Brazil has made notable advances in soil mapping compared to other Latin American countries, driven by collaborative efforts from federal agencies, research institutions, and universities. However, detailed soil maps remain limited, particularly at local scales. This study explored soil spatial variability in a coffee plantation located in the Brazilian Cerrado, assessing the effectiveness of integrating proximal and remote sensing data to create detailed soil maps that support precision agriculture. Soil samples were collected from multiple depths across the study area and analyzed for texture, fertility, and elemental composition using portable X-ray fluorescence (pXRF). Additionally, terrain attributes derived from a digital elevation model were examined to understand their relationship with soil properties. Our results demonstrated that elements associated with parent material, such as Fe, Si, Ti, Al, and Ca, were more reliable indicators for distinguishing soil classes than topographic features. The dominance of Cambissolos Háplicos (CX), which have lower clay content and contain gravel, suggested a reduced need for soil amendments compared to Latossolos Vermelhos (LV), leading to potential cost savings for producers. These findings underscore the utility of pXRF in detecting soil variability and emphasize that combining proximal and remote sensing data can enhance the efficiency and sustainability of agricultural management.

Citriculture has worldwide importance, and monitoring the nutritional status of plants through leaf analysis is essential. Recently, proximal sensing has supported this process, although there is a lack of studies conducted specifically for citrus. The objective of this study was to evaluate the application of portable X-ray fluorescence spectrometry (pXRF) combined with machine learning algorithms to predict the nutrient content (B, Ca, Cu, Fe, K, Mg, Mn, P, S, and Zn) of citrus leaves, using inductively coupled plasma optical emission spectrometry (ICP-OES) results as a reference. Additionally, the study aimed to differentiate 15 citrus scion/rootstock combinations via pXRF results and investigate the effect of the sample condition (fresh or dried leaves) on the accuracy of pXRF predictions. The samples were analyzed with pXRF both fresh and after drying and grinding. Subsequently, the samples underwent acid digestion and analysis via ICP-OES. Predictions using dried leaves yielded better results (R² from 0.71 to 0.96) than those using fresh leaves (R² from 0.35 to 0.87) for all analyzed elements. Predictions of scion/rootstock combinations were also more accurate with dry leaves (Overall accuracy = 0.64, kappa index = 0.62). The pXRF accurately predicted nutrient contents in citrus leaves and differentiated leaves from 15 scion/rootstock combinations. This can significantly reduce costs and time in the nutritional assessment of citrus crops.

Soils are fundamental to support ecosystem functions. Therefore, understanding soil organic carbon stock (SCS) as a strategic environmental component for climate regulation, sustainable agricultural practices, and maximizing fertilizer application is essential. This study aimed to predict and map the SCS at Palmital Farm, Minas Gerais state, using land use, parent materials, soil classes, proximal and remote sensing data as predictor variables. A total of 76 soil samples were collected in the surface (A) and subsurface (B) horizons using a regular grid design (200 × 200 m). Carbon content was determined using the Walkley–Black method. Prediction models were created via Random Forest algorithm and validated through leave-one-out cross-validation. Environmental and anthropogenic influences on SCS were analyzed statistically and spatially. The highest mean SCS values were found at the surface horizon under native vegetation and pasture areas (~ 33 Mg ha⁻¹), due to continuous organic matter deposition and lack of disturbance. Charnockite-derived soils showed higher SCS (31.24 Mg ha⁻¹, surface). Among soil classes, Argissolos stood out reaching the highest SCS (35.86 Mg ha⁻¹). Remote sensing data and terrain attributes yielded the best SCS prediction (R²=0.43). The main factors influencing SCS were land use/management and soil texture. Also, strong correlations were observed between vegetation indices, some terrain attributes and SCS. This approach, combining multiple variables, indicated a moderate potential to achieve accurate SCS predictions using these data, demonstrating the high variability and complexity of SCS spatial variation in detailed-scale mapping.

Soil parent material information is essential for defining potential soil uses. However, intense tropical weathering makes direct identification of parent material challenging considering their great depths. Proximal sensing can support the detailed prediction and mapping of parent material through soil analysis. The objective of this study was to map parent material on an experimental farm in Southeastern Brazil using a portable X-ray fluorescence spectrometer (pXRF) and the Random Forest algorithm. A total of 114 soil samples (0-20 and 60-80 cm) were collected, representing multiple soil classes, landforms, and land uses. In addition, 17 rock fragments (gabbro and granite-gneiss) were collected for model calibration. All samples (soil and rock) were analyzed using pXRF. Parent material prediction models were then developed using the Random Forest algorithm and applied to spatialized pXRF-derived chemical data to generate parent material maps. The accuracy of spatial predictions was evaluated using the Kappa index (KI), overall accuracy (OA), and user’s and producer’s accuracies. The 60-80 cm dataset yielded the highest prediction accuracy (KI = 0.52; OA = 0.76). The most important variables were Mn, Fe, P, K, and Ca contents. These results demonstrate that pXRF data can be used to predict and map the spatial distribution of soil parent material with reasonable accuracy, even across different soil classes and land uses.