Seasonal Impact of Deer Browsing on the Demography of Fagus Japonica Seedlings in a Cool-Temperate Forest

日緑工誌，J. Jpn. Soc. Reveget. Tech., 39（2）, 232―240,（2013）論文 ORIGINAL ARTICLE

SPECIAL ISSUE “Protection and restoration of vegetation damaged by deer grazing”

Seasonal impact of deer browsing on the demography of Fagus japonica seedlings in a cool-temperate forest

ISHIZUKA, Wataru＊1）,KAJI,Mikio2） and GOTO, Susumu2）

1）Graduate School of Arts and Sciences, the University of Tokyo 東京大学大学院総合文化研究科 2）Graduate School of Agricultural and Life Sciences, the University of Tokyo 東京大学大学院農学生命科学研究科

Abstract: Severe browsing pressure from deer may have great effects on the regeneration dynamics of woody seedlings by affecting seedling mortality and seedling demography. This study quantified the seasonal impact of deer browsing on the relationship between seedling survival and the timing of seedling germination. We established a fenced deer exclosure in a cool-temperate forest in Chichibu that experienced browsing pressure from deer. Newly germinated seedlings of the dominant beech, Fagus japonica, were marked and monitored frequently in the fenced plot and outside the deer fence (open plot). To compare the demography of current- year seedlings, the effects of the timing of germination, light conditions, and total seedling density in each monitoring quadrat were analyzed considering their differences between the two types of plots. The survival rate at the end of the growth period was significantly lower in the open plot (2.0％)thaninthefencedplot(9.4％), mainly because of deer browsing (73.8％ of the total mortality in the open plot). Intensive browsing was observed only during the early growth season, while damping off was the major cause of mortality in the fenced plot (79.0％ of total) during the middle season. Statistical analysis detected opposite effects in the open and fenced plots; early germination was disadvantageous and advantageous for survival, respectively. Seedling demography relating to the advantage of the timing of germination may change depending on the impact of deer browsing, which has clear seasonality. Key word: early emergence advantage, deer browsing, Fagus japonica, seedling demography

石塚航・梶幹男・後藤晋：冷温帯林におけるイヌブナ実生動態への季節的なシカ摂食害の影響摘要：シカの高い摂食圧は樹木の実生生残数のみならず，生残条件をも変え，更新動態に大きな影響を与えるとみられる。本研究では，実生発生タイミングと当年生残との関係における，シカの季節的な摂食害の影響を評価するため，シカ害のみられる秩父地方の冷温帯林にてシカ防護柵を設置し，柵内外の調査枠内に発生した優占種イヌブナの当年生実生を調査対象とした。個体識別して実生の消長を追跡し，当年生残に対する発生タイミングの効果を，生育地の光条件や実生密度の影響とともに分析し，柵内外で比較した。柵外区の実生当年生残率（2.0 ％）は柵内区の生残率（9.4％）よりも有意に低く，死亡実生の 73.8％がシカ摂食害によるものだった。死亡要因は顕著な季節性を示し，シカ害は実生発生初期に，柵内区の主たる死亡要因である菌害（全実生の 79.0％）は生育中期に集中していた。統計解析によって，柵内外で実生の発生タイミングが当年の生残に影響していることがわかったが，柵外区では早い発生が不利に，柵内区では早い発生が有利になっており，対照的な影響が検出された。早い発生の有利性はシカの存在下では失われ，季節性のあるシカの摂食によって実生生残の動態が変わることもあると示唆された。キーワード：早い発生の有利性，シカ食害，イヌブナ，Fagus japonica，実生生残過程

is an important and well-known issue3，26，30）. Many studies 1. Introduction have shown that deer significantly affect the survival of rela- The impact of deer browsing on various forest ecosystems tively large current-year seedlings by direct brows-

＊Corresponding author：3―8―1 Komaba, Meguro-ku, Tokyo, 153―8902 Japan E-mail：[email protected] Seasonal impact of deer browsing on the demography of Fagus japonica seedlings in a cool-temperate forest 233 ing4，16，14，20，24，30，31）. In cool-temperate forests, low seedling re- one of the dominant species in climax forest, together with cruitment because of increased deer populations may affect Fagus crenata Blume, which is another Japanese beech spe- forest regeneration negatively14，19）. Therefore, deer browsing cies, and a coniferous tree, Tsuga sieboldii Carrière. These is regarded as a severe limiting factor for seedling survival two beech species are well-known of exhibiting a habitat of in these forests. However, seedling germination and subse- mast seedling at intervals of several years. Mature seeds are quent establishment are the most vulnerable life stages for mainly dispersed by gravity and germinate during the next temperate forest trees6，12）. Therefore, seedling survival for spring. After germination, the seedling establishment proc- such trees is affected strongly by deer browsing, damping ess of beech is strongly affected by multiple limiting factors, off, small mammals, and other predators such as in- especially during the initial growth period. Seedling mortal- sects1，22，27）. ity seems to be high under a closed canopy or dwarf bam- In addition to those mortality factors, the conditions of boo layer because of their light demand and susceptibility to each seedling may also affect its survival, e.g., the germina- pathogenic fungi. Predation by small mammals, other verte- tion phase, light conditions, and vegetation density. Several brates, and insect larvae also increase the mortality of beech studies have demonstrated an advantage of early germina- seedlings. Therefore, the seedling recruitment of F. japonica tion on seedling survival over late germination1，2，10，27，32）.Inde- is highly limited in natural stands8）. ciduous forests, the forest canopy becomes closed during 2.2 Study area and the two experimental plots leaf flushing; therefore, there is a seasonal limitation on the Our study was conducted on the southwestern steep slope available resources in the understory (i.e., ephemeral light). of Mt. Hakutai (1,794 m a.s.l.) in the University of Tokyo Early germinating seedlings, therefore, can utilize these lim- Chichibu Forest (35°55′N, 138°50′E, 1,210―1,260 m a.s.l., with ited resources efficiently and grow sufficiently during the a slope inclination of approximately 33°), which is located in early growing season28）. An advantage of early germination central Japan. The mean temperature and annual precipita- may be detected if the timing of seedling germination is re- tion between 1993 and 2001 at the nearest meteorological lated to the resistance to mortality caused by pathogens station (200 m apart from the study area on 1,160 m a.s.l.) and／or the preferences of predators such as small ro- were 8.9°C and 1,416 mm, respectively. This region belongs dents1，13，27）. Madsen reported17） that emerged beech seed- to the cool-temperate zone and is categorized as having a Pa- lings (Fagus sylvatica L.) were browsed immediately by cific Ocean-type climate because there is more precipitation deer, which suggests that the advantage of early germina- in the summer than in the winter. The study area was cov- tion may not be maintained if deer browse severely. How- ered by an old growth forest with a total basal area of 48.7 ever, the relationship between the effect of deer browsing m2／ha, and no management activities were done to facilitate and the timing of germination on seedling survival remains monitoring the ecosystem during long-term research. The unclear. relative abundances of three dominant species, F. japonica, Observational study was, then, conducted to trace the de- F. crenata,andT. sieboldii, in this area were 24.9％, 23.1％, mography of woody seedlings at cool-temperate natural for- and 26.5％, respectively7）. Although the dwarf bamboo est in central Japan. Browsing pressure from Sika deer (Cer- (Sasamorpha borealis (Hack.) Nakai) was widely distributed vus Nippon Temminck) has increased since, at least, the last in this region5）, retrogression occurred associated with deer two decades5，11，20）. Using a dominant tree species, Fagus ja- browsing at our study site (M. Kaji, pers. obs.). In 2006, no ponica Maxim., we monitored the date of seedling germina- dwarf bamboo but several shrub and herbal species (such as tion and its subsequent fate inside and outside deer exclo- Pieris japonica (Thunb.) D. Don ex G. Don, Pertya glabres- sure. We tested the advantage of early germination consider- cens Sch. Bip. ex Nakai, Ainsliaea acerifolia Sch. Bip. var. ing the effect of the presence of deer. We asked the follow- subapoda Nakai, and Chloranthus serratus Thunb.) can be ing questions. What was the limiting factor for germinated seen on the forest floor. Only P. japonica was an evergreen seedlings and what was its effect on seedling survival？ and unpalatable species for deer, which was the first major Was seedling demography, especially the advantage of early plants among the vegetation of the forest floor, despite its germination, affected by deer browsing？ patch distribution. In the growth season (from late spring to autumn), forage biomass increased with the emergence of 2. Materials and Methods the saplings of the upper tree species. Based on the periodic 2.1 Study species deer route-census survey by the Univ. Tokyo Forest, the av- Japanese beech, F. japonica, is a deciduous broad-leaved erage number of observed deer around our study site was tree. It is mainly distributed in the cool-temperate forests of 1.0／km in 2006 and varied with season; nearly 0.0／km during Pacific Ocean-type regions. In central Japan, F. japonica is December to February15）. 234 ISHIZUKA・KAJI・GOTO

In this area, a mast year for F. japonica occurred during cotyle (or root in some cases) of the seedling was cut off, 2005 and a number of seedlings were expected to emerge in and／or that the apical meristem of the seedling was cut by 2006. In early April 2006, before the emergence of F. japon- herbivory. The former case was presumably attributed to ro- ica seedlings, we established a fenced exclosure to prevent dent attack, whereas the latter case was attributed to insect the effect of deer. We used 1.8 m height fencing poles made herbivory, such as larvae. Seedlings killed by fatal disease from steel pipes and polyethylene net to make a 28 × 14 m caused by pathogenic infections were classified as“damping plot in the fenced deer exclosure (i.e., the fenced plot). The off,”i.e., necrosis in the axis. Seedlings killed by drought or fenced plot was established immediately adjacent to a 50 × desiccation during the early root elongation phase were clas- 50 m plot, which was established in 1991 and divided into 25 sified as“desiccation.”The number of seedlings killed by subplots of 10 × 10marrangedin5× 5matrixes7）.Thelat- each mortality factor was calculated daily on the basis of the ter plot was outside the deer fence (i.e., the open plot). The number of killed seedlings and the period between two se- lack of plot replication was mainly due to the difficulty of rial surveys (i.e., days). The time series changes in these erecting the fence structure in a natural forest. Therefore, daily mortality rates during the seedling census period rep- we conducted a fine-scale observational study on the basis of resented the seasonality of each mortality factor. comprehensive tracing of seedling demography in such for- We measured the light conditions in each subplot as a ests. possible environmental condition related to seedling sur- Newly germinated beech seedlings were marked and vival. Hemispherical photographs were taken 1 m above the monitored to assess their demography in the fenced and ground three times during the spring to correspond to the openplots.Weused50× 50 cm (0.25 m2) quadrats, which timing of bud flush in the canopies of the tree species, i.e., were divided by steel wires into 10-cm sections, for monitor- from mid-April to early May during 2006. Photographs were ing. Sixteen and one hundred quadrats were installed in the taken at the center of each subplot in the open plot (n ＝ 25) fenced and open plots, respectively, to cover approximately 1 and at the same position in each quadrat in the fenced plot ％ of each plot area. All of the quadrats were fixed to the (n ＝ 16). The degree of canopy openness (％) for each ground using steel pegs for identifying the locations of seed- measurement was calculated using CanopOn 2 (http:／／ lings accurately. In the open plot, four quadrats were set takenaka-akio.cool.ne.jp／etc／canopon 2／). We also took pho- around the center of each subplot. Data from these four tographs after leaf unfolding in the tree canopy and calcu- quadrats were combined and treated as the subplot data lated openness, but we did not use these data in the analysis with four replications. In the fenced plot, the location of each because the forest canopy was closed evenly without any quadrat was selected at random and was sufficiently apart gaps in the two study plots. from other quadrats. Thus, we treated one quadrat as one 2.4 Data analysis subplot in the fenced plot. We tested the differences in total mortality at the end of 2.3 Seedling census and measurement initial growth season in the open and fenced plots using a We conducted a seedling census by visiting each quadrat chi-square test. to monitor newly emerged seedlings and the seedling condi- Before the seedling survival was assessed, the difference tions, i.e., survival／death and the growth stage transition of in spring light conditions between the two plots was each seedling in the quadrats. This census was performed checked by analysis of variance (ANOVA), as a proxy of the more than once per week during April and May, every 2 comparison of the abiotic environments between them. Can- weeks from June to August, and every 4 weeks from Sep- opy openness was used for ANOVA considering the differ- tember onward. The seedling census was continued until ences in the sites and the dates when the data were re- December 4, 2006. During each seedling census, seedling corded. As a representative value of the spring light condi- mortality was classified into four categories: (i) total loss, (ii) tions, the average canopy openness during three measure- cut off, (iii) damping off, and (iv) desiccation.“Total loss”in- ments was defined as“spring canopy openness”in the pre- dicated that the whole seedling was lost or vanished without sent study. trace, presumably attributed to deer browsing because this We assessed the relationship between the timing of seed- process was dominant only in the open plot not in the fenced ling germination and its survival and its difference between plot. There was no considerable explanation for total loss the two plots. We considered the effect of the timing of ger- other than deer browsing. Among four categories, a small mination on seedling survival (i.e., whether early emergence number of vanished seedlings were observed in the fenced was advantageous) together with the effects of other possi- plot, indicating there was a little unknown mortality in total ble factors, i.e., spring canopy openness and total density of loss category.“Cut off”indicated that the hypocotyle or epi- emerged seedlings. A generalized linear model (GLM) was Seasonal impact of deer browsing on the demography of Fagus japonica seedlings in a cool-temperate forest 235 constructed using the dependent variable of seedling sur- Table 1 Number of surviving and dead F. japonica seedlings vival (survived or dead) throughout the initial year. Thus, in the open and the deer-fenced plots. GLM was employed using a binomial error structure with a Dead (％) Survived Total logit link function. The independent variables were the tim- Plot Total Damp- Desicca- (％) Cut off (％) ing of germination of each seedling (the first census date re- loss ing off tion corded), spring canopy openness in the corresponding sub- Open 41 1,479 116 275 102 2,003 plot, total density of emerged seedlings in the correspond- (25 m2) (2.0) (73.8) (5.3) (13.7) (5.1) (100.0) ing subplot, and category of the plot (open or fenced). The Fenced 67 15 63 565 5 715 following function was constructed as the full model: (4 m2) (9.4) (2.1) (8.8) (79.0) (0.7) (100.0) 2 2 Survivalijk ～ logistic｛［datei ＋ (datei) ＋ lightj ＋ (lightj) 2 ＋ densityj ＋ (densityj) ］× plotk｝, where Survivalijk is the probability of survival throughout the initial year for the ithseedlinginthejth subplot of the open／fenced plot, datei is the timing of germination for the ith seedling, lightj is the spring canopy openness in the jth subplot, densityj is the seedling density of the jth subplot, and plotk indicated whether the plot was the open or fenced plot. We considered the possibility of optimum values for datei, lightj, and densityj for seedling survival; therefore, squared values were included in the full model. We assumed that the effect of the plot on seedling survival would be included in“the differences of the effects”of these candidate variables; therefore, plotk was set as the interactions of all variables. Overall, all variables in the full model (six in total, including the squared values) had interactions with the type of plot. To exclude variables and interactions that did not af- Fig. 1 Survival curves for the first growing season of F. fect the goodness-of-fit for the model, model selection was japonica seedlings in the open plot (filled circle) and deer-fenced plot (open circle). The propor- performed using a backward stepwise procedure from the tions indicate the total number of surviving seed- full model. We used Akaike Information Criterion (AIC) to lings in the subjected plot compared with the total 9 assess the goodness-of-fit during model selection ）. Finally, number of germinated seedlings on the census the model with the lowest AIC value was selected as the dates in the subjected plot. Each point is a census date. best-fit model. In the best-fit model, if the interaction of plotk was not excluded for variable x, this indicated that variable x was estimated as better with different parameters in each 0.0001). In the open plot, the first two mortality factors in the plot, rather than the common parameter in the two plots. Us- order of importance were total loss (73.8％) and damping off ing the output of the best-fit model, we assessed the relative (13.7％), whereas in the fenced plot, they were damping off effects of the existence of deer on the advantage of early (79.0％), cut off (8.8％) (Table 1). The observed major mor- germination, as well as on the relationships of the light con- tality factors and their effects on seedlings differed signifi- dition and density with seedling survival. All statistical analy- cantly between the two plots. Although total loss mortality ses were performed using R 2.12.225）. was also observed in the fenced plot, the proportion (2.1％ of all seedlings) was quite lower than that of the major mor- 3. Results tality in this plot. 3.1 Demography of both plots The temporal survival rate in each plot was calculated on The total numbers of current-year seedlings that emerged each census date, using the total number of surviving seed- in the open and fenced plots were 2,003 and 715, respec- lings and total number of germinated seedlings in each plot. tively (Table 1). Among these, 41 (cumulative survival rate, The seasonal patterns of the survival rates were entirely dif- 2.0％)and67(9.4％) seedlings survived during the initial ferent in the open and fenced plots (Fig. 1). In the open plot, year of 2006 in the open and fenced plots, respectively. A sig- survival rate declined sharply from mid-April to mid-May and nificant difference was detected in the survival rate for the then again from mid-June to early July. In the fenced plot, whole season between the two plots (χ2 ＝ 72.16, p ＜ however, survival rate decreased sharply from mid-June to 236 ISHIZUKA・KAJI・GOTO

May 20 and August 25, 2006). 3.2 Effects of limiting factors on seedling survival In both plots, the seedlings emerged continuously for over 2 months from mid-April until the latter part of June. The observed final seedling emergence differed between the two plots, i.e., June 19 and 29, 2006, in the open and fenced plots, respectively, but the peak emergence time was the same in both plots, i.e., the first 3 weeks of the emergence period. Of the 41 surviving seedlings in the open plot, half (21 seedlings) germinated by mid-May, 19 seedlings germinated between April 24 and May 10, 2006, while one seedling germinated on April 17, 2006. No seedling that germinated after mid-May survived, except for one that germinated on June 8, 2006. Of the 67 surviving seedlings in the fenced plot, most (64 seedlings) had germinated by the first monitoring time point. The other three seedlings germinated during early May,andthelaterarrivaldidnotsurviveinthefencedplot. ANOVA used to compare canopy openness at the different sites on three dates detected no significant differences between the two plots (mean square ＝ 15.1, F value ＝ 2.3, d.f. ＝ 1, p ＝ 0.14). However, a significant difference was detected between the dates (mean square ＝ 761.8, F value ＝ 115.0, d.f. ＝ 1, p ＜ 0.001). Canopy openness declined with time in the spring, but the degree of openness was not different between the open and fenced plots. We summarized the measured data as spring canopy openness, which ranged from 18.6％ to 28.6％. The mean values (±standard Fig. 2 The daily mortality rate of seedlings for each deviation) of spring canopy openness in the open and fenced mortality factor and its change throughout the plots were 27.4 ± 1.9％ and 21.5 ± 2.2％, respectively. first growing season in (a) the open plot and (b) the deer-fenced plot. The length and width GLM and the model selection analyzed the relationships of the bar indicate the average number of seed- among the timing of germination, light conditions, and seedlings killed by mortality factors (／m2)andthe ling density with seedling survival, and their differences be- period between two consecutive surveys, re- tween the open and fenced plots (Table 2). The best-fit spectively. model did not exclude any independent variables, but it excluded some interactions with the plot during the model se- early July (Fig. 1). The difference in the overall period of ma- lection procedure. The best-fit model had interactions with jor mortality in the two plots was also reflected in the sea- the plot in the intercept and the following variables: the sonal patterns of the daily mortality rates (Fig. 2). Each of squared value of timing of germination, the spring canopy the seasonal patterns was represented on the basis of the openness and its squared value, and the squared value of daily mortality rates and plots (Fig. 2). Total loss mortality seedling density. All three factors (timing, light, and density) prevailed from initial emergence of seedlings until mid-June had interactions, at least with the squared values; therefore, in the open plot. For other types of mortality in the two plots, the effects of these factors on seedling survival were esti- no differences were observed in the prevailing period be- mated to be different between the plots. However, the effect tween the two plots, i.e., cut off mortality from mid-May to of light on the survival of seedlings in the fenced plot was mid-June and damping off from the end of May until August, not important because it has no significant coefficients with while desiccation was observed only at the beginning of the spring canopy openness or the squared value (Table 2). growing season (Fig. 2). In the open plot, even though total To quantify the effect of each factor on seedling survival loss mortality was main cause of mortality throughout the and to identify its pattern, we calculated the probability of year, half of the seedlings died from damping off during the survival for the range of the corresponding factor (Fig. 3). prevailing period (48.6％ of surviving seedlings between Different effects were estimated for all factors; therefore, the Seasonal impact of deer browsing on the demography of Fagus japonica seedlings in a cool-temperate forest 237

Table 2 Result of the generalized linear model and model selection to estimate the survivability of current-year seedlings. Interaction (factor × plot type) Factors Open plot Fenced plot

Intercept ○－119.489 (±50.892)＊ 24.442 (±27.218)n.s. Date of emergence ― 2.902×10－1 (±8.932×10－2)＊＊ (Date of emergence)2 ○－3.627×10－3(±1.275×10－3)＊＊－1.531×10－2(±4.204×10－3)＊＊＊ Light ○ 8.315(±3.785)＊－1.532(±2.575)n.s. (Light)2 ○－1.546×10－1(±7.021×10－2)＊ 3.849×10－2(±5.901×10－2)n.s. Density of seedlings ―－2.303×10－2 (±1.065×10－2)＊ (Density of seedlings)2 ○ 1.258×10－4(±4.619×10－5)＊＊ 5.661×10－5(±2.683×10－5)＊

Significance level: ＊＊＊p ＜ 0.001, ＊＊p ＜ 0.01, ＊p ＜ 0.05, n.s. p ＞ 0.05 The estimated coefficients for the best-fit model are shown with the standard errors in parentheses. Model selection was conducted starting from a full model that contained three types of factors with the squared values and interactions with the type of plot (open／ fenced plots). No factors were excluded by the procedure of model selection, however the interactions were. As shown in the second column, when the interactions were excluded in each factor (as shown in the second column in the table as“―”), one common coefficient was estimated for the two plots. The model equation is described in the Materials and Methods section.

Fig. 3 Changes in the estimated survival probabilities of F. japonica based on three factors, according to GLM. The probabilities in the open plot (solid line) and the deer-fenced plot (dashed line) are shown separately. (a), (b), and (c) show the probability changes based on timing of germination (date), spring canopy openness of each subplot, total density of germinated beech seedlings in each quadrat, respectively. In each graph, the x-axis has been adjusted to the minimum and maximum values, while the other two factors that were not of interest were fixed as the average values (e.g., in graph a, the average values were used for light and density). probabilities of all factors were calculated separately for the highest in the open plot. It should be noted that the coeffi- open and fenced plots. For the timing of germination (Fig. 3 cients of the light conditions in the fenced plot were not sig- a), a distinct difference was observed in the pattern of prob- nificant although the estimated probability was shown based ability change between the two plots. An advantage of early on the light conditions in the fenced plot (Fig. 3b). For seed- germination was detected in the fenced plot, whereas early ling density (Fig. 3c), the effect was apparent in the open and late timing of germination were disadvantageous com- plot but not in the fenced plot. The survival rate in the open pared with intermediate timing of germination in the open plot was estimated to be high in the order of high, low, and plot. For the light conditions (Fig. 3b), the optimum condi- intermediate amount of seedling densities. tion was estimated when the probability of survival was 238 ISHIZUKA・KAJI・GOTO

beech seedlings derived from the mast year were relatively 4. Discussion large size because of the large size of seeds1）; therefore, they Mostly emerged seedlings of Japanese beech (F. japon- were familiar vegetation during this season. The lack of ica) died during the first growing season. In our observa- dwarf bamboo cover in our study area also meant that the tional study, the distinct difference in seedling demography germinated seedlings were completely exposed to preda- was detected between the open and fenced plots. Although tors. Thus, the seedlings were browsed severely by deer, the plot design in the present study did not contain true rep- and they were valuable food items in the early season, as lication, our research supplemented this shortcoming by us- noted by Madsen17）. During May-June, leaf unfolding (flushing monitoring quadrats, which were placed to cover each ing) led to an increase in the available forage biomass, which plot thoroughly, and these two plots were located on the also reduced the relative importance of germinated beech same forest floor on a mountain slope. The fence did not ex- seedlings. Thus, deer shifted their food preferences from clude small rodents and／or insects but deer. Thus, there is beech seedlings to more plentiful vegetation types23）.There- no reason to interpret the observed differences between the fore, a clear pattern of deer browsing was observed during open and fenced plots as the presence or absence of the im- the early growth season in the present study. pact of deer. We discuss the impact of deer on seedling de- In the open and fenced plots, damping off mortality pre- mography as follows, possessing the reproducibility of our dominated during the rainy season (June to August), and the findings in a future study. temporal mortality rates were 77.7％ and 86.5％,respec- 4.1 Impact and seasonality of mortality tively. Despite the high mortality rate during the rainy sea- The survival rate of beech seedlings in the first growing son, the relative importance of damping off in total mortality season was 4.7 times higher in the fenced plot (9.4％)than rate throughout the year was different between two plots in the open plot (2.0％; Table 1). The main form of mortality (Table 1), and the impact of deer browsing might have ac- in the open plot was total loss mortality, which was presum- counted for this difference. The damping off was mainly ably due to deer browsing. This indicated that the presence caused by pathogenic infection via fungus Colletotrichum de- of deer and their direct browsing strongly affected seedling matium (Pers.) Grove. (according to the isolation of the my- establishment as a severe limiting factor, even with the plen- celium by Dr. T. Yamada). The similar results to our study tiful input of masting (production of many sound seeds) was reported in a previous study of F. crenata 27）.Thisisa from beech trees. These results were consistent with those well-known mortality factor because of the high humidity in of previous studies using deer exclosures, most of which ex- the rainy season, which allows the pathogenic fungus to amined seedlings of Fagus species4，16，21，24，30，31）. Beech seed- spread rapidly among the beech seedlings18）, causing fatal lings are preferred food items for deer, and the seedling re- necrosis27,29）. In the present study, damping off caused by cruitment of beech species is hampered by constant her- pathogenic fungi was the major limiting factor for current- bivory by deer24）. Our previous work also indicated the vul- year seedlings during the rainy season, especially at the site nerable recruitment of two beech seedlings of F. japonica under the absence of deer. and F. crenata at the deer dominant sites8）. Desiccation mortality also had the seasonality at the early Moreover, the survival curves and daily mortality rates germination phase; however the relative impact as a total throughout the growing season demonstrated clear seasonal mortality showed the disparity between two plots (Table 1). differences between the plots (Figs. 1 and 2). The frequent This desiccation occurred due to the failure of root elonga- monitoring data showed that the mortality factors in both tion depend on the condition of forest floor as well as plots, including deer-related mortality in the open plot, had weather condition. It was suggested that increasing the seasonal trends. For example, total loss mortality predomi- number of desiccation in the open plot might relate to the in- nated during the early growth season (from April to mid- direct effect of deer, because the surface of forest floor May; Fig. 2), whereas damping off mortality was most com- tended to be disturbed by the frequent deer activities where mon in the rainy season (from June to August; Fig. 2). the slope was relatively steep (approximately 33°in our stud- A concentrated occurrence of total loss mortality during ied area). the early growing season appeared to be unreasonable be- 4.2 Survival advantage of early germination cause deer were still present after June in the study area, ac- The GLM analysis detected a distinct difference between cording to route-census surveys15）. However, it would be rea- the two plots for the effect of the timing of seed germination sonable when we consider the relative attraction of germi- (Table 2, Fig. 3), which was associated with the seasonality nated seedlings as food for deer. Shrub vegetation was poor of each dominant mortality factor. In the fenced plot, the on the forest floor in the early spring, and the germinated early timing of germination was significantly advantageous Seasonal impact of deer browsing on the demography of Fagus japonica seedlings in a cool-temperate forest 239 for survival (Fig. 3). This advantage for woody seedlings was ing the first growing season was compared inside and out- explained by avoiding mortality due to the pathogenic infec- side a fenced deer exclosure, which was established before tions2，10，28） and predators, such as small rodents1）. The light seedling germination. In the presence of deer, seedling sur- conditions of the forest floor are reduced from spring to vival decreased and seedling demography also changed summer in the temperate deciduous forests; therefore, the compared with that in the absence of deer. Seedling mortal- available resources for seedlings are highly ephemeral. ity due to deer browsing was the major cause of mortality, Seedlings grow at a higher rate and make their transitions and it was concentrated in the early growth season. There- among growth stages (develop to the true leaf stage) earlier fore, the advantage of the timing of germination for seedling under highly ephemeral light conditions than under poor survival would be affected by deer; an advantage of early light conditions28）. Sufficient growth before leaf unfolding in germination was not observed outside the fence, whereas an the canopy may allow seedlings to resist infections by patho- advantage was observed inside the fence. More experiments genic fungi27，28）. Therefore, the timing of germination af- are needed to verify our observations, but the seasonal fected seedling survival via their growth during the early browsing habit of deer was indicated to affect the seedling spring season and following pathogenic infection. In our bank quantity and the condition of surviving seedlings. study, early germinating seedlings in the fenced plot may have avoided pathogenic mortality in the same manner. Acknowledgments: We special thank to Shin-Ichi Morinaga In contrast to the results of previous studies1，2，10，28）and the for encouraging the manuscript writing. We are also grateful results obtained in the fenced plot, there was no early emer- to Toshihiro Yamada for identifying the fungus, C. dema- gence advantage in the open plot (Fig. 3). However, the dis- tium, Keiko Kitamura and Thwe Thwe Win for useful com- advantage of later timing of emergence detected in the open ments on the manuscript. This work was partly supported by plot was consistent with the result of the fenced plot, be- JSPS. cause of the higher susceptibility to fungal infection as de- References scribed above. It was suggested that the impact of deer browsing was severe in the early growth season for flushed 1）Abe, M., Honda, A., Hoshizaki, K. and Miguchi, H. (2008) seedlings and／or saplings in the cool-temperate forests17）;the Advantage of early seedling emergence in Fagus crenata: importance of cotyledon stage for predator escape and browsing pressure was high in the early growth season (pri- pathogen avoidance. Ecol. Res., 23: 681―688. marily from April to May) at the studied area where there 2）Arrieta, S. and Suárez, F. (2005) Spatial patterns of seed- was small vegetation in winter. Therefore, early germinated ling emergence and survival as a critical phase in holly seedlings were exposed to a high browsing pressure for a (Ilex aquifolium L.) woodland recruitment in Central Spain. longer period than late germinated seedlings. We found that For. Ecol. Manag., 205: 267―282. total loss mortality was the highest in the early growth sea- 3）Côté,S.D.,Rooney,T.P.,Tremblay,J.P.,Dussault,C.and Waller, D.M. (2004) Ecological impacts of deer overabun- son (Fig. 2), which may have overwhelmed the advantage of dance. Ann. Rev. Ecol. Evol. Syst., 35: 113―147. growth during the early season. Late germinated seedlings 4）Farris, E. and Filigheddu, R. (2008) Effects of browsing in released from deer browsing, but they experienced the high relation to vegetation cover on common yew (Taxus bac- risks related to other mortality factors, especially damping cata L.) recruitment in Mediterranean environments. Plant off caused by the pathogenic fungi. Beech seedlings in the Ecol., 199: 309―318. 5）Furubayashi, K. and Yamane, M. (1997) Dynamics of sika open plot appeared to be faced by the dilemma of the most deer and suzutake populations after the clear cutting of for- suitable timing for germination, relating to a risk of preda- ests in the Tanzawa Mountains (in Japanese with English tion and a chance of growth. As a result, the disadvantages summary). Wildl. Cons. Japan, 2: 195―204. of early and late emergence allowed intermediate emerging 6）Harcombe, P.A. (1987) Tree life tables: simple birth, seedlings to survive at significantly higher rates (Fig. 3). growth, and death data encapsulate life histories and ecological roles. BioScience, 37: 557 568. Therefore, beech seedlings would experience no survival ad- ― 7）Ishizuka, W. and Kaji, K. (2010) Extraction of topographical vantages of early germination unless there is a low abun- features in a natural forest in Chichibu using DEM and dis- dance of deer in the area or a plentiful alternative food sup- tributional properties of Fagus crenata and F. japonica ply for deer in the early growth season of April and May. within a local scale (in Japanese with English summary). Bull. Tokyo Univ. For., 123: 56―69. 5. Conclusions 8）Ishizuka, W., Goto, S. and Kaji, M. (2012) Factors responsi- ble for co-dominance of two beech species in a cool tem- We performed frequent monitoring of current-year beech perate forest in central Japan: interspecific comparison of seedlings in a cool-temperate natural forest with deer brows- spatial distribution and growth traits. Ann. For. Res., 55: ing pressure. Seedling demography from germination dur- 240 ISHIZUKA・KAJI・GOTO

93―103. summary). Wildl. Cons. Japan, 8: 63―77. 9）Johnson, J.B. and Omland, K.S. (2004) Model selection in 21）Nomiya, H., Suzuki, W., Kanazashi, T., Shibata, M., Tanaka, ecology and evolution. Trend. Ecol. Evol., 19: 101―108. H. and Nakashizuka, T. (2003) The response of forest floor 10）Jones, R.H., Allen, B.P. and Sharitz, R.R. (1997) Why do vegetation and tree regeneration to deer exclusion and dis- early-emerging tree seedlings have survival advantages？: turbance in a riparian deciduous forest, central Japan. Plant AtestusingAcer rubrum (Aceraceae). Am. J. Bot., 84: Ecol., 164: 263―276. 1714―1718. 22）Ohkubo, T., Niwa, A., Kaji, M. and Hamaya, T. (1989) Nut 11）Kaji, M. (1997) The forest damages by the deer -a case fall production and seedling survival in a natural Fagus ja- study of the University Forests of Tokyo- (in Japanese). ponica forest in the Chichibu mountains central Japan (in Hoppô Ringyô, 49: 169―172. Japanese with English summary). Jap. J. Ecol., 39: 17―26. 12）Kitajima, K. and Fenner, M. (2000) Ecology of seedling re- 23）Oi, T. and Suzuki, M. (2001) Damage to sugi (Cryptomeria generation. In: Seeds. the ecology of regeneration in plant japonica) plantations by sika deer (Cervus nippon)in communities, 2nd (ed. Fenner, M.) pp. 331―360, CABI Pub, northern Honshu, Japan. Mamm. Study, 26: 9―15. New York. 24）Olesen, C.R. and Madsen, P. (2008) The impact of roe deer 13）Kudo, G. (1996) Herbivory pattern and induced responses (Capreolus capreolus), seedbed, light and seed fall on natu- to simulated herbivory in Quercus mongolica var grosseser- ral beech (Fagus sylvatica) regeneration. For. Ecol. rata. Ecol. Res., 11: 283―289. Manag., 255: 3962―3972. 14）Kumar, S., Takeda, A. and Shibata, E. (2006) Effects of 13- 25）R Development Core Team (2011) R: A language and envi- year fencing on browsing by sika deer on seedlings on Mt. ronment for statistical computing. R Foundation for Statisti- Ohdaigahara, central Japan. J. For. Res., 11: 337―342. cal Computing, Vienna, Austria. 15）Kurita, N., Igarashi, Y., Ohmura, K., Nishiyama, N., 26）Rooney, T.P. (2001) Deer impacts on forest ecosystems: a Yoshida, Y. and Shibano, S. (2007) Monthly change of the North American perspective. Forestry, 74: 201―208. observed number of deer, using light census. Ann. Rep. 27）Sahashi, N., Kubono, T. and Shoji, T. (1994) Temporal oc- Tech. Train. Res. in Univ. Tokyo Forest, H19, 36―39. currence of dead seedlings of Japanese beech and associ- 16）Liang, S.Y. and Seagle, S.W. (2002) Browsing and micro- ated fungi. J. Jap. For. Soc., 76: 338―345. habitat effects on riparian forest woody seedling demogra- 28）Seiwa, K. (1998) Advantages of early germination for phy. Ecology, 83: 212―227. growth and survival of seedlings of Acer mono under differ- 17）Madsen, P. (1995) Effects of seedbed type on wintering of ent overstory phenologies in deciduous broad-leaved for- beech nuts (Fagus sylvatica) and deer impact on sprouting ests. J. Ecol., 86: 219―228. seedlings in natural regeneration. For. Ecol. Manag., 73: 29） Shibata, M. and Nakashizuka, T. (1995) Seed and seedling 37―43. demography of 4 cooccurring carpinus species in a temper- 18）Maeda, T. (1988) Studies on natural regeneration of beech ate deciduous forest. Ecology, 76: 1099―1108. (Fagus crenata Blume) (in Japanese with English sum- 30）Takatsuki, S. and Gorai, T. (1994) Effects of Sika-deer on mary). Bul. Coll. Agric. Utsunomiya Univ., 46: 1―79. the regeneration of a Fagus crenata forest on Kinkazan Is- 19）Murata, I., Saruki, S., Kubota, K., Inoue, S., Tashiro, N., land, northern Japan. Ecol. Res., 9: 115―120. Enoki, T., Utsumi, Y. and Inoue, S. (2009) Effects of sika 31）Tsujino, R. and Yumoto, T. (2004) Effects of sika deer on deer (Cervus nippon) and dwarf bamboo (Sasamorpha bore- tree seedlings in a warm temperate forest on Yakushima Is- alis) on seedling emergence and survival in cool-temperate land, Japan. Ecol. Res., 19: 291―300. mixed forests in the Kyushu Mountains. J. For. Res., 14: 32）Verdú, M. and Traveset, A. (2005) Early emergence en- 296―301. hances plant fitness: A phylogenetically controlled meta- 20）Ninomiya, S. and Furubayashi, K. (2003) Spatial sturucture analysis. Ecology, 86: 1385―1394. and regeneration of beech Forest affected by overbrowsing (Accepted Aug. 15, 2013) by sika deer (Cervus nippon) (in Japanese with English