Early Warning Signals for Tree Mortality: A Review of Indicators, Limitations, and Emerging Opportunities

Tree mortality is increasing worldwide, intensifying the need for forest monitoring systems that can detect vulnerability before irreversible damage occurs. Predicting mortality remains difficult because tree death results from interacting predisposing conditions, inciting disturbances, and contributing agents that unfold over time. These drivers interact with recovery processes in highly non-linear ways, creating cascading stress trajectories in which trees may cross tipping points beyond which recovery becomes impossible.

This review synthesizes the literature on resilience indicators and early warning signals for assessing tree mortality risk. Existing approaches span a broad range, from threshold-based indicators and model-based risk predictions to disturbance-focused recovery metrics and indicators derived from changes in time-series dynamics. Despite their conceptual diversity, most approaches share two key limitations. First, many overlook the tree-level physiological stress history, such that similar drought events may lead to minimal or catastrophic damage depending on prior stress exposure. Second, many indicators are often too late for operational resilience monitoring, relying on retrospective data or signals that typically emerge only after substantial structural damage has already occurred.

Physiological theory and empirical evidence indicate that stress responses relevant to mortality risk arise at the level of hormonal regulation and stomatal control, affecting photosynthesis, transpiration, and leaf reflectance well before structural damage occurs. Recent advances in sensor technology now enable high-frequency observation of these processes through a growing range of in-situ and remote measurements. Yet few frameworks exist to interpret such time series as early warning signals. While resilience theory offers promising concepts for understanding critical transitions, it has rarely been applied to real-time forest monitoring. We highlight this gap and emphasize the role of controlled experiments in validating which physiological signals reliably precede mortality, enabling the translation of high-frequency measurements into actionable early warning indicators.