Fabrício Do Carmo Farias

Context: Pedological mapping in the Amazon is essential but highly challenging, particularly in hydromorphic environments where soil properties and boundaries are difficult to delineate. Improving mapping accuracy is critical for sustainable land-use planning and ecosystem management, yet conventional approaches are often time-consuming and costly. Digital soil mapping (DSM) offers a promising alternative by integrating machine learning algorithms with geospatial data to predict soil mapping units. Aims: This study evaluates DSM approaches in the Eastern Amazon using legacy soil data, environmental covariates, and multiple machine learning algorithms, with emphasis on classification accuracy and spatial uncertainty. Methods: We compared the performance of eight machine learning algorithms for predicting soil mapping units using predictors derived from a digital elevation model, Landsat-9 imagery, and legacy soil maps. A total of 244 spatial sampling units (30 m × 30 m pixels), corresponding to the centroids of soil mapping units and supported by field observations across nine soil classes, were used, with 70% allocated for training and 30% for validation. Recursive Feature Elimination was applied to identify the most relevant predictors. Model performance was assessed using overall accuracy, Kappa index, user’s accuracy, and producer’s accuracy. Key results: The random forest implementation algorithm (ranger) achieved the best performance, with an overall accuracy of 0.74 and a Kappa index of 0.71. Other algorithms showed moderate agreement with the conventional pedological map (accuracy = 0.56; Kappa = 0.49), whereas the artificial neural network showed low performance (accuracy = 0.26; Kappa = 0.14). Conclusions: Results indicate that the density and spatial configuration of soil mapping units influence the agreement between conventional and digital maps, particularly in hydromorphic environments where classification uncertainty is higher. Implications: These findings highlight the potential of decision tree-based algorithms for DSM in complex tropical landscapes, while emphasizing the need for further research to address limitations related to data availability, spatial resolution, and the quality of reference soil maps.

Predictive Digital Soil Mapping (PDSM) in Eastern Amazonia faces challenges due to its environmental complexity, difficult access, and scarce legacy data. While legacy soil maps contain valuable tacit knowledge, updating them requires methods that can handle uncertainty. This study evaluates the integration of old soil maps with machine learning to update soil information in Tracuateua, Pará, with a specific focus on the performance of ensemble learning and the explicit incorporation of uncertainty metrics in soil mapping units under hydromorphic influence, which, in addition to being difficult to access, are influenced by complex pedogenetic processes. We combined 270 sampling points, equivalent to the total pixels that captured the variability of soil mapping units, with environmental covariates and historical data. Several algorithms were tested, including an ensemble approach, to predict mapping units and quantify uncertainty through entropy and confusion indices. The ensemble model demonstrated improved stability and reduced classification uncertainty compared to single models, particularly in challenging hydromorphic environments. Although accuracy gains were modest, the models captured soil–environment relationships, with climate as: Annual Mean Temperature 22,000 years ago (Tmean_22k), relief: Channel Network Base Level (CNBL and altitude) and organism variables: Land Surface Temperature (LST) emerging as the main predictors. Spatialized uncertainty estimates, expressed through entropy and the confusion index, provide a practical decision-support tool for guiding field surveys and identifying areas of low mapping reliability. By explicitly transferring the pedologist’s mental model—encoded as tacit knowledge in legacy soil maps—into ensemble learning, this study presents a robust and transferable framework for updating soil maps in data-scarce tropical regions, balancing predictive performance, spatial consistency, and uncertainty-aware interpretation.