Country-wide mapping of harvest areas and post-harvest forest recovery using Landsat time series data in Japan

Abstract

The characterization of forest harvesting and subsequent vegetation recovery provides valuable insights for effective forest management. Although Landsat time series data offers spatially explicit information regarding large-area forest disturbance and recovery, the detailed characterization of country-wide harvest and post-harvest recovery is insufficient. Despite the importance of planting in harvest areas, the mapping of areas replanted by forest management after harvest are not usually considered. This study investigated an approach to detect harvest and other forest disturbance areas country-wide and to characterize post-harvest recovery using annual Landsat time series data from 1984 to 2020. To do so, a random forest algorithm was used to classify disturbance agents and stable land cover classes using predictor variables derived from the LandTrendr temporal segmentation of five spectral indices and topographic and climate data. Post-harvest recovery was characterized as forest species composition (i.e., coniferous/broadleaved forests) and then used to link replanted areas in harvest areas. The disturbance agents/stable land cover classification achieved producer’s and user’s accuracies of 80.1% (±4.8%) and 93.8% (±3.8%), respectively, for the forest harvest class. The overall accuracy of post-harvest recovery was high (83.9%) and a comparison with statistical data of replanted areas for entire Japan showed good agreement in trends and estimated area with a root mean square error of 2687.8 ha (15.2%). Overall, harvested forest area accounted for 4.6% of the total land area in Japan during the past 35 years. The results indicated that approximately 60.0% of coniferous plantation forests were recovered as coniferous forests. The spatial and temporal distribution of coniferous forests after harvest presumably represented the replanting activity of forest management in harvest areas. The approach presented in this study has the potential to provide valuable information on country-wide replantation activity in harvest areas using Landsat time series data.