Monitoring Forest Degradation in Temperate Regions Using Google Earth Engine

Abstract:

Traditionally, monitoring forest degradation has been challenging and the area estimates of forest degradation have been associated with large uncertainties. Recent advancements in availability of remote sensing data and cloud-based platforms, e.g. Google Earth Engine, provide new opportunities for monitoring forest degradation. Although there are several studies in forest degradation in the tropics recently, studies on temperate forest degradation are rare. This study presents an approach on monitoring abrupt and gradual forest degradation and area estimation in the country of Georgia, which is a temperate region with complex forest types and climate conditions. The method combines spectral mixture analysis and continuous change detection and classification. In this study, we found that fraction of different endmembers performed differently, and using different metrics for different types of forests (coniferous forest, deciduous forest, and mixed forest) would improve the accuracy. Based on our tests on different metrics, we developed the optimal metric for monitoring forest degradation. We mapped deforestation, forest degradation, stable forest and non-forest from 1987 to 2019 in Georgia, and conducted rigorous accuracy assessment. The overall accuracy of the map is 0.901. The user’s and producer’s accuracy of forest degradation class are 0.675 and 0.793, respectively. Based on preliminary unbiased area estimates within 95% of confidence interval, forest degradation accounts for 3604.98 ± 594.94 km², which is much larger than the area estimates of deforestation 291.00 ±180.96 km² . This study presents an effective approach of monitoring forest degradation in temperate forests and reducing the uncertainty of area estimates of forest degradation.