Oak Resprouting Survival and Competition for 19 Years after Wildﬁre in the Republic of Korea

: In the Republic of Korea, forest ﬁre occurred at a rate of more than 400 events yr − 1 from 2000 to 2016, and the burned areas are concentrated in the eastern coastal region. In this region, pine forest is extensively distributed, and pine forest is converted to oak forest through natural succession or stump-sprout regeneration following forest ﬁres. However, there is a lack of research on oak competition and survival during the regeneration process in burned areas. Therefore, this research was carried out to evaluate the e ﬀ ects of oak survival ratio of Quercus serrata (glandbearing oak), Q. variabilis (oriental oak), and Q. mongolica (Mongolian oak) in burned areas. The investigation plots were set to survey tree height, diameter at breast height, sprout survival, and composition in Goseong-gun, Kangwon province, Republic of Korea. The mean tree heights of Q. serrata and Q. variabilis were approximately 9.8 m and 9.1 m, respectively, which were higher than the approximately 5.8 m heights of Q. mongolica stands 18 years after a forest ﬁre, and the trend for tree diameter at breast height was the same for all three species stands. In the early stage of regeneration after the forest ﬁre, the survival probability of the Q. serrata sprouts decreased at a rate of 1.7–2.0 times higher than that of other oak species in the Q. serrata -dominant stands, and this tendency was found to the survival probability in Q. variabilis -dominant stands as well. The median survival time of Q. variabilis sprouts in the Q. variabilis -dominant stands was approximately 10.1 years, which was similar to Q. serrata -dominant stands. However, the dominant stand of Q. mongolica was di ﬀ erent from that of other dominant stands since the stand mostly located in the ridge and the upper part of the mountain where topography and soil conditions were considerably dissimilar from the other stands. The decision of dominant species seems to be determined by the survival and occurrence of sprouts during the early stage in post-ﬁre regenerated oak forests. Therefore, it would be more desirable to coppice sprout for dominant species productivity and rapid dominance after a forest ﬁre. ,


Introduction
In East Asia, forest fires are a very important cause of disturbance in forest ecosystems due to their influence on the forest and its environment, and fires are known to be a main cause of secondary succession [1][2][3]. Especially in the Republic of Korea, it is a common phenomenon that forests of Pinus densiflora (Korean red pine), P. thunbergii (black pine), and other conifers are replaced by oak dominant forests after forest fire [4].
fire-damaged areas in secondary succession processes. Stump sprouts are important in Quercus spp. reforestation in the burned area. However, the sprouts have growth limitations [18,27].
The competition of naturally regenerated Quercus spp. is divided into between-individual and within-individual. In the case of between-individual competition, oak outcompetes Pinus densiflora except in barren areas, such as ridges. Oak can grow under the canopy of red pine. However, red pine is shade intolerant [28,29]. In the case of multiple sprouts from a tree, a within-individual relationship occurs in the oak species. Stump sprout is a secondary reboot in oak species. Some studies have shown that resprouts in the early stage of forest development after a disturbance gain a competitive advantage by utilizing the stump [30] and hence grew disproportionately more height than sprouts emerging from seeds [31]. However, in the later stages of forest development, the resprouts grow disproportionately less than single-stem sprouts because they experience more intense competition with dense polycormons [32,33]. According to the previous studies, the coppice management of multiple stump-sprouts influences the improvement of volume, height growth, and survival [34,35]. In particular, vertical position is important because of species composition and growth. As the upper layer develops, the depth of the middle and lower vegetation decreases [36]. In other words, biotic factors mainly consist of competitions between-species or between-individuals and relationships with microbes.
The objective of this study was to understand regeneration characteristics of naturally regenerated oak forest in the Eastern coast of the Korean peninsula during 19 years after the forest fire. Specifically, to figure out and which Quercus species were dominant in the oak forest as time passed, sprouting patterns including overstory composition, growth, and development of the major oak species in the study sites were monitored. Since most of the burned areas in the Republic of Korea have been restored through artificial regeneration, it will help not only to understand post-fire regeneration processes in these oak forests, but also to develop alternative post-fire management options for these regions.

Study Site Description and Survey to Tree Growth
In the Goseong region on the eastern coast of the Republic of Korea, many forests were lost by the "Goseong forest fire (3834 ha)" in 1996 and the "Eastern coast forest fire (23,794 ha)" in 2000. These were accidental fires caused by human action. In the Goseong region, the annual average temperature was approximately 12.2 • C, and the annual average precipitation was 1402.2 mm from 1981 to 2010 by the table of climatological standard normals [37]. The altitude of the study sites ranged from 51 to 166 m. In this region, the pine forest has wide distribution, and many pine forests mainly consist of Pinus densiflora in overstory (crown layer) and broadleaf trees such as Q. mongolica, Q. variabilis, Q. serrata, Q. dentata, Alnus japonica (Japanese alder), Fraxinus rhynchophylla (ash), Castanea crenata (Korean chestnut), etc. in the understory of the unburned area [11]. In contrast, the overstory of the study sites was composed of naturally regenerated Quercus species, mainly Q. mongolica Fisch. Ex Ledeb., Q. variabilis Blume, Q. serrata Murray, Q. acutissima Carruth., etc. on the naturally regenerated stand in the burned area [38]. This experiment was performed to confirm the overstory dominant and competition of species in naturally regenerated stands after two forest fires (1996 and 2000).
The three study sites were selected to monitor the growth of naturally regenerated plants as soon as forest fire occurrence. One study site consisted of five-square survey plots (10 × 10 m, 100 m 2 plot −1 ). Initially, the five survey plots were located in the valley, but one survey plot in the valley was turned into a mountain stream three years later because the waterway had been changed due to landslide. Therefore, a total of 14 survey plots sites has been remained ( Figure 1; valley: 4 plots, mountainous facing: 5 plots, ridge: 5 plots).
The number of individual oaks (stump sprouts per ground area) was measured ten times (2001, 2002, 2003, 2004, 2005, 2006, 2008, 2010, 2010, and 2018) from 2001 to 2018. Sprout height, root collar diameter (RCD), and diameter at breast height (DBH) of Quercus spp. were measured only 2 and 18 years after the forest fire (2002 and 2018) in 14 study plots during the fall season (from September to November). In each survey plot, the dominant species was determined by the number of sprouts, that consisted of the forest crown layer after 18 years from the forest fire. The criteria for the survival ratio were based on a study after forest fire damage [27]. The survival and composition ratios were calculated by stump sprout number through elapsed time after the forest fire. For example, the composition ratio was calculated to specified species per total sprout number of Quercus spp. by Equations (1) (1) and (2). (1) where SNsp yr denotes sprouts number of each species and each year within plot, SNsp 2001 is sprout number of each species at 2001, while SNtotal yr is sprouts number of the total each year within plot.

Analysis
Each subpopulation and the time-series viability data of each Quercus spp. were analyzed by Kaplan-Meier survival analysis and Gehan-Breslow methods. The Gehan-Breslow test is a nonparametric test using the chi-square test. Further, it assumes that early survival times are known more accurately than later survival times, and the required sample number is less than log-rank and the Tarone-Ware method [39,40]. We selected this method because the change of survival was important in the initial stage after a forest fire.
Multiple comparisons of survival curves among oak species were confirmed by the Holm-Sidak adjustment method [41] using Sigmaplot 12.5 (SYSTAT Inc., Chicago, Illinois, USA). The probability density function was calculated as the survival function multiplexed by the hazard function. Further, the differences in tree growth and tree density among dominant-species were verified by nonparametric methods (Kruskal-Wallis test and Dunn-Bonferroni method). All data analyses used SPSS 18.0 (IBM, North Castle, New York, USA), except for survival analysis.

Sprout Growth and Composition over Time
The mean tree heights of the dominant species in each stand, eighteen years after a forest fire, are shown in Table 1. The Q. serrata-and Q. variabilis-dominant stands were similar at 9.8 (Standard deviation, SD: 3.8) and 9.1 (SD: 4.3) m in height, however these values were higher than 6.0 (SD: 2.8) m height in Q. mongolica (p < 0.05, H-value = 28.703). The mean DBH of Q. serrata was the highest at 14.3 (SD: 4.1) cm, followed by Q. variabilis at 11.6 (SD: 4.3) cm and Q. mongolica at 5.9 (SD: 2.5) cm (p

Analysis
Each subpopulation and the time-series viability data of each Quercus spp. were analyzed by Kaplan-Meier survival analysis and Gehan-Breslow methods. The Gehan-Breslow test is a non-parametric test using the chi-square test. Further, it assumes that early survival times are known more accurately than later survival times, and the required sample number is less than log-rank and the Tarone-Ware method [39,40]. We selected this method because the change of survival was important in the initial stage after a forest fire.
Multiple comparisons of survival curves among oak species were confirmed by the Holm-Sidak adjustment method [41] using Sigmaplot 12.5 (SYSTAT Inc., Chicago, IL, USA). The probability density function was calculated as the survival function multiplexed by the hazard function. Further, the differences in tree growth and tree density among dominant-species were verified by non-parametric methods (Kruskal-Wallis test and Dunn-Bonferroni method). All data analyses used SPSS 18.0 (IBM, North Castle, NY, USA), except for survival analysis.

Sprout Growth and Composition over Time
The mean tree heights of the dominant species in each stand, eighteen years after a forest fire, are shown in Table 1. The Q. serrataand Q. variabilis-dominant stands were similar at 9.8 (Standard deviation, SD: 3.8) and 9.1 (SD: 4.3) m in height, however these values were higher than 6.0 (SD: 2.8) m height in Q. mongolica (p < 0.05, H-value = 28.703). The mean DBH of Q. serrata was the highest at 14.3 (SD: 4.1) cm, followed by Q. variabilis at 11.6 (SD: 4.3) cm and Q. mongolica at 5.9 (SD: 2.5) cm (p < 0.05, H-value = 63.364). According to the Korean yield table [42] the DBH of Q. mongolica ranged from 8 to 15 cm at 20 years old, which is higher than the growth of Q. mongolica in the study site. Otherwise, the tree density of the dominant species was approximately 20,300 sprouts ha −1 in the Q. mongolica stand, which was the highest, followed by approximately 9100 sprouts ha −1 in the Q. serrata stand and approximately 8700 sprouts ha −1 in the Q. variabilis stand (p < 0.05, H-value = 8.409). In the trees above 2 m height, the composition ratio of dominant species occupying approximately 0.781 was Q. mongolica, followed by 0.371 in Q. serrata, and 0.216 in Q. variabilis.
In the year following the forest fire, most study sites began stump sprouting of oak species (Q. mongolica, Q. variabilis, and Q. serrata) at a rate of more than two hundred per plot, although the composition ratio was different among species. More than 90% of oak sprouts were in the range from 10 to 160 cm height after two years ( Figure 2). Nine years after the forest fire, the number of Q. variabilis sprouts rapidly decreased from four to eight years, more than in other oak species (Figure 3). In total, the relative growth of Q. variabilis height was 1.1 m per year nine years after, which is higher than the 0.7 m of Q. serrata and 0.3 m of Q. mongolica p < 0.01). In the year following the forest fire, most study sites began stump sprouting of oak species (Q. mongolica, Q. variabilis, and Q. serrata) at a rate of more than two hundred per plot, although the composition ratio was different among species. More than 90% of oak sprouts were in the range from 10 to 160 cm height after two years (Figure 2). Nine years after the forest fire, the number of Q.

Survival Analysis of Oak Sprouts
The sprout survival ratio (SSR) of Quercus variabilis was the highest (0.162) in the five Q. variabilis-

Survival Analysis of Oak Sprouts
The sprout survival ratio (SSR) of Quercus variabilis was the highest (0.162) in the five Q. variabilis-dominant plots, followed by Q. mongolica (0.045) and Q. serrata site (0.014) 18 years after the forest fire. In Q. variabilis-dominant stands, the estimated survival probability of Q. variabilis sprouts ranged from 0.974 at 2 years to 0.0378 at 18 years, with a decrease of approximately 0.5 points during the first 8 years after the forest fire. The total censored percentage of this species was 0.038 over 18 years since the forest fire. The mean survival time of sprouts was approximately 10.062 (9.575-10.550 95% CI) years in Q. variabilis, the lowest of the oak species, followed by approximately 12.875 (12.479-13.271 95% CI) years in Q. mongolica and approximately 13.070 (12.858-13.283 95% CI) years in Q. serrata (p < 0.01).
In the case of Quercus serrata-dominant stands, the SSR of Q. serrata was the highest (0.150) in the four Q. serrata-dominant sites, followed by Q. serrata (0.130) and Q. mongolica (0.052) 18 years after the forest fire. The survival ratio of Q. serrata sprouts decreased by approximately 0.7 points during the first eight years after the forest fire (0.95-0.27). The total censored percentage of this species was 0.038. The mean survival time of Q. serrata sprouts was approximately 10.062 (9.575-10.550 95% CI) years, which was the lowest of the oak species (p < 0.01). The sprout survival probability in Q. serrata-dominant stands was the same as in Q. variabilis-dominant stands.
Meanwhile, in Q. mongolica-dominant stands, the SSR of Q. mongolica was 0.628 higher than 0.132 in Q. serrata. The mean survival times of Quercus spp. sprout were approximately 14 (14.228-15.958 95% CI) years in Q. serrata. Furthermore, the sprout survival probability in Q. mongolica-dominant stands was different than that in Q. variabilisand Q. serrata-dominant stands, which were caused not only by sprouts of Q. serrata occurring in only one site but also by the number of Q. serrata sprouts occurring in Q. mongolica-dominant stands 18 years after the forest fire (see Figure 2). In addition, the Q. mongolica-dominant stand was located on the ridge and upper part of the mountain, and the soil physical and chemical properties were more substandard than found at other sites.

Discussion
In the Q. variabilis-dominant stands, the average height of sprouts of Q. variabilis 18 years after the forest fire was approximately 9.1 m, higher than other species; however, tree density was approximately 870 sprouts ha −1 , lower than Q. serrata (1370 sprouts ha −1 ). According to the Korean yield table [42,43], the average tree height of Q. variabilis and Q. serrata are in the range of 8.4-12.4 m (Q. variabilis) and 6.8-8.5 m (Q. serrata) at 20 years old, which are similar to these results. However, the tree height of Q. mongolica (average: 6.0 m) is lower than the yield table that ranges from 8.2 to 10.8. In the initial stage, the sprouting number of Quercus spp. was the highest at approximately 9870 sprouts ha −1 in Q. serrata, followed by Q. mongolica and Q. variabilis. During the first 8 years after the forest fire, the tree survival ratio of Quercus spp. rapidly decreased, which was fastest in Q. variabilis, followed by Q. serrata, and Q. mongolica. In the case of Q. serrata-dominant stands, the Q. serrata sprouts had within-individual competitions during the first eight years. As a result, the tree height growth of Q. serrata was higher than that of other species. Q. serrata covered an overstory 15 years after the forest fire. The density probability consisted of survival probability multiplied by hazard probability. Thus, the density probability means snapshot probability at a specific point in time. In addition, density changes hazard and survival, which are less sensitive than hazards (see Figure 4). The high increase of density function means high hazard at a specific time. Thus, the SSR of the dominant species rapidly decreased in the early stage (before 10 years), while the SSR of the suppressed trees rapidly changed after 10 years.
The height growth in low density stump sprouts was higher than that in high density [44]. In addition, the coppice management effect in the early stage after thinning on Q. variabilis forest increased productivity more than in the later stage [45]. In particular, the growth of the suppressed trees rapidly decreased 10 years after the forest fire in Q. variabilisand Q. serrata-dominant stands, which appears to be caused by the fact that the dominant trees formed the overstory. Consequently, it seems resources such as water, light, and nutrients were not sufficiently supplied to all sprouts of the suppressed trees [46,47]. In addition, the precipitation levels were 1264 mm and 1283 mm in 2007 and 2010, respectively, lower than the climatological normal (1402 mm, 1981-2010) which corresponded to the period of the rapid decrease in the number of sprouts of the suppressed trees. The previous study showed that the quantity of oak sprouts was rapidly reduced when resources were not supplied properly [48,49]. Meanwhile, the tendency of Q. mongolica-dominant stands was different than that of Q. serrata and Q. variabilis stands. The height and DBH growth in Q. mongolica-dominant stand were lower than documented in the Korean yield table [43], smaller stature and smaller DBH may be due to geographical location. The Q. mongolica-dominant stands were located on the ridge or upper part of the mountain. Considering the work of Keim et al. [48], these topographical features and soil conditions may have negative effects not only on stump sprouting, but also on the height growth of other Quercus spp. with low competition.
It has been reported that Q. variabilis, Q. acutissima and Q. serrata are rapidly selected as the main sprouts rather than Q. mongolica and Q. dentate, and the dynamics are inversely correlated [50,51]. Additionally, the topography is a major factor in the growth of Quercus species that are dependent on dominant coverage [52]. In this process, the difference in sprouting ability of each oak species was affected by geographical location which is consistent with previous articles [36,44]. In the case of Q. mongolica forest after the forest fire, the tree density of Q. mongolica was continuously the highest. This result is similar to that in a California black oak forest where, after disturbances, stump sprouts regenerated more rapidly than other sprouts, and seeds generated 13 years after a forest fire [53].
topographical features and soil conditions may have negative effects not only on stump sprouting, but also on the height growth of other Quercus spp. with low competition. It has been reported that Q. variabilis, Q. acutissima and Q. serrata are rapidly selected as the main sprouts rather than Q. mongolica and Q. dentate, and the dynamics are inversely correlated [50,51]. Additionally, the topography is a major factor in the growth of Quercus species that are dependent on dominant coverage [52]. In this process, the difference in sprouting ability of each oak species was affected by geographical location which is consistent with previous articles [36,44]. In the case of Q. Meanwhile, the effects of topographic conditions on each species stand might be unexpectedly stronger than their survival and growth characteristics on species competition. It was consistent with the previous study that Q. mongolica forests are more widely distributed in ridge areas or higher elevation of Republic of Korea than Q. serrata and Q. variabilis forests [54,55]. In order to investigate the intact effect of species competition, further studies should be conducted under similar site conditions for each species. The mean tree height of the Q. mongolica stand was lower than that of the other stands, and it was difficult for certain species and individuals to grow vertically. The diameter before thinning was affected not only growth limitation of stump sprouts but also the relative growth rate after thinning in the oak forest [22,35], and clump growth and survival of Quercus nigra (water oak) were increased with overstory thinning intensity in a 28-year-old Q. nigra forest [56]. Stump sprouts have a low occurrence because oak trees are slow to grow and strive for survival compared to inter-individual competition [33]. Meanwhile, the high density of sprouts was reduced due to competition, but after the determination of the dominant relationship with the competition pressure, the growth increased, and this phenomenon led to vertical differentiation [30,31]. In the case of the Q. mongolica stand, competition between individuals was not cut-throat, but the growth was insufficient due to the soil inferiority, such that vertical differentiation seemed to be reduced. However, many studies have shown that oak growth caused by stump sprouting has marginal growth. Therefore, further studies and analysis of stump sprouting are needed.

Conclusions
Stump sprouting of oak is a common phenom and natural regeneration process in the Republic of Korea after a forest fire. The space available to each individual stump sprout influences not only the intensity of inter-and intra-specific competition, but also the growth of oak sprouts during the regeneration process after a forest fire. For this reason, this study was performed to determine the relationship between sprout density and species dominance in natural regeneration. The authors confirmed that high mortality due to competition in the first eight years is associated with dominance later in the burned area. As a result, when natural sprouting occurred with such high probability of density and the decrease of SSR at an early stage in Q. serrata and Q. variabilis, they lead to dominance. Thus, during the first eight years after a forest fire, it would be more desirable to stump sprouts of certain species for their productivity and rapid dominance. These results were used to predict stand change, which was natural domination selection and sprout growth through oak sprout viability in the early stage. Meanwhile, this result did not consider environmental conditions, such as topography, terrain location, and soil properties. Therefore, further research is needed to figure out the effects of these conditions. Author Contributions: J.K. was responsible for study design, data analysis, and drafting this manuscript. J.-H.L. was responsible for initial design, and supervised the field survey. M.S. and S.-H.H. was responsible for field data collection, providing feedback, and revision to manuscript. W.K. was responsible for initial design and provided a major contribution to securing funding support, and providing feedback and revision to manuscript drafts. All authors have read and agreed to the published version of the manuscript.