers of 0–10 and 10–20 cm are statistically insignificant. In the Retisols of the five-year-old hayfield with perennial grasses and of the 5- to 20-year-old unmanages fallow lands, a sod layer of 3–10 cm in thickness is formed, and the SOC content displays an accumulative distribution with a maximum in the surface layer. However, the SOC stocks in the upper and lower parts of the former plow horizon do not display statistically significant differences. Similar regularities have been found for the sandy Podzols against the background of generally lower contents and stocks of SOC.

Soil organic carbon (C _org ) maintains the fundamental properties, regimes, and functions of soils, being a key indicator of soil quality and sustainability. In recent years, complex statistical models have increasingly been used to model and map the C _org content in soils. The problem of these models is their low interpretability. In this paper, we consider interpretable machine learning approaches by the example of thematic modeling of organic carbon content in soils in the south of the Bugul’minsko-Belebeevskaya Upland (Samara oblast). To interpret the results of a random forest model, several methods have been applied including permutation feature importance, partial dependence plots (PDPs), individual conditional expectation (ICE) plots, Shapley values, and interaction analysis using H-statistics. The key factors controlling the spatial variation of C _org in soils have been identified. The Shapley value analysis permitted us to assess both the global and local contribution of features to the prediction results and also to find the functional form of their influence. Significant spatial variability in the contribution of individual factors emphasizes the importance of local characteristics for interpreting modeling results.

A stationary system for automated measurement of soil hydrothermal regimes has been theoretically substantiated and developed. This system was used to conduct regular experimental studies of layer-by-layer moisture and temperature, as well as the groundwater depth, in the Tver Region (VNIMZ Experimental Field) on agrosod-podzolic gleyed light loamy soils. These data were used to construct chronoisopleth maps of moisture and temperature using digital methods. Using layer-by-layer values of soilhydrological constants (SHC), determined from the main hydrophysical characteristics, the soil water regime was analyzed, identifying periods of gravitational runoff, swampiness and soil drought in the annual cycle. Throughout the year, the studied soil layer experiences periods of excess and deficiency of moisture for plants, as well as periods of rapid changes in moisture and temperature throughout the profile due to meteorological conditions, particularly in late October, early November, and late April. This requires appropriate agronomic and reclamation measures for the studied soils. Obtaining, accumulating, and processing quantitative digital data on soil water and temperature regimes enables the application of modern methods for analyzing dynamic soil properties within intelligent digital platforms, establishing strict quantitative correlation with meteorological data, identifying and predicting the onset of critical periods, and scientifically managing soil reclamation processes.

This study examines the possibility for detecting bare soil surface (BSS), using remote sensing data and for applying this approach to model texture and total soil carbon in the topsoil. The study area comprises agricultural fields within the North Caucasus Federal Agricultural Research Center in the northeastern spurs of the Stavropol Upland. Landsat 5/8 imagery and spectral indices (NDVI, NBR2, and BSI) were used to classify BSS. Optimal discrimination of BSS from green and non-photosynthetic vegetation was achieved, using a BSI threshold of >0.18 (F1-score = 0.98). Spatial modeling of topsoil properties was conducted by comparing five regression models (Linear Regression, K-Nearest Neighbors, Random Forest, Extreme Gradient Boosting, and Cubist). A linear regression model demonstrated the highest performance; key predictors included the mean and standard deviation of SWIR2. The model exhibits strong predictive capability for sand, silt, clay, and total carbon (adjusted R ² = 0.82, 0.75, 0.78, and 0.74, respectively). The topsoil is characterized by the range of clay content from 5 to 30%, of silt content from 15 to 60%, of sand content from 20 to 80%, and of total carbon from 0.5 to 3%. Predictive maps of texture and total carbon derived from BSS reveal considerable local heterogeneity, with soils consisting of Chernozems and Phaeozems, influenced by parent rock variability. The modeling results highlight the high predictive power of the BSS spectral characteristics, even when used as a single predictor for digital soil mapping.