Topographic wetness index (TWI) is used as a proxy for soil moisture, but how well it performs across varying timescales and methods of calculation is not well understood. To assess the effectiveness of TWI, we examined spatial correlations between in situ soil volumetric water content (VWC) and TWI values over 5 years in soils at 42 locations in an agroforestry catena in Fayetteville, Arkansas, USA. We calculated TWI 546 ways using different flow algorithms and digital elevation model (DEM) preparations. We found that most TWI algorithms performed poorly on DEMs that were not first filtered or resampled, but DEM filtration and resampling (collectively called generalization) greatly improved the TWI performance. Seasonal variation of soil moisture influenced TWI performance which was best when conditions were not saturated and not dry. Pearson correlation coefficients between TWI and grand mean VWC for the 5-year measurement period ranged from 0.18 to 0.64 on generalized DEMs and 0.15 to 0.59 for on DEMs that were not generalized. These results aid management of crop fields with variable moisture characteristics.

Broad-base agricultural terraces can be difficult to delineate in flat landscapes, particularly when covered by crops, due to subtle changes in elevation over relatively wide distances. In northeastern Oklahoma, these terraces are usually less than half a meter high and 15 to 20 m wide. The objective of this research was to develop and test a technique for identifying and classifying terraces using computer vision applied to terrain derivatives calculated from digital elevation models at five sites. We tested 38 terrain-derivative grid combinations or sets that represented 19 terrain characteristics, calculated from elevation models after two Gaussian smoothing strategies to provide some degree of generalization and a removal of excess noise. The best subsets achieved a 98% classification accuracy (kappa 0.96) and consisted of derivatives representing hydrology, morphometry, and visibility categories. Inaccuracies occurred primarily at the edges of some of the study sites, where agricultural fields bordered incised drainage areas where changes in elevation were similar to those for the terraces. Further study will elucidate the relationships between terrace “borrow” and “deposition” areas in the terrace areas and their relationships to yield and salinity issues. This work seeks to automate terrace identification for digital soil mapping on terraced fields for the improved delivery of soil information for resource conservation and land use.

Soil nutrient content varies spatially across agricultural fields in hard-to-predict ways, particularly in floodplains with complex fluvial depositional history. Satellite reflectance data from the Sentinel-2 (S2) mission provides spatially continuous land reflectance data that can aid model development when used with point observations of nutrients. Reflectance from vegetation is assumed to obstruct land reflectance of bare soil, such that researchers have masked vegetation in models. We developed a routine for masking vegetation within Google Earth Engine (GEE) using Random Forest classification for iterative application to libraries of S2-images. Using gradient boosting, we then developed soil nutrient models for surface soils at a 250-ha agricultural site using S2 images. Soils were sampled at 2145 point locations to a 23-cm depth and analyzed for Ca, K, Mg, P, pH, S, and Zn. Results showed that masking vegetation improved model performance for models from subsets of the data (80% of samples used for model development, 20% validation), but full data sets did not require masking to achieve accuracy. Models of Ca, K, Mg, and S were successful (validation R² > 0.60 to 0.96), but models for pH, P, and Zn failed. Bare soil composite images from S2 data are helpful in predicting soil fertility in low-relief floodplains.

Morro de Seis Lagos, a region in the Brazilian Amazon, contains a small (less than 1%) formation of siderite carbonatites which is considered to be one of the world’s largest niobium reserves. This highly weathered geological and pedological occurrence makes the site ideal for studying the pedogenetic process of lateralization and the spatial variability of chemical elements. The aim of this study was to investigate the influences of various sampling combinations (scenarios) derived from three sampling designs on the spatial predictions associated with chemical compounds (Al₂O₃, Fe₂O₃, MnO, Nb₂O₅, TiO₂, and SiO₂), using multiple machine learning algorithms combined with remotely sensed imagery. The dataset comprised 341 samples from the Geological Survey of Brazil (CPRM). Covariates included remotely sensed data collected from Sentinel-2 MSI, Sentinel-1A, and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), and topographic attributes were calculated from a 20 m digital elevation model derived from hydrologic data (HC-DEM). The machine learning algorithms (Generalized Linear Models with Elastic Net Regularization (GLMNET), Nearest Neighbors (KNN), Neural Network (NNET), Random Forest (RF) and Support Vector Machine (SVMRadial) were used in combination with covariates and measured elements at point locations to spatially map the concentrations of these chemical elements. The optimal covariates for modeling were selected using Recursive Feature Elimination (RFE), processing 10 runs for each chemical element. The RF, SVMRadial, and KNN models performed best, followed by the models from the Neural Network group (NNET). The sampling scenarios were not significantly different, based on root mean square error (F = 1.7; p-value = 0.15) and mean absolute error (F = 0.4; p-value = 0.79); however, significant differences were observed in the coefficient of determination (F = 41.2; p-value < 0.00) across all models. Overall, the models performed poorly for all elements, with R² ranging from 0.07 to 0.27, regardless of sampling scenario (F = 1.6; p-value = 0.08). Relatively, RF, GLMET, and KNN performed better, compared to other models. The terrain attributes were significantly more successful as to the spatial predictions of the elements contained in laterites than were the remote sensing spectral indices, likely due to the fact that the underlying spatial structures of the two formations (laterite and talus) occur at different elevations.

Soil salinity is a serious threat to crop productivity and is anticipated to increase in the coming decades, particularly in semi‐arid to arid agricultural regions. Accessible geospatial data and data mining techniques can enable high‐resolution mapping of soil salinity to improve predictions and soil management. In this study, 64 features derived from Sentinel‐1, Sentinel‐2, land surface temperature, apparent electrical conductivity, and auxiliary geospatial datasets were used to classify salinity in the Sonoran Basin and Range (SBR) ecoregion. As ground truth data, 361 soil samples were collected, and electrical conductivity was measured using the saturated paste method. Recursive feature elimination selected 12 variables for the optimized model (OM), while inferential analysis further reduced them to six for the simplified model (SM). Cross‐validation analysis of OM and SM yielded comparable accuracy (0.68 ±0.04 and 0.68 ±0.05). Based on this and additional analysis (classification report, confusion matrix, etc.), the SM was utilized to predict salinity and non‐salinity maps (10 m resolution) in two areas within the SBR ecoregion (121,110 ha). Based on uncertainty analysis, the non‐saline predicted class comprised larger areas with uncertainties <0.2 compared to the saline class. Overall, the aridity index and Sentinel‐2 features were the most relevant predictors. Findings provide valuable insights for identifying non‐saline areas potentially suitable for agriculture and highlight the importance of salinity mapping and predictions for soil and water resource planning for sustaining crop and range production in arid regions.