Microhabitat selection in relation to edge
Do bush-crickets avoid the habitat edge? Microhabitat selection in the wart-biter (Decticus verrucivorus)
Dag Ø. Hjermann
Department of Biology, Division of Zoology, PO Box 1050 Blindern,
N-0316 Oslo, Norway
Abstract
1. During a 3-4 week period, I recorded the movements of individually marked adult wart-biters (Decticus verrucivorus (L.), Orthoptera: Tettigoniidae) in two small habitat "islands" lying in cereal fields in SE Norway.
2. Wart-biters of both sexes used a quite wide range of relatively short vegetation, only avoiding bare ground and vegetation with height 30-80 cm and taller.
3. Microhabitats lying 1-2 m from the habitat edge were used less than expected, and microhabitats less than 1 m from the edge were clearly avoided. In the largest habitat island, this edge avoidance was independent of vegetation type; in the small patch, edge avoidance seemed to differ between vegetation types. During the study, wart-biters in the smallest habitat island tended to move out of the island center and into the 1-2 m zone. The sexes did not differ with statistic significance in their habitat use.
4. Wart-biters were also marked and tracked in two road verge patches. Here, both sexes preferred microhabitats close to the edges.
Introduction
The wart-biter Decticus verrucivorus (Linneus 1758) is a large, practically flightless bush-cricket living in sun-exposed grasslands with low vegetation. Because habitat destruction and changing practices in animal husbandry (especially increasing use of fertilizer), the species has declined in many parts of Northern Europe the last decades (Ingrisch, 1979; Holst, 1986). In England, it is on the verge of extinction (Cherrill & Brown, 1990a, 1990b). In Norway, it is typically found in small populations in small "islands" of remnant habitat, in addition to some roadside verges.
In this paper, I analyse this species' microhabitat choice within small habitat islands. I especially emphasize microhabitat selection relative to the position of the microhabitat patch within the habitat island, testing whether these bush-crickets select microhabitat patches within these islands only according to the vegetation of the microhabitat, or whether the position of the microhabitat patch (its distance to the island edge) also influences microhabitat selection. I focus on this aspect because the smaller the habitat islands, the larger is the potential influence of edge-effects. E.g., in the study area (see below), the median area of habitat islands is 213 m2, which means that about 43 % of the typical habitat island is < 2 m from the habitat edge, assuming that it is circular. The analysis is performed on two temporal scales in order to obtain a fuller picture of habitat selection (Johnson, 1980).
Study system and field work
In a study of a 2 x 3 km agricultural area in Frogn, southern Norway, Hjermann & Ims (1996) found that its distribution among 70 habitat islands was largely determined by the area and isolation of the habitat islands, as expected from a metapopulation model. The present study was conducted from July 23 toAugust 26, 1993, in two of those habitat islands (the species and the study area is described in Hjermann and Ims, 1996). Both were islets of semi-natural pasture on shallow soil, surrounded by cereal crops; their vegetation was fairly similar. Their areas were 914 m2 and 390 m2, and they will be referred to as the large and the small habitat island, respectively. The habitat islands were divided into a 0.25x0.25m grid using a network of sticks and strings, and their vegetation were mapped by classifying each cell in the grid according to six vegetation classes based on vegetation height and structure (Tab. 1).
Adult wart-biters were captured by hand and marked by gluing one end of a red 80 x 4-mm plastic tape to the top of the pronotum. To avoid “panic movements” after marking, the insects were placed in an open-bottomed wire cage (ca 15 x 15 cm) on the site of capture for the rest of the day and the following night, and released by removing the top of the cage early (ca 8 am) the following morning. Subsequent “recaptures” were done by reading numbers written on the tapes, which usually could be seen from a distance of 1-2 m. Thus, disturbance of the insect’s behaviour was minimized. The weight of the plastic tape was 30-35 mg or 1-2 % of the weight of the insect, and did not seem to inhibit movement, stridulation, mating or egg-laying. For each recapture, the position of the insect relative to the 0.25 x 0.25 m grid was recorded. Surveys were done with one- to two-hour intervals (half-hour intervals in the large patch during the first week). Not all wart-biters were recovered at each survey, and longer intervals (up to 24 hours) were used in rainy or cold weather, when the wart-biters moved very slowly or hardly at all.
In addition to the two islands of seminatural pasture, I also marked and tracked wart-biters in two 2 m wide road verges, lying between a tarred road and a cereal field. Most of the marked bush-crickets lived in a 40 m long (80 m2) section of one of the road verges. The wart-biter positions were recorded to a 0.25 x 0.25 m grid, but the vegetation was not mapped because it was spatially homogenous and periodically cut, varying much more in time than in space. The road verge bush-crickets were analysed as a single group.
Analysis of microhabitat selection
The habitat selection of the wart-biters in the habitat islands were analysed using three methods, differing in scale and assuptions. The first and simplest analysis (hereafter referred to as "Analysis 1") was done by classifying the grid cells of the large and small habitat islands in 18 (6 x 3) categories: according to the six vegetation categories, and according to three edge proximity (EP) classes based on the distance to nearest edge between habitat and non-habitat: 0 1 m, 12 m, and >2 m. For each individual animal, a selection indices Si (i=1..18) for each of the 18 categories were calculated using the following formula:
where Ui is the number of times an animal visited a habitat category and Ai is the area of that habitat category. Thus, the indices sum to 1 for each animal.
In Analysis 2 and 3, I quantified the selection indices of each EP category by assuming that the effects of vegetation and EP on selection were independent. I used four EP categories instead of three by distinguishing between 2-4 m and > 4 m. The two sets of selection indices (vj for vegetation type j=1 to 6, ek for EP category k=1 to 4) were estimated simultaneously by finding (using Proc Nonlin, SAS 6.12) the index values that minimize the deviance = 2 log likelihood. Assume that an individual has been observed making D movements (habitat choices), and that its choice of habitat by the i’th movement was vegetation type j’ and EP category k’. Let Ajki denote the availability of the combination of vegetation type j and EP category k when this choice was made. Modifying Eq. 3 in Arthur et al. (1996), the deviance (2LL) was calculated from the following formulae:
, where(Eq. 2)
The difference between Analysis 2 and 3 lies in the availabilities Ajki, which again reflect differences in the time-scale of habitat selection. In Analysis 2 (as in Analysis 1), the entire patch was assumed to be available: Ajki was set equal to the proportion of the patch area covered by the combination of vegetation type j and EP category k. Thus, Ajki was equal for all observations i. In Analysis 3, I assumed that the animals made a choice from the parts of the habitat relatively close to its position by the previous observation, rather than from the entire habitat. The availabilities therefore vary between observations (Arthur et al., 1996). The goal of this method was to analyse habitat selection on a smaller temporal and spatial scale and to reduce interdependence between subsequent observations. For instance, one can assume that the entire area within a certain radius of the previous position is available to the animal (Arthur et al. 1996). Here, I chose to use the observed distribution of movement lengths (r) between subsequent observations to estimate an availability function which describes the availability of a given grid square, depending on its distance from the previous observation (Hjermann, 2000). The availability function is based on the assumption that log(r2) is normally distributed and linearly dependent on the logarithm of the time interval between observations and on biotic and environmental variables that may influence movement speed (see Hjermann, 2000, for more details about this method). The variables considered here were Hour (hour of the day), Hour2, Sunny (weather type: sunny or cloudy), Sex, Ind (one category for every individual insect), and vegetation and edge distance category before movement. The analyses were done for each patch separately, using Proc GLM (SAS Institute 1996) with Mallow’s Cp (Kleinbaum et al., 1988, p. 420) as a criterion to find the best set of variables. The resulting availability function was then combined with the habitat maps to find the availabilities for each EP-vegetation combination prior to each choice of habitat. Note that in contrast with Analysis 1 and 2, areas outside the habitat island were also considered available (for these areas, vegetation type was set to 6 and ek in Eq. 2 was set to 1).
The result of Analysis 2 and 3 were a set of vegetation and EP selection indices for each individual. The EP selection indices were used as dependent variables in a MANOVA
Figure 1. Movement of the wart-biters in the large (this page) and small (nex page) habitat islands throughout the study period, shown separately for females (top) and males (bottom). Individual animals are distiguished by use of different symbols. The x- and y-axes show spatial extent in meters.
(using Pillai's Trace) to look for differences between sexes and between the two habitat islands.
The road verges were not classified with respect to vegetation, so wart-biter habitat were classified only with respect to distance from the road in 0.5 m intervals, which yields four within-habitat categories (since the road verge was 2 m broad) and two outside-habitat categories (the road and the field). Thus, I performed only the following analyses on the road verge animals: Analysis 1 (with six categories instead of 18) and the equivalent of Analysis 3, replacing Eq. 2 with the unmodified Eq. 3 of Arthur et al. (1996).
Results
Altogether, 41 bush-crickets were successfully marked and "recaptured" (spotted after marking) at least once (for background statistics, see Table 2). Movement varied quite much between individuals and during the study period; while it happened that a single animal moved across most of the island in the course of a few hours, some of the wart-biters could stay within a couple of square meters for several days (Fig. 1). For the analysis of habitat use, only those animals "recaptured" at least ten times were used (13 animals in the large, 7 in the small habitat island, and 7 in the road verge). Analysis 1 (Fig. 2) shows that in both seminatural habitat islands, a fairly wide range of vegetation types (types 2, 3 and 4) were used about equally much. In the large habitat island, habitat use is lowest close to the island edge and highest >2 meters from the edge for almost all vegetation types (type 5 being an exception). In the small island, the pattern is more complex: the center of the island is preferred in the case of vegetation type 4, while for vegetation types 2 and 3, the areas between 1 and 2 m from the edge is most used.
Analysis 2 quantifies the amount of habitat use relative to edge proximity (EP). In a MANOVA using the EP selection indices as dependent variables, EP selection was found not to differ statistically between neither sexes nor vegetation islands (sex: P=0.71, island: P=0.81, interaction: P=0.39). On both islands, EP does not seem to influence habitat use when EP is more than 2 m, but use declines drastically when EP becomes less than 2 m (Fig. 3a). Note, however, that the effect of assuming that vegetation and EP does not interact is to "average out" a seemingly high interaction between vegetation and EP in the case of the small habitat island.
In Analysis 3, habitat preferences within the study period (rather than in the study period as a whole) is emphasized. Again, EP selection was found not to differ statistically between vegetation islands (sex: P=0.23, sex*island interaction: P=0.18). However, EP selection differed between vegetation islands (F=5.89, P=0.005). In the large habitat island, habitat selection on this smaller scale resembled larger-scale selection (Fig. 3b). In contrast, in the small habitat island, also the center (>4 m) of the habitat island was avoided, indicating that although the center of the island was much used during the study period as a whole, animals tended to move out of the center during the study period.
In the road verges, the edges were avoided in both Analysis 1 and 3, i.e., on both a long and short time-scales (Fig. 4). MANOVA did not reveal significant differences between the sexes neither on the large (P=0.42) or small scale (P=0.36).
Discussion
The results presented here show relatively weak microhabitat selection on the basis of microhabitat vegetation (Fig. 2a,b). In contrast, the effect of edge proximity on habitat use seems stronger and more consistent between the two habitat islands; habitat use declines when the distance to the habitat edge becomes smaller than 2 m (Fig. 3a). However, selection of vegetation types would of course be stronger if the habitat islands were more heterogeneous. Thus, there is no reason to conclude that vegetation generally has a smaller effect on microhabitat selection than edge proximity. The wart-biter's vegetation preferences have been discussed at length elsewhere (e.g., Ingrisch, 1979, Cherrill & Brown 1990a, 1990b), so I will mainly discuss the effect of edge proximity on microhabitat selection.
I can think of three alternative explanations for the apparent edge avoidance observed in this study. First, the edge zone microhabitat may have small but important differences in vegetation, or differ in some unmeasured aspect of habitat quality. However, at least on the large habitat island, areas close to the edge are avoided over a wide array of vegetation types (Fig. 2a). Second, the animals may avoid the edge-zone per se, for instance as an anti-predation strategy. Third, the observed pattern may be a by-product of social interactions. It is well known that wart-biter behaviour is strongly influenced by social interactions: singing males attract other males at long distances (positive phonotaxis), but if they come too close together they engage in a “singing contest” until one or both males move away (Keuper et al., 1986). In concordance with this, Weidemann et al. (1990) observed that wart-biter males were clumped and regularly spaced within the clumps; they suggested that this was a “resource-based lek” due to both passive (clumping of the resource, i.e., tall plants suitable as singing perches) and active (movement towards other males) factors. In habitat islands that are so small that most individuals can hear each other, the aggregation formed by positive phonotaxis is more likely to form somewhere in the center of the patch than along the edges. An aggregation formed somewhat off-center can result from chance events, such as the position of especially attractive males (the volume of the song varies quite much among individuals; pers. obs.). In the small habitat island, the wart-biters seemed to aggregate east of the patch center (Fig. 1c,d); this was the reason for the apparent "center-avoidance" on a short time-scale (Fig. 3b). The position of this aggregation seems not to be fully explained by preference for certain vegetation types (Fig. 2b), and positive phonotaxis may be a more plausible explanation.
I found no evidence that microhabitat selection differed between females and males. According to Cherrill & Brown (1990a, 1990b), both sexes need access both to open spots close to the ground (for basking in the sun) and to tall, dense vegetation (for hiding from predators). In addition, males prefer some tall vegetation as singing perches, and females need open, sandy spots for egg-laying (Ingrisch & Boekholt, 1982; Cherrill et al., 1991). Among wart-biters, females provides most of the parental investment (Wedell & Arak, 1989; Wedell, 1993), so the males' singing positions are expected to be influenced by the females' habitat preferences. On the other hand, females usually mate several times (with 3-5 days between each mating) and therefore benefit from not being too far away from males, at least not beyond hearing range (Keuper et al. (1986) reckon that wart-biters seldom can hear each other from distances over 6-10 m). Thus, it is likely that both males and females influence each other's habitat use and as a result have seemingly similar microhabitat selection patterns.
With respect to edge response, the selection pattern in the road verges seems to be quite the opposite of the pattern in the habitat islands (Fig. 4). This is probably a result of different vegetation patterns and the wart-biters' aforementioned need for open spots for basking and oviposition. The vegetation of the road verges was monotonously dense, in contrast to the more heterogeneous seminatural pasture of the habitat islands, where spots with little or no vegetation were interspersed throughout most of the habitat. Therefore, wart-biters in the road verge might stay close to the edges because they need access to this type of microhabitat.
Patch edge avoidance can have a quite large influence on a species' ecology, especially when it inhabits small habitat islands. One consequence may be inefficient use of habitat; the animals experience a higher density of conspecifics than the overall density in the entire habitat patch. For this reason, the density of animal species living in habitat islands may be influenced by the form and area of the habitat islands. For instance, Keuper et al. (1986) have suggested that one consequence of clumping caused by acoustic communication may be that wart-biter populations often occupies only part of larger, suitable areas. In addition, edge avoidance can possibly influence population processes such as emigration from habitat islands, which is a particularly important process in metapopulations.