EUVIMEDEUVIMED entdecken
Crossref

Leveraging machine learning for predictive environmental geology: A comprehensive review

V. O. Binitie, O. Andre-Obayanju

Ife Journal of Science · 05.05.2026 · 28 (1) · 1-14

DOI: 10.4314/ijs.v28i1.1 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Ife Journal of Science, online erschienen am 05.05.2026. Verfasst von V. O. Binitie und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This review is a synthesis of machine learning (ML) applications in predictive environmental geology by conducting a systematic review of 157 peer-reviewed papers published within the period of 2015 to 2025. The study examines ML applications in six key areas (landslide susceptibility mapping, groundwater resource assessment, soil erosion prediction, seismic hazard analysis, mineral prospectivity modeling, and contamination transport prediction). The review has compared the performance properties of supervised learning algorithms (Random Forest, Support Vector Machines, neural networks), unsupervised clustering approaches, and deep learning networks, especially on validation methodologies and spatial prediction issues. Quantitative synthesis indicates that ensemble approaches, particularly Random Forest and gradient boosting, generally outperform traditional geostatistical methods. For example, across 47 landslide studies, median AUC values ranged from 0.89–0.92 for ensemble models, compared to 0.75–0.82 for logistic regression baselines. Convolutional Neural Networks were shown to be more effective on spatial prediction of high-resolution imagery (median R2 = 0.84-0.91 using 23 studies), but at much greater computational cost. Such critical issues as the quality of data (only 34% of the studies reviewed used spatial cross-validation), the model interpretability (especially with deep learning methods), and benchmarking datasets are still missing. The review outlines three research priorities, including (1) the creation of hybrid physics-guided ML architectures that can make use of process-based insights, (2) the creation of robust uncertainty quantification strategies that could be applied to spatial predictions, and (3) the development of transfer learning solutions that can facilitate knowledge transfer across geological contexts. Our synthesis provides evidence-based advice to the practitioners and researchers who are interested in taking advantage of data-driven methodology in environmental geology, and who are also critical of the methodological needs and constraints.

Zitieren

APA

Binitie, V. O., & Andre-Obayanju, O. (2026). Leveraging machine learning for predictive environmental geology: A comprehensive review. Ife Journal of Science, 28(1), 1-14. https://doi.org/10.4314/ijs.v28i1.1

Vancouver

Binitie VO, Andre-Obayanju O. Leveraging machine learning for predictive environmental geology: A comprehensive review. Ife Journal of Science. 2026;28(1):1-14. doi:10.4314/ijs.v28i1.1

BibTeX
@article{Binitie2026,
  author = {Binitie, V. O. and Andre-Obayanju, O.},
  title = {Leveraging machine learning for predictive environmental geology: A comprehensive review},
  journal = {Ife Journal of Science},
  year = {2026},
  volume = {28},
  number = {1},
  pages = {1--14},
  doi = {10.4314/ijs.v28i1.1}
}
Kostenlos starten →