Cascade effects of slope gradient on ground vegetation and small-rodent populations in a forest ecosystem


In this study, we set out to determine the cascade effects of slope gradient on ground vegetation and small-rodent populations in a forest ecosystem. We focused on two forest-dwelling small rodents with different habitat requirements, the striped field mouse Apodemus agrarius (preferring dense ground vegetation) and the Korean field mouse A. peninsulae (dense forest and woodland). The study area comprised natural deciduous forests and Japanese larch Larix kaempferi plantations in South Korea. The abundance of A. agrarius but not that of A. peninsulae was related to slope gradient. There was a negative effect of slope gradient on ground vegetation coverage and a positive effect of ground vegetation on A. agrarius populations. Our results highlight that the population of A. agrarius was indirectly influenced by the negative effects of slope gradient on ground vegetation. Slope gradient can, therefore, be a limiting factor in the microhabitats occupied by small rodents. This study reveals a critical role for slope gradient since it can modify not only microhabitat conditions, but also small-rodent populations, and this finding can contribute to improved microhabitat management.


Introduction
Forests form complex ecosystems, in which biotic assemblages and physicochemical environments interact (Pickett & Cadenasso, 2002). Within these ecosystems, the distribution and abundance of wildlife and vegetation are, therefore, closely linked to various and complex features of the climate and topography, promoting various effects through the ecosystem (Monzón et al., 2011;Eom et al., 2018). Factors related to topography, such as elevation and location, direction, gradient of a slope, directly or indirectly affect habitat characteristics, such as solar radiation, temperature, water, and nutrients (Aguilar et al., 2010;Kunkel et al., 2011). A steep slope gradient facilitates soil erosion in conditions of heavy precipitation, with consequent negative effects on fine materials and soil nutrients in forest ecosystems (El Kateb et al., 2013;Sepúlveda & Carrillo, 2015). This process in turn negatively affects vegetation growth through changes to the soil water-holding capacity and soil nutrient content, such as nitrogen and phosphorus (Kapolka & Dollhopf, 2001;Jiao et al., 2009;Lou et al., 2016;Ma et al., 2016). Therefore, the unfertilized condition of the soil on steep slopes makes it difficult for vegetation to grow or survive decreases in vegetation coverage with increasing slope gradient (Parsakhoo et al., 2009).
This effect of slope gradient on vegetation cover has implications for the suitability of habitats for wildlife, as vegetation provides both food and shelter (Krebs, 2009). Different species form distinct habitat preferences based on vegetation characteristics. Ground vegetation, for example, provides important resources for small rodents, such as shelter from predators and a diverse range of food (Kang et al., 2013a). It is, therefore, a key factor influencing small-rodent populations, and the abundance and species richness of their populations can be regulated by vegetation characteristics in a particular habitat.
Small rodents are ubiquitous components in forest ecosystems, where they are an important part of the food web (Orrock & Connolly, 2016). They are prey to terrestrial and flying predators and consumers of fungi, invertebrates, plants, and seeds (Sullivan & Sullivan, 2001). They are also seed dispersers, through either larderor scatter-hoarding (Campos et al., 2017). As a result, the species composition and abundance of plants in a particular habitat is influenced by the hoarding behaviors of small-rodent populations.
Various studies have considered the relationships between slope gradient and vegetation and between vegetation and small-rodent populations (Xu et al., 2008;Dong et al., 2014;Lee et al., 2019). However, there is a lack of studies focusing on the relationships between all three. Negative effects of soil erosion on vegetation succession have been reported in studies on the relationship between slope gradient and vegetation (Wang et al., 2015;Yuan et al., 2019). In contrast, other studies have reported a positive influence of ground vegetation on small-rodent populations in diverse environments because of its functions of providing food and cover resources (Coda et al., 2014;Crego et al., 2018;Lovera et al., 2019). In this study, we set out to determine the cascade effects of slope gradient on ground vegetation and small-rodent populations, given the possibility that slope gradient may negatively affect small-rodent populations indirectly because of its negative effects on ground vegetation.
We conducted three analyses to determine: (1) the characteristics of small-rodent populations according to their responses to habitat variables; (2) the effect of slope gradient on habitat variables preferred by small-rodent populations; and (3) the relationship between slope gradient and small-rodent populations. We used the analysis to test three hypotheses: (1) ground vegetation is a key factor for smallrodent populations; (2) slope gradient has a negative effect on ground vegetation; and (3) small-rodent populations are negatively influenced by slope gradient.

Experimental design and data collection
Four study plots were randomly established to survey slope gradient, habitat variables and small-rodent populations. Each study plot was 90 × 90 m (0.81 ha) and contained a total of 49 trapping points set up at intervals of 15 m using a grid (7 × 7).
Small-rodent data were collected using the capture-recapture method during three consecutive nights in each month from May to November, 2014. Sherman live traps (7.62 × 8.89 × 22.86 cm), baited with peanuts, were set up at all trapping points on each study plot. Each trap was checked every morning when traps were active. The following data were recorded for captured individuals: species and trap location. Each individual was given a specific identification (ID) using toe clipping, and this was also recorded. On the basis of the different species captured, we decided to focus on two: the striped field mouse Apodemus agrarius (Pallas 1771), and the field mouse A. peninsulae (Thomas 1907). After data recording was completed, captured individuals were immediately released at the capture point. We also recorded trapping points where no animals were captured, for input to the statistical analysis. The experimental protocols for the treatment and care of animals were reviewed and approved according to the guidelines of the local ethics committee (Institutional Animal Care and Use Committee, Chung-Ang University; approval number: CAU 2014-005).

Data analysis
We used whole data collected and merged from both natural deciduous forests and Japanese larch plantations for all statistical procedures. If small-rodent populations had a preference for stands, this might be a bias because small rodents could be affected by stands, and it might affect final results. Preferences of small rodents for stands, therefore, were evaluated using a Mann-Whitney U -test. As a result, there were no significant differences in the number of small rodents between natural deciduous forest and Japanese larch plantation (A. agrarius, Z = 0.856, p = 0.392; A. peninsulae, Z = −0.263, p = 0.793). The whole data set, therefore, was used for all statistical procedures.
To test the three relationships among slope gradient, ground vegetation, and small-rodent populations, we used a generalized linear mixed model (GLMM) and a Mann-Whitney U -test, after testing the normality of all variables using a Shapiro-Wilk test. Then, a Spearman rank sum test was performed to assess multicollinearity among all variables. In case of high correlation between two variables (r 0.7), the variable more highly correlated with the dependent variables or with more ecological meaning was selected (Carrilho et al., 2017). As a result, two variables were removed from the analysis: the percentage of bare area and number of downed trees.
Thereafter, we performed three GLMM procedures with the stands and sites as random factors (R package: lme4; Bates et al., 2015), as follows. First, we evaluated the responses of small-rodent populations to ground cover percentage. We also evaluated responses to the volume of downed trees, which has been reported to have positive effects on small-rodent populations (Kang et al., 2013b). Second, we analyzed the responses of ground vegetation to the slope gradient. In addition, we added the percentage coverage of each vertical layer (>1 m) and number of standing trees to this second model because ground vegetation growth can be inhibited by light deprivation because of vegetation extent higher in the canopy (Bolen & Robinson, 2003). Third, we assessed the relationship between slope gradient and small-rodent populations (table 1) again using a Mann-Whitney U test, by measuring the gradient at traps where animals were captured, and those where they were not (slope gradient in the 'capture trapping points' versus 'non-capture trapping points'). In each GLMM procedure, model selection was based on the corrected Akaike information criterion (AICc). We selected models according to AICc < 2, and then conducted a model averaging procedure (R package: MuMIn; Bartoń, 2016). All statistical analyses were performed using the R software program (R Core Team, 2017).

Results
In the  Two models explaining hypothesis 2 (the relationship between ground vegetation and slope gradient) were selected using the GLMM procedure (table 4). These models incorporated five variables: standardized slope gradient, coverage of understory, mid-story, overstory vegetation, and number of standing trees. Three of these had significant effects on the percentage of ground vegetation (table 5). Ground vegetation decreased with increasing slope gradient (coefficient = −0.0145, Z = 15.911, p < 0.001) and increasing understory vegetation (coefficient = −0.1189,  Abbreviations: VDT, volume of downed trees; %VEG, percentage of ground vegetation; %WD, percentage of woody debris. Z = 7.082, p < 0.001). In contrast, vegetation coverage in the overstory had a positive effect on percentage of ground vegetation (coefficient = 0.1007, Z = 6.999, p < 0.001).

Discussion
In this study, we demonstrated cascade effects caused by the direct influence of slope gradient on ground vegetation, which in turn affects A. agrarius populations. Our analysis showed that, firstly, A. agrarius preferred microhabitats with dense ground vegetation; secondly, the negative effect of slope gradient on ground vegetation; and, finally, slope gradient had a negative effect on A. agrarius abundance.
Small-rodent species utilize microhabitats for survival and reproduction based on requirements such as food, water, den, shelter, and space; these habitat requirements are species-specific (Mallegowda et al., 2015;Coda et al., 2016). Microhabitat occupancies may therefore differ among sympatric species. In this study, we captured individuals of two small-rodent species, A. agrarius and A. peninsulae, which use different microhabitats in the same area. In this study, it was shown that A. agrarius, but not A. peninsulae, preferred microhabitats with dense ground vegetation. This agrees with findings that A. agrarius mainly inhabits areas of dense ground vegetation in order to access seeds and new growth vegetation as food, whereas A. peninsulae prefers to access tree seeds and acorns in dense forests and woodland (Jo, 2015;Lee et al., 2018). Hence, dense ground vegetation provides both food and shelter for A. agrarius, while this vegetation does not provide A. peninsulae with its preferred food resources.
The growth and survival of plants are influenced by abiotic factors, such as soil nutrients, water, and sunlight (Richards & Coley, 2007;Zhu, 2016). In our study, we found that the percentage cover of ground vegetation was affected by slope gradient, and the under-and overstory vegetation. As Pimentel (2006) points out, soil erosion can negatively modify soil nutrients and water. Increases in slope gradient can accelerate soil erosion. Therefore, we concluded that slope gradient has a negative effect on ground vegetation abundance.
Competition among plant species for sunlight affects the structure and function of vegetation in forest ecosystems (Clark & Bullock, 2007). In our study, we determined that under-and overstory vegetation cover negatively and positively affected ground vegetation, respectively. A dense understory inhibits the growth of ground vegetation. Similarly, a dense overstory will have a negative impact on understory vegetation; this yields indirect positive influences on ground vegetation cover.
Apodemus agrarius preferred a microhabitat with dense ground vegetation. However, ground vegetation decreased as slope gradient increased, and A. agrarius avoided high slope gradients (i.e., steep slopes). This species was therefore indirectly influenced by increasing slope gradient. This finding enabled us to confirm the cascade effect of slope gradient on A. agrarius populations through its negative effect on ground vegetation.
In conclusion, ground vegetation, given its function in providing food and cover resources, had a significant influence on A. agrarius abundance. However, the percentage of ground vegetation decreased with increasing slope gradient because of soil erosion. Therefore, slope gradient had an indirect and negative effect on the A. agrarius population. These results suggest that gradient can be a limiting factor in the occupation of microhabitats by small rodents. Slope gradient, therefore, should be considered in conservation policies that aim to promote both ground vegetation and small-rodent populations.