Long-term tree mortality prediction following stem breakage in loblolly pine (Pinus taeda L.) plantations: implications for management in the southeastern United States

Introduction: Loblolly pine (Pinus taeda L.) plantations in the southeastern United States are increasingly vulnerable to climate-driven disturbances that cause stem breakage and elevate tree mortality. Consequently, it is imperative to develop adaptive management strategies following stem breakage for forest managers and landowners to mitigate tree mortality and sustain forest productivity. However, predictive frameworks for post-breakage mortality remain underdeveloped, hampering adaptive management efforts.

Methods: We conducted a split-plot experiment in four mid-rotation plantations, applying two whole-plot silvicultural treatments (no thinning vs. thinning) and five subplot treatments defined by the proportion of trees damaged (PTD: 0, 25, 50, 75, 100%). Stem breakage was simulated by rifle shooting of main stems, and individual trees were monitored for 12 years. We developed a multiple logistic regression model including thinning, plot-level damage (PTD), tree-level damage severity (PSD), diameter at breast height (DBH), total height (HT), and crown height (CRNHT), and used parametric bootstrapping (10,000 iterations) to quantify uncertainty.

Results: Our findings highlight three key patterns: (1) Short-term mortality (<4 years) remained low but increased sharply over the long term following stem breakage; (2) Larger DBH and greater tree height reduced mortality risk; and (3) Plot-level damage proportion had a stronger influence on mortality than individual-tree damage severity.

Discussion: These results highlight the dominant role of plot-level damage regimes and tree size in long-term survival after stem breakage. Incorporating these insights into adaptive thinning and disturbance management frameworks will help sustain productivity and resilience in loblolly pine plantations under escalating climate extremes.

1 Introduction

Demand for pine (Pinus spp.) plantations in the southeastern United States has increased steadily to supply pulpwood and sawtimber markets (Zhao et al., 2023). Loblolly pine (Pinus taeda L.) is a significant contributor to the timber industry, covering approximately 14.4 million hectares across the region (South and Harper, 2016). However, the frequency of severe weather events, including wildfires and hurricanes, has been continuously increasing due to climate change, threatening the timber industry in this region by injuring either the entire or partial crown of a tree (Sun, 2016). Given that tree crowns and stems carry out essential functions, such as photosynthesis and transpiration, these functions, as well as the structural stability of the tree, are hindered after crown injury (Varner et al., 2021; Needham et al., 2022). This reduced stability and photosynthetic rates affect not only growth and reproductive success but also result in tree mortality and consequently reduced stand productivity (Tanner et al., 2014; Pope et al., 2023). Therefore, the development of adaptive management strategies to mitigate the impacts of stem breakage is essential for enhancing resilience and sustaining long-term timber yields in southeastern loblolly pine plantations.

Thinning reduces inter- and intra-specific competition, thereby enhancing growth conditions and promoting tree development (Hwang et al., 2018; Hwang et al., 2020). Numerous studies have demonstrated that these improved conditions enhance forest productivity by stimulating growth and lowering mortality after severe disturbances, such as wildfires and drought (Ritchie et al., 2007; Sohn et al., 2016), ultimately accelerating stand resilience. For instance, Thomas et al. (2024) found that thinning alleviates plant hydraulic stress and enhances carbon and water fluxes in residual trees, resulting in higher stand productivity and reduced drought-induced mortality. Moreover, the positive effects on growth performance intensify with greater thinning intensity (Sohn et al., 2016). While numerous studies have delved into tree mortality following natural disturbances under varying thinning intensities, the extent to which thinning mitigates tree mortality across different levels of stem breakage remains unclear. As climate change amplifies the frequency of extreme weather events, understanding how thinning efficacy interacts with stem breakage severity is crucial for developing adaptive management strategies that reconcile commercial objectives with conservation goals under changing climatic conditions.

Given tree mortality has binary responses (alive and dead), it is predicted using logistic regression, which estimates the probability of mortality using the logit link function as follows:

$\mathit{\text{Prob}}\left(\mathit{\text{Tree dead}}\right)=\frac{1}{1+e^{-({\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+{\beta }_3{x}_3\dots +{\beta }_n{x}_n)}}$

where x₁ through x_n are explanatory variables and β₁ through β_n are coefficients estimated from observed data using maximum likelihood (Shearman et al., 2019). Previous studies have identified key drivers of post-disturbance mortality across diverse events, including fires (Hood et al., 2018; Cansler et al., 2020), ice storms (Ryall and Smith, 2005; Deschênes et al., 2019), hurricanes and other severe wind events (Johnsen et al., 2009; Crosby, 2020; Fortuin et al., 2022), and drought (Wang et al., 2012; Meir et al., 2015), with pre-disturbance tree characteristics and health status emerging as consistent predictors (Shortle et al., 2003; Woolley et al., 2011), with pre-disturbance tree characteristics and health status emerging as consistent predictors (Shortle et al., 2003; Woolley et al., 2011). Beyond classic logistic regression, time-explicit survival models (Maringer et al., 2021) and Bayesian logistic frameworks (Berdanier and Clark, 2016) have been advanced to capture delayed mortality and uncertainty. However, their application to mid-rotation stem breakage in loblolly pine is limited by the infrequent, tree-level assessment of breakage severity and the demanding data requirements of longitudinal or finely parameterized approaches (Metcalf et al., 2009). Bootstrapping circumvents these constraints and offers a practical pathway to robustly predict mortality in stands where stem damage occurred years or even decades prior (Roberts and Martin, 2010).

To address the knowledge gap above, we compiled forest survey data from experimental plots in loblolly pine plantations. By analyzing this data, we aimed to answer the following specific questions: (1) How do tree mortality rates following the stem breakage differ between short- and long-term observations? (2) Which factors significantly affect tree mortality following the stem breakage? (3) How does the effectiveness of thinning on tree mortality vary with the severity of stem breakage? We addressed these questions using classical logistic regression with a bootstrapping approach to examine the uncertainty of our predictions.

2 Materials and methods

2.1 Site description and experimental design

In March 2008, four mid-rotation loblolly pine plantations (14–16 years), were selected in McCurtain County, southeastern Oklahoma (Figure 1). Our study is a long-term follow-up on the experiment established in Dipesh et al. (2015), which provides more detailed location information, site characteristics, and soil assessments. These plantations, owned by Weyerhaeuser Co. (Federal Way, WA) and managed by the Kiamichi Tree Farm (Broken Bow, OK), share similar soil conditions owing to their proximity. Given that one of the stands was replaced in 2009 due to accidental damage by a logging crew, all subsequent measurements for that stand were shifted by 1 year for the first three assessment periods. The long-term climate averages of the study area (2008–2020) include an average daily minimum temperature of −1.6 °C in January and an average daily maximum temperature of 33.3 °C in August. The area received approximately 152.4 cm of annual precipitation, with an average wind speed of 4.8 km/h. During the study period, the average total solar radiation recorded was 15.7 MJ (Oklahoma Climatological Survey, 2023).

Figure 1

Figure 1. Map of the study area in southeastern Oklahoma. The points indicate the locations of the three experimental sites (Eagletown, Hotchtown, and Union Valley) in McCurtain County. More detailed information is available in Appendix 1.

A split-plot design was implemented to evaluate the effects of thinning and stem breakage severity on tree mortality in 2008. Two replicate stands each were assigned to either an unthinned control (CTRL) or a thinned treatment (THIN). Prior to thinning, all stands were established at 1,074 trees per hectare, and thinning operations were completed less than a year before the study, resulting in a post-thinning density of 321–348 trees per hectare. Chemical release and fertilization were applied uniformly across all stands throughout the study. Each whole plot was subdivided into five split-plots (221 to 537 m²), each containing 28–68 trees. Within each split-plot, five proportions of damaged trees (PTD: 0, 25, 50, 75, and 100%) were randomly assigned, and stem breakage was simulated by shooting the main stem of selected trees with a rifle.

2.2 Measurement and calculation

Prior to stem breakage, every tree was tagged and measured for diameter at breast height (DBH), tree height (HT), and crown height (CRNHT) (the height from the ground to the base of the live crown). DBH was measured to the nearest 0.1 cm using diameter tapes. HT and CRNHT were measured to the nearest 0.1 m using a Haglöf Vertex IV Hypsometer with Transponder T3 (Haglöf, Längsele, Sweden). Following each simulated stem breakage, the diameter at the breakpoint and length of the broken top were measured for damaged trees. The proportion of stem damage (PSD) was calculated for each tree as: PSD = [length of a broken section/(HT − CRNHT)], ranging from 4 to 52%.

Visual assessments of post-stem breakage mortality were conducted at five intervals: 2008 (less than a year following the damage), 2010 (2 years following the damage), 2012 (4 years following the damage), 2015 (7 years following the damage), and 2020 (12 years following the damage). For the replacement stand, the first three assessments occurred 1 year later (2009, 2011, 2013), after which all stands were surveyed concurrently.

2.3 Statistical analyses

All analyses were performed in R 4.4.0 (R Core Team, 2024). The dataset comprised 896 trees, split randomly into training (80%) and validation (20%) subsets for model validation. Parametric bootstrapping (B = 10,000) was used to assess model performance under a binomial distribution assumption for unknown parameters. For each bootstrap sample, a multiple logistic regression model predicting the probability of post-breakage tree mortality was fitted via maximum likelihood estimation (MLE) via the glm function in R (Zuur et al., 2009).

Six predictors were included for model development: presence of thinning (THIN), damage severity at the plot-level (PTD) and individual tree level (PSD), and tree characteristics (DBH, HT, and CRNHT). Interaction terms between THIN × PSD and THIN × PTD were tested, but the preliminary analysis revealed no significant effects for these interaction terms, so they were excluded from the model. The final model form is

$\mathit{\text{Prob}}\left(\mathit{\text{Tree dead}}\right)=\frac{1}{1+e^{-\left(\begin{array}{l}{\beta }_0+{\beta }_1\cdot \mathit{\text{THIN}}+{\beta }_2\cdot \mathit{PTD}+{\beta }_3\cdot \mathit{PSD}\\ +{\beta }_4\cdot \mathit{\text{CRNHT}}+{\beta }_5\cdot \mathit{HT}+{\beta }_6\cdot \mathit{DBH}\end{array}\right)}}$

where THIN is 1 when thinning was applied to the plot, 0 otherwise; PTD represents the proportion of damaged trees in the plot and was treated as a categorical variable with five levels (0, 25, 50, 75, and 100%). The 0% damage level served as the reference category in the model; PSD is the proportion of stem damage; CRNHT is the crown height; HT is the total tree height; and DBH is the diameter at breast height.

The significance of each variable was determined using bootstrap confidence intervals (CI). Bootstrap estimates of variance, bias, and root mean square error (RMSE) for each parameter were calculated following (Hwang et al., 2023):

${\widehat{\mathit{Var}}}_B\left(\theta \right)=\frac{1}{B-1}\sum _{i=1}^B{({\theta }^{\ast \left(i\right)}-{\overline{\theta }}_B^{\ast })}^2$

${\widehat{\mathit{\text{Bias}}}}_B\left(\theta \right)=\frac{1}{B}\sum _{i=1}^B{\theta }^{\ast \left(i\right)}-\theta$

$\widehat{\mathit{\text{RMSE}}}\left(\theta \right)=\sqrt{\frac{1}{B}\sum _{i=1}^B{({\theta }^{\ast \left(i\right)}-\theta )}^2}$

where ${\widehat{\mathit{Var}}}_B\left(\theta \right)$ is the sample variance, ${\theta }^{\ast \left(1\right)},{\theta }^{\ast \left(2\right)},\dots {\theta }^{\ast \left(B\right)}$ , are bootstrap estimates of the sample parameter $\theta$ , and ${\overline{\theta }}_B^{\ast }$ is the mean of the bootstrap samples. A mortality threshold of 0.50 was set to diagnose tree mortality for the model, and all hypothesis tests were performed with a significance level (α) of 0.05. Both mean and bootstrap regression lines were visualized to predict tree mortality in relation to significant variables. Since it is impractical to display 10,000 lines from bootstrapping, we randomly selected 100 lines to illustrate the mortality rates of loblolly pines.

3 Results

3.1 Cumulative tree mortality following stem breakage

Of the 896 loblolly pine trees monitored, 147 trees (16%) died over the 12-year post-damage period. In the first year after simulated stem breakage (2009), overall mortality was minimal, and no deaths occurred at the Union Valley stands (THIN) (Figure 2). Cumulative mortality then increased steadily, with a marked surge observed in the fourth-year post-breakage (2012) at all sites except Eagletown. By 2020, cumulative mortality rates, regardless of thinning treatment, ranged from 7 to 26%, with the most pronounced increase occurring between 2015 and 2020.

Figure 2

Figure 2. Cumulative tree mortality rates following simulated stem breakage across four mid-rotation loblolly pine stands in southeastern Oklahoma from 2009 to 2020. The four stands represent two replicates of two whole-plot silvicultural treatments: control (CTRL; solid lines, solid circles) and thinned (THIN; dashed lines, solid triangles).

3.2 Factors affecting the probability of tree mortality following stem breakage

The results of the multiple logistic regression model predicting post-stem breakage mortality are presented in Table 1. Tree size, specifically DBH and HT, emerged as significant predictors, whereas CRNHT did not. Thinning treatment (THIN) also had no detectable effect on mortality risk.

Table 1

Table 1. Parameter estimates for the multiple logistic regression model predicting post-stem breakage mortality in loblolly pine.

The effects of damage regimes varied by scale: tree-level damage (PSD) was not significantly associated with mortality, but plot-level tree damage (PTD) significantly influenced tree mortality although its impact was not proportional to the severity of the damage. Conversely, larger trees exhibited lower odds of death post-breakage (DBH: OR = 0.84; HT: OR = 0.68), reflecting the greater resilience of more mature individuals. When applied to the validation dataset, the model achieved an overall accuracy of 0.821 (95% CI: 0.757–0.874), with sensitivity of 0.993 (correctly identifying surviving trees) and specificity of 0.114 (poorly discriminating dead trees, which were the minority class).

Bootstrap-derived estimates reinforced the variable significance identified by the classical logistic regression (Table 2). Across 10,000 resampling iterations, RMSE for all predictors remained low, ranging from 0.040 to 1.214, indicating that coefficient estimates are both precise and stable under repeated sampling.

Table 2

Table 2. Bootstrap-derived parameter estimates for the logistic regression model predicting post-stem breakage mortality in loblolly pine.

The model demonstrated that the probability of tree mortality following stem breakage decreased with increasing DBH and HT (Appendix 2). Trees in plots with any level of damage (i.e., PTD > 0%), irrespective of the PSD level of individual tree, exhibited higher survival probabilities than those in undamaged plots. For instance, a 20 cm-DBH tree in a plot with 25% PTD had approximately half the probability of tree death compared to a tree with the same DBH in undamaged plots (i.e., PTD = 0%). However, no significant difference in the probability of tree mortality emerged between thinned and unthinned plots when other predictors were held constant. Comparisons between fitted and bootstrap regression curves revealed increasing divergence at lower DBH (Figure 3) and lower HT (Figure 4), indicating greater uncertainty in predicted mortality for smaller or shorter trees.

Figure 3

Figure 3. Predicted probability of post-stem breakage mortality in loblolly pine as a function of diameter at breast height (DBH) and proportion of damaged trees (PTD), based on multiple logistic regression with bootstrapping. The upper panel illustrates predictions for non-thinned stands (CTRL), while the lower panel shows predictions for thinned stands (THIN). Solid black lines indicate logistic regression predictions based on observed data, while dashed brown lines represent predictions generated from 100 out of 10,000 bootstrapping simulations for visualization clarity.

Figure 4

Figure 4. Predicted probability of post-stem breakage mortality in loblolly pine as a function of tree height (HT) and proportion of damaged trees (PTD), based on multiple logistic regression with bootstrapping. The upper panel illustrates predictions for non-thinned stands (CTRL), while the lower panel shows predictions for thinned stands (THIN). Solid black lines indicate logistic regression predictions based on observed data, while dashed brown lines represent predictions generated from 100 out of 10,000 bootstrapping simulations for visualization clarity.

4 Discussion

4.1 Validation of the logistic regression model predicting post-stem tree mortality of loblolly pines

Our model’s validation (Section 3.3) confirmed a key limitation: a very low specificity (0.1143) and a resulting poor ability to discriminate true tree deaths. This is a bias likely driven by the predominance of live trees in the raw data, a well-documented challenge in logistic regression for mortality studies. Hwang et al. (2023) explored four class-balancing techniques in bootstrapped logistic models of post-fire mortality and found that appropriate class-balancing scenarios can improve model parsimony although their success depends on the specific method and dataset. Shearman et al. (2019) proposed balanced random forests to mitigate imbalance in tree mortality studies, particularly under fire-induced scenarios. Therefore, future improvements to our model should explore class-balancing strategies, such as weighted sampling, synthetic minority oversampling (SMOTE), or ensemble classifiers, to enhance the correct identification of tree deaths.

In addition, bootstrapping (B = 10,000) consistently confirmed the significance and stability of our predictors (PTD, DBH, and HT). However, predicted mortality rates displayed greater variability at lower DBH and HT values (Figures 3, 4). For visualization, we randomly selected 100 bootstrap samples out of 10,000, however, this subset disproportionately included smaller trees, which were underrepresented in the full dataset, exaggerating variability in the low-size class predictions in turn. These results underscore the importance of addressing class imbalance and ensuring representative sampling across tree-size distributions to generate reliable mortality projections for adaptive management in loblolly pine plantations.

4.2 Tree mortality is more pronounced in long-term periods than in short-term periods

We monitored loblolly pine mortality following simulated stem breakage over a 12-year period to evaluate thinning effects and inform adaptive management under climate change. Across all four study sites, regardless of the presence of thinning operations, tree mortality remained low through 2012 before accelerating sharply thereafter, with the exception of the Eagletown stand (Figure 2). These dynamics align with Lloret et al. (2022), suggesting that short-term forest resilience to disturbance reflects disturbance severity. For instance, red pine (Pinus densiflora) typically recovers from surface-fire damage within 5 years (Hwang et al., 2023). Although loblolly pines can tolerate up to 70% crown loss without immediate die-off (Bragg et al., 2003), the simulated stem damage in our study (4%–52% PSD) was unlikely to cause rapid mortality.

However, long-term resilience is shaped by cumulative effects of environmental stresses, including repeated disturbances (e.g., fires, storms, drought, and pest outbreaks) (Seidl et al., 2014; Johnstone et al., 2016) and evolving competition dynamics (Yi et al., 2022). For example, Lutz and Halpern (2006) observed severe biomass declines over 14–38 years following episodic windthrow in the western Cascade Range of Oregon, USA. Similarly, we suspect several temperature peaks during summer and unexpected winter storms after 2012 likely interacted with initial stem injuries, creating synergistic stress that drove abrupt increases in tree mortality across the study sites. This observed pattern of delayed mortality strongly aligns with the previous studies (Berdanier and Clark, 2016). They demonstrate that severe drought can set in motion protracted morbidity leading to eventual death. The synergistic stress in our study is likely a manifestation of this same mechanism (i.e., the initial stem breakage acted as a primary stressor, which was then compounded by subsequent climatic events, such as prolonged droughts). This leads to both protracted morbidity and delayed mortality. Moreover, mechanical damage-induced tree mortality may enhance structural complexity through canopy gap formation (Lutz and Halpern, 2006; Arellano et al., 2019), potentially complicating long-term mortality assessment in unmanaged forests. These contrasting short-term resistance and delayed mortality mechanisms highlight the need for nuanced management strategies. For example, immediate post-breakage interventions, such as salvage logging, may be necessary to remove severely damaged stems. Concurrently, future thinning regimes must account for the longer-term stress interactions we observed (e.g., increased vulnerability to drought or pests) by prioritizing the removal of high-risk, damaged trees to strengthen the long-term resilience of the remaining stands. Practitioners should integrate both immediate post-breakage responses and longer-term stress interactions when designing thinning regimes and other interventions to mitigate stem breakage impacts effectively.

4.3 Initial tree size is highly associated with tree mortality following stem breakage

Bootstrapped logistic regression demonstrated a strong negative relationship between tree size and mortality following stem breakage. The finding aligns with numerous studies that have investigated tree mortality caused by mechanical injuries, showing that larger trees exhibit enhanced resilience to stressors, such as windthrow and fire (Ribeiro et al., 2016; Moreau et al., 2022; Ma et al., 2023). This is attributed to their thicker trunks and more extensive root systems, and greater carbohydrate reserves, which collectively support water uptake and physiological recovery following severe weather events (Piper and Paula, 2020).

Contrary to our expectation that thinning would indirectly reduce mortality by promoting size growth, thinning effects on long-term survival were transient. To evaluate this, we examined the DBH distribution of surviving trees each year in our study area. Early in the experiment (through 2012), surviving cohorts were skewed toward larger DBH classes in thinned plots, allowing consistent initial gains in resource (Dipesh et al., 2015). Beyond this period, mortality rates in thinned and unthinned treatments converged, suggesting that other factors, such as inter-tree competition resurgence, pathogen outbreaks, or subsequent climatic extremes, overrode the initial advantage conferred by thinning. Moreau et al. (2022) similarly report that drought-resistance benefits from thinning can diminish over time, particularly when secondary stressors (e.g., invasive species colonization) emerge. Our results thus highlight the limitation of one-off thinning interventions for sustaining long-term resilience. We recommend that adaptive thinning regimes incorporate periodic reassessments of stand structure, pest and disease pressures, and site-specific disturbance risks. By integrating variable thinning intensities and complementary measures (e.g., targeted pest control), forest managers can better maintain the growth-mortality balance and mitigate the legacy effects of mechanical damage over extended periods.

4.4 Plot-level damage exerts a greater influence on mortality risk than tree-level damage

The fitted logistic regression model revealed that PTD had a more substantial effect on post-stem breakage mortality than PSD. Although previous work by Dipesh et al. (2015) showed that greater tree-level injury suppresses growth, our findings indicate a counterintuitive negative association between PTD and mortality. Elevated PTD creates canopy openings that alter microenvironmental conditions, such as increasing light penetration and reducing competition for soil moisture (Beaudet et al., 2004; Senécal et al., 2018). In addition, this pattern parallels the benefits of thinning where reduced stand density enhances the survival of remaining stems with moderate damage (Weiskittel et al., 2011; Knapp et al., 2021; Moreau et al., 2022). Our results demonstrate that stand-level disturbance regimes can reconfigure resource availability and competitive dynamics in ways that encourage long-term resilience more profoundly than the severity of damage inflicted on individual trees. These insights suggest that adaptive management should incorporate strategic stand-level interventions, such as controlled damage simulations paired with density adjustments, to maximize post-breakage survival and sustain forest productivity.

5 Conclusion

This study underscores the long-term dynamics of tree mortality following stem breakage in loblolly pine plantations, highlighting the dominant role of plot-level damage (PTD) over tree-level damage (PSD) in predicting tree mortality. Although thinning initially enhanced tree size and reduced early mortality, its protective effect declined over time as inter-tree competition and climatic stresses became the dominant forces shaping survival. In stands with higher PTD, canopy gaps alleviated resource limitation, thereby encouraging the resilience of residual trees. These findings carry several management implications. Adaptive thinning regimes should integrate both tree- and stand-level disturbance dynamics, balancing objectives for timber production with the maintenance of forest health and resilience. Management prescriptions must also account for the cumulative impacts of climate-induced events, such as droughts, storms, and pest outbreaks, on stand structure and composition. Future research should examine how varying thinning intensities interact with different disturbance regimes to influence short- and long-term patterns of tree mortality and growth across diverse forest types. By incorporating these insights into silvicultural planning, forest managers can refine thinning strategies to sustain productivity and enhance resilience in the face of ongoing environmental change.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

KH: Methodology, Conceptualization, Data curation, Validation, Investigation, Writing – review & editing, Formal analysis, Writing – original draft, Supervision, Software. WK: Writing – review & editing, Supervision, Funding acquisition.

Funding

The author(s) declare that financial support was received for the research and/or publication of this article. This study was carried out with the support of R&D Program for Forest Science Technology (Project No. “RS-2025-02213493”) provided by Korea Forest Service (Korea Forestry Promotion Institute).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The authors declare that no Gen AI was used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/ffgc.2025.1708548/full#supplementary-material

References

Arellano, G., Medina, N. G., Tan, S., Mohamad, M., and Davies, S. J. (2019). Crown damage and the mortality of tropical trees. New Phytol. 221, 169–179. doi: 10.1111/nph.15381

PubMed Abstract | Crossref Full Text | Google Scholar

Beaudet, M., Messier, C., and Leduc, A. (2004). Understorey light profiles in temperate deciduous forests: recovery process following selection cutting. J. Ecol. 92, 328–338. doi: 10.1111/j.0022-0477.2004.00869.x

Crossref Full Text | Google Scholar

Berdanier, A. B., and Clark, J. S. (2016). Multiyear drought-induced morbidity preceding tree death in southeastern US forests. Ecol. Appl. 26, 17–23. doi: 10.1890/15-0274

PubMed Abstract | Crossref Full Text | Google Scholar

Bragg, D. C., Shelton, M. G., and Zeide, B. (2003). Impacts and management implications of ice storms on forests in the southern United States. For. Ecol. Manag. 186, 99–123. doi: 10.1016/S0378-1127(03)00230-5

Crossref Full Text | Google Scholar

Cansler, C. A., Hood, S. M., Varner, J. M., van Mantgem, P. J., Agne, M. C., Andrus, R. A., et al. (2020). The fire and tree mortality database, for empirical modeling of individual tree mortality after fire. Sci Data 7:194. doi: 10.1038/s41597-020-0522-7

PubMed Abstract | Crossref Full Text | Google Scholar

Crosby, M. K. (2020). Winds of change: Tornado and hurricane impacts on Louisiana forests. E–gen. Tech. Rep. SRS–253. Asheville, NC: US Department of Agriculture, Forest Service, Southern Research Station 253, 177–181.

Google Scholar

Deschênes, É., Brice, M. H., and Brisson, J. (2019). Long-term impact of a major ice storm on tree mortality in an old-growth forest. For. Ecol. Manag. 448, 386–394. doi: 10.1016/j.foreco.2019.06.018

Crossref Full Text | Google Scholar

Dipesh, K. C., Will, R. E., Hennessey, T. C., Lynch, T. B., Heinemann, R. A., Holeman, R. T., et al. (2015). Effects of simulated ice storm damage on midrotation loblolly pine stands. For. Sci. 61, 774–779. doi: 10.5849/forsci.14-060

Crossref Full Text | Google Scholar

Fortuin, C. C., Montes, C. R., Vogt, J. T., and Gandhi, K. J. (2022). Predicting risks of tornado and severe thunderstorm damage to southeastern US forests. Landsc. Ecol. 37, 1905–1919. doi: 10.1007/s10980-022-01451-7

Crossref Full Text | Google Scholar

Hood, S. M., Varner, J. M., Van Mantgem, P., and Cansler, C. A. (2018). Fire and tree death: understanding and improving modeling of fire-induced tree mortality. Environ. Res. Lett. 13:113004. doi: 10.1088/1748-9326/aae934

Crossref Full Text | Google Scholar

Hwang, K., Han, H. S., Marshall, S. E., and Page-Dumroese, D. S. (2018). Amount and location of damage to residual trees from cut-to-length thinning operations in a young redwood forest in northern California. Forests 9:352. doi: 10.3390/f9060352

Crossref Full Text | Google Scholar

Hwang, K., Han, H. S., Marshall, S. E., and Page-Dumroese, D. S. (2020). Soil compaction from cut-to-length thinning operations in young redwood forests in northern California. Can. J. For. Res. 50, 185–192. doi: 10.1139/cjfr-2019-0225

Crossref Full Text | Google Scholar

Hwang, K., Kang, W., and Jung, Y. (2023). Application of the class-balancing strategies with bootstrapping for fitting logistic regression models for post-fire tree mortality in South Korea. Environ. Ecol. Stat. 30, 575–598. doi: 10.1007/s10651-023-00573-8

Crossref Full Text | Google Scholar

Johnsen, K. H., Butnor, J. R., Kush, J. S., Schmidtling, R. C., and Nelson, C. D. (2009). Hurricane Katrina winds damaged longleaf pine less than loblolly pine. South. J. Appl. For. 33, 178–181. doi: 10.1093/sjaf/33.4.178

Crossref Full Text | Google Scholar

Johnstone, J. F., Allen, C. D., Franklin, J. F., Frelich, L. E., Harvey, B. J., Higuera, P. E., et al. (2016). Changing disturbance regimes, ecological memory, and forest resilience. Front. Ecol. Environ. 14, 369–378. doi: 10.1002/fee.1311

Crossref Full Text | Google Scholar

Knapp, E. E., Bernal, A. A., Kane, J. M., Fettig, C. J., and North, M. P. (2021). Variable thinning and prescribed fire influence tree mortality and growth during and after a severe drought. For. Ecol. Manag. 479, 1–12. doi: 10.1016/j.foreco.2020.118595

Crossref Full Text | Google Scholar

Lloret, F., Jaime, L. A., Margalef-Marrase, J., Pérez-Navarro, M. A., and Batllori, E. (2022). Short-term forest resilience after drought-induced die-off in southwestern European forests. Sci. Total Environ. 806:150940. doi: 10.1016/j.scitotenv.2021.150940

PubMed Abstract | Crossref Full Text | Google Scholar

Lutz, J. A., and Halpern, C. B. (2006). Tree mortality during early forest development: a long-term study of rates, causes, and consequences. Ecol. Monogr. 76, 257–275. doi: 10.1890/0012-9615(2006)076[0257:TMDEFD]2.0.CO;2

Crossref Full Text | Google Scholar

Ma, Q., Su, Y., Niu, C., Ma, Q., Hu, T., Luo, X., et al. (2023). Tree mortality during long-term droughts is lower in structurally complex forest stands. Nat. Commun. 14:7467. doi: 10.1038/s41467-023-43083-8

PubMed Abstract | Crossref Full Text | Google Scholar

Maringer, J., Hacket-Pain, A., Ascoli, D., Garbarino, M., and Conedera, M. (2021). A new approach for modeling delayed fire-induced tree mortality. Ecosphere 12:e03458. doi: 10.1002/ecs2.3458

Crossref Full Text | Google Scholar

Meir, P., Mencuccini, M., and Dewar, R. C. (2015). Drought-related tree mortality: addressing the gaps in understanding and prediction. New Phytol. 207, 28–33. doi: 10.1111/nph.13382

PubMed Abstract | Crossref Full Text | Google Scholar

Metcalf, C. J. E., McMahon, S. M., and Clark, J. S. (2009). Overcoming data sparseness and parametric constraints in modeling of tree mortality: a new nonparametric Bayesian model. Can. J. For. Res. 39, 1677–1687. doi: 10.1139/X09-083

Crossref Full Text | Google Scholar

Moreau, G., Chagnon, C., Achim, A., Caspersen, J., D’Orangeville, L., Sánchez-Pinillos, M., et al. (2022). Opportunities and limitations of thinning to increase resistance and resilience of trees and forests to global change. Forestry 95, 595–615. doi: 10.1093/forestry/cpac010

Crossref Full Text | Google Scholar

Needham, J. F., Arellano, G., Davies, S. J., Fisher, R. A., Hammer, V., Knox, R. G., et al. (2022). Tree crown damage and its effects on forest carbon cycling in a tropical forest. Glob. Change Biol. 28, 5560–5574. doi: 10.1111/gcb.16318

PubMed Abstract | Crossref Full Text | Google Scholar

Oklahoma Climatological Survey. 2023. Past data & files. Available online at: https://www.mesonet.org/index.php/weather (Accessed March 16, 2023).

Google Scholar

Piper, F. I., and Paula, S. (2020). The role of nonstructural carbohydrates storage in forest resilience under climate change. Curr. Forestry Rep. 6, 1–13. doi: 10.1007/s40725-019-00109-z

Crossref Full Text | Google Scholar

Pope, C. A., Cannon, J. B., Bigelow, S. W., and Sharma, A. (2023). Effects of hurricane canopy gaps on longleaf pine and upland oak sapling growth. For. Ecol. Manag. 529:120684. doi: 10.1016/j.foreco.2022.120684

Crossref Full Text | Google Scholar

R Core Team (2024). R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing.

Google Scholar

Ribeiro, G. H. P. D. M., Chambers, J. Q., Peterson, C. J., Trumbore, S. E., Marra, D. M., Wirth, C., et al. (2016). Mechanical vulnerability and resistance to snapping and uprooting for Central Amazon tree species. For. Ecol. Manag. 380, 1–10. doi: 10.1016/j.foreco.2016.08.039

Crossref Full Text | Google Scholar

Ritchie, M. W., Skinner, C. N., and Hamilton, T. A. (2007). Probability of tree survival after wildfire in an interior pine forest of northern California: effects of thinning and prescribed fire. For. Ecol. Manag. 247, 200–208. doi: 10.1016/j.foreco.2007.04.044

Crossref Full Text | Google Scholar

Roberts, S., and Martin, M. A. (2010). Bootstrap-after-bootstrap model averaging for reducing model uncertainty in model selection for air pollution mortality studies. Environ. Health Perspect. 118, 131–136. doi: 10.1289/ehp.0901007

PubMed Abstract | Crossref Full Text | Google Scholar

Ryall, K. L., and Smith, S. M. (2005). Patterns of damage and mortality in red pine plantations following a major ice storm. Can. J. For. Res. 35, 487–493. doi: 10.1139/x04-180

Crossref Full Text | Google Scholar

Seidl, R., Rammer, W., and Spies, T. A. (2014). Disturbance legacies increase the resilience of forest ecosystem structure, composition, and functioning. Ecol. Appl. 24, 2063–2077. doi: 10.1890/14-0255.1

PubMed Abstract | Crossref Full Text | Google Scholar

Senécal, J. F., Doyon, F., and Messier, C. (2018). Tree death not resulting in gap creation: an investigation of canopy dynamics of northern temperate deciduous forests. Remote Sens 10:121. doi: 10.3390/rs10010121

Crossref Full Text | Google Scholar

Shearman, T. M., Varner, J. M., Hood, S. M., Cansler, C. A., and Hiers, J. K. (2019). Modelling post-fire tree mortality: can random forest improve discrimination of imbalanced data? Ecol. Model. 414:108855. doi: 10.1016/j.ecolmodel.2019.108855

Crossref Full Text | Google Scholar

Shortle, W. C., Smith, K. T., and Dudzik, K. R. (2003). Tree survival and growth following ice storm injury. Newtown Squre, PA: US Department of Agriculture, Forest Service, Northeastern Research Station, 4.

Google Scholar

Sohn, J. A., Saha, S., and Bauhus, J. (2016). Potential of forest thinning to mitigate drought stress: a meta-analysis. For. Ecol. Manag. 380, 261–273. doi: 10.1016/j.foreco.2016.07.046

Crossref Full Text | Google Scholar

South, D. B., and Harper, R. A. (2016). A decline in timberland continues for several southern yellow pines. J. For. 114, 116–124. doi: 10.5849/jof.15-006

Crossref Full Text | Google Scholar

Sun, C. (2016). Timber market recovery after a hurricane. For. Sci. 62, 600–612. doi: 10.5849/forsci.15-123

Crossref Full Text | Google Scholar

Tanner, E. V., Rodriguez-Sanchez, F., Healey, J. R., Holdaway, R. J., and Bellingham, P. J. (2014). Long-term hurricane damage effects on tropical forest tree growth and mortality. Ecology 95, 2974–2983. doi: 10.1890/13-1801.1

Crossref Full Text | Google Scholar

Thomas, A., Kolb, T., Biederman, J. A., Venturas, M. D., Ma, Q., Yang, D., et al. (2024). Mitigating drought mortality by incorporating topography into variable forest thinning strategies. Environ. Res. Lett. 19:034035. doi: 10.1088/1748-9326/ad29aa

Crossref Full Text | Google Scholar

Varner, J. M., Hood, S. M., Aubrey, D. P., Yedinak, K., Hiers, J. K., Jolly, W. M., et al. (2021). Tree crown injury from wildland fires: causes, measurement and ecological and physiological consequences. New Phytol. 231, 1676–1685. doi: 10.1111/nph.17539

PubMed Abstract | Crossref Full Text | Google Scholar

Wang, W., Peng, C., Kneeshaw, D. D., Larocque, G. R., and Luo, Z. (2012). Drought-induced tree mortality: ecological consequences, causes, and modeling. Environ. Rev. 20, 109–121. doi: 10.1139/a2012-004

Crossref Full Text | Google Scholar

Weiskittel, A. R., Hann, D. W., Kershaw, J. A. Jr., and Vanclay, J. K. (2011). Forest growth and yield modeling. Oxford: John Wiley & Sons.

Google Scholar

Woolley, T., Shaw, D. C., Ganio, L. M., and Fitzgerald, S. (2011). A review of logistic regression models used to predict post-fire tree mortality of western North American conifers. Int. J. Wildland Fire 21, 1–35. doi: 10.1071/WF09039

Crossref Full Text | Google Scholar

Yi, C., Hendrey, G., Niu, S., McDowell, N., and Allen, C. D. (2022). Tree mortality in a warming world: causes, patterns, and implications. Environ. Res. Lett. 17:030201. doi: 10.1088/1748-9326/ac507b

Crossref Full Text | Google Scholar

Zhao, D., Bullock, B. P., and Wang, M. (2023). Long-term dynamics of aboveground carbon stocks in managed loblolly pine plantations in the Southeast United States. For. Ecol. Manag. 546:121384. doi: 10.1016/j.foreco.2023.121384

Crossref Full Text | Google Scholar

Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., and Smith, G. M. (2009). “GLM and GAM for count data” in Mixed effects models and extensions in ecology with R. Eds. A. F. Zuur, E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith (New York, NY: Springer New York), 209–243.

Google Scholar