Next Article in Journal
From Nose to Heart: Introducing Large Language Models to Explore How Olfactory Experiences Influence Forest Visitors’ Emotional Resilience
Previous Article in Journal
Forest Carbon Sinks in Chinese Provinces and Their Impact on Sustainable Development Goals
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Forests
Volume 16
Issue 1
10.3390/f16010084
forests-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Modeling Tree Mortality Induced by Climate Change-Driven Drought: A Case Study of Korean Fir in the Subalpine Forests of Jirisan National Park, South Korea

by
Wontaek Lim
1,
Hong-Chul Park
2,
Sinyoung Park
3,
Jeong-Wook Seo
4,
Jinwon Kim
2 and
Dongwook W. Ko
5,6,*
1
Industry-Academic Cooperation Foundation, Kookmin University, Seoul 02707, Republic of Korea
2
Korea National Park Research Institute, Wonju 26441, Republic of Korea
3
Department of Forest Resources, Graduate School, Kookmin University, Seoul 02707, Republic of Korea
4
Department of Wood & Paper Science, Chungbuk National University, Chungbuk 28644, Republic of Korea
5
Department of Forest Environment and Systems, Kookmin University, Seoul 02707, Republic of Korea
6
Forest Carbon Graduate School, Kookmin University, Seoul 02707, Republic of Korea
*
Author to whom correspondence should be addressed.
Forests 2025, 16(1), 84; https://doi.org/10.3390/f16010084
Submission received: 10 December 2024 / Revised: 31 December 2024 / Accepted: 31 December 2024 / Published: 6 January 2025
(This article belongs to the Special Issue Forest Dieback, Tree Mortality Risks and Post-disturbance Resilience to Environmental Changes)
Browse Figures
Versions Notes

Abstract

:
Increased drought frequency due to climate change is intensifying tree mortality, a critical issue in forest ecosystem management, especially in vulnerable subalpine ecosystems. Korean fir (Abies koreana E.H. Wilson), an endemic species of South Korea that grows in subalpine areas, is threatened by climate change-induced drought. However, our understanding of drought’s impact on tree mortality, particularly its seasonal and legacy effects, remains limited. To better understand drought-related mortality of Korean fir, we conducted annual mortality surveys, starting in 2012, at 10 fixed transects in Jirisan National Park, identified seasonal and legacy effects using redundancy analysis, and modeled Korean fir mortality, incorporating biotic and abiotic factors, using random forests. We found that early growing season drought had the greatest impact on Korean fir mortality, with legacy effects extending up to three years, while late growing season drought affected mortality only in the previous year. The mortality model achieved high predictive accuracy (94%) and revealed significant site- and size-dependent mortality patterns. These findings provide critical insights into the complex interactions between biotic and abiotic factors affecting tree mortality and offer valuable guidance for conservation strategies aimed at preserving climate-sensitive species in the face of ongoing climate change.

1. Introduction

Tree mortality, particularly that related to drought, is being increasingly reported due to rising temperatures and shifting precipitation patterns driven by climate change [1,2]. Temperature and precipitation are crucial factors that define climate zones and are directly linked to drought stress in ecosystems [3,4], which, defined as the deficit of available water relative to potential evapotranspiration, is a key factor influencing climate-induced vegetation distributions [5,6,7]. Currently, climate change is causing unprecedented temperature increases and altering the water cycle, resulting in changes in precipitation patterns. These changes are intensifying competition for resources within vegetation communities and increasing the vulnerability of forest ecosystems [8,9], contributing to accelerated tree mortality rates [10,11]. With predicted rises in the frequencies of droughts and high-temperature events due to climate change, drought-related tree mortality is expected to increase [1,2,8,12]. This mortality poses major challenges for forest ecosystem management, as it adversely impacts forest biomass production, biodiversity, and carbon storage and uptake [13]. Consequently, drought-related tree mortality has emerged as a critical issue in forest ecosystem management.
Subalpine areas are particularly vulnerable to drought-related tree mortality [14,15,16,17]. These areas, often referred to as glacial refugia, host boreal tree species that migrated to cooler environments following the last glacial maximum [18,19,20]. Increasing global temperatures and drought pose a threat to these refugia, leading to higher mortality rates among boreal species [21,22,23]. Korean fir (Abies koreana E.H. Wilson), an endemic tree species native to the subalpine areas of Korea, has experienced a catastrophic dieback [24,25,26,27,28], leading to its classification as Endangered on the IUCN Red List. To conserve this species and its associated biodiversity, various stakeholders, including government agencies, academia, and private research organizations, are investigating the causes of the Korean fir dieback to develop effective conservation plans. However, the causes remain contentious [29]. While several studies suggest that drought is a primary factor contributing to mortality across most of the distribution range of Korean fir, including Jirisan National Park (hereafter referred to as Jirisan) [28,30,31,32,33,34,35], excessive soil moisture was suggested as an important mortality factor in Hallasan National Park [24]. Therefore, the causes of Korean fir mortality may vary by region, and it is necessary to more clearly determine when and where drought is a contributing factor to guide region-specific management for Korean fir.
In the context of drought, the impact on tree mortality is not equal across seasons. Droughts occurring during the growing season have a greater effect on tree mortality than those occurring during the dormant season, and within the growing season, the impact may vary depending on the timing [14,36]. Seasonal differences in drought effects may be related to growth characteristics and stages, as the growth rates of fixed-growth conifers, such as Korean fir, typically increase until the summer solstice and decline afterward [37,38]. This pattern may be associated with the xylem growth process, which involves cell division and expansion in the cambium layer until the summer solstice, followed by xylem maturation through secondary cell wall formation and lignification [39,40,41]. The seasonal effects of drought may influence long-term tree mortality rates, and this legacy effect has been reported to vary depending on species and the timing of drought [14]. The legacy effect is associated with the carbon starvation mechanism, which explains tree mortality as a result of exhaustion of non-structural carbohydrates caused by prolonged drought events [12]. Therefore, to understand and accurately predict drought-related tree mortality, it is essential to clarify the species-specific seasonal and legacy effects of drought.
Drought-related tree mortality can also be influenced by various biotic and abiotic factors, particularly vegetation and topographic characteristics [42,43]. Generally, drought-related tree mortality is higher in stands with high tree densities and basal areas [44,45], where vulnerability to mortality is high due to intensified competition for water [8,46,47]. Tree size is also linked to drought vulnerability [48,49,50], with larger trees typically experiencing higher mortality due to greater solar exposure [51,52,53]. However, these relationships are complex and can show both linear and non-linear patterns, even within the same species [54,55]. Topographic features such as elevation, slope, and aspect, which are related to temperature and solar radiation, significantly affect the spatial patterns of drought-related mortality by influencing soil water availability and evapotranspiration [56,57,58,59]. Therefore, understanding and predicting drought-related tree mortality requires careful consideration of the interactions between biotic and abiotic factors that impact water availability and demand.
Unfortunately, our understanding of the impact of drought on tree mortality remains limited, making it difficult to accurately analyze and simulate how drought impacts tree mortality [60]. Deriving a unified mechanism for drought-related tree mortality is challenging due to the physiological complexities and uncertainties of tree’s drought responses, as well as variations across species, seasons, populations, and individuals [60,61,62,63,64]. Empirical data, statistical analysis, and modeling are crucial tools for identifying the patterns of tree mortality linked to climate change [62,65], which requires analytical techniques capable of handling both linear and non-linear complex relationships.
Using a single model to analyze and make predictions from complex data characterized by both linear and non-linear relationships is highly challenging. Random forests, a machine learning technique [66], has demonstrated robust performance with such complex data and is widely used in ecology to examine intricate interactions between biotic and abiotic factors [67,68]. It has also been applied to predict tree mortality [65,69,70,71], however, few studies have incorporated and assessed seasonal and legacy effects.
In this study, we developed an individual-level mortality prediction model using random forests for Korean fir in Jirisan. The model predicts individual tree mortality based on weather, topographic, community, and individual characteristics, with a particular focus on the seasonal and legacy effects of drought on Korean fir. To confirm the seasonal and legacy effects of drought, we employed a multivariate statistical analysis comparing multi-year mortality rates with drought information. The results aim to offer insights into the analysis and prediction of mortality for climate-sensitive species, such as Korean fir, in response to climate change.

2. Materials and Methods

2.1. Study Sites

We investigated the impact of drought on the mortality of Korean fir in the subalpine region of Jirisan, a mountainous national park located in south-central South Korea. We also developed a model to predict Korean fir mortality caused by drought. Jirisan’s highest peak, Cheonwangbong, reaches 1915 m above sea level, with a long ridge extending from east to west maintaining elevations above 1300 m (Figure 1). The elevation gradient results in diverse climatic conditions, supporting a variety of vegetation communities at different elevations. The subalpine zone of Jirisan functions as an interglacial refugia for boreal tree species such as Korean fir, spruce, and yew [72,73].
Recently, local mean temperatures in study sites have been rising due to climate change, while total yearly precipitation has remained relatively stable. However, there have been notable decreases in monthly precipitation during May and June (Figure 2).

2.2. Field Data

In response to the severe mortality of Korean fir, the Korea National Park Research Institute began monitoring Korean fir mortality in Jirisan in 2012 by establishing seven fixed transects, with three additional transects added in 2017 (Figure 1, Table 1). Surveys were conducted annually in April and May, just before the start of the growing season, to record the diameter at breast height (DBH) and mortality status of all tree species with a DBH greater than 6 cm. Newly dead trees identified during these surveys were considered to have died in the previous year. By 2020, 238 out of 472 Korean firs in the initial transects and 15 out of 135 trees in the additional transects were confirmed dead. However, no Korean fir mortality has been observed in one transect, JR_03.

2.3. Drought Information

Drought information was quantified at the local scale using the climatic water deficit (CWD), defined as the difference between potential evapotranspiration and actual evapotranspiration [7]. Monthly CWDs were calculated in R using the “CWD and AET function” [74,75,76], which is based on the Thornthwaite-Mather soil-water balance model [77,78]. Weather data for CWD were sourced from the Automated Synoptic Observing System of the Korea Meteorological Administration. In this study, Namwon and Sancheong stations, located near Jirisan at elevations of 133 m and 138 m above sea level, respectively, were selected to compare long-term drought conditions since the 1980s.
To analyze the seasonal effects of drought, monthly CWDs were divided into three groups: the early growing season CWD (EG_CWD), late growing season CWD (LG_CWD), and dormant season CWD (D_CWD). The growing season of Korean fir near Jirisan typically begins in May and ends in September [79] and was separated into the early and late growing seasons based on the summer solstice, considering the different physiologic phases of the growth cycle. Consequently, the early growing season was defined as May–June, the late growing season as July–September, and the dormant season as the previous October–April, with CWD accumulated for each period.

2.4. Analysis of Drought’s Impacts on Korean Fir Mortality

The seasonal and legacy effects of drought on Korean fir mortality was assessed using redundancy analysis (RDA), a multivariate statistical technique [80,81]. For this analysis, a species mortality matrix was constructed based on annual mortality rates for each tree species, calculated based on basal area:
M R i j = D B A i j B A i j
in which MRij, DBAij, and BAij represent the annual mortality rate, dead basal area, and total basal area of species i in year j, respectively.
The explanatory variable matrix used the EG_CWDs, LG_CWDs, and D_CWDs from the past 10 years to capture both the seasonal and legacy effects of drought.
The RDA was conducted using the “rda” function of the vegan package [82,83] in the R version 4.1.1 [84] environment. Significant explanatory variables were identified through stepwise selection using the “ordistep” function. Multicollinearity among the selected variables was assessed using the variance inflation factor (VIF). The significance of the variances explained by the axes and the selected variables was evaluated through permutation tests [85,86].

2.5. Development of Random Forest-Based Korean Fir Mortality Models

An individual-level mortality model for Korean fir was developed using the random forest algorithm, incorporating weather-, topography-, individual-, and stand-related variables (Table 2). Each surveyed Korean fir was labeled as either dead or alive at each year. The dataset included 253 individuals labeled as dead and 2643 labeled as alive.
Weather variables included the EG_CWD, LG_CWD, and D_CWD based on the results of the RDA of drought impacts on Korean fir mortality. Topographic variables included elevation, slope, and aspect, with aspect expressed as northness, ranging from 0 to 180°, where 0° represents North and 180° represents South. For individual variables, DBH was used to represent tree sizes, and relative DBH (RDBH), the ratio of DBH to the 95th percentile of all trees across all monitoring sites, was used to represent the relative size within the population. Stand variables reflected both Korean fir population and community characteristics, using density and basal area to represent each.
Tree mortality data are inevitably class-imbalanced because they are a relatively rare event. This imbalance can present challenges for machine learning models, including random forests, as rare classes are often underrepresented during the learning process [87,88,89]. To mitigate bias caused by class imbalance, sampling techniques such as over-sampling and under-sampling are commonly used [65,90]. In this study, the number of individuals categorized as dead was notably smaller than the number categorized as alive. Using only under-sampling would result in a substantial loss of information from the alive samples, while over-sampling could lead to overfitting due to the replication of identical samples. Therefore, we applied a hybrid approach that combined over-sampling for the minority class and under-sampling for the majority class [65,90]. The survival class was reduced to 1327 samples (50% of the original 2643 samples), while the dead class was increased to 673 samples (265% of the original 253 samples).
The Korean fir mortality models were trained using the randomForest package [91] in the R version 4.1.1 [84] environment. Each decision tree in the random forest model was trained with bootstrapped samples and cross-validated through out-of-bag sampling. Hyperparameter tuning was performed through a grid search approach, optimizing ntree, mtry, and nodesize. The relative importance of each explanatory variable was assessed using mean decrease accuracy, which quantifies the contribution of each variable to model performance by measuring the reduction in accuracy when the variable is permuted. Model performance was evaluated based on producer accuracy, user accuracy, overall accuracy, and Cohen’s Kappa. Additionally, predicted mortality rates were compared to the mortality rates observed during field monitoring by site and diameter class.

2.6. Confirming the Effect of Climate Change on Korean Fir Mortality

The impact of climate change on Korean fir mortality was evaluated using the Korean fir mortality models based on a recent weather scenario and two historical weather scenarios. For this, mortality rates were simulated and compared under the weather scenarios at the initial monitoring sites. To consider legacy effects on Korean fir mortality, the weather scenarios reflected drought conditions for specific time frames. The time frame of the recent weather scenario was determined based on the Korean fir mortality monitoring period at Jirisan and the legacy effect periods of drought identified in this study. The time frames of the two historical weather scenarios were set to 20 and 30 years prior to the recent weather scenario’s time frame, respectively.

3. Results

3.1. CWD of Growing Stages in the Study Area

The mean annual and growing season CWDs in the Jirisan area were higher during the 2011–2020 period than they were during the 1981–2010 period (Figure 3). Although the maximum CWD values have not increased over time, extremely high CWD values have recently been recorded frequently and continuously. For the annual CWD, extremely high values were recorded from 2013 to 2019, while during the growing seasons, extremely high EG_CWDs were recorded from 2012 to 2017, and extremely high LG_CWDs were recorded from 2013 to 2019. On the other hand, D_CWDs from 2011 to 2020 were lower than the 1981–2010 mean.

3.2. Seasonal and Legacy Effects of Drought on Korean Fir Mortality

In the RDA of Korean fir mortality at the monitoring sites, the EG_CWDs from the past three years and the previous year’s LG_CWD were found to influence tree mortality (Table 3). However, D_CWD was found to have no influence. The VIFs of the selected variables were all lower than 5, indicating no problems with multicollinearity. All selected variables had a significant effect, explaining a combined 91.5% of the variance in tree mortality. The EG_CWD from two years ago explained the largest portion, accounting for 24.8% of the variance (p < 0.001). The current year’s EG_CWD explained 21.8% of the variance (p < 0.001), the EG_CWD from three years ago explained 15.9% (p < 0.01), the previous year’s LG_CWD explained 15.3% (p < 0.01), and the previous year’s EG_CWD explained 13.7% of the variance (p < 0.01).
The biplot of the RDA explains 58.5% of the variance in tree mortality (Figure 4), with the first and second axes explaining 35.5% (p < 0.01) and 23.0% (p < 0.05), respectively. Korean fir mortality shows a negative relationship with the first axis and a positive relationship with the second axis. This was interpreted as indicating that mortality is most closely related to the EG_CWDs of the current and previous years, which have a negative relationship with the first axis and a positive relationship with the second axis. The EG_CWD from two years ago and LG_CWD from one year ago can be linked to mortality only in the second axis, showing a positive relationship with the second axis. However, the EG_CWD from three years ago exhibited a positive relationship with the first axis and a negative relationship with the second axis, indicating a lower correlation with Korean fir mortality. Therefore, EG_CWDs from the past two years and the previous year’s LG_CWD were used as weather variables in the Korean fir mortality model.

3.3. Evaluation of the Korean Fir Mortality Model

The random forests Korean fir mortality model was trained using annual monitoring data and predicted mortality by incorporating weather-, topography-, individual-, and stand-related variables. The model achieved an overall accuracy of 94.7%, with a Cohen’s Kappa of 0.732 (Table 4). The producer accuracy was 96.0% for mortality and 94.6% for survival, while the user accuracy was 63.0% for mortality and 99.6% for survival. Among the explanatory variables, DBH and RDBH were found to be the most important for predicting mortality, followed by slope, stand basal area, stand density, and the current year’s EG_CWD (Figure 5).
The model showed reasonable performance, with a root mean square error of 9.6% (Figure 6). The regression coefficient between the observed and predicted mortality rates by site and year was 1.39, indicating a tendency to overestimate mortality (Figure 6). However, a strong correlation between observed and predicted mortality rates (R2 = 0.81) suggests that the model provides a sufficiently accurate explanation of Korean fir mortality.
To compare the observed and predicted mortality rates by site and diameter class, the mortality rates were refined accordingly (Figure 7). The observed and predicted mortality rates were highly correlated across all sites, and the model accurately predicted no mortality in JR_03, where no mortality was observed. Overall, the model tended to overestimate mortality, though the degree of overestimation varied by site. In JR_05 and JR_10, the predicted mortality rates closely matched the observed mortality, while only a slight overestimation was seen in JR_06. Similar levels of overestimation were seen for sites JR_01, JR_02, JR_07, and JR_09. However, mortality was highly overestimated for JR_04 and JR_08, with regression coefficients exceeding 2.

3.4. Effect of Climate Change on Korean Fir Mortality

Before analyzing the effect of climate change, a recent weather scenario and two historical weather scenarios for comparing Korean fir mortality due to drought conditions were established considering the legacy effects of drought on Korean fir mortality. The recent weather scenario reflected drought conditions from 2010 to 2019, and the two historical weather scenarios reflected drought conditions from 20 years ago (1990–1999) and 30 years ago (1980–1989), respectively. The mean EG_CWD was 166 mm for the recent, 82 mm for the 20-years-ago, and 109 mm for the 30-years-ago scenarios. The mean LG_CWD was 337 mm for the recent, 293 mm for the 20-years-ago, and 211 mm for the 30-years-ago scenarios. The mean D_CWD was 7 mm for the recent, 26 mm for the 20-years-ago, and 32 mm for the 30-years-ago scenarios.
We compared the predicted Korean fir mortality rates using the recent and historical weather scenarios to confirm the impact of drought conditions on Korean fir mortality (Table 5). The prediction using the recent weather scenario presented an annual mean mortality rate of 12.8% over an 8-year period. In contrast, predictions using historical weather scenarios, which had lower mean CWDs during the growing season, presented lower mortality rates compared to the recent weather scenario: an annual mean mortality rate of 6.4% for weather data from 20 years ago and 8.2% for those from 30 years ago. However, the impact of CWD on Korean fir mortality varied by site. In JR_01, JR_02, JR_04, and JR_06, the predictions presented substantially lower predicted mortality rates at the lower growing season CWDs of the historical scenarios, while in JR_05 and JR_07, differences were minimal.

4. Discussion

The results of this study provide notable insights into the drought influencing Korean fir mortality in Jirisan, with particular attention to the seasonal and legacy effects of drought. The high accuracy of the random forest model in predicting mortality suggests that the model effectively captures the complex interactions between individual, stand, topographic, and weather factors influencing tree mortality.

4.1. Drought’s Impact on Korean Fir Mortality

We analyzed the impact of drought on Korean fir mortality using annual monitoring data from Jirisan. Typically, forest monitoring data are collected at multi-year intervals, making it challenging to precisely determine the dieback timing [14,16]. However, the monitoring data in this study were collected annually, providing a high temporal resolution. This enabled a clearer analysis of the impacts of drought, the severity of which fluctuates year by year.
The results of the RDA revealed that Korean fir mortality is strongly linked to drought stress during the growing season, with EG_CWD, representing drought stress during the early growing season, showing a particularly pronounced effect [25,31,33,35]. Cell division and growth in the cambium layer of conifers generally occur during the early growing season [39,40,41], which may increase their vulnerability to drought stress [92]. Increases in EG_CWD, related to precipitation decreases and temperature increases during the early growing season (May–June), are believed to have been a major factor in Korean fir mortality in Jirisan.
Moreover, legacy effects of drought on Korean fir mortality were confirmed [14,93], with notable effects depending on the period in the growing season [36,94]. The EG_CWDs from the preceding three years had a significant impact on mortality, while only the previous year’s LG_CWD had a significant impact. Compared to previous studies on similar species, this response period is notably short, indicating that Korean firs in Jirisan are particularly vulnerable to climate change [14]. This finding underscores the importance of both current and past drought conditions when assessing drought’s impacts on forest ecosystems and predicting drought-related mortality. Additionally, further investigation into carbon starvation in Korean firs is necessary to clarify the physiological mechanisms underlying drought-related mortality [12].

4.2. The Korean Fir Mortality Model

The mortality model in this study simulated the mortality and survival of Korean fir using a small number of variables, including weather-, topography-, individual-, and stand-related characteristics, achieving a high prediction accuracy of 94%. The primary factors predicting mortality and survival of Korean fir during the monitoring period were two relating to tree size, indicating that sensitivity to drought stress varies significantly with tree size, as has been shown in other studies [49,55,65]. Larger trees may experience a greater reduction in net productivity under drought stress because they must allocate more resources to foliage than smaller trees and have lower hydraulic conductivity [95,96,97]. Furthermore, larger tree’s crowns receive direct sunlight, leading to higher transpiration rates and, thus, an increased sensitivity to drought stress compared to understory trees [52,53]. At the stand level, mortality rates and tree diameter class have shown both linear [48,49,50,54,98] and nonlinear [55] relationships. The model successfully captured these relationship types between tree size and mortality, which varied by year and stand.
Among the topographic variables, slope was identified as an important variable in predicting Korean fir mortality, with trees in steeper areas simulated as being more sensitive to drought [28,48,93]. In areas with steep slopes, water runoff occurs more easily, leading to relatively lower soil water availability [99,100]. However, while Korean fir in Hallasan National Park has been reported to exhibit higher mortality rates on gentle slopes [24], this pattern was not observed in Jirisan. At JR_03, the site with the gentlest slope, no Korean fir mortality was observed, and the model also predicted no mortality. Additionally, JR_06 and JR_07, sites with steep slopes, showed high mortality rates both in monitoring and model predictions.
Other topographic variables, such as aspect and elevation, were found to have the lowest importance. Although aspect and elevation influence potential evapotranspiration, a factor in the calculation of ecosystem water demand, through solar radiation and temperature, respectively [101,102], the gradient range of the monitoring sites in this study appears too narrow to capture the impact of variables effectively (Table 1). To better analyze the effects of aspect and elevation, further spatial statistical research using remote sensing and GIS techniques and study sites over a broader spatial scale is necessary.
Stand variables, such as basal area and density, representing both the community and Korean fir populations were also important variables in predicting Korean fir mortality. Drought-related mortality has shown both linear [50,55,98] and non-linear [54] relationships with stand density and basal area, likely due to intensified competition for water under drought conditions [8,47,103,104]. However, based on the results of this study, it is difficult to determine the precise impacts of competition intensity on drought-related mortality. Resolving uncertainties regarding the role of stand density and basal area in Korean fir mortality will require further research, including long-term monitoring and the direct quantification of water competition, particularly as it shifts with climate change at the landscape scale [15].
Weather variables were assessed as having less importance in predicting Korean fir mortality than other variables. The CWD was used as a weather variable in this study, representing local scale drought conditions across the Jirisan area, and site-specific and individual variables that contribute to differences in drought intensity were more important in the model. Among weather variables, the current year’s EG_CWD had the highest importance, followed, in descending order, by the previous year’s EG_CWD, the previous year’s LG_CWD, and the EG_CWD from two years ago. Showing consistency with the findings of previous studies, drought immediately preceding Korean fir mortality was assessed as having the greatest impact [31,33,35,105].

4.3. Application and Limitations of the Korean Fir Mortality Model

The mortality model is also applicable for assessing the impact of climate change on Korean fir mortality. Despite the relatively low importance of weather variables, simulations using the lower growing season CWDs from two historical weather scenarios (1980–1989 and 1990–1999) predicted lower mortality rates. At JR_01, JR_02, JR_04, and JR_06, simulations using both historical weather scenarios predicted substantially lower mortality rates than the simulation using the recent weather scenario, suggesting that Korean fir mortality at these locations may be closely associated with climate change-induced rising EG_CWDs. Notably, simulations at JR_06, even though lower under historical weather conditions, showed high mortality rates, suggesting that additional factors may be contributing to the mortality rates at this site. In contrast, simulations for JR_05 and JR_07 showed few differences in mortality rates between the recent and historical scenarios, indicating that Korean fir mortality at these sites may be driven by species or stand characteristics, such as the short lifespan of Korean fir [25] or the dieback and regeneration wave often observed in subalpine Abies species [106,107].
This study confirmed that increased Korean fir mortality rates are associated with extreme changes in local precipitation patterns, with reduced precipitation during May and June identified as a key factor. At the JR_03 site, no Korean fir mortality was observed, suggesting this site may serve as a refugia for Korean fir. However, the long-term sustainability of this refugia remains uncertain. To evaluate the sustainability of potential refugia, climate scenarios incorporating extreme weather pattern changes are necessary. However, climate scenarios based on global circulation models often simplify future climate by presenting mean weather conditions, which fail to capture extreme weather events at local scales [108,109,110]. This is especially problematic for precipitation, for which local and regional biases can be pronounced [111,112,113]. To assess the potential of future refugia for Korean fir, it will be essential to adjust the local and regional biases of climate change scenarios [114,115].

5. Conclusions

Tree mortality results from the complex interactions of numerous factors, making it extremely challenging to predict accurately. In this study, we confirmed that increases in drought frequency influence Korean fir mortality in the subalpine area of Jirisan, highlighting both the seasonal and legacy effects of drought. The impact of drought on Korean fir mortality was found to exhibit dependency by season, with drought during the growing season influencing dieback, particularly during the early growing season, which was found to be most detrimental. Drought during the early growing season had a significant impact on mortality up to three years later, and drought during the late growing season had a significant effect on mortality in the next year. Based on the seasonal and legacy effects, our mortality model effectively explained Korean fir mortality using a simplified set of variables, including weather, topography, stand, and individual characteristics, and successfully predicted the complex relationships between these variables and tree mortality, which vary by site and year. By comparing simulations using both recent and historical weather data, the model demonstrated its potential to assess the effects of climate change on Korean fir mortality.
These results provide valuable insights into Korean fir mortality and offer essential information for conservation efforts. Additionally, the methodology can contribute to a detailed analysis of the drivers of mortality and the development of mortality models for trees experiencing severe dieback due to climate change. We also suggest where future research should focus in order to address uncertainties related to Korean fir conservation in Jirisan. Although elevation and aspect are important variables affecting the spatial patterns of mortality, their influence could not be confirmed within the scope of this study, necessitating spatial statistical research at a larger scale. Furthermore, the study was unable to explain the relationship between intensified water competition under drought conditions and mortality. To better interpret these dynamics, efforts directly quantifying water competition will be required.

Author Contributions

Conceptualization, W.L., H.-C.P., J.-W.S., and D.W.K.; methodology, W.L., H.-C.P., and S.P.; software, W.L. and S.P.; validation, W.L. and H.-C.P.; formal analysis, W.L. and S.P.; investigation, W.L., H.-C.P., and J.K.; resources, W.L., J.K., and D.W.K.; data curation, W.L., S.P., and J.K.; writing—original draft preparation, W.L.; writing—review and editing, W.L., H.-C.P., J.-W.S., and D.W.K.; visualization, W.L. and S.P.; supervision, H.-C.P. and D.W.K.; project administration, H.-C.P. and J.K.; funding acquisition, J.K. and D.W.K. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the Korea National Park Research Institute (KNPRI), Korea National Park Service (KNPS), Ministry of Environment (MOE) of the Republic of Korea, as the “The monitoring project of Ecosystem in National Park according to climate change (NPRI 2017-36; 2018-14; 2021-31)”.

Data Availability Statement

The datasets presented in this article are not readily available because they play a role in policy decision-making. Requests to access the datasets should be directed to the corresponding author. However, public access is expected to be granted at a subsequent stage.

Acknowledgments

We are grateful to all contributors for the subalpine forest monitoring program of the Korea National Park Research Institute.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Allen, C.D.; Macalady, A.K.; Chenchouni, H.; Bachelet, D.; McDowell, N.; Vennetier, M.; Kitzberger, T.; Rigling, A.; Breshears, D.D.; Hogg, E.H.; et al. A Global Overview of Drought and Heat-Induced Tree Mortality Reveals Emerging Climate Change Risks for Forests. For. Ecol. Manag. 2010, 259, 660–684. [Google Scholar] [CrossRef]
  2. Allen, C.D.; Breshears, D.D.; McDowell, N.G. On Underestimation of Global Vulnerability to Tree Mortality and Forest Die-off from Hotter Drought in the Anthropocene. Ecosphere 2015, 6, art129. [Google Scholar] [CrossRef]
  3. Feddema, J.J. A Revised Thornthwaite-Type Global Climate Classification. Phys. Geogr. 2005, 26, 442–466. [Google Scholar] [CrossRef]
  4. Thornthwaite, C.W. An Approach toward a Rational Classification of Climate. Geogr. Rev. 1948, 38, 55–94. [Google Scholar] [CrossRef]
  5. Mather, J.R.; Yoshioka, G.A. The Role of Climate in the Distribution of Vegetation. Ann. Assoc. Am. Geogr. 1968, 58, 29–41. [Google Scholar] [CrossRef]
  6. Stephenson, N. Actual Evapotranspiration and Deficit: Biologically Meaningful Correlates of Vegetation Distribution across Spatial Scales. J. Biogeogr. 1998, 25, 855–870. [Google Scholar] [CrossRef]
  7. Stephenson, N.L. Climatic Control of Vegetation Distribution: The Role of the Water Balance. Am. Nat. 1990, 135, 649–670. [Google Scholar] [CrossRef]
  8. Jump, A.S.; Ruiz-Benito, P.; Greenwood, S.; Allen, C.D.; Kitzberger, T.; Fensham, R.; Martínez-Vilalta, J.; Lloret, F. Structural Overshoot of Tree Growth with Climate Variability and the Global Spectrum of Drought-Induced Forest Dieback. Glob. Change Biol. 2017, 23, 3742–3757. [Google Scholar] [CrossRef]
  9. Tilman, D. Resources: A Graphical-Mechanistic Approach to Competition and Predation. Am. Nat. 1980, 116, 362–393. [Google Scholar] [CrossRef]
  10. Anderegg, W.R.L.; Schwalm, C.; Biondi, F.; Camarero, J.J.; Koch, G.; Litvak, M.; Ogle, K.; Shaw, J.D.; Shevliakova, E.; Williams, A.P.; et al. Pervasive Drought Legacies in Forest Ecosystems and Their Implications for Carbon Cycle Models. Science 2015, 349, 528–532. [Google Scholar] [CrossRef] [PubMed]
  11. Senf, C.; Buras, A.; Zang, C.S.; Rammig, A.; Seidl, R. Excess Forest Mortality Is Consistently Linked to Drought Across Europe. Nat. Commun. 2020, 11, 6200. [Google Scholar] [CrossRef] [PubMed]
  12. McDowell, N.; Pockman, W.T.; Allen, C.D.; Breshears, D.D.; Cobb, N.; Kolb, T.; Plaut, J.; Sperry, J.; West, A.; Williams, D.G.; et al. Mechanisms of Plant Survival and Mortality during Drought: Why Do Some Plants Survive While Others Succumb to Drought? New Phytol. 2008, 178, 719–739. [Google Scholar] [CrossRef] [PubMed]
  13. IPBES. Global Assessment Report on Biodiversity and Ecosystem Services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services; IPBES Secretariat: Bonn, Germany, 2019; p. 56. [Google Scholar]
  14. Bigler, C.; Gavin, D.G.; Gunning, C.; Veblen, T.T. Drought Induces Lagged Tree Mortality in a Subalpine Forest in the Rocky Mountains. Oikos 2007, 116, 1983–1994. [Google Scholar] [CrossRef]
  15. Krumm, F.; Kulakowski, D.; Risch, A.C.; Spiecker, H.; Brändli, U.-B.; Bebi, P. Stem Exclusion and Mortality in Unmanaged Subalpine Forests of the Swiss Alps. Eur. J. For. Res. 2012, 131, 1571–1583. [Google Scholar] [CrossRef]
  16. Smith, J.M.; Paritsis, J.; Veblen, T.T.; Chapman, T.B. Permanent Forest Plots Show Accelerating Tree Mortality in Subalpine Forests of the Colorado Front Range from 1982 to 2013. For. Ecol. Manag. 2015, 341, 8–17. [Google Scholar] [CrossRef]
  17. Van Mantgem, P.J.; Stephenson, N.L. Apparent Climatically Induced Increase of Tree Mortality Rates in a Temperate Forest. Ecol. Lett. 2007, 10, 909–916. [Google Scholar] [CrossRef] [PubMed]
  18. Gentili, R.; Bacchetta, G.; Fenu, G.; Cogoni, D.; Abeli, T.; Rossi, G.; Salvatore, M.C.; Baroni, C.; Citterio, S. From Cold to Warm-Stage Refugia for Boreo-Alpine Plants in Southern European and Mediterranean Mountains: The Last Chance to Survive or an Opportunity for Speciation? Biodiversity 2015, 16, 247–261. [Google Scholar] [CrossRef]
  19. Hewitt, G.M. Genetic Consequences of Climatic Oscillations in the Quaternary. Philos. Trans. R. Soc. Lond. B Biol. Sci. 2004, 359, 183–195. [Google Scholar] [CrossRef]
  20. Provan, J.; Bennett, K.D. Phylogeographic Insights into Cryptic Glacial Refugia. Trends Ecol. Evol. 2008, 23, 564–571. [Google Scholar] [CrossRef]
  21. Dobrowski, S.Z. A Climatic Basis for Microrefugia: The Influence of Terrain on Climate. Glob. Change Biol. 2011, 17, 1022–1035. [Google Scholar] [CrossRef]
  22. Gonzalez, P.; Neilson, R.P.; Lenihan, J.M.; Drapek, R.J. Global Patterns in the Vulnerability of Ecosystems to Vegetation Shifts Due to Climate Change. Glob. Ecol. Biogeogr. 2010, 19, 755–768. [Google Scholar] [CrossRef]
  23. Harsch, M.A.; Hulme, P.E.; McGlone, M.S.; Duncan, R.P. Are Treelines Advancing? A Global Meta-Analysis of Treeline Response to Climate Warming. Ecol. Lett. 2009, 12, 1040–1049. [Google Scholar] [CrossRef]
  24. Ahn, U.-S.; Yun, Y.-S. Causes of Decline in the Korean Fir Based on Spatial Distribution in the Mt. Halla Region in Korea: A Meta-Analysis. Forests 2020, 11, 391. [Google Scholar] [CrossRef]
  25. Seo, J.-W.; Choi, E.-B.; Park, J.-H.; Kim, Y.-J.; Lim, H.-I. The Role of Aging and Wind in Inducing Death and/or Growth Reduction in Korean Fir (Abies Koreana Wilson) on Mt. Halla, Korea. Atmosphere 2021, 12, 1135. [Google Scholar] [CrossRef]
  26. Chun, Y.-M.; Kim, S.-M.; Park, E.-H.; Park, S.-H.; Lee, H.-Y.; Kim, J.-W. Monitoring on the Vegetation Structure and Dynamics of Abies Koreana Populations in Jirisan National Park. Korean J. Ecol. Environ. 2021, 35, 408–423. [Google Scholar] [CrossRef]
  27. Kim, J.-W.; Jeon, J.-Y.; Park, H.-C. Survey on the Annual Mortality of Evergreen Conifers (Abies koreana, Abies nephrolepis) in the Major National Park: A Case Study on Seoraksan, Deogyusan, Jirisan National Park. J. Nat. Park Res. 2021, 12, 79–84. [Google Scholar] [CrossRef]
  28. Park, H.-C.; Moon, G.-S.; Lee, H.; Lee, N.-Y. A Study on the Spatial Information and Location Environment of Dead Coniferous Tree in Subalpine Zone in Jirisan National Park. Korean J. Ecol. Environ. 2020, 34, 42–54. [Google Scholar] [CrossRef]
  29. Koo, K.-A.; Kim, D.-B. Review Forty-Year Studies of Korean Fir (Abies Koreana Wilson). Korean J. Ecol. Environ. 2020, 34, 358–371. [Google Scholar] [CrossRef]
  30. Kong, W.-S. Species Composition and Distribution of Korean Alpine Plants. J. Korean Geogr. Soc. 2002, 37, 357–370, (In Korean with English abstract). [Google Scholar]
  31. Koo, K.-A.; Park, W.-K.; Kong, W.-S. Dendrochronological Analysis of Abies Koreana W. at Mt. Halla, Korea: Effects of Climate Change on the Growths. Korean J. Ecol. 2001, 24, 281–288. (In Korean) [Google Scholar]
  32. Lim, C.-H.; Kim, H.-J. Machine Learning Application for Identifying Habitat Suitability Changes of Indicator Tree Species against Recent Climate Change. J. Clim. Change Res. 2020, 11, 793–805. [Google Scholar] [CrossRef]
  33. Park, W.-K.; Seo, J.-W. A Dendroclimatic Analysis on Abies Koreana in Cheonwang-Bong Area of Mt. Chiri, Korea. Korean J. Quat. Res. 1999, 13, 25–33, (In Korean with English abstract). [Google Scholar]
  34. Yoo, S.; Lim, C.-H.; Kim, M.; Song, C.; Kim, S.-J.; Lee, W.K. Potential Distribution of Endangered Coniferous Tree Species under Climate Change. J. Clim. Change Res. 2020, 11, 215–226. [Google Scholar] [CrossRef]
  35. Lim, J.-H.; Woo, S.-Y.; Kwon, M.-J.; Chun, J.-H.; Shin, J.-H. Photosynthetic Capacity and Water Use Efficiency under Different Temperature Regimes on Healthy and Declining Korean Fir in Mt. Halla. J. Korean For. Soci. 2006, 95, 705–710, (In Korean with English abstract). [Google Scholar]
  36. Anderegg, L.D.L.; Anderegg, W.R.L.; Berry, J.A. Not All Droughts Are Created Equal: Translating Meteorological Drought into Woody Plant Mortality. Tree Physiol. 2013, 33, 672–683. [Google Scholar] [CrossRef]
  37. Rathgeber, C.B.K.; Cuny, H.E.; Fonti, P. Biological Basis of Tree-Ring Formation: A Crash Course. Front. Plant Sci. 2016, 7, 734. [Google Scholar] [CrossRef]
  38. Seo, J.-W.; Eckstein, D.; Jalkanen, R.; Schmitt, U. Climatic Control of Intra- and Inter-Annual Wood-Formation Dynamics of Scots Pine in Northern Finland. Environ. Exp. Bot. 2011, 72, 422–431. [Google Scholar] [CrossRef]
  39. Malik, R.; Rossi, S.; Sukumar, R. Variations in the Timing of Different Phenological Stages of Cambial Activity in Abies Pindrow (Royle) along an Elevation Gradient in the North-Western Himalaya. Dendrochronologia 2020, 59, 125660. [Google Scholar] [CrossRef]
  40. Rossi, S.; Deslauriers, A.; Anfodillo, T. Assessment of Cambial Activity and Xylogenesis by Microsampling Tree Species: An Example at the Alpine Timberline. IAWA J. 2006, 27, 383–394. [Google Scholar] [CrossRef]
  41. Rossi, S.; Deslauriers, A.; Anfodillo, T.; Carrer, M. Age-Dependent Xylogenesis in Timberline Conifers. New Phytol. 2008, 177, 199–208. [Google Scholar] [CrossRef] [PubMed]
  42. Hollunder, R.K.; Mariotte, P.; Carrijo, T.T.; Holmgren, M.; Luber, J.; Stein-Soares, B.; Guidoni-Martins, K.G.; Ferreira-Santos, K.; Scarano, F.R.; Garbin, M.L. Topography and Vegetation Structure Mediate Drought Impacts on the Understory of the South American Atlantic Forest. Sci. Total Environ. 2021, 766, 144234. [Google Scholar] [CrossRef]
  43. Schwartz, N.B.; Feng, X.; Muscarella, R.; Swenson, N.G.; Umaña, M.N.; Zimmerman, J.K.; Uriarte, M. Topography and Traits Modulate Tree Performance and Drought Response in a Tropical Forest. Front. For. Glob. Change 2020, 3, 596296. [Google Scholar] [CrossRef]
  44. Crouchet, S.E.; Jensen, J.; Schwartz, B.F.; Schwinning, S. Tree Mortality After a Hot Drought: Distinguishing Density-Dependent and -Independent Drivers and Why It Matters. Front. For. Glob. Change 2019, 2, 21. [Google Scholar] [CrossRef]
  45. He, F.; Duncan, R.P. Density-Dependent Effects on Tree Survival in an Old-Growth Douglas Fir Forest. J. Ecol. 2000, 88, 676–688. [Google Scholar] [CrossRef]
  46. Bottero, A.; D’Amato, A.W.; Palik, B.J.; Bradford, J.B.; Fraver, S.; Battaglia, M.A.; Asherin, L.A. Density-Dependent Vulnerability of Forest Ecosystems to Drought. J. Appl. Ecol. 2017, 54, 1605–1614. [Google Scholar] [CrossRef]
  47. Young, D.J.N.; Stevens, J.T.; Earles, J.M.; Moore, J.; Ellis, A.; Jirka, A.L.; Latimer, A.M. Long-Term Climate and Competition Explain Forest Mortality Patterns under Extreme Drought. Ecol. Lett. 2017, 20, 78–86. [Google Scholar] [CrossRef]
  48. Bennett, A.C.; McDowell, N.G.; Allen, C.D.; Anderson-Teixeira, K.J. Larger Trees Suffer Most during Drought in Forests Worldwide. Nat. Plants 2015, 1, 1–5. [Google Scholar] [CrossRef]
  49. Etzold, S.; Ziemińska, K.; Rohner, B.; Bottero, A.; Bose, A.K.; Ruehr, N.K.; Zingg, A.; Rigling, A. One Century of Forest Monitoring Data in Switzerland Reveals Species- and Site-Specific Trends of Climate-Induced Tree Mortality. Front. Plant Sci. 2019, 10, 307. [Google Scholar] [CrossRef]
  50. Macalady, A.K.; Bugmann, H. Growth-Mortality Relationships in Piñon Pine (Pinus edulis) during Severe Droughts of the Past Century: Shifting Processes in Space and Time. PLoS ONE 2014, 9, e92770. [Google Scholar] [CrossRef]
  51. Ma, Q.; Su, Y.; Niu, C.; Ma, Q.; Hu, T.; Luo, X.; Tai, X.; Qiu, T.; Zhang, Y.; Bales, R.C.; et al. Tree Mortality during Long-Term Droughts Is Lower in Structurally Complex Forest Stands. Nat. Commun. 2023, 14, 7467. [Google Scholar] [CrossRef]
  52. Nepstad, D.C.; Tohver, I.M.; Ray, D.; Moutinho, P.; Cardinot, G. Mortality of Large Trees and Lianas Following Experimental Drought in an Amazon Forest. Ecology 2007, 88, 2259–2269. [Google Scholar] [CrossRef] [PubMed]
  53. Roberts, J.; Cabral, O.M.R.; De Aguiar, L.F. Stomatal and Boundary-Layer Conductances in an Amazonian Terra Firme Rain Forest. J. Appl. Ecol. 1990, 27, 336–353. [Google Scholar] [CrossRef]
  54. Floyd, M.L.; Clifford, M.; Cobb, N.S.; Hanna, D.; Delph, R.; Ford, P.; Turner, D. Relationship of Stand Characteristics to Drought-Induced Mortality in Three Southwestern Piñion-Juniper Woodlands. Ecol. Appl. 2009, 19, 1223–1230. [Google Scholar] [CrossRef]
  55. Looney, C.E.; Sullivan, B.W.; Kolb, T.E.; Kane, J.M.; Hart, S.C. Pinyon Pine (Pinus edulis) Mortality and Response to Water Addition across a Three Million Year Substrate Age Gradient in Northern Arizona, USA. Plant Soil 2012, 357, 89–102. [Google Scholar] [CrossRef]
  56. Allen, C.D.; Breshears, D.D. Drought-Induced Shift of a Forest–Woodland Ecotone: Rapid Landscape Response to Climate Variation. Proc. Natl. Acad. Sci. USA 1998, 95, 14839–14842. [Google Scholar] [CrossRef]
  57. Cartwright, J.M.; Littlefield, C.E.; Michalak, J.L.; Lawler, J.J.; Dobrowski, S.Z. Topographic, Soil, and Climate Drivers of Drought Sensitivity in Forests and Shrublands of the Pacific Northwest, USA. Sci. Rep. 2020, 10, 18486. [Google Scholar] [CrossRef] [PubMed]
  58. Dorman, M.; Svoray, T.; Perevolotsky, A.; Moshe, Y.; Sarris, D. What Determines Tree Mortality in Dry Environments? A Multi-Perspective Approach. Ecol. Appl. 2015, 25, 1054–1071. [Google Scholar] [CrossRef]
  59. Resler, L.M.; Butler, D.R.; Malanson, G.P. Topographic Shelter and Conifer Establishment and Mortality in an Alpine Environment, Glacier National Park, Montana. Phys. Geogr. 2005, 26, 112–125. [Google Scholar] [CrossRef]
  60. Trugman, A.T.; Anderegg, L.D.L.; Anderegg, W.R.L.; Das, A.J.; Stephenson, N.L. Why Is Tree Drought Mortality so Hard to Predict? Trends Ecol. Evol. 2021, 36, 520–532. [Google Scholar] [CrossRef]
  61. McDowell, N.G.; Fisher, R.A.; Xu, C.; Domec, J.C.; Hölttä, T.; Mackay, D.S.; Sperry, J.S.; Boutz, A.; Dickman, L.; Gehres, N.; et al. Evaluating Theories of Drought-Induced Vegetation Mortality Using a Multimodel–Experiment Framework. New Phytol. 2013, 200, 304–321. [Google Scholar] [CrossRef]
  62. Meir, P.; Mencuccini, M.; Dewar, R.C. Drought-Related Tree Mortality: Addressing the Gaps in Understanding and Prediction. New Phytol. 2015, 207, 28–33. [Google Scholar] [CrossRef]
  63. Parolari, A.J.; Katul, G.G.; Porporato, A. An Ecohydrological Perspective on Drought-Induced Forest Mortality. J. Geophys. Res. Biogeosci. 2014, 119, 965–981. [Google Scholar] [CrossRef]
  64. Sala, A.; Piper, F.; Hoch, G. Physiological Mechanisms of Drought-Induced Tree Mortality Are Far from Being Resolved. New Phytol. 2010, 186, 274–281. [Google Scholar] [CrossRef]
  65. McNellis, B.E.; Smith, A.M.S.; Hudak, A.T.; Strand, E.K. Tree Mortality in Western U.S. Forests Forecasted Using Forest Inventory and Random Forest Classification. Ecosphere 2021, 12, e03419. [Google Scholar] [CrossRef]
  66. Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
  67. Cutler, D.R.; Edwards, T.C., Jr.; Beard, K.H.; Cutler, A.; Hess, K.T.; Gibson, J.; Lawler, J.J. Random Forests for Classification in Ecology. Ecology 2007, 88, 2783–2792. [Google Scholar] [CrossRef] [PubMed]
  68. Fox, E.W.; Hill, R.A.; Leibowitz, S.G.; Olsen, A.R.; Thornbrugh, D.J.; Weber, M.H. Assessing the Accuracy and Stability of Variable Selection Methods for Random Forest Modeling in Ecology. Environ. Monit. Assess. 2017, 189, 316. [Google Scholar] [CrossRef]
  69. Baguskas, S.A.; Peterson, S.H.; Bookhagen, B.; Still, C.J. Evaluating Spatial Patterns of Drought-Induced Tree Mortality in a Coastal California Pine Forest. For. Ecol. Manag. 2014, 315, 43–53. [Google Scholar] [CrossRef]
  70. Cho, N.; Agossou, C.; Kim, E.; Lim, J.-H.; Seo, J.-W.; Kang, S. Machine-Learning Modeling on Tree Mortality and Growth Reduction of Temperate Forests with Climatic and Ecophysiological Parameters. Ecol. Modell. 2023, 483, 110456. [Google Scholar] [CrossRef]
  71. Cordeiro, N.G.; Pereira, K.M.; Terra, M.D.; de Oliveira Silveira, E.M.; de Oliveira, I.M.; Júnior, F.W.; van den Berg, E.; de Mello, J.M. The Role of Environmental Filters in Brazilian Savanna Vegetation Dynamics. For. Ecol. Manag. 2021, 500, 119645. [Google Scholar] [CrossRef]
  72. Kong, W.-S.; Kim, K.; Lee, S.; Park, H.; Cho, S.-H. Distribution of High Mountain Plants and Species Vulnerability Against Climate Change. J. Environ. Impact Assess. 2014, 23, 119–136. [Google Scholar] [CrossRef]
  73. Korea National Arboretum. Ecology of Woody Plants in South Korea (I): Conifers; Korea National Arboretum: Pocheon, Republic of Korea, 2015. (In Korean)
  74. Andrus, R.A.; Chai, R.K.; Harvey, B.J.; Rodman, K.C.; Veblen, T.T. Increasing Rates of Subalpine Tree Mortality Linked to Warmer and Drier Summers. J. Ecol. 2021, 109, 2203–2218. [Google Scholar] [CrossRef]
  75. Bentz, B.J.; Millar, C.I.; Vandygriff, J.C.; Hansen, E.M. Great Basin Bristlecone Pine Mortality: Causal Factors and Management Implications. For. Ecol. Manag. 2022, 509, 120099. [Google Scholar] [CrossRef]
  76. Redmond, M.D. CWD and AET Function, V1.0.2; Zenodo: Geneva, Switzerland, 2019. Available online: https://zenodo.org/records/4490031 (accessed on 12 September 2024). [CrossRef]
  77. Thornthwaite, C.W.; Mather, J.R. Publication in Climatology, Volume 10, Number 3: Instructions and Tables for Computing Potential Evapotranspiration and the Water Balance; Laboratory of Climatology: Centerton, AR, USA, 1957; pp. 185–311. [Google Scholar]
  78. Thornthwaite, C.W.; Mather, J.R. Publication in Climatology, Volume 8, Number 1: The Water Balance; Laboratory of Climatology: Centerton, AR, USA, 1955; pp. 1–104. [Google Scholar]
  79. Park, J.-H.; Choi, E.-B.; Park, H.-C.; Lee, N.-Y.; Seo, J.-W. Intra-Annual Dynamics of Cambial and Xylem Phenology in Subalpine Conifers at Deogyusan National Park in the Republic of Korea. J. Wood Sci. 2021, 67, 22. [Google Scholar] [CrossRef]
  80. Rao, C.R. The Use and Interpretation of Principal Component Analysis in Applied Research. Sankhyā Ser. A 1964, 26, 329–358. [Google Scholar]
  81. van den Wollenberg, A.L. Redundancy Analysis an Alternative for Canonical Correlation Analysis. Psychometrika 1977, 42, 207–219. [Google Scholar] [CrossRef]
  82. Dixon, P. VEGAN, a Package of R Functions for Community Ecology. J. Veg. Sci. 2003, 14, 927–930. [Google Scholar] [CrossRef]
  83. Oksanen, J.; Blanchet, F.G.; Friendly, M.; Kindt, R.; Legendre, P.; McGlinn, D.; Minchin, P.R.; O’Hara, R.B.; Simpson, G.L.; Solymos, P.; et al. Vegan: Community Ecology Package. R package version 2.5-7. 2020, p. 295. Available online: https://cran.r-project.org/web/packages/vegan/vegan.pdf (accessed on 25 September 2024).
  84. R Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2021; p. 3802. [Google Scholar]
  85. Borcard, D.; Gillet, F.; Legendre, P. Canonical Ordination. In Numerical Ecology with R; Borcard, D., Gillet, F., Legendre, P., Eds.; Springer International Publishing: Cham, Switzerland, 2018; pp. 203–297. ISBN 978-3-319-71404-2. [Google Scholar]
  86. Legendre, P.; Legendre, L. Numerical Ecology, 3rd ed.; Elsevier: Oxford, UK, 2012; p. 990. [Google Scholar]
  87. Chawla, N.V.; Bowyer, K.W.; Hall, L.O.; Kegelmeyer, W.P. SMOTE: Synthetic Minority Over-Sampling Technique. J. Artif. Intell. Res. 2002, 16, 321–357. [Google Scholar] [CrossRef]
  88. He, H.; Garcia, E.A. Learning from Imbalanced Data. IEEE Trans. Knowl. Data Eng. 2009, 21, 1263–1284. [Google Scholar] [CrossRef]
  89. Menardi, G.; Torelli, N. Training and Assessing Classification Rules with Imbalanced Data. Data Min. Knowl. Disc. 2014, 28, 92–122. [Google Scholar] [CrossRef]
  90. Mohammed, R.; Rawashdeh, J.; Abdullah, M. Machine Learning with Oversampling and Undersampling Techniques: Overview Study and Experimental Results. In Proceedings of the 2020 11th International Conference on Information and Communication Systems (ICICS), Irbid, Jordan, 7–9 April 2020; Institute of Electrical and Electronics Engineers, Inc.: Danvers, MA, USA, 2020; pp. 243–248. [Google Scholar]
  91. Liaw, A.; Wiener, M. Classification and Regression by randomForest. R News 2002, 2, 18–22. [Google Scholar]
  92. Restaino, C.M.; Peterson, D.L.; Littell, J. Increased Water Deficit Decreases Douglas Fir Growth throughout Western US Forests. Proc. Natl. Acad. Sci. USA 2016, 113, 9557–9562. [Google Scholar] [CrossRef]
  93. Guarín, A.; Taylor, A.H. Drought Triggered Tree Mortality in Mixed Conifer Forests in Yosemite National Park, California, USA. For. Ecol. Manag. 2005, 218, 229–244. [Google Scholar] [CrossRef]
  94. Foster, T.E.; Schmalzer, P.A.; Fox, G.A. Timing Matters: The Seasonal Effect of Drought on Tree Growth. J. Torrey Bot. Soc. 2014, 141, 225–241. [Google Scholar] [CrossRef]
  95. McDowell, N.G.; Allen, C.D. Darcy’s Law Predicts Widespread Forest Mortality under Climate Warming. Nat. Clim. Change 2015, 5, 669–672. [Google Scholar] [CrossRef]
  96. Ryan, M.G.; Phillips, N.; Bond, B.J. The Hydraulic Limitation Hypothesis Revisited. Plant Cell Environ. 2006, 29, 367–381. [Google Scholar] [CrossRef]
  97. Zhang, Y.J.; Meinzer, F.C.; Hao, G.Y.; Scholz, F.G.; Bucci, S.J.; Takahashi, F.S.C.; Villalobos-Vega, R.; Giraldo, J.P.; Cao, K.F.; Hoffmann, W.A.; et al. Size-Dependent Mortality in a Neotropical Savanna Tree: The Role of Height-Related Adjustments in Hydraulic Architecture and Carbon Allocation. Plant Cell Environ. 2009, 32, 1456–1466. [Google Scholar] [CrossRef] [PubMed]
  98. Greenwood, D.L.; Weisberg, P.J. Density-Dependent Tree Mortality in Pinyon-Juniper Woodlands. For. Ecol. Manag. 2008, 255, 2129–2137. [Google Scholar] [CrossRef]
  99. Moore, I.D.; Grayson, R.B. Terrain-Based Catchment Partitioning and Runoff Prediction Using Vector Elevation Data. Water Resour. Res. 1991, 27, 1177–1191. [Google Scholar] [CrossRef]
  100. Sunada, K.; Hong, T.F. Effects of Slope Conditions on Direct Runoff Characteristics by the Interflow and Overland Flow Model. J. Hydrol. 1988, 102, 323–334. [Google Scholar] [CrossRef]
  101. Chen, J.; Zhang, J.; Peng, J.; Zou, L.; Fan, Y.; Yang, F.; Hu, Z. Alp-Valley and Elevation Effects on the Reference Evapotranspiration and the Dominant Climate Controls in Red River Basin, China: Insights from Geographical Differentiation. J. Hydrol. 2023, 620, 129397. [Google Scholar] [CrossRef]
  102. Peng, H.; Jia, Y.; Zhan, C.; Xu, W. Topographic Controls on Ecosystem Evapotranspiration and Net Primary Productivity under Climate Warming in the Taihang Mountains, China. J. Hydrol. 2020, 581, 124394. [Google Scholar] [CrossRef]
  103. Bradford, J.B.; Shriver, R.K.; Robles, M.D.; McCauley, L.A.; Woolley, T.J.; Andrews, C.A.; Crimmins, M.; Bell, D.M. Tree Mortality Response to Drought-Density Interactions Suggests Opportunities to Enhance Drought Resistance. J. Appl. Ecol. 2022, 59, 549–559. [Google Scholar] [CrossRef]
  104. Condés, S.; Vallet, P.; Bielak, K.; Bravo-Oviedo, A.; Coll, L.; Ducey, M.J.; Pach, M.; Pretzsch, H.; Sterba, H.; Vayreda, J.; et al. Climate Influences on the Maximum Size-Density Relationship in Scots Pine (Pinus sylvestris L.) and European Beech (Fagus sylvatica L.) Stands. For. Ecol. Manag. 2017, 385, 295–307. [Google Scholar] [CrossRef]
  105. Seo, J.-W.; Jeong, H.-M.; Lee, K.-H.; Park, H.-C. Dating the Dead Years of Korean Fir (Abies Koreana E.H. Wilson) at Imgeollyeong in Jirisan National Park. J. Nat. Park Res. 2019, 10, 219–223, (In Korean with English abstract). [Google Scholar]
  106. Bekker, M.F.; Malanson, G.P. Linear Forest Patterns in Subalpine Environments. Prog. Phys. Geogr. Earth Environ. 2008, 32, 635–653. [Google Scholar] [CrossRef]
  107. Sato, K.; Iwasa, Y. Modeling of Wave Regeneration in Subalpine Abies Forests: Population Dynamics with Spatial Structure. Ecology 1993, 74, 1538–1550. [Google Scholar] [CrossRef]
  108. Li, H.; Sheffield, J.; Wood, E.F. Bias Correction of Monthly Precipitation and Temperature Fields from Intergovernmental Panel on Climate Change AR4 Models Using Equidistant Quantile Matching. J. Geophys. Res. Atmos. 2010, 115, D10. [Google Scholar] [CrossRef]
  109. Mehrotra, R.; Sharma, A. A Resampling Approach for Correcting Systematic Spatiotemporal Biases for Multiple Variables in a Changing Climate. Water Resour. Res. 2019, 55, 754–770. [Google Scholar] [CrossRef]
  110. Wehrli, K.; Guillod, B.P.; Hauser, M.; Leclair, M.; Seneviratne, S.I. Assessing the Dynamic Versus Thermodynamic Origin of Climate Model Biases. Geophys. Res. Lett. 2018, 45, 8471–8479. [Google Scholar] [CrossRef]
  111. Hagemann, S.; Chen, C.; Haerter, J.O.; Heinke, J.; Gerten, D.; Piani, C. Impact of a Statistical Bias Correction on the Projected Hydrological Changes Obtained from Three GCMs and Two Hydrology Models. J. Hydrometeorol. 2011, 12, 556–578. [Google Scholar] [CrossRef]
  112. Johnson, F.; Sharma, A. A Nesting Model for Bias Correction of Variability at Multiple Time Scales in General Circulation Model Precipitation Simulations. Water Resour. Res. 2012, 48, 1. [Google Scholar] [CrossRef]
  113. Wood, A.W.; Leung, L.R.; Sridhar, V.; Lettenmaier, D.P. Hydrologic Implications of Dynamical and Statistical Approaches to Downscaling Climte Model Outputs. Clim. Change 2004, 62, 189–216. [Google Scholar] [CrossRef]
  114. Cannon, A.J.; Sobie, S.R.; Murdock, T.Q. Bias Correction of GCM Precipitation by Quantile Mapping: How Well Do Methods Preserve Changes in Quantiles and Extremes? J. Clim. 2015, 28, 6938–6959. [Google Scholar] [CrossRef]
  115. Kum, D.; Lim, K.-J.; Jang, C.-H.; Ryu, J.; Yang, J.-E.; Kim, S.-J.; Kong, D.-S.; Jung, Y. Projecting Future Climate Change Scenarios Using Three Bias-Correction Methods. Adv. Meteorol. 2014, 2014, e704151. [Google Scholar] [CrossRef]
Forests 16 00084 g001
Figure 1. Location of Jirisan National Park and the Korean fir mortality monitoring transects within the park. Namwon and Sancheong stations are the locations where the weather data used in this study were collected by the Korea Meteorological Administration. (Source: Esri, NASA, NGA, USGS).
Figure 1. Location of Jirisan National Park and the Korean fir mortality monitoring transects within the park. Namwon and Sancheong stations are the locations where the weather data used in this study were collected by the Korea Meteorological Administration. (Source: Esri, NASA, NGA, USGS).
Forests 16 00084 g001
Forests 16 00084 g002
Figure 2. Comparisons of weather conditions in the 2011–2020 period with those in the reference period (1981–2010) around Jirisan: (a) differences between the mean annual temperatures and precipitations for each year from 2011 to 2020 and the overall means of the reference period, and (b) mean monthly temperatures and precipitations for the 2011–2020 period and the reference period.
Figure 2. Comparisons of weather conditions in the 2011–2020 period with those in the reference period (1981–2010) around Jirisan: (a) differences between the mean annual temperatures and precipitations for each year from 2011 to 2020 and the overall means of the reference period, and (b) mean monthly temperatures and precipitations for the 2011–2020 period and the reference period.
Forests 16 00084 g002
Forests 16 00084 g003
Figure 3. Time series of cumulative climatic water deficit (CWD), around Jirisan, including the annual (A_CWD), early growing season (EG_CWD), late growing season (LG_CWD), and dormant season (D_CWD) CWDs. The dashed lines represent the overall means of the cumulative CWDs before the baseline (2010) and the dotted lines represent the overall means of the cumulative CWDs after the baseline.
Figure 3. Time series of cumulative climatic water deficit (CWD), around Jirisan, including the annual (A_CWD), early growing season (EG_CWD), late growing season (LG_CWD), and dormant season (D_CWD) CWDs. The dashed lines represent the overall means of the cumulative CWDs before the baseline (2010) and the dotted lines represent the overall means of the cumulative CWDs after the baseline.
Forests 16 00084 g003
Forests 16 00084 g004
Figure 4. RDA biplot of Korean fir mortality (orange dashed arrow) and the selected weather variables (blue solid arrows) related to tree mortality in Jirisan. The first and second axes explain 35.5 and 23.0% of the variance in tree mortality, respectively. See Table 2 for variable abbreviations.
Figure 4. RDA biplot of Korean fir mortality (orange dashed arrow) and the selected weather variables (blue solid arrows) related to tree mortality in Jirisan. The first and second axes explain 35.5 and 23.0% of the variance in tree mortality, respectively. See Table 2 for variable abbreviations.
Forests 16 00084 g004
Forests 16 00084 g005
Figure 5. Relative importance of the explanatory variables in the Korean fir mortality model. See Table 2 for variable abbreviations.
Figure 5. Relative importance of the explanatory variables in the Korean fir mortality model. See Table 2 for variable abbreviations.
Forests 16 00084 g005
Forests 16 00084 g006
Figure 6. Performance assessment of the Korean fir mortality model. The root mean square error (RMSE) and coefficient of determination (R2) were used to assess model performance. The solid black line represents the regression line between observed and predicted mortality and the solid gray line represents the line of equality (y = x).
Figure 6. Performance assessment of the Korean fir mortality model. The root mean square error (RMSE) and coefficient of determination (R2) were used to assess model performance. The solid black line represents the regression line between observed and predicted mortality and the solid gray line represents the line of equality (y = x).
Forests 16 00084 g006
Forests 16 00084 g007
Figure 7. Performances of the Korean fir mortality model by site and DBH class. The solid black line represents the regression line between observed and predicted mortality and the solid gray line represents the line of equality (y = x).
Figure 7. Performances of the Korean fir mortality model by site and DBH class. The solid black line represents the regression line between observed and predicted mortality and the solid gray line represents the line of equality (y = x).
Forests 16 00084 g007
Table 1. Descriptions of the fixed transects used to monitor Korean fir mortality in Jirisan.
Table 1. Descriptions of the fixed transects used to monitor Korean fir mortality in Jirisan.
SiteLocal NameQuadrat Size (m2)Elevation
(m a.s.l.)		Slope
(Degrees)		Aspect
(Degrees)		Survey Start Year
JR_01Jeseokbong400178612.72482012
JR_02Jangteomok900165110.2862012
JR_03Seseokpyeongjeon90015498.72412012
JR_04Youngsinbong600160018.82642012
JR_05Byeoksoryeong400134217.6552012
JR_06Banyabong1600165630.2862012
JR_07Banyabong1600164224.92072012
JR_08Norumok1600134326.41772017
JR_09Dwaejipyeongjeon160013559.6332017
JR_10Yimgeolryeong400138713.92582017
Table 2. Explanatory variables used in the Korean fir mortality models.
Table 2. Explanatory variables used in the Korean fir mortality models.
TypeVariableAbbreviationUnit
WeatherEarly growing season CWD i years agoEG_CWD_imm
Late growing season CWD i years agoLG_CWD_imm
Dormant season CWD i years agoD_CWD_imm
TopographySlopeSlope%
ElevationElevationm
Aspect-NorthnessAspect-
IndividualDiameter at breast heightDBHcm
Relative DBH by 95th percentileRDBHratio
StandDensity of Korean firDensityKn/ha
Basal area of Korean firBAKm2/ha
Density of all treesDensityAn/ha
Basal area of all treesBAAm2/ha
Table 3. The selected weather variables related to the seasonal and legacy effects of drought on tree mortality in Jirisan. Altogether, the selected variables explain 91.5% of the variance in tree mortality in Jirisan. See Table 2 for variable abbreviations.
Table 3. The selected weather variables related to the seasonal and legacy effects of drought on tree mortality in Jirisan. Altogether, the selected variables explain 91.5% of the variance in tree mortality in Jirisan. See Table 2 for variable abbreviations.
VariableVIFVarianceProportion
EG_CWD_01.50331.5278 ***0.2183
EG_CWD_12.50530.9606 **0.1372
EG_CWD_23.05881.7325 ***0.2475
EG_CWD_31.22611.1112 **0.1587
LG_CWD_11.49841.0719 **0.1531
Variance explained by significant variables0.9149
** (0.001 ≤ p < 0.01), *** (p < 0.001).
Table 4. Confusion matrix of the Korean fir mortality model.
Table 4. Confusion matrix of the Korean fir mortality model.
ModelReferenceUser’s
Accuracy
ClassDeadAlive
Dead2431430.630
Alive1025000.996
Producer’s accuracy0.9600.946
Table 5. Differences in the predicted mean mortality rates at each site between the recent and historical weather scenarios. The “Recent scenario” column presents predicted mean mortality rates based on the recent weather conditions from 2010 to 2019. The “Scenario-based difference” columns present the differences in predicted mean mortality rates between the recent and historical weather scenarios (recent-historic). The historical weather scenarios are based on weather conditions from 1990 to 1999 (20 years ago) and 1980 to 1989 (30 years ago).
Table 5. Differences in the predicted mean mortality rates at each site between the recent and historical weather scenarios. The “Recent scenario” column presents predicted mean mortality rates based on the recent weather conditions from 2010 to 2019. The “Scenario-based difference” columns present the differences in predicted mean mortality rates between the recent and historical weather scenarios (recent-historic). The historical weather scenarios are based on weather conditions from 1990 to 1999 (20 years ago) and 1980 to 1989 (30 years ago).
SiteRecent Scenario (%)Scenario-Based Difference (%)
20-Years-Ago30-Years-Ago
All site12.86.44.6
JR_0111.510.08.8
JR_0211.79.95.8
JR_030.00.00.0
JR_0413.411.79.0
JR_058.11.50.0
JR_0625.58.38.6
JR_0722.73.04.6
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Lim, W.; Park, H.-C.; Park, S.; Seo, J.-W.; Kim, J.; Ko, D.W. Modeling Tree Mortality Induced by Climate Change-Driven Drought: A Case Study of Korean Fir in the Subalpine Forests of Jirisan National Park, South Korea. Forests 2025, 16, 84. https://doi.org/10.3390/f16010084

AMA Style

Lim W, Park H-C, Park S, Seo J-W, Kim J, Ko DW. Modeling Tree Mortality Induced by Climate Change-Driven Drought: A Case Study of Korean Fir in the Subalpine Forests of Jirisan National Park, South Korea. Forests. 2025; 16(1):84. https://doi.org/10.3390/f16010084

Chicago/Turabian Style

Lim, Wontaek, Hong-Chul Park, Sinyoung Park, Jeong-Wook Seo, Jinwon Kim, and Dongwook W. Ko. 2025. "Modeling Tree Mortality Induced by Climate Change-Driven Drought: A Case Study of Korean Fir in the Subalpine Forests of Jirisan National Park, South Korea" Forests 16, no. 1: 84. https://doi.org/10.3390/f16010084

APA Style

Lim, W., Park, H.-C., Park, S., Seo, J.-W., Kim, J., & Ko, D. W. (2025). Modeling Tree Mortality Induced by Climate Change-Driven Drought: A Case Study of Korean Fir in the Subalpine Forests of Jirisan National Park, South Korea. Forests, 16(1), 84. https://doi.org/10.3390/f16010084

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop