18 AI-Driven and Machine Learning Approaches in Geospatial Modeling for Hydrologic and Geologic Mapping

This session features GIS, remote sensing, and AI-driven/machine learning approaches to geologic and hydrologic mapping, emphasizing predictive modeling, cross-scale analysis, and innovative workflows that extend beyond traditional cartographic methods.

Tuesday, 10 March 2026
8:30 AM - 9:30 AM
RCC, 104

Machine Learning for Supporting Digital Mapping of Soil Profile to Monitor Copper Distribution in the Brazilian Cerrado

Paper number: 18-0
Program: Technical Sessions
Time: 8:30 AM – 8:45 AM

Authors

Abstract

The distribution of soil attributes in profiles can be studied innovatively using digital mapping of soil profile (DMSP). To support that novel approach, the right combination of machine learning algorithms and non-destructive soil analysis obtained by low-cost sensors is essential to spread the adoption of DMSP properly. The aim of this research was to evaluate the accuracy of the combinations between machine learning algorithms with data obtained from Nix Spectro 2™ to predict Copper (Cu) concentration in soils under agricultural use for different periods in the Brazilian Cerrado. This study was carried out in the southwest of the state of Piauí, Brazilian Northeast, at four locations within the same farm, including virgin soil and soil cultivated for 12, 18, and 24 years, respectively. A 1 m² grid was designed to collect 100 equidistant soil samples in each studied soil profile, making a total of 400. All samples were scanned with Nix Spectro 2™ sensor and analyzed through atomic absorption spectrometry to obtain Cu concentrations. The data set was randomly divided into modeling (120 samples) and predicting groups (280 samples). The modeling group was also randomly subdivided into training (94 samples) and validation (24 samples) groups to support the comparison between decision tree (DT), random forest (RF) and gradient boosting (GB) algorithms for predicting Cu. The algorithm with best performance was applied to predict Cu. Observed and Predicted values were used in geostatistical modeling to produce the DMSP. The results showed that GB was more accurate to predict Cu. The linear regression between observed and predicted values show an R2 of 0.63, which can be considered good regarding the purpose of the model. The predictors more relevant for predicting Cu were wavelength information between 470 and 510 nm. The DMSP shows strong similarities between observed and predicted Cu in all studied soils. It is concluded that GB is efficient in predicting Cu through data obtained by Nix Spectro 2™, providing significant support for promoting accurate DMSP. 

Geological Society of America Abstracts with Programs. Vol. 58, No. 1

doi: 10.1130/abs/2026SE-411512

© Copyright 2026 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.