Dr Å. Berggren, Department of Conservation Biology, PO Box 7002, Swedish University of Agricultural Sciences, SE-750 07, Uppsala, Sweden. Fax: + 46 (0)18 67 35 37. E-mail:Asa.Berggren@nvb.slu.se
1Fragmentation and habitat loss affects both existing and introduced populations. Small habitat areas may have harsher biotic and abiotic conditions, as well as restricting population sizes. Loss of connectivity reduces the opportunities for individuals to move between patches to rescue populations or to re-colonize patches. Knowledge of how landscape composition affects the introduced populations is therefore essential for successful management and future re-introductions.
2To study the effect of landscape composition and structure on the success of colonization, population growth and dispersal distances, we introduced Roesel’s bush-crickets Metrioptera roeseli to 70 habitat islands in areas previously unoccupied by the species. The introduction sites differed in habitat area and connectivity. The population survival and dispersal were then studied for 5 years after initial introductions.
3In addition to results showing the importance of suitable habitat for population persistence, connectivity in form of linear landscape elements and nodes was also crucial. Linear landscape elements and/or nodes were important for colonization success, growth and dispersal. Linear landscape elements and nodes also reduced the negative effects of unsuitable habitat (matrix) and isolation from suitable habitat and on the populations.
4These results stress the importance of connectivity in the landscape for population survival and establishment. Consideration of this should be taken into account in both management and re-introductions of bush-crickets and other invertebrates with similar population characteristics and behaviour.
Connectivity in an anthropogenic landscape often occurs as linear elements. The linear landscape elements often facilitate dispersal in both mammals and insects (Henderson, Merriam & Wegner 1985; Munguira & Thomas 1992; Mauritzen et al. 1999; Niemelä & Spence 1999). Linear elements may consist of many different habitats such as: riparian zones, remnant habitat patches, urban greenways, forest remnants, fencerows, road verges, ditches, banks, lanes and hedgerows (Dowdeswell 1987; Petit & Burel 1998). Linear elements are assumed to mitigate the negative effects of habitat fragmentation by increasing landscape connectivity (Dunning, Danielson & Pulliam 1992). Evidence from some studies suggest that they are valuable as conservation tools and that they increase population persistence for some species (Beier & Noss 1998). Small habitat patches in agricultural landscapes provide habitats for many invertebrates species, but low connectivity may compromise the usefulness of those habitats for species with limited mobility. Landscape connectivity depends not only on the composition and structure of the habitat, but also on the habitat requirements and dispersal abilities of different species (den Boer 1990; Lord & Norton 1990; With & Crist 1995; Tischendorf & Fahrig 2000). Highly vagile species may see a landscape as functionally connected, while species with restricted dispersal may see the same landscape as totally disconnected due to a lower fragmentation threshold (Klein 1989; With & Crist 1995). Linear landscape elements may also be connected to each other in intersection areas, so-called nodes (Forman & Godron 1986). These may be important because they provide additional ways to move in for the individuals that use the linear elements for dispersal.
While connectivity is important for all populations, it may be particularly important for small populations after colonization, and therefore for recently introduced populations. Introductions, re-introductions and translocations of species for conservation purposes are becoming increasingly important (Abbott 2000; Richard-Hansen, Vié & de Thoisy 2000). Knowledge of how landscape composition affect introduced populations is therefore important (Hanski & Thomas 1994; Thomas & Hanski 1997).
In this study we assess the effect of landscape composition and structure on introduced populations of Roesel’s bush-cricket (Metrioptera roeseli) (Hagenbach) (Orthoptera: Tettigoniidae), a species inhabiting 70 areas in a farmland landscape. We tested the effects of connectivity, isolation from suitable habitats and habitat size on colonization success, population growth and dispersal distance of the populations.
Materials and methods
Roesel’s bush-cricket is a small species, 12–18 mm in length (Bellman 1985), common in south and central Europe, as well as in Finland and Latvia. In Sweden it occurs mainly around Lake Mälaren in the south-east. However, today there is an ongoing range expansion (Pettersson 1996). The species is found in ungrazed moist high-grass areas, where it feeds on grass, grass seeds and small insects (Marshall & Haes 1988). The eggs are laid during summer and autumn in grass stems and hatch in May one or two years later (Ingrisch 1986). The nymphs thereafter go through 5–6 instars before they become adults. The species has one generation per year (Marshall & Haes 1988).
Adult males stridulate in July–October. As long as the temperature is high or the weather is sunny the males stridulate almost continuously. The song is characteristic, which makes the species easy to census. A small proportion of the individuals are macropterous. There is some variation between years, but the proportion of macroptery is less than 1% (Vickery 1965).
Study areas and introductions of bush-crickets
Different-sized propagules of M. roeseli were introduced on 70 habitat islands, previously uninhabited by the species, in a large-scale experiment in 1994–95. All introduced bush-crickets were in the last nymphal stage and hence virginal. The experimental areas were situated in the agricultural landscapes in the counties of Uppland and Stockholm, located in south-eastern Sweden. The habitat islands consisted of patches of ungrazed semi-natural grasslands of varying sizes (264–8642 m2) within arable fields. Due to the common rotation system used by the farmers, the use of fields changed every season and included cereal crops, fallows and leys. Some of the selected habitat islands were connected to other suitable habitat patches (unoccupied by M. roeseli), while others were isolated by up to 85 m (see Table 1). The minimum distance between introductions was 2 km. The surrounding matrix consisted of arable fields, forests and human settlements (see Table 1), and included potential barriers such as creeks and roads (de Jong & Kindvall 1991; Kindvall & Ahlén 1992). The minimum distance from the edge of the current distribution was 17 km (details of the experiment are presented in Berggren 2001). The groups of propagules were then distributed across these patches in a random order. There were no differences between the five propagule sizes regarding the landscape composition or structure into which they were released (Pearson product-moment correlation, r < 0·5), therefore effects of propagule size (Berggren 2001) were not included in this study.
Table 1. Descriptive statistics for independent variables used in analyses of habitat features affecting Roesel’s bush-cricket population dynamics
Linear landscape element (ha)
Patch size (ha)
Uninhabitable land (ha)
Maps (1·5 × 1·5 km) centred around the introduction patch of all introduction sites were digitized. Five different types of landscape elements were defined based on information found on land use maps (scale 1 : 10 000) and habitat mapping during field visits: arable fields, forests, semi-natural grasslands, uninhabitable land (housing areas, streams, lakes) and linear landscape elements (ditches, road verges). Areas of all landscape elements were measured on the digitized maps. Data on patch size, their isolation from other suitable habitats and number of nodes (intersections of linear landscape elements) were also estimated from the maps. A few of the landscape variables were correlated. Arable field area and forest area were negatively correlated (Pearson product-moment correlation, r = −0·78, P < 0·0001), arable field area and degree of isolation were negatively correlated (Pearson product-moment correlation, r = −0·51, P < 0·0001), and the number of nodes was positively correlated with the amount of linear landscape elements (Pearson product-moment correlation, r = 0·69, P < 0·0001). The other correlations between landscape variables were only weakly correlated (all r < 0·5). Due to the correlation to both area of forest and degree of isolation from suitable habitat, the variable areas of the arable fields were excluded from the analyses, mainly because there is largely an either/or presence of arable field or forest, and forests seem to show biologically interesting features for the bush-cricket. Although nodes and linear elements show a positive correlation we decided to keep both variables in the analyses due to their evident importance and effects on population survival and distribution. In the analyses, landscape data within a radius of 238 m from the centre of the habitat island was used. The choice of 238 m is based on the 95% limit of dispersal data from all individuals in the first year after introduction. This distance was used to obtain a suitable estimation of the landscape that might be encountered by the individuals (and affecting them). Because of a change in use of the field surrounding the introduction patch from initially cereal crops to ley or fallow in some localities, the landscape data from the first year will represent all years.
A minimum area of 30 ha around the introduction patch was censused annually at the end of the reproductive season (i.e. August and September). Within this area more than 95% of individuals are expected to be found (de Jong & Kindvall 1991). With the dispersal of individuals, the radius of the censused area increased through the years up 78 ha around the introduction patch. The censuses were made by listening for stridulating males. An ultrasound detector was used to detect individuals more effectively at longer distances. A distance from the release points about 2 km along the roads was censused from a car when entering and leaving the sites. The inventories were made only in warm (c. > 18 °C), dry and sunny weather between 9 a.m. and 4 p.m. (details are given in Berggren 2001). Numbers of males found at each site and their locations were entered on the digitized maps. The presence of one or more than one stridulating individual was our criterion for successful colonization. To determine whether the groups of propagules had colonized the areas successfully, the census of 1999 was used. As a measure of population increase or decrease in the patches we used the ratio of the number of individuals introduced and number of individuals recorded in 1999 for each population separately. This ratio will hereafter be called population growth. Dispersal distances from the centre of the habitat patches were measured for each individual to the nearest 1·0 m. A population mean was then calculated from this data.
All independent variables were tested for normality by Shapiro–Wilk W-test and if not normally distributed, transformed to normal or near-normal distribution. To analyse the effects of patch size, isolation (to nearest preferred habitat), landscape composition (areas of different habitats) and landscape structure (linear landscape elements and nodes) on colonization success we used a multiple nominal logistic regression. Effects of the same parameters on population growth and dispersal distance were analysed by multiple regression models (a forward stepwise model). Correlations of variables were analysed by Pearson product-moment correlation. The residuals did not deviate from normal distributions when tested by the Kolmogorov–Smirnov test. All analyses were carried out using jmp (jmp 1995), except for the Kolmogorov–Smirnov test which was performed using systat (systat 1992).
Of the 70 introductions, 42 (60%) had successfully colonized the areas in 1999. Populations in landscapes with many linear landscape elements showed a higher colonization success than populations with few (Fig. 1, Wald χ2 = 6·45, d.f. = 1, P = 0·011). This was the only variable that had an effect on colonization success.
Growth rate declined with increasing isolation from suitable habitat and more forest in the landscape (Tables 2, 3). However, growth rate increased with the area of grassland, the number of linear landscape elements, the number of nodes (Fig. 2) and patch size (Tables 2, 3). There were also significant effects of interactions between area forest and grassland and patch size (Tables 2, 3), with higher population growth in landscapes with larger areas of grassland and larger patch sizes, but with less forest. There were also significant interactions between isolation from suitable habitat and linear landscape elements and nodes (Tables 2, 3), with lower population growth in more isolated patches and with few linear landscape elements and nodes.
Table 2. Summary of the forward, stepwise, linear multiple regression models identifying predictors of population growth and dispersal distance. Terms with more than one component, separated by × indicate interactions between component terms
n = 70, R2 adj = 0·237 n = 50, R2 adj = 0·211.
Linear landscape element
−0·48 × 10−3
Forest × grassland
−14·4 × 10−3
Forest × nodes
Forest × patch size
−10·6 × 10−3
Isolation × linear landscape element
0·18 × 10−3
Isolation × nodes
Table 3. Overview of impact of habitat variables (positive and negative denotes significant impact, NS = not significant) on Roesel’s bush-cricket population dynamics
Linear landscape elements
Forest × grassland
Forest × nodes
Forest × patch size
Isolation × nodes
Isolation × linear landscape element
Dispersal distance was higher in all years for individuals from populations with higher population growth (F = 8·30, d.f. = 1, P = 0·005). There was a significant negative effect of the area of forest on dispersal distance and a significant positive effect of nodes on dispersal distance (Tables 2, 3, Fig. 3). There was also a significant effect of the interaction between the number of nodes and the area of forest on dispersal distance (Tables 2, 3), with longer dispersal distances in landscapes with many nodes and small areas of forest.
The main findings of this study were that: (1) colonization success was higher in landscapes with more linear elements, (2) population growth rate was reduced with increased amount of forest and isolation from suitable habitat, whereas patch size, nodes, linear elements and grassland all increased population growth, and (3) whereas forest reduced dispersal distance, the number of nodes seemed to facilitate the dispersal ability of the species through the landscape.
In an earlier study we found that the size of the propagule of introduced M. roeseli is important for successful colonization (Berggren 2001). Larger propagule sizes have a higher probability of colonization success than smaller ones, but this shows only one of the important variables in determining the outcome of an introduction. What is also important for colonization success is the surrounding environment. This effect of landscape structure on colonization success is not well understood. For M. roeseli, linear landscape elements, i.e. ditches and road verges, seem to be important. We think that linear elements facilitate dispersal in this insect that walks from patch to patch, and that this connectivity increases the access of the species to a greater spectrum of habitats. The structure of linear elements also results in sharp gradients (both vertically and longitudinally) of biotic and abiotic conditions. These environmental factors can change within and between seasons, and habitat heterogeneity can reduce the risk of population extinction (Kindvall 1996). As M. roeseli lay its eggs in grass stems in relatively moist areas, a gradient of humidity in the linear elements and a good opportunity to disperse to moist areas may be fundamental for the development of the eggs. The value of linear landscape elements, both for dispersal and reproduction, has also been shown in several other studies (de Maynader & Hunter 1999; Laurance & Laurence 1999).
Large areas of forest had a negative effect on population growth, with higher growth rate in landscapes with smaller forest areas. Apart from the fact that the forest itself is an unsuitable habitat (de Jong & Kindvall 1991), the vegetation in road verges and in ditches in the forests is also different from the vegetation in road verges and ditches in farmland areas, where grasses are more productive and diverse. The trees also cast shadows on the surrounding ground during parts of the day which lowers the temperature, perhaps to suboptimal levels for the bush-cricket. The area of grassland had a positive effect on population increase, with a higher growth rate where there were larger grassland areas. This confirms grassland as being a good habitat for the species and shows the importance of large grasslands for population increase and thereby population viability.
Isolation had a negative effect on population growth rate, with larger growth rates in populations that were initially less isolated from good habitat. This isolation effect may be a result of decreased possibilities for individuals to reach good habitats and colonize them (Hjermann & Ims 1996; Haddad 1999), which is crucial if the vegetation on the introduction patch deteriorates. The isolation may also have depleted the original patch of individuals by emigration and thereby decreased the possibility of survival on the patch, if emigrants did not return from unsuccessful explorations through the matrix (Sih, Jonsson & Luikart 2000). This would be especially fatal in patches that had few individuals.
There was a higher growth rate in the populations in landscapes with many linear landscape elements than in those with few. The linear element offers both suitable habitat in which the populations can grow but also offers, as discussed earlier, an opportunity to move through the landscape and find suitable habitat when there are changes in weather and vegetation.
Populations in landscapes with many nodes had a higher population increase than populations in landscapes with few. Although the nodes are slightly broader than the linear elements, it is not likely that this fact alone increases the growth rate. The nodes probably slow the dispersal rate of the individuals momentarily (Forman & Godron 1986). This causes the individuals to aggregate in the nodes, a mechanism that may increase the chances of successful reproduction. Such a decrease in dispersal rate has been recorded in a forest beetle Pterostichus melanarius (Illiger), which first expands into rural areas along road verges and then colonizes adjacent deciduous forest more slowly (Niemelä & Spence 1999). The nodes also give individuals the opportunity to move further into the landscape and into new linear landscape elements, with a higher chance of finding suitable habitats. This enables the individuals to move in new directions and cover new areas and reach more locations further away, which increases the chances of finding patches of suitable habitat and linear elements to colonize.
It is interesting to note that population growth in M. roeseli is greater in larger patches. This pattern is expected as a consequence of patch-size-dependent emigration rates. Several empirical studies have shown that the emigration rate is negatively correlated with patch size (Kareiva 1985; Hill et al. 1996; Kuussaari et al. 1996). If this were the case here, the growth rate would decrease with decreasing patch size (Thomas & Hanski 1997). The large patches also offer areas of expansion but, probably more importantly, also a greater possibility of heterogeneity in vegetation that offers suitable areas for feeding and reproduction (Rosenberg, Noon & Meslow 1997). It is also possible that returning immigrants increased in numbers with patch area, and supported the population (Hill et al. 1996; Kuussaari et al. 1996).
There was a significant effect of interactions between isolation from suitable habitat and nodes and linear elements on population increase. The interactions had a higher degree of explanation than the separate variables. The negative effect of isolation was accentuated in populations that lived in landscapes that also had few nodes and linear elements. The isolation of the populations results in a longer time to reach and colonize grassland areas and linear landscape elements for the individuals (Haddad 1999); this is emphasized further if there are limited good habitat areas to find and few directions in which to move. A high degree of isolation and few suitable elements for dispersal may also reduce the likelihood of dispersing individuals to find their way back to the introduction patch. The negative effect of isolation on population growth can therefore be reduced if the landscape has many linear elements and nodes. The negative effect of forest was greater in populations in areas with smaller areas of grassland and smaller patch sizes. This interaction also had a higher degree of explanation than the separate variables, probably being mainly an effect of available habitat. Landscapes with large areas of forest and small patch size and small grassland areas provide few good habitats in which the populations can increase their population, and the available habitat will govern the level of population size.
Populations with higher growth rate dispersed longer. This suggests that dispersal occurred in suitable and productive habitat (i.e. not due to unsuitable habitat). The increasing populations need larger habitats and this is shown by an expansion of the population’s distribution range. Longer dispersal distances are therefore a result of successful colonizations by the species, especially if the growth rate of the populations is high. There was a significant negative effect of area of forest on dispersal distance. Apart from being an unsuitable habitat for the species, this results in low connectivity and little availability to move through the landscape that is also seen in other studies (Roland, Keyghobadi & Fownes 2000).
The amount of nodes had a positive significant effect on mean dispersal distance from the introduction patch. Although initially they may slow down the movement rate of individuals by offering a greater patch for feeding and reproduction, they will soon be saturated and then have a secondary effect with enhancing movement rates. The nodes give the landscape a two-dimensional grid system (Forman & Godron 1986), which enables the individuals to move in new directions, cover new areas and reach more locations further away. This greatly increases the chance of finding patches of suitable habitat and new linear elements to colonize. This effect of enhanced dispersal in linear elements through matrix has been seen earlier in studies of many other different species, such as birds (Desrochers & Hannon 1997; Brooker, Brooker & Cale 1999), insects (Munguira & Thomas 1992; Haddad 1999), amphibians (Rosenberg et al. 1998) and mammals (Henderson et al. 1985).
In landscapes with more nodes and less forest, the dispersal distance was longer. This interaction had a higher degree of explanation than the variables alone, which shows that large areas of forest combined with few nodes have a strong negative effect on dispersal in the populations. The negative effect of unsuitable habitat (forest) (de Jong & Kindvall 1991) can therefore be reduced if connectivity of the landscape is increased.
Importance of connectivity for introduced populations
The suitability of sites proposed for re-introductions and translocations is fundamental for the success of conservation work. The site suitability needs to be considered at a range of spatial scales such as size, availability of good habitat and connectivity in the surrounding landscape (Lindenmayer 1994). Understanding population dynamics in a modified landscape is therefore the key to successful management of the world’s species for survival, regardless of whether the species exist at their more or less original places or are introduced into new areas. This study shows that the possibility to move between patches is a very important feature for introduced populations. The linear landscape elements and nodes are a vital structure in landscape management, when they offer an increase in area of suitable habitat as well as the higher possibility of interchange of individuals between populations.
Other variables in the landscape that might be correlated with the ones studied and effecting colonization success, population growth and dispersal distance, are unknown and regarded as unlikely. Landscapes that have approximately the same amount of, for instance, linear landscape elements still look very different in other features. The introduction sites were distributed over a large area, which result in a great variety of farm size, farm practices and local climate.
Nor do we think that there was a correlation with detection probability and certain landscape variables, so that in some sites individuals were harder to find. The characteristic high call of the males makes them easy to detect at far distances and are easily heard through different types of vegetation. Earlier studies of this species have shown that there is a very high detectability in different kinds of biotops (such as grasslands, fields and forests) (de Jong & Kindvall 1991; Berggren 2001).
It has been posed that linear landscape elements may reduce the average distance the species can move, since individuals become exhausted by moving back and forth in the elements. This should affect adversely the rate of survival in metapopulations (Mader, Schell & Kornacker 1990). However, this does not seem to be the case for M. roeseli. Instead, the linear landscape elements provide both good habitat and a means of dispersal through the landscape, and increases the chances for the populations to survive.
Nodes for increasing dispersal and thereby the chance of finding suitable habitat is the second important component of connectivity. The nodes make it possible for individuals to re-direct their movements and offer an opportunity for faster dispersal over larger areas.
Many linear landscape elements, such as hedgerows and road verges, work both as habitats for reproduction and for dispersal for many other species (Telleria & Santos 1995; Petit & Burel 1998). Some authors stress the importance of determining the use of the linear elements (Rosenberg et al. 1997). However, for management purposes this is often not meaningful, especially when one use does not exclude the other. In fact, this double function merely increases the value of these habitats (Henderson et al. 1985; Dowdeswell 1987).
Arthropods respond rapidly to environmental changes (Kremen et al. 1993). This is why studies on the individual level of invertebrates can reveal important features of the landscape and identify important features of the ecosystem. Knowledge of this becomes increasingly essential, as human activities result in homogenization of habitats in landscapes that formerly had networks of linear elements rich in biodiversity (Ruuska & Helenius 1996).
We thank Doug Armstrong, Åke Berg, Lennart Hansson, Bo Söderström, and two anonymous referees for valuable comments on earlier drafts of the manuscript. We also thank Staffan Roos for logistic support. The research was supported by King Carl XVI Gustaf’s 50-year anniversary fund to Åsa Berggren.
Received 25 October 2000; revision received 13 March 2001