Global field observations of tree die-off reveal hotter-drought fingerprint for Earth's forests

Forests are crucial for global ecosystem services and human economies. However, they face increasing threats from climate change, particularly hotter droughts, leading to widespread forest die-off events. The uncertainty surrounding the fate of forests in a warming world necessitates a quantitative understanding of the common climate anomalies associated with these mortality events. Previous studies have lacked a global, quantitative assessment of climate conditions triggering forest die-off, relying on limited data sources such as experimental manipulations, single-site or regional-scale observations, or process-based projections that often lack realistic representation of tree mortality processes. This study aims to address this gap by creating a comprehensive database of field-observed tree mortality events, quantifying a global "hotter-drought fingerprint," and projecting the frequency of these conditions under future warming scenarios.

The authors reviewed 230 peer-reviewed studies on drought and heat-induced tree mortality, selecting 154 that met specific criteria: on-the-ground observations of mortality pulses attributable to climate, and precise geolocation. This literature review revealed a strong association between hotter droughts and tree mortality events, though previous studies lacked the global scope and quantitative analysis of the current study. The review highlights the complexities of tree responses to drought and heat, including the triple threat of amplified atmospheric drought, intensified soil drought, and direct heat stress, and the deadly trade-off between using water for evaporative cooling versus survival during drought. The increased frequency, severity, and intensity of chronic soil droughts due to anthropogenic warming is also reviewed. Existing studies, while informative, suffer from limitations in geographic scope, experimental methodology, and the representation of tree mortality processes in models.

The study developed a global database of precisely geo-referenced observations of climate-induced tree mortality from 154 peer-reviewed publications, covering all forested continents and encompassing 1303 plots between 1970 and 2018. Data were obtained from publications and author data requests. Georeferencing was achieved using precise coordinates or site descriptions. Six climate metrics (monthly average maximum temperature, vapor pressure deficit, climatic water deficit, soil moisture, monthly total precipitation, and the Palmer Drought Severity Index) from TerraClimate were analyzed for each location. Monthly anomalies (z-scores) relative to the period of record (1958–2019) were calculated for the 4 years preceding, the year of, and the 4 years following mortality onset. A "hotter-drought fingerprint" was defined as the concurrent occurrence of significantly warmer and drier than average conditions across all six metrics during the typically warmest/driest months. Projected climate for two warming scenarios (+2°C and +4°C warming relative to pre-industrial levels) were superimposed onto observed climate (1985–2015) to assess the future frequency of these conditions. Whittaker biomes were used to classify the plots based on mean annual temperature and precipitation. Statistical analysis included linear regressions to examine trends and assessing the frequency of mortality-year conditions under various warming scenarios.

The study found a globally consistent "hotter-drought fingerprint" for tree mortality events, which showed that climate conditions in the mortality year were significantly warmer and drier than the long-term average for all six climate variables. This fingerprint was evident across various biomes, except for tropical rainforests where data were more limited. Specifically, during the mortality year, the warmest/driest months showed significantly higher z-scores for TMAX, VPD, and CWD, and lower z-scores for PPT, SOIL M, and PDSI. The year preceding and following mortality also showed a similar, but weaker, signal. The frequency of mortality-triggering climate conditions increases nonlinearly with warming; a 22% increase under +2°C and a 140% increase under +4°C warming compared to observed climate (1985-2015). Analyses also showed that TMAX anomalies during mortality years have been increasing faster than background warming. The study observed that many conditions during mortality years became warmer and drier (Supplementary Fig. S7), with TMAX, VPD, and CWD increasing while SOIL M and PDSI decreased.

The study's findings support the hypothesis that hotter droughts are a major driver of global tree mortality, even across diverse biomes. The consistent "hotter-drought fingerprint" highlights the shared climatic triggers of mortality across different forest types. The observed trend toward hotter and drier conditions during mortality years, along with the projected increase in frequency under future warming, emphasizes the vulnerability of forests to climate change. The limitations in data for tropical rainforests highlight the need for future studies in these regions. While the study focuses on acute climate extremes, chronic drought and other factors could also contribute to mortality. The observed three-year window (one year before, during, and one year after mortality) suggests legacy effects of drought and the imprecise nature of identifying the precise onset and duration of mortality. The potential for underestimation of future mortality risks due to multiple interacting factors (insects, pathogens, wildfire, etc.) is acknowledged.

This study quantifies a global hotter-drought fingerprint associated with tree mortality events, demonstrating the accelerating risk of climate change to Earth's forests, particularly historical old-growth forests. The frequency of lethal climate conditions is projected to increase nonlinearly with warming. The database created provides a valuable resource for future research to improve models, validate remote-sensing technologies, and guide monitoring efforts. Future work should focus on addressing the limitations of the current database, integrating additional climate signals, and improving the representation of tree mortality in Earth system models.

The database is inherently biased due to the reliance on available peer-reviewed studies, resulting in under-representation of certain regions, particularly tropical and boreal forests. The focus on acute climate extremes may neglect the importance of chronic drought and other interacting factors. The imprecise identification of mortality onset could influence the temporal analysis. The study acknowledges the potential underestimation of future mortality risk due to the exclusion of other disturbance processes.