Answers are grounded in this paper and cite their sources.
Given the difficulties in carrying out pedological surveys using conventional methods, digital mapping of soil classes (DMSC) has the potential to support the resumption of systematic pedological surveys. For this, strategies that use legacy information from conventional surveys are desirable. However, it is necessary to understand what information can be predicted in the DMSC of new areas with satisfactory quality. Object-based DSMC strategies have outperformed pixel-based approaches, although it is unclear whether this is true in any context. This study aimed to evaluate the use of legacy information from systematic soil surveys, comparing pixel-based and object-based DMSC approaches. The study area comprises the São Paulo municipalities of Cordeirópolis, Santa Gertrudes, Iracemápolis, Rio Claro and Ipeúna. Elementary relief units were used to delimit polygons in the construction of databases for DMSC based on objects. The results were evaluated considering the global accuracy, Kappa index, in addition to the true positive rates and the precision observed in the predicted classes. Elementary relief units were effective when used in an object-based DMSC approach. The tests employed showed satisfactory results for the extrapolation of legacy information from systematic pedological surveys in the state of São Paulo in pixel and object DMSC approaches, reaching an accuracy of 61 % and a Kappa index of 0.54. The low performance in predicting specific classes was not directly related to their greater or lesser occurrence in the study area.
No takes yet. Share an insight, caveat, or question.
Valadares et al. (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: