Near Real-Time Detection of EVI Time-Series Breakpoints Using Bayesian Inference for Deforestation Monitoring in the Chaco Forest
Grings, Francisco González Bianco, Francisco Roitberg, Esteban Morandeira, Natalia Arellana, Javier Gayol, Maira
Grings, Francisco
González Bianco, Francisco
Roitberg, Esteban
Morandeira, Natalia
Arellana, Javier
Gayol, Maira
Type
Ponencia en congreso
Abstract
Deforestation poses a significant threat to natural ecosystems, particularly in Argentina’s Chaco region—one of the world’s most rapidly changing forest areas. This study focuses on the detection of sudden deforestation events, where forest cover is rapidly removed within a few months. Monitoring such changes across vast areas requires the use of satellite-based vegetation indices, such as the Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI) from MODIS. However, accurately identifying deforestation events is challenging due to seasonal variability, sensor noise, data gaps, and algorithmic inconsistencies. These factors can obscure true deforestation signals or generate false positives. To address these issues, a robust detection approach must explicitly model time-series dynamics, capturing trends, seasonality, and uncertainty, to reliably distinguish genuine deforestation breakpoints from natural variation and noise. In this paper, three models for the detection of breakpoints in EVI time series were proposed: a simple z-score anomaly detector, and two fully Bayesian models; one temporally uncorrelated and one fully correlated. Results indicate that the Bayesian schemes significantly improve over the naive approach (zscore: AUC=0.921, F1-score=0.870, Bayes: AUC=0.959, F1-score=0.925), for a reasonable cost in computing time ×1000.
Description
Date
2026-03-16
Journal Title
Journal ISSN
Volume Title
Publisher
ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Collections
Research Projects
Organizational Units
Journal Issue
Keywords
DEFORESTATION, BAYESIAN INFERENCE, CHACO FOREST, REAL-TIME
Citation
Grings, F., González Bianco, F., Roitberg, E., Morandeira, N., Arellana, J., and Gayol, M.: Near Real-Time Detection of EVI Time-Series Breakpoints Using Bayesian Inference for Deforestation Monitoring in the Chaco Forest, ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., X-3/W4-2025, 191–196, https://doi.org/10.5194/isprs-annals-X-3-W4-2025-191-2026