Digital soil mapping including additional point sampling in Posses ecosystem services pilot watershed, southeastern Brazil

Enhancing the accuracy of digital soil mapping using the surface and subsurface soil characteristics as continuous diagnostic layers

Digital soil mapping relies on relating soils to a particular set of covariates, which capture inherent soil spatial variation. In digital mapping of soil classes, the most commonly used covariates are topographic attributes, RS attributes, and maps, including geology, geomorphology, and land use; in contrast, the subsurface soil characteristics are usually ignored. Therefore, we investigate the possibility of using soil diagnostic characteristics as covariates in a mountainous landscape as the main aim of this study. Conventional covariates (CC) and a combination of soil subsurface covariates with conventional covariates (SCC) were used as covariates, and random forest (RF), Multinomial Logistic Regression (LR), and C5.0 Decision Trees (C5) were used as different machine learning algorithms in digital mapping of soil family classes. Based on the results, the RF model with the SCC dataset had the best performance (KC = 0.85, OA = 90). In all three models, adding soil covariates to the sets of covariates increased the model performance. Soil covariates, slope, and aspect were selected as the principal auxiliary variables in describing the distribution of soil family classes.

Disentangling Jenny's equation by machine learning

The so-called soil-landscape model is the central paradigm which relates soil types to their forming factors through the visionary Jenny's equation. This is a formal mathematical expression that would permit to infer which soil should be found in a specific geographical location if the involved relationship was sufficiently known. Unfortunately, Jenny's is only a conceptual expression, where the intervening variables are of qualitative nature, not being then possible to work it out with standard mathematical tools. In this work, we take a first step to unlock this expression, showing how Machine Learning can be used to predictably relate soil types and environmental factors. Our method outperforms other conventional statistical analyses that can be carried out on the same forming factors defined by measurable environmental variables.

Spatial prediction of soil depth using environmental covariates by quantile regression forest model

Prediction of soil depth for larger areas provides primary information on soil depth and its spatial distribution that becomes vital for land resource management, crop, nutrient, and ecosystem modeling. The present study assessed the spatial distribution of soil depth over 160,205 km² of Andhra Pradesh, India, using 20 covariables by quantile regression forest (QRF). An aggregate of 2854 soil datasets compiled from various physiographic units were randomly partitioned into 80:20 ratio for calibration (2283 samples) and validation (571 samples). Landsat imagery, terrain datasets (8), and bioclimatic factors (11) were utilized as covariates. The QRF model outputs signified that precipitation, multi-resolution index of valley bottom flatness (MrVBF), mean diurnal range, isothermality, and elevation were the most important variables influencing soil depth variability across the landscape. Spatial prediction of soil depth by QRF model yielded a ME of - 1.81 cm, RMSE of 34 cm, PICP of 90.2, and a R² value of 42% as compared to ordinary kriging which results in a ME of - 0.14 cm, a RMSE of 37 cm, and a R² value of 32%. As soil depth is spatially dynamic and has significant correlation with terrain and environmental covariates, better prediction was possible by the QRF model. However, high-density bioclimatic variables could be utilized along with high-resolution terrain variables to improve the predictive accuracy.

Soil Color and Mineralogy Mapping Using Proximal and Remote Sensing in Midwest Brazil

Soil color and mineralogy are used as diagnostic criteria to distinguish different soil types. In the literature, 350–2500 nm spectra were successfully used to predict soil color and mineralogy, but these attributes currently are not mapped for most Brazilian soils. In this paper, we provided the first large-extent maps with 30 m resolution of soil color and mineralogy at three depth intervals for 850,000 km2 of Midwest Brazil. We obtained soil 350–2500 nm spectra from 1397 sites of the Brazilian Soil Spectral Library at 0–20 cm, 20–60, and 60–100 cm depths. Spectra was used to derive Munsell hue, value, and chroma, and also second derivative spectra of the Kubelka–Munk function, where key spectral bands were identified and their amplitude measured for mineral quantification. Landsat composites of topsoil and vegetation reflectance, together with relief and climate data, were used as covariates to predict Munsell color and Fe–Al oxides, and 1:1 and 2:1 clay minerals of topsoil and subsoil. We used random forest for soil modeling and 10-fold cross-validation. Soil spectra and remote sensing data accurately mapped color and mineralogy at topsoil and subsoil in Midwest Brazil. Hematite showed high prediction accuracy (R2 > 0.71), followed by Munsell value and hue. Satellite topsoil reflectance at blue spectral region was the most relevant predictor (25% global importance) for soil color and mineralogy. Our maps were consistent with pedological expert knowledge, legacy soil observations, and legacy soil class map of the study region.

Classification of Soils into Hydrologic Groups Using Machine Learning

Hydrologic soil groups play an important role in the determination of surface runoff, which, in turn, is crucial for soil and water conservation efforts. Traditionally, placement of soil into appropriate hydrologic groups is based on the judgement of soil scientists, primarily relying on their interpretation of guidelines published by regional or national agencies. As a result, large-scale mapping of hydrologic soil groups results in widespread inconsistencies and inaccuracies. This paper presents an application of machine learning for classification of soil into hydrologic groups. Based on features such as percentages of sand, silt and clay, and the value of saturated hydraulic conductivity, machine learning models were trained to classify soil into four hydrologic groups. The results of the classification obtained using algorithms such as k-Nearest Neighbors, Support Vector Machine with Gaussian Kernel, Decision Trees, Classification Bagged Ensembles and TreeBagger (Random Forest) were compared to those obtained using estimation based on soil texture. The performance of these models was compared and evaluated using per-class metrics and micro- and macro-averages. Overall, performance metrics related to kNN, Decision Tree and TreeBagger exceeded those for SVM-Gaussian Kernel and Classification Bagged Ensemble. Among the four hydrologic groups, it was noticed that group B had the highest rate of false positives.