Assessing Forest Recovery Following Selective Logging-related Tropical Forest Disturbances using Sentinel-1 Radar Data

Abstract
Selective logging has emerged as the second leading cause of forest disturbance in West Papua, Indonesia. Understanding forest recovery is essential in order to support land management planning and conservation efforts. Studies investigating large-scale forest recoveries commonly rely on optical sensors, while the potential of radar data remains understudied. Contrary to optical data, the radar signal is capable of penetrating parts of the tree foliage, allowing for an improved assessment of the status of the vegetation structure.

In this study, we investigate the signal recovery of Sentinel-1 (C-band radar) data for selective logging-related forest disturbances in West Papua, Indonesia. Forest disturbances were based on the RADD alerts and further manually classified into four classes (road horizontal, road vertical, large-sized, and small-sized logging events). A method was developed to assess the signal recovery of backscatter for each disturbance type, initially focusing on temporal recovery and incorporating spatial analysis. Furthermore, the recovery process was characterized using three commonly employed recovery metrics for optical data.

We presented a method capable of assessing the signal recovery of Sentinel-1. Road verticals exhibit a longer recovery time than road horizontals, which is attributed to radar shadow effects. Conversely, small logging events demonstrate faster recovery times than large ones. The findings show that the edge of the logged area recovers faster than the interior, whether it is a logging road or logging events. Lastly, the study demonstrates that three spectral recovery metrics, originally applied in optical imagery, can effectively be used in radar data, providing diverse insights into signal recovery.