Mapping Forest Disturbances Across the Southwestern Amazon: Evaluation and Comparison of Optical Remote Sensing Algorithms

Abstract:

People living in the Peru-Brazil borderlands rely on the provisioning and regulating services provided by the tropical rainforest. Understanding the land-use and land-cover dynamics in this region is essential for analyzing the effects of vegetation changes on ecosystem services, but the remoteness and frequent cloudiness of the region have hindered an accurate and continuous land-use and land-cover monitoring. This study evaluates three forest disturbance algorithms—LandTrendr (Landsat-based detection of Trends in Disturbance and Recovery), CODED (Continuous Degradation Detection), and MLb (Machine Learning-based)—and seeks to determine which of them performs best to map historical deforestation and degradation across the Peru-Brazil borderlands. The algorithms are implemented using a variety of spectral bands and indices, such as Normalized Difference Vegetation Index (NDVI), Normalized Burn Ratio (NBR), Normalized Difference Water Index (NDWI), Soil-Adjusted Vegetation Index (SAVI), and Normalized Difference Fraction Index (NDFI). Preliminary results identify LandTrendr as a good algorithm when NDFI or NBR are used as input data. Final results will then demonstrate not only which algorithm but also what spectral indices or bands are best in mapping historical deforestation and degradation in the Peru-Brazil borderlands. This research confirms that the systematic and periodic collection of forest reference data becomes increasingly important to make possible the implementation of forest disturbance algorithms, which are key tools for generating historical maps needed to make informed decisions for sustainable development practices.