Sakamoto, S.H., Suzuki, S.N., Koshimoto, C. et al. J For Res (2015) 20: 151. doi:10.1007/s10310-014-0447-0
Abstract
Intensive monitoring of multiple taxonomic groups is essential to understand ecosystem structure and function. In such studies, plant ecologists desire to monitor animal communities as efficiently as possible. We examined the efficiency of two trapping methods with different trap distances and total trapping areas for monitoring forest-floor rodent communities. We especially targeted two endemic, terrestrial, and semi-arboreal species dominating in forests throughout Japan. The long-distance (25 m) and the short-distance (10 m) methods assumed less than and more than one trap per female territory, respectively. Irrespective of species, capture efficiency tended to be higher in the long-distance than in the short-distance method. Recapture probabilities of two species were more similar in the short-distance than in the long-distance method in 2013. Observed densities for both species in the short-distance method were relatively similar to the estimated densities based on the long-distance method in 2011. In contrast, observed densities were lower than the estimated densities in the short-distance method in 2013. Observed densities of semi-arboreal species tended to be underestimated when only terrestrial traps were used. From these results, we propose that the short-distance method can efficiently estimate densities and community composition when researchers plan a small-effort study for a short time with a small area. In contrast, the long-distance method will allow researchers to estimate population densities robustly, when taking a large effort of a 3- to 4-day session with a large area. Our results also suggest that using both terrestrial and arboreal traps will improve the estimation.
Introduction
The structures and functioning of ecosystems have recently been widely investigated, but comprehensive understanding of ecosystems remains elusive. Long-term ecological research (LTER) has increased since the 1980s, investigating ecosystems over long time scales and broad spatial scales (Hobbie et al. 2003; Biodiversity Center of Japan, Ministry of the Environment 2012; Japan Long-Term Ecological Research Network 2012; US Long Term Ecological Research Network 2012). However, the intensive monitoring of multiple taxonomic groups within one site is relatively rare due to the difficulty of such monitoring. Plant ecologists who attempt long-term vegetation censuses often lack skills in the measurement of animal communities, and vice versa animal ecologists. Forest ecosystems are inevitably influenced by plant and animal communities (Dyer et al. 2010; Swanson et al. 2011; Fontaine and Kennedy 2012). A convenient method allowing non-zoologists to easily and accurately estimate animal communities with their vegetation censuses would be highly desirable for the monitoring of multiple taxonomic groups at one site.
Within a forest, small mammals utilize various types of plant resources, including fruits, seeds, tree cavities, roots, stems, and bark, for their food and nests (Hamilton 1941; Laurance 1994; Baxter and Hansson 2001; Holloway and Malcolm 2007). Moreover, they also indirectly affect plant communities by eating insects, burrowing in soils, transferring and burying seeds, and providing food for large predators that disturb forests (e.g., Criddle 1930; Vander Wall 1990; Focardi et al. 2000). These direct and indirect effects of small mammals are crucial in maintaining forest ecosystems and ecosystem functioning (Ostfeld and Canham 1993; Weltzin et al. 1997).
To estimate the species abundance and composition of small mammal communities, we rely on trapping methods (e.g., Dueser and Shugart 1978) because most small mammals are nocturnal or crepuscular and use holes in trees and/or underground, making direct observations of them generally difficult. Continuous improvements have been made in trapping techniques. Most studies of such improvements have focused on the “accuracy” of trapping in estimating the number of individuals (e.g., Efford 2004). However, “efficiency” of trapping (that is the balance of effort and outcome) becomes of greater concern to plant ecologists who attempt to monitor small mammal communities as conveniently as possible (Lambert et al. 2005).
In this study, we examine the efficiency of two trapping methods for the forest-floor rodent community. Throughout Japan, two endemic Apodemus species, the terrestrial A.speciosusand the semi-arboreal A. argenteus, dominate in forests (Miyao et al. 1963; Murakami 1974; Doi and Iwamoto 1982; Oka 1992). Both species often use the same habitats competitively (Sekijima 2004; Sakamoto et al. 2012) and strongly impact the fates of tree seeds (Soné and Kohno 1999; Shimada 2001; Abe et al. 2008; Takahashi and Shimada 2008; Tamura and Hayashi 2008; Mizuki and Takahashi 2009). We tested two trapping methods differing in trap distances and the total areas of trapping (hereafter defined as long- and short-distance methods). The long-distance method assumed that we placed less than one trap per female mouse territory. The short-distance method assumed that we placed more than one trap per female territory. Based on this assumption, we hypothesized that, when the individual density of mice is high and territories are saturated within the habitat, the long-distance method would not be able to capture all rodent individuals in the study area due to trap deficiency and would thus underestimate rodent densities, whereas the short-distance method could capture all individuals. Conversely, when individual density is low, both methods would return empty traps and the long-distance method could capture relatively more individuals. This scenario will differ between mice if intra-specific competition contributes to a species’ response to the two methods.
We compared whether and if so how the capture efficiency per trap, the recapture probability, the sex ratio of mice, and individual density differed between the two methods. We calculated the estimated population density per area using a Bayesian estimation based on the Peterson method of mark–recapture. The estimated population density was also compared with the densities observed using the two methods.
Materials and Methods
Study site
The study site was located at the Sugadaira Montane Research Center, University of Tsukuba (36°31′N, 138°21′E), Ueda-shi, Nagano Prefecture, Central Japan, on the Sugadaira plateau at an altitude of about 1,300 m. The annual mean temperature at the site was 6.5 °C and the average monthly temperature ranged from 19.4 °C in August to −5.6 °C in February. The mean annual rainfall was 1,226 mm and the annual mean of maximum snow depth was 102 cm for the years 1971–2006.
At the research center, different types of vegetation are maintained along a successional gradient (i.e., grassland, Japanese red pine forest, red pine, and broad-leaved mixed forest, and deciduous broad-leaved forest) (Sugadaira Montane Research Center, University of Tsukuba 2012). The dominant plant within the grassland is Japanese pampas grass (Miscanthus sinensis), which is typical of mountainous areas in Japan. Every autumn, the facilities’ managers remove all of the above-ground plant parts to prevent vegetative succession from grassland to forest. The Japanese red pine forest is dominated by Pinus densiflora, for which the above-ground removal ceased approximately 45 years ago. P. densiflora, Betula platyphylla var. japonica and Quercus crispula dominate the red pine and broad-leaved mixed forest, where above-ground removal has not been conducted for approximately 60 years. Dwarf bamboo dominates large areas of the understorey of the red pine forest and the mixed forest, while various tree saplings and shrubs, such as Schisandra chinensis and Rhus trichocarpa, occupy the understorey areas that lack dwarf bamboo. The deciduous broad-leaved forest is composed of B. platyphylla var. japonica, Q. crispula, and many additional tree species. The broad-leaved forest is located in a valley in which no above-ground removal has occurred. Dwarf bamboo dominates most areas of the understorey in this forest.
Study species and trapping procedures
Although we targeted all small rodents that can be captured by Sherman traps (SFA Folding Trap; 6 × 7 × 16 cm; HB Sherman Traps, Tallahassee, FL, USA), almost all individuals captured at our study site belonged to two Apodemus species. Therefore, we analyzed the population structures of these two species in this study. A.speciosus (AS) and A. argenteus(AA) often co-exist and dominate in forests throughout Japan. They have similar morphologies, but several traits show clear inter-specific differences: body weight, 20–60 g in AS and 10–20 g in AA; hind foot length, 22–28 mm in AS and 17–21 mm in AS; face shape, round for AS and relatively long and thin for AA; eyes, large and conspicuous for AS and small for AA (Nakata et al. 2009).
We evaluated the efficiency of two trapping methods with differing trap distances in August and October 2011 and October 2013: the long- and short-distance methods. Trapping points were placed 25 m apart on the x- and y-axes in the long-distance method and 10 m apart on the x- and y-axes in the short-distance method. Females of these species are more site-tenacious than males and territorial (e.g., Oka 1992). The average home range size of the female of both species was from 24 to 26 m radius (Oka 1992), and territory generally was smaller than the home range. Hence, the long-distance method assumed that we placed less than one trap per female territory. In contrast, the short-distance method assumed that we placed more than one trap per female territory. The two trapping methods in each census were conducted in two vegetation types (red pine forest and red pine and broad-leaved mixed forest).
We conducted 4-day trapping sessions from 10 to 13 August and 26 to 29 October 2011. For the long-distance method, three plots were established: one plot with a 5 × 3 grid within a 100 × 50 m area in a red pine forest (15 trap stations) and two plots with a 3 × 3 grid within a 50 × 50 m area in red pine and broad-leaved mixed forests (9 trap stations at each plot). For the short-distance method, 36 and 36 trap stations (a 6 × 6 grid within a 50 × 50 m area) were established in the two vegetation types, respectively. In both censuses in 2011, we conducted the long-distance method during the first 3 days of each census period and the short-distance method during the last day of each census period. In the long-distance method, we set one trap on the ground and one at a height of 1 m on a tree at each trap station. In the short-distance method, to minimize trapping effort, we set one trap on the ground at each trap station. The number of traps was maintained during each census. Therefore, the total number of trap-nights was 396 trap-nights in the long-distance method (66 traps × 3 days × 2 census period) and 144 trap-nights in the short-distance method (72 traps × 1 day × 2 census period).
From 8 to 11 October 2013, we conducted a more comprehensive census with a 4-day trapping session in the two trapping methods. For both trapping methods, one plot was established in a red pine forest and one in red pine and broad-leaved mixed forests (in total, four plots). For the long-distance method, each plot had a 5 × 5 grid within a 100 × 100 m area (25 trap stations at each plot). For the short-distance method, each plot had a 5 × 5 grid within a 40 × 40 m area (25 trap stations at each plot). We set one trap on the ground and one at a height of 1 m on a tree at each trap station in both methods. The traps that captured mice were replaced on successive days during each session, and thus the number of traps was maintained in each trapping day. Therefore, the total number of trap-nights in 2013 was 800 trap-nights (50 ground traps and 50 tree traps × 4 days × 2 methods).
Each trap was wrapped with a polyethylene film to stabilize the internal temperature and to exclude rain. The trap was baited with four peanuts and a small piece of raw sweet potato. To reduce stress in the captured mice, the traps were opened at 1500 hours and checked the next morning from 0500 to 0800 hours. The species, sex, reproductive stage, weight, and hind-foot length of the mice, as well as the trap location, were recorded before release. Moreover, we marked all mice by cutting their hair in different body regions to identify individuals.
Analysis
Capture efficiency was represented as the number of captured individuals per trap-night. Based on the multiple capture–recapture data, the total population size n and recapture probability α (=the number of captured individuals/n) were estimated by the Lincoln–Peterson model (Petersen 1896; Lincoln 1930) expanded by a Bayesian method as follows:
Ct,p∼Binom(Np,αm)
Rt,p∼Binom(Mt,p,αm)
where Ct and Rt, p are the total number of individuals caught and the number of individuals already marked (recaptures) when caught, respectively, in sample t in plot p; Mt,p is the number of marked individuals in the population before the tth sample is taken in plot p. Y ~ Binom (X,p) means that the number of observations Y follows a binomial distribution with parameters (number of trials X, probability of success on each trial p). The Np is a latent variable for the number of individuals in plot p, and assumed to following a poisson distribution, Pois(ApD), where Ap is the area of plot p and D is the density of individuals; the prior of D is an inverse Gamma distribution, InvGamma (0.01,0.01). Prior distributions for αm, recapture probability for method m (long- or short-distance method), is a beta distribution, β(1, 1).
For 2013, individual density D was estimated by using only data from each of the two methods, and by using data from both of them.
Posterior distributions of the parameters were inferred by Markov Chain Monte Carlo (MCMC) methods using the WinBUGS (Lunn et al. 2000) via R2WinBUGS package (Sturtz et al. 2005) of R v.2.11.1 (R Development Core Team 2010). The convergence of three independent chains of 4000 iterations after a burn-in of 1,000 iterations, thinned to every fourth iterations, was assessed via the Gelman–Rubin’s diagnostic to ensure satisfactory sampling of the posterior distributions (Gelman and Rubin 1992).
Results
For terrestrial traps, both of the long-distance and short-distance methods captured more individuals of AS than AA throughout our censuses except in the case of the short-distance method in October 2013 (Table 1). Sex ratios of AS showed similar trends; however, those of AA differed between the two methods. Sex ratios of AS were biased toward males in August 2011 but toward females in October of both years. Sex ratios of AA were relatively biased toward males in the long-distance method whereas they were toward females in the short-distance method in August and October 2011 (Table 1). Arboreal traps captured more individuals of AA than AS in the long-distance method in August and October 2011 and in the short-distance method in October 2013. However, arboreal traps captured no individuals in the long-distance method in October 2013. Two individuals of Eothenomys smithii were captured by terrestrial traps in the short-distance method in August 2011, and seven and one Urotrichus talpoides were captured in the long- and short-distance methods, respectively, in October 2013.
Table 1
Number of individuals of Apodemus speciosus (AS) and A. argenteus (AA) captured during three censuses by the two methods
Long-distance method (25-m intervals)
Short-distance method (10-m intervals)
Day 1
Day 2
Day 3
Day 4
Total
Sex ratio
Day 1
Day 2
Day 3
Day 4
Total
Sex ratio
August 2011
33 terrestrial and 33 arboreal traps in 1.32 ha
72 terrestrial traps in 0.72 ha
AS
Terrestrial
19
12 (6)
15 (8)
–
46 (14)
0.63
–
–
–
23
23
0.65
Arboreal
0
0
1
–
1 (0)
0
–
–
–
–
–
AA
Terrestrial
2
2 (0)
2 (0)
–
6 (0)
0.67
–
–
–
3
3
0.33
Arboreal
1
1 (1)
2 (2)
–
4 (2)
0
–
–
–
–
–
October 2011
33 terrestrial and 33 arboreal traps in 1.32 ha
72 terrestrial traps in 0.72 ha
AS
Terrestrial
6
9 (5)
8 (3)
–
23 (8)
0.21
–
–
–
14
14
0.43
Arboreal
0
1
0
–
1 (0)
0
–
–
–
–
–
AA
Terrestrial
3
4 (0)
4 (3)
–
11 (3)
0.50
–
–
–
4
4
0.25
Arboreal
0
2 (0)
1 (0)
–
3 (0)
0.67
–
–
–
–
–
October 2013
50 terrestrial and 50 arboreal traps in 2.42 ha
50 terrestrial and 50 arboreal traps in 0.5 ha
AS
Terrestrial
3
8 (2)
12 (6)
11 (6)
34 (14)
0.35
1
1 (0)
1 (0)
7 (2)
10 (2)
0.43
Arboreal
0
0
0
0
0
–
0
0
0
0
0
–
AA
Terrestrial
1
0
0
2 (0)
3 (0)
0.67
1
1 (1)
3 (1)
7 (1)
12 (3)
0.67
Arboreal
0
0
0
0
0
–
0
1
0
1 (0)
2 (0)
1
Number of recaptured individuals are shown in parentheses. Total number of individuals includes all individuals (new and recaptured) in each method in each census
The number of captured individuals per trap-night (=capture efficiency) tended to be higher for AS than AA throughout the three censuses (Fig. 1). Capture efficiency estimated by terrestrial traps for AS was highest on day 1 in August 2011, and capture efficiency for both species was higher in the long-distance method than in the short-distance method in most censuses, except in the case for AA in October 2013 (Fig. 1). Capture efficiency for both species estimated in the short-distance method sharply increased at day 4 in October 2013 (Fig. 1).
Fig. 1
Capture efficiency (no. of captured individuals/trap-night) of Apodemus speciosus (AS) and A. argenteus (AA) in each census. Circle terrestrial traps; triangle all traps (terrestrial + arboreal); closed symbol long-distance method; open symbol short-distance method
For all Bayesian models, the potential scale reduction factor of Gelman-Rubin’s diagnostic was <1.1 for all parameters, indicating MCMC chains were successfully converged. The recapture probabilities by terrestrial traps and all (terrestrial and arboreal) traps for AS were equivalent in August and October 2011, and for AA in October 2011 and 2013 (Fig. 2). The recapture probability for AS based on terrestrial traps tended to be higher in October 2011 and 2013 than in August 2011, and in the long-distance method than in the short-distance method in October 2013 (Fig. 2). In contrast, the recapture probability for AA based on all traps tended to be higher than that based on terrestrial traps in August 2011 (Fig. 2).
Fig. 2
Recapture probability of A. speciosus (AS) and A. argenteus (AA) in each census. Circle terrestrial traps; triangles all traps (terrestrial + arboreal); closed symbol long-distance method; open triangle short-distance method. Bold and thin vertical bars indicate 50 and 95 % credible intervals, respectively
With increasing census days, the observed densities (=n/ha) in the long-distance method were gradually close to the estimated densities in the long-distance method (Fig. 3). In 2011, observed densities in the short-distance method were relatively similar to the estimated densities in the long-distance method despite the 1-day trapping session, irrespective of species and census month. In contrast, the 4-day trapping session in October 2013 revealed large credible intervals of the estimated densities for both species in the short-distance method, which were caused by low recapture probabilities (Fig. 2) and a rapid increase in observed densities at day 4 (Fig. 3). The observed densities of both species in the short-distance method were generally lower than the estimated densities in the short-distance method, while similar to the estimated densities calculated using data from both methods (Fig. 3). There was no large difference between the observed densities by terrestrial and by all traps (Fig. 3).
Fig. 3
Observed densities of A. speciosus (AS) and A. argenteus (AA) based on the cumulative number of individuals captured during 3- or 4-day censuses and estimated densities of those calculated using the capture–recapture model in each census. Circlewithsolid line observed density by only terrestrial traps; trianglewithdotted line by all traps; open symbol the short-distance method; closed symbol the long-distance method; gray symbol estimated density by using data of both methods. Bold and thin vertical bars indicate 50 and 95 % credible intervals, respectively. Some of the upper limits of the credible intervals were outside the plot region; 65.2 and 60.6 for AS by all and terrestrial traps, respectively, in August 2011; 133.2 for AS in the short-distance method with terrestrial traps in October 2013; 90.6 and 124.8 for AA in the short-distance method with all and terrestrial traps, respectively, and 66.5 and 100.5 for AA by both of the methods with all and terrestrial traps, respectively, in October 2013
Discussion
Our results demonstrated that the short-distance method showed higher efficiency for estimating population densities compared to the long-distance method, especially when researchers planned a small-effort study for a short time with a small area. By comparing day 1 of both methods, the densities observed in the short-distance method tended to be higher than those in the long-distance method, while they were more similar to the estimated densities based on the long-distance method by using the Bayes estimation based on the Peterson method (Fig. 3). In addition, recapture probabilities for both species were more similar in the short-distance method than the long-distance method (Fig. 2). Moreover, the short-distance method also captured a non-dominant rodent species, Eothenomys smithii.Furthermore, the short-distance method required the researchers to lower trapping efforts with smaller trapping areas. However, trapping in small areas can be greatly affected by short- distance immigration and emigration of each individual. This can lead to large fluctuations in capture efficiency (the number of captures per trap) among days (Fig. 1) and a low recapture probability (Fig. 2), and result in overestimation of the estimated densities with large credible intervals in the short-distance method. From these results, we suggest that the short-distance method is more efficient for estimating, without any mark–recapture information, population densities and community composition of forest-floor small rodents when conducting a small effort session by 1 day and a small area, but simultaneously this method risks leading to large fluctuations in observed population density, especially when the trapping area is too small.
In contrast, the long-distance method showed a higher capture efficiency by terrestrial traps than that in the short-distance method. This pattern was because the long-distance method generally captured a higher number of new and marked individuals of the dominant species, A.speciosus, per trap than the short-distance method (Table 1). This suggests that the long-distance method produces a more conservative and robust estimation for the density of dominant species if researchers can trap for consecutive days covering a broad study area. The long-distance method can investigate the study area containing variable habitats, and hence it will allow researchers to obtain small mammals other than rodents, such as Urotrichus talpoides. However, a large difference between the estimated and observed density of A.speciosus was observed in August 2011 when its population density was highest, suggesting that the method seems not to be suitable when the population density is high because the number of traps is insufficient.
Studies intentionally evaluating the efficiency of trapping methods for estimating small mammal density, especially those including trap distance, are very rare. Most previous studies have empirically applied trapping at 10-m intervals using a single type of trap without verification of the trap distances according to the focus animal. In other cases, studies have tested the effects of trap types (e.g., Wiener and Smith 1972; Williams and Braun 1983; O’Farrell et al. 1994; Lambert et al. 2005), heights (Lambert et al. 2005), grid sizes (e.g., Steele et al. 1984), or the spatial arrangements of traps (e.g., Weihong et al. 1999; Pearson and Ruggiero 2003) on trapping outcome. Weihong et al. (1999) examined two spatial arrangements of traps and found that trap grids appeared to be better than trap lines for detecting the presence/absence of rodent species when two species co-exist and one appears subordinate to the other. However, they placed traps at 25-m intervals and did not examine the effects of trap distance. The present study evaluated the efficiency of the 10- and 25-m interval methods and clarified the differences in costs and benefits between the two interval methods.
A. speciosus usually inhabits terrestrial environments, whereas a large part of the activity of A. argenteus occurs in trees, especially in summer (Sekijima 1997; Sakamoto et al. 2012). Therefore, censusing without arboreal traps, as we did for short-distance method in 2011, can lead to an underestimation of A. argenteus density; the number of captures was ca. 60 and ca. 20 % larger by all traps than only by terrestrial traps in August (long-distance methods in 2011) and October (long-distance method in 2011 and short-distance method in 2013), respectively (Table 1).
It is unclear why the sex ratio of A. argenteus on the ground was relatively biased toward males in the long-distance method but biased toward females in the short-distance method in the two censuses in 2011 (Table 1). One possibility is that a territorial female has small home range and is site-tenacious, whereas a non-territorial male has a wider range in both species (Oka 1992). The trap areas used in the long-distance method might have included areas that were unsuitable for females to form their territories. Territory acquisition of female A. argenteus would be influenced by the presence of competitive superior A.speciousfemales. In contrast, adult and sub-adult males move across wider areas compared to females (Oka 1992). In this study, both methods in 2011 were conducted within areas in which the two Apodemus species co-occurred, based on the results of our previous study (Sakamoto et al. 2012), and thus the long-distance method that contained less than one trap per female territory could not capture many females of the competitive inferior species, while the short-distance method that contained more than one trap per female territory might enable us to capture many females whose range size was smaller than that of the males. In 2013, plots were placed in new areas where it was unknown whether the two Apodemusspecies co-occurred. Consequently, population densities of both species were relatively low, and the bias in sex ratio might be relaxed in 2013, which resulted in both methods obtaining similar sex ratios.
This study has several limitations. In 2011, our study design used the long-distance method in the first 3 days and then switched to the short-distance method on the subsequent fourth day, producing potential problems because some individuals might become attracted to the free food available and be captured more easily with the progression of days during a census. However, the increase in capture efficiency with the progression of days was only observed in the short-distance method in 2013 but not in the long-distance method. In addition, the capture efficiency on the ground did not increase but fluctuated: 58, 36, 45, and 32 % in August 2011, and 18, 27, 24, and 19 % in October 2011 on days 1, 2, 3, and 4 of each census, respectively (Table 1). We also conducted only one 1-day census of the short-distance method, on 29 September 2012, for which the capture rate on ground was 40 % (data not shown). These results suggest that, even when a stand-alone, 1-day short-distance trapping session is conducted, an adequate number of rodent individuals can be captured. An additional limitation is the absence of sufficient replications from a number of study sites. Our previous study over 2 years (Sakamoto et al. 2012) demonstrated that seasonal and spatial distributions of two Apodemus species were generally consistent between years in our study site, suggesting the results of this study being robust. Our comprehensive census in 2013 that conducted two distance methods with identical numbers of traps in independent areas demonstrated more robust results in which there were no controversial results among species and sex.
Our results suggest that the short-distance method, placing more than one trap per female territory, can be useful for non-expert researchers, such as plant and insect ecologists, when they plan a small effort session to estimate, easily and efficiently, the abundance and composition of the small mammal community. In contrast, the long-distance method has a benefit for detecting a rodent community when rodents distribute within restricted areas within study sites. Although the long-distance method requires researchers to use a sufficient number of traps, study areas, and study periods, the method is available for expert researchers, such as rodent ecologists, when they very conservatively and robustly estimate population demography, especially for dominant species. For both methods, it will be preferable to add arboreal traps when we focus on communities including semi-arboreal species.
Acknowledgements
We acknowledge the Sugadaira Montane Research Center, University of Tsukuba, for permission to conduct research within the site. We also thank Takahiro Ogai, Shigekazu Tomizuka, Kouji Nagaoka, and Mariko Katsuyama for their research assistance and a member of SMRC for valuable advice on the field research. This work was supported by Grant-in-Aid for Challenging Exploratory Research from JSPS (Grant Numbers 24657018 to Shinsuke H. Sakamoto; 23650236 to Chihiro Koshimoto), and by grants from the University of Miyazaki (Support Program for Integrated Research Project for Human and Veterinary Medicine).
References
Abe M, Honda A, Hoshizaki K, Miguchi H (2008) Advantage of early seedling emergence in Fagus crenata: importance of cotyledon stage for predator escape and pathogen avoidance. Ecol Res 23:681–688CrossRef
Baxter R, Hansson L (2001) Bark consumption by small rodents in northern and southern hemispheres. Mamm Rev 31:47–59CrossRef
Criddle S (1930) The prairie pocket gopher, Thomomys talpoides rufescens. J Mamm 11:265–280CrossRef
Doi T, Iwamoto T (1982) Local distribution of two species of Apodemus in Kyushu. Res Pop Ecol 24:110–122CrossRef
Dueser RD, Shugart HH Jr (1978) Microhabitats in a forest-floor small mammal fauna. Ecology 59:89–98CrossRef
Dyer LA, Letourneau DK, Chavarria GV, Amoretti DS (2010) Herbivores on a dominant understory shrub increase local plant diversity in rain forest communities. Ecology 91:3707–3718PubMedCrossRef
Efford M (2004) Density estimation in live-trapping studies. Oikos 106:598–610CrossRef
Focardi S, Capizzi D, Monetti D (2000) Competition for acorns among wild boar (Sus scrofa) and small mammals in a Mediterranean woodland. J Zool 250:329–334CrossRef
Fontaine JB, Kennedy PL (2012) Meta-analysis of avian and small-mammal response to fire severity and fire surrogate treatments in U.S. fire-prone forests. Ecol Appl 22:1547–1561PubMedCrossRef
Gelman A, Rubin D (1992) Inference from iterative simulation using multiple sequences. Stat Sci 7:457–511CrossRef
Hamilton WJ Jr (1941) The food of small forest mammals in eastern United States. J Mamm 22:250–263CrossRef
Hobbie JE, Carpenter SR, Grimm NB, Gosz JR, Seastedt TR (2003) The US long term ecological research program. BioSci 53:21–32CrossRef
Holloway G, Malcolm JR (2007) Nest trees used by northern and southern flying squirrels in central Ontario. J Mamm 88:226–233CrossRef
Lambert TD, Malcolm JR, Zimmerman BL (2005) Variation in small mammal species richness by trap height and trap type in southeastern Amazonia. J Mamm 86:982–990CrossRef
Laurance WF (1994) Rainforest fragmentation and the structure of small mammal communities in tropical Queensland. Biol Conserv 69:23–32CrossRef
Lincoln FC (1930) Calculating waterfowl abundance on the basis of banding returns. USDA Circ 118:1–4
Lunn DJ, Thomas A, Best N, Spiegelhalter D (2000) WinBUGS—a Bayesian modelling framework: concepts, structure, and extensibility. Stat Comp 10:325–337CrossRef
Miyao T, Morozumi W, Morozumi M, Hanamura H, Sato N, Akahane H, Sakai A (1963) Small mammals on Mt Yatsugatake. I. Small mammals in the subalpine forest zone on Mt Yatsugatake. Zool Mag 72:133–138 (in Japanese with English summary)
Mizuki I, Takahashi A (2009) Secondary dispersal of Dioscorea japonica (Dioscoreaceae) bulbils by rodents. J For Res 14:95–100CrossRef
Murakami O (1974) Growth and development of the Japanese wood mouse (Apodemus speciosus). I. The breeding season in the field. Jpn J Ecol 24:194–206 (in Japanese with English summary)
Nakata T, Saitoh T, Iwasa MA (2009) Apodemus speciosus, Apodemus argenteus. In: Ohdachi SD, Ishibashi Y, Iwasa MA, Saitoh T (eds) The wild mammals of Japan. Mammalogical Society of Japan, Kyoto, pp 169–173
O’Farrell MJ, Clark WA, Emmerson FH, Juarez SM, Kay FR, O’Farrell TM, Goodlett TY (1994) Use of a mesh live trap for small mammals: are results from Sherman live traps deceptive? J Mamm 75:692–699CrossRef
Oka T (1992) Home range and mating system of two sympatric field mouse species, Apodemus speciosus and Apodemus argenteus. Ecol Res 7:163–169CrossRef
Ostfeld RS, Canham CD (1993) Effects of meadow vole population density on tree seedling survival in old fields. Ecology 74:1792–1801CrossRef
Pearson DE, Ruggiero LF (2003) Transect versus grid trapping arrangements for sampling small-mammal communities. Wildl Soc Bull 31:454–459
Petersen CGJ (1896) The yearly immigration of young plaice into the Limfjord from the German Sea. Rep Dan Biol Station 6:1–48
Sakamoto SH, Suzuki SN, Degawa Y, Koshimoto C, Suzuki RO (2012) Seasonal habitat partitioning between sympatric terrestrial and semi-arboreal Japanese wood mice, Apodemus speciosus and A. argenteus in spatially heterogeneous environment. Mamm Stud 37:261–272CrossRef
R Development Core Team (2010) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. http://www.R-project.org
Sekijima T (1997) Evaluation of vertical habitat use in Apodemus argenteus and A. speciosus, by the footprint recording method. Jpn J Ecol 47:151–158 (in Japanese with English abstract)
Sekijima T (2004) Does interspecific competition affect the vertical habitat segregation of Apodemus argenteus and Apodemus speciosus? Experimental evaluation by removal and food supplementation. Mamm Stud 29:97–104CrossRef
Shimada T (2001) Hoarding behaviors of two wood mouse species: different preference for acorns of two Fagaceae species. Ecol Res 47:127–133CrossRef
Soné K, Kohno A (1999) Acorn hoarding by the field mouse, Apodemus speciosus Temminck (Rodentia: Muridae). J For Res 4:167–175CrossRef
Steele BB, Bayn RL Jr, Grant CV (1984) Environmental monitoring using populations of birds and small mammals: analyses of sampling effort. Biol Conserv 30:157–172CrossRef
Sturtz S, Ligges U, Gelman A (2005) R2WinBUGS: a package for running WinBUGS from R. J Stat Soft 12:1–16
Swanson ME, Franklin JF, Beschta RL, Crisafulli CM, DellaSala DA, Hutto RL, Lindenmayer DB, Swanson FJ (2011) The forgotten stage of forest succession: early-successional ecosystems on forest sites. Front Ecol Env 9:117–125CrossRef
Takahashi A, Shimada T (2008) Selective consumption of acorns by the Japanese wood mouse according to tannin content: a behavioral countermeasure against plant secondary metabolites. Ecol Res 23:1033–1038CrossRef
Tamura N, Hayashi F (2008) Geographic variation in walnut seed size correlates with hoarding behaviour of two rodent species. Ecol Res 23:607–614CrossRef
Vander Wall SB (1990) Food hoarding in animals. University of Chicago Press, Chicago
Weihong J, Veitch CR, Craig JL (1999) An evaluation of the efficiency of rodent trapping methods: the effect of trap arrangement, cover type, and bait. NZ J Ecol 23:45–51
Weltzin JF, Archer S, Heitschmidt RK (1997) Small-mammal regulation of vegetation structure in a temperate savannah. Ecology 78:751–763CrossRef
Wiener JG, Smith MH (1972) Relative efficiencies of four small mammal traps. J Mamm 53:868–873CrossRef
Williams DF, Braun SE (1983) Comparison of pitfall and conventional traps for sampling small mammal populations. J Wildl Manag 47:841–845CrossRef
No comments:
Post a Comment