Trait‐Based Tree Mortality Risk Assessment From the Perspective of Imaging Spectroscopy

Droughts can lead to large-scale tree mortality and thus undermine the ability of ecosystems to sequester carbon and heighten risks for wildfire and flood. Despite its importance, drought-induced tree mortality remains difficult to predict due to complex interactions among physiological and ecological processes that govern plant responses to drought. These interactions vary widely across space and time due to high variation in drought exposure and traits that mediate mortality-inducing stressors during drought, such as cavitation or/and hydraulic failure (plant hydraulic traits), carbon starvation (photosynthetic and allocation traits), and secondary biotic attacks (defense traits) (Anderegg et al. 2015). Developing trait-based frameworks to predict tree mortality risk has been recommended to disentangle these interactions and inform better forest management before actual wilting or mortality (Xu and Trugman 2021). However, acquiring large-scale plant trait information is difficult, leaving the performance of trait-based mortality predictions beyond forest stands largely untested. Remote sensing has enabled spatially continuous mapping of plant traits. Previous studies that explore trait-based tree mortality have focused on canopy structure (Stovall et al. 2019) and hydraulic traits (Liu et al. 2021). Although it has been recognized that leaf functional traits provide insights into plant survival/growth strategies, which influence plant response to drought, few studies have mapped traits in the leaf economic spectrum (Díaz et al. 2016) and fully explored their potential as a precursor to tree mortality. Lower leaf mass per area (LMA) has been found to be associated with lower drought tolerance and higher mortality at the plot scale (Greenwood et al. 2017), but landscape-scale studies are still missing and there have been debates about whether LMA is consistently associated with drought tolerance (McGregor et al. 2021). In their newly published research, Queally et al. (2025) present a pioneering study that links pre-mortality plant functional traits with pixelwise percent mortality. Using random forest models, Queally et al. (2025) analyzed the main physiological and abiotic drivers for drought mortality. This study was conducted in the context of the California megadrought (2012–2016) at a 30-m resolution across two California sites in the southern Sierra Nevada, which represent contrasting biotic and abiotic conditions—the Soaproot Saddle site is water-limited at lower elevation with more resource-acquisitive plants, while the Lower Teakettle site is located at the transition zone of water-limited to energy-limited forest at higher elevation with more resource-conservative plants. One of the biggest highlights of Queally et al. (2025) is that they mapped nine foliar traits—including nitrogen, LMA, lignin, cellulose, sugars, starch, non-structural carbohydrates (NSCs), chlorophyll, and phenolic—using an airborne imaging spectrometer at the early stage of drought (2013–2014). They explored whether these leaf traits can serve as key predictors of actual mortality (2016/2017) and found distinct trait-mortality linkages across sites. At the site or community scale, mean canopy height, LMA, and soil moisture were consistently ranked as top predictors of mortality across years and sites. Although higher LMA is consistently linked with higher mortality within both sites, the relative importance of LMA was higher at the Soaproot site, representing a stronger link between foliage traits and mortality. Soaproot is dominated by broadleaf evergreen species like canyon and coast live oak, which are found at lower altitudes and latitudes and therefore likely better adapted to drought. In addition, since Soaproot is a water-limited site, previous drought may have contributed to an existing ecosystem/plant community with enhanced drought tolerance and stronger trait-mortality relationships. The Teakettle site is dominated by white and red fir and Lodgepole pine—species that tend to be found at higher altitudes and latitudes. The weaker signals in LMA or other leaf traits linked to mortality at the Lower Teakettle site could, therefore, be because the site experiences less water stress during normal years and thus plant species found there are less adapted to drought. These results demonstrate the importance of both species composition and community assembly in predicting drought mortality patterns. In addition, changes in canopy water content and sugars showed high variable importance for Soaproot but not for Teakettle, implying a higher contribution of plant water stress to mortality at the Soaproot site. Interestingly, at the species level, the positive LMA-mortality relationship was also observed across most groups, with foliar nitrogen negatively associated with mortality for needleleaf species in both sites. This finding is somewhat counterintuitive since lower LMA and/or higher leaf nitrogen and chlorophyll are typically associated with greater carbon acquisition, and thus greater transpiration and weaker drought-resistant strategies, according to the plant economic spectrum (Díaz et al. 2016). Queally et al. (2025) explained that trees with lower LMA and higher nitrogen and chlorophyll content were potentially more likely to survive because of a photosynthetic advantage under water-limited conditions during early drought stress. In addition to lower LMA, higher leaf cellulose—an indicator of tougher leaves—was also found to be associated with lower mortality for broadleaf species. In contrast, higher leaf sugar was linked to greater percentage mortality in both broadleaf and needleleaf species at the Soaproot site, likely reflecting the physiological process of leaf sugar accumulation due to starch breakdown under drought stress. Yet, leaf traits played less important roles as mortality predictors for needleleaf species at the Teakettle site. These results imply that while these canopy foliage traits are more readily mapped via spectroscopy, their mechanistic link to drought-induced mortality remains uncertain. Exploring the linkage between leaf functional traits and hydraulic traits, which are directly related to tree mortality (Anderegg et al. 2015), or comparing the trait composition before/after the drought event (Trugman et al. 2020) may provide further insight on mechanisms of tree-level drought response. In addition to leaf traits, Queally et al. (2025) also included canopy structure—mean canopy height, height SD, and tree cover in their assessment of pre-drought predictors of forest mortality. In this study, mean canopy height was found to be one of the top predictors of tree mortality, with taller trees generally linked with higher mortality. This is consistent with the findings of previous studies where larger trees suffered most during drought primarily because of higher water stress, due to both greater hydraulic constraints and higher evaporative demand, with the coupling between drought stress and insect attacks further strengthening the correlation between size and mortality (Bennett et al. 2015). Additionally, greater tree cover was linked to lower mortality at the Soaproot site, since greater leaf area may improve microclimate buffering and thus reduce soil evaporation. Canopy height SD also contributed to mortality prediction, though its effects varied across sites and species since it influences both local abiotic stress and susceptibility to insect attacks. The linkage between plant traits and mortality found in this study provides insights into process-based mortality modeling, suggesting the potential for improving the simulation of forest dynamics by incorporating trait maps and canopy structure into vegetation dynamic models. Classic ecosystem models assign constants to plant traits based on coarse plant functional types. However, plant traits can vary widely across space and time, even within the same species. Thus, representing key trait plasticity in response to stress (Xu and Trugman 2021) can provide a more realistic framework for modeling forest response to droughts than the traditional parameterization scheme. The study by Queally et al. (2025) also reveals two broad insights for studying tree mortality. First, proper image pre-processing is essential for accurately mapping plant information. This study identified high sensor zenith angles in imagery as a significant source of discrepancies between the two mortality datasets. Second, one of the key challenges in trait-based tree mortality prediction is the temporal lag between drought events, plant physiological responses, and the eventual onset of mortality. This study linked 2014 trait maps with 2016/2017 tree mortality datasets since 2014 was reported as the most severe year of the prolonged mega-drought. However, different leaf traits respond to drought on varying temporal scales. For example, NSCs may change relatively quickly since stress triggers starch decomposition. In contrast, traits such as LMA typically reflect more chronic, longer-term responses to environmental conditions, suggesting that extended time series of traits may be necessary to better understand trait-mortality relationships. In summary, Queally et al. (2025) provide a prototype for exploring the early-stage detection of plant mortality risk at the site- and species-scale by utilizing open-access airborne imaging spectroscopy and lidar along with abiotic factors. Connections between foliar traits and mortality risk could vary depending on local abiotic conditions, species, and canopy structure. Therefore, transferability and generalization of relationships and approaches proposed in this study to regional and global scales across climatic regimes require further investigations, especially along aridity gradients, which can benefit from compiling longer time series of collocated plant economic spectrum traits, hydraulic traits, canopy structure, and mortality data. Ongoing collection of airborne imagery like AVIRIS and the NEON Airborne Observation Platform (Kampe et al. 2010), along with the future launch of the Surface Biology and Geology hyperspectral mission (SBG; Cawse-Nicholson et al. 2021), microwave remote sensing instruments such as the NASA-ISRO Synthetic Aperture Radar mission (NISAR; Kellogg et al. 2020), and others, will facilitate such long-term studies and opportunities to expand the work by Queally et al. (2025). Meicheng Shen: conceptualization, writing – original draft, writing – review and editing. Kyla Dahlin: conceptualization, writing – review and editing. Xiangtao Xu: conceptualization, writing – review and editing. Zachary Butterfield: writing – review and editing. This work was supported by the US National Science Foundation (award #2044818). The authors declare no conflicts of interest. This article is a Invited Commentary on Natalie Queally et al., https://doi.org/10.1111/gcb.70246. The authors have nothing to report.