Optimizing SAR data processing and thresholding for forest change detection: an application for early deforestation warnings on eastern Amazonia

2020/05/15 by Doblas, Juan
#Applications (stat.AP) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2005.07334

Abstract

The present work proposes a prototype for an operational method for early deforestation detection of cloudy tropical rainforests. The proposed methodology makes use of Sentinel-1 SAR data processed into the Google Earth Engine platform for flag the areas where the probability of recent deforestation is high. The evaluation of the results over a region on the Eastern Amazon basin showed that copolarized data (VV band) offers the best results in terms of producer's accuracy (95,4% for a 5% significance, 88,9% for 1% significance), while crosspolarized data (VH band) offered excellent results in terms of user's accuracy (86% for a 5% significance, 100% for 1% significance).

Related

• Detecting Deforestation from Sentinel-1 Data in the Absence of Reliable Reference Data
• MultiEarth 2022 Deforestation Challenge -- ForestGump
• Combining recurrent and residual learning for deforestation monitoring using multitemporal SAR images
• deFOREST: Fusing Optical and Radar satellite data for Enhanced Sensing of Tree-loss
• Near real-time deforestation detection in Malaysia and Indonesia using change vector analysis with three sensors