Emerging signals of declining forest resilience under climate change

Forest ecosystems, covering roughly 30% of the land surface, are vital for the global carbon cycle and provide essential ecosystem services. Their resilience—the capacity to withstand and recover from disturbances—is crucial for their persistence and the continued provision of these services. However, forests face increasing threats from natural and anthropogenic disturbances, raising concerns about their resilience under climate change. This study aims to assess how forest resilience has evolved globally over the past two decades, particularly in the context of intensifying disturbance regimes and the importance of sustained carbon sinks for climate change mitigation. Understanding these dynamics is crucial for developing effective conservation and management strategies. Previous research has shown that as systems approach a tipping point, they lose resilience, exhibiting critical slowing down (CSD) of system processes. Increased temporal autocorrelation (TAC) in ecosystem state can be used as an indicator of CSD and declining resilience. While spatial patterns of forest resilience have been studied, the temporal evolution at large scales has been challenging due to data limitations. The increasing availability of long-term satellite data now allows for a global-scale assessment of this issue.

Theoretical studies and prior research have established the link between ecosystem resilience, tipping points, and critical slowing down (CSD). Indicators like increased temporal autocorrelation (TAC) in ecosystem state variables signal a decline in resilience and an increased risk of regime shifts. Previous work has examined spatial patterns of forest resilience using these indicators, but large-scale temporal analyses have been limited by data availability and methodological challenges. Studies have highlighted the importance of considering both natural and anthropogenic drivers of forest resilience, such as climate variability, forest management practices and land use change. Existing literature has shown varying responses across different biomes, with some indicating potential benefits from warming and CO2 fertilization in certain regions while others show a clear decline in resilience due to environmental stress.

This study uses a global dataset of kernel normalized difference vegetation index (KNDVI) derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor (2000-2020) at 0.05° spatial resolution. KNDVI serves as a proxy for ecosystem productivity and represents the state of forest ecosystems. The 1-lag TAC of KNDVI is calculated as a CSD indicator related to resilience. A random forest (RF) regression model is developed to identify relationships between long-term TAC (2000–2020) and various forest and climate metrics. To analyze temporal dynamics, TAC is computed using 3-year rolling windows. Factorial simulations of the RF model are used to separate the contribution of environmental factors and filter out confounding signals from climate drivers. The resulting time series of annual TAC and its temporal trend (STAC) are used to detect changes in forest resilience over time. The study compares average TAC over two decades (2000–2010 and 2011–2020) and analyzes the relationship between declining resilience and abrupt declines (ADs) in forest primary productivity (GPP), focusing on intact forests to minimize the influence of land management practices. Extensive sensitivity analyses are conducted to assess the robustness of the findings.

The study reveals a widespread and significant decline in forest resilience in tropical, temperate, and arid regions, characterized by an increase in TAC and a negative STAC. Boreal forests show more varied local patterns, but on average, exhibit an increasing trend in resilience. These trends are consistent across both managed and intact forests, suggesting that large-scale climate drivers are predominantly responsible. The decline in resilience is statistically linked to abrupt declines in forest primary productivity. Roughly 23% of intact undisturbed forests, representing 3.32 Pg C of gross primary productivity, have reached a critical threshold and are experiencing further degradation. While vegetation greening, possibly due to CO2 fertilization, has had a positive effect on resilience, the intensification of water limitations and extreme climate events, especially in tropical, arid, and temperate regions, outweigh this positive effect, resulting in a net loss of resilience. Intact forests show higher resilience (lower TAC) than managed forests, indicating the negative impact of forest management on resilience. However, the temporal trends (STAC) are comparable between managed and intact forests, again highlighting the dominant role of large-scale climate drivers. Abrupt declines (ADs) in forest productivity are strongly associated with declining resilience (high STAC) in boreal forests but not in tropical forests, suggesting different mechanisms driving ADs in these regions. In boreal forests, the ADs may follow a slow drift towards a critical resilience threshold, potentially triggered by changes in environmental drivers and factors like insect outbreaks. In tropical forests, rapid and intense disturbances like fires or droughts may induce ADs independently of long-term trends in resilience.

The findings highlight a concerning global trend of declining forest resilience, primarily driven by climate change impacts such as increasing water limitations and climate variability. While CO2 fertilization and warming have positive effects in certain regions (like boreal forests), these are often outweighed by the negative effects of extreme climatic events. The similar trends observed in both managed and intact forests underscore the pervasive influence of large-scale climate drivers. The strong association between declining resilience and abrupt declines in forest productivity points to the importance of monitoring resilience as an early warning signal for potential ecosystem shifts. The study's use of satellite-based data and advanced statistical techniques offers a robust global perspective on this critical issue. The distinct mechanisms driving abrupt declines in boreal and tropical forests suggest the need for region-specific strategies for forest conservation and management. The results reinforce the urgency for mitigation and adaptation plans that account for the reduced capacity of forests to withstand perturbations.

This study provides compelling evidence for a widespread decline in forest resilience across many of the world's biomes, primarily due to climate change. The observed strong link between declining resilience, abrupt declines in productivity, and increasing temporal autocorrelation highlights the importance of incorporating resilience thinking into forest management practices. Further research should focus on refining region-specific predictions of resilience loss, investigating the interaction between climate change, forest management, and disturbances, and developing strategies for enhancing forest resilience under future climate scenarios. The findings call for proactive adaptation and mitigation strategies to protect and restore forest ecosystems.

The study relies on satellite-derived vegetation indices as proxies for ecosystem states, which may not capture the full complexity of forest dynamics. The analysis focuses primarily on large-scale patterns and may not fully reflect local variations in resilience. While the study accounts for forest management, the impact of specific management practices on resilience could warrant further investigation. Future research could explore alternative methods for identifying and quantifying resilience indicators and incorporate more detailed data on various forest disturbances.