Temporal Observation Fusion for Deforestation Detection in Amazonian Legal Reserves: Insights from Reserve- and Pixel-Level Evaluation
Authors:
- Marcelo Vitor Gualberto Santo ChavesMCMarcelo Vitor Gualberto Santo Chaves
- Marcelo Lourençoni PaulettiMPMarcelo Lourençoni Pauletti
- Mateus Gabriel de SouzaMSMateus Gabriel de Souza
- Francisco Kimerling CamposFCFrancisco Kimerling Campos
- Kalill José Viana da PáscoaKPKalill José Viana da Páscoa
- Thiza Falqueto AltoéTAThiza Falqueto Altoé
- Lucas Resende GomideLGLucas Resende Gomide
- Samuel José SilvaSSSamuel José Silva
- Luis Marcelo Tavares de CarvalhoLCLuis Marcelo Tavares de Carvalho
Preprint
- 17 October 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Monitoring deforestation within Legal Reserves (LRs) requires methods capable of integrating spectral, spatial, and temporal information over large areas. This study developed and evaluated a semantic segmentation model based on a U-Net++ architecture with an EfficientNet-B0 backbone to identify deforestation in LRs declared in the Brazilian Rural Environmental Registry (CAR) in the state of Pará. Temporal differences in NDVI, NBR, and NDRE derived from Harmonized Sentinel-2 MSI imagery were stacked as three input channels. The model-development dataset comprised 6,949 LRs, divided into 70% for training and 30% for validation. The best model was selected according to the lowest validation loss, and the final decision threshold was calibrated using the Matthews Correlation Coefficient (MCC) at the LR level. During validation, the model achieved an IoU of 0.6820 and a Dice coefficient of 0.8110 at the pixel level, as well as an F1-score of 0.9743 and an MCC of 0.9492 at the LR level. The independent test included 2,500 LRs balanced between deforested and non-deforested classes, each represented by two distinct temporal combinations (CT1 and CT2), in addition to a Maximum Fusion strategy based on pixel-wise probabilities. At the LR level, Maximum Fusion achieved the best overall performance, with an accuracy of 0.9216, F1-score of 0.9183, MCC of 0.8460, and sensitivity of 0.8808. At the pixel level, however, CT1 achieved the highest IoU (0.2294), indicating lower geometric agreement with PRODES polygons during independent testing. Overall, the model was more consistent in identifying the occurrence of deforestation within LRs than in reproducing its exact spatial geometry. The results also highlight the influence of temporal image selection and the potential of integrating multiple observations to support environmental monitoring.Keywords
This publication has 43 references indexed in Scilit:
- DASAM-CD: Dual-attention adaptive SAM for change detection on remote sensing imagesJournal of King Saud University Computer and Information Sciences, 2026
- Slow Progress Under Brazil’s Native Vegetation Protection Law in the Southeastern AmazonEnvironmental Management, 2026
- Explosive growth of secondary roads is linked to widespread tropical deforestationCurrent Biology, 2025
- Leveraging remote sensing for transparency and accountability in Amazonian commodity supply chainsOne Earth, 2024
- Forest disturbance characterization in the era of earth observation big data: A mapping reviewInternational Journal of Applied Earth Observation and Geoinformation, 2024
- Trends and applications of google earth engine in remote sensing and earth science research: a bibliometric analysis using scopus databaseEarth Science Informatics, 2023
- An attention-based U-Net for detecting deforestation within satellite sensor imageryInternational Journal of Applied Earth Observation and Geoinformation, 2022
- A review of deep learning methods for semantic segmentation of remote sensing imageryExpert Systems with Applications, 2021
- Deforestation and Carbon Loss in Southwest Amazonia: Impact of Brazil’s Revised Forest CodeEnvironmental Management, 2017
- Mapping properties to monitor forests: Landholder response to a large environmental registration program in the Brazilian AmazonLand Use Policy, 2016