Insights into forest vegetation changes and landscape fragmentation in Southeastern China: From a perspective of spatial coupling and machine learning

Abstract

• Develop a comprehensive index for assessing forest fragmentation. • Theil-Sen and Mann-Kendall has been applied to examine the forest vegetation change dynamics during 2000–2020. • Identify 15 spatial coupling modes of forest vegetation change and fragmentation dynamics. • There are forests experienced recovery, but severe landscape fragmentation. • RF models reveals distinct driver patterns between forest vegetation change and fragmentation dynamics. A comprehensive quantification of the relationships between forest vegetation changes and forest landscape fragmentation is urgently required to provide valuable insights for informed forest management decisions. Previous efforts often focused on these two aspects separately, overlooking their intricate relationships, restricts the formulation of precise forest protection and recovery measures. To address this gap, taking Southeast China as a case, this study employed the Google Earth Engine (GEE) platform to calculate the annual Normalized Difference Vegetation Index (NDVI) from 2000 to 2020, a forest fragmentation comprehensive index (FFCI) was constructed to evaluate the static and dynamic forest landscape fragmentation over the study period, then the Theil-Sen and Mann-Kendall, the two-dimensional framework and the bivariate spatial autocorrelation, and a machine learning algorithm (random forest, RF) were employed to explore the forest vegetation change and forest landscape fragmentation, and their spatial coupling relationships and driving patterns, respectively. The results showed that the temporal trend of the forest vegetation can be categorized into three distinct phases, with the pattern of “decreasing-rising-decreasing”, 78.2% of the area improved in forest vegetation, while 11.0% degraded. Approximately 6.0% of forest landscape showed a decline in fragmentation, while 9.3% experienced increased fragmentation trends, including 3.8% of forest landscape with moderate to high static fragmentation and 5.5% of forest landscape with low to very low static fragmentation. Throughout the study period, approximately 66.8% of forest landscape remained stability with the improved forest vegetation; however, there were cases (more than 7% of forests) with improved forest vegetation, but experienced severe landscape fragmentation. The RF outcome revealed that the forest vegetation change dynamics were influenced by the multiple factors, including distance to forest edge, distance to road, and elevation; while distance to forest edge emerged as the overwhelming dominant factor driving the forest landscape fragmentation. This research offers significant insights into the intricate interconnections between the forest vegetation changes and the forest landscape fragmentation dynamics.