MACHINE LEARNING ON THE DETAILED MAPPING OF SOIL LIMESTONE REQUIREMENT

Soil acidity is a significant constraint to agricultural productivity in Brazil, which makes the precise application of lime necessary to improve soil fertility. This study aimed to obtain a detailed lime requirement (LR) map, indicating the necessary quantities of lime for soil layers up to 20 cm deep. Chemical attribute maps were generated using the Random Forest (RF) model in a Google Earth Engine and Google Collab environment. These parameter maps were applied in the empirical LR formula for the Piracicaba region, and through map algebra, a spatial LR map was created. Model performance was evaluated through 10-fold cross-validation, using the metrics coefficient of determination (R²), RMSE, and RPIQ. In the study region, the lime requirement ranged from 0 to 4 t. ha^-1.

Keywords - Lime requirement mapping, pedometrics, detailed mapping, soil fertility, agricultural decision support.