Relative impacts of land‐use and climate change on grasshopper range shifts have changed over time

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Global Ecology and Biogeography published by John Wiley & Sons Ltd 1Department of Biodiversity and Landscape Ecology, Osnabrück University, Osnabrück, Germany 2, Herten, Germany 3Department of Conservation Ecology and Entomology, Stellenbosch University, Matieland, South Africa 4Institute of Biodiversity and Landscape Ecology (IBL), Münster, Germany


| INTRODUC TI ON
The global decline in biodiversity is one of the greatest challenges facing humanity today (Rockström et al., 2009). Among insects, it is assumed that up to 500,000 species have become extinct since the beginning of the industrial era . As species loss continues, there is an urgent need for large-scale studies disentangling the effects of the most important drivers of recent biodiversity range shifts Sirami et al., 2017;Titeux et al., 2017).
However, owing to the longstanding history of species recording across Central Europe, extensive occurrence data are available for some of these groups from unsystematically recorded observations. Although these opportunistic distribution data were frequently used for national atlas projects (e.g., Maas, Detzel, & Staudt, 2002), they have rarely been used for scientific evaluations of species range shifts because of concerns about potential bias owing to unstandardized species recording. However, advanced modelling techniques addressing these concerns have recently become available and thus have enabled analyses of such opportunistic data (Carvalheiro et al., 2013;Isaac, van Strien, August, de Zeeuw, & Roy, 2014;van Strien et al., 2019).
Here, we apply spatial point pattern analysis (SPPA) to quantify range shifts based on a data set of > 100,000 aggregated observational records on the distribution of 58 well-sampled Central European Orthoptera species (hereafter referred to as grasshoppers). This method enables modelling of the relative probability of occurrence of a species based on opportunistic distribution data and local sampling intensity. The main advantage of SPPA for the purpose of our study is that it reduces potential biases attributable to differences in sampling effort across space and time (cf. Bivand, Pebesma, & Gomez-Rubio, 2013;van Strien, van Swaay, & Termaat, 2013).
Grasshoppers are excellent indicators of changes in both land use and climate (e.g., Bazelet & Samways, 2011;Fartmann et al., 2012;Löffler, Poniatowski, & Fartmann, 2019). Earlier Central European work on large grasshopper species sets found that it was mainly habitat specialists that had declined historically, owing to change in land use in the past (Maas et al., 2002;Reinhardt, Köhler, Maas, & Detzel, 2005;Schuch, Bock, Leuschner, Schaefer, & Wesche, 2011). Yet these studies found no effects of climate change on species range shifts. However, more recent studies of regional scope provide clear evidence that thermophilic habitat generalists have recently expanded their ranges in response to climate warming (Fumy, Löffler, Samways, & Fartmann, 2020;Löffler et al., 2019;Poniatowski, Heinze, & Fartmann, 2012). Despite these findings, to date there are no large-scale studies that evaluate the relative impact of changes in both land use and climate on distributional changes in a comprehensive set of grasshopper species.
Here, we calculate range shifts for comparison between two time periods: a historical one, with two comparative time intervals embedded within it (pre-1990 vs. 1990-1999), and a recent one, also with two embedded comparative time intervals (1990-1999 vs. 2000-2017). Owing to rapid habitat loss before 1990 (e.g., Carvalheiro et al., 2013), we hypothesize that severe range retractions took place in several Central European grasshopper species in that early period. However, climate change has probably become an increasingly important factor driving grasshopper range shifts in recent years, as has been shown for butterflies (Pöyry et al., 2009).
In order to test the relative impacts of changes in land use and climate on both early and recent distributional changes of Central European grasshoppers, observed range shifts were related to: (a) K E Y W O R D S biodiversity loss, dispersal ability, distribution modelling, functional trait, global warming, habitat specificity, insect conservation, Orthoptera, species farmland index, species temperature index the species temperature index (STI); and (b) the species farmland index (SFI) (cf. Devictor et al., 2012;Fumy et al., 2020). In addition, we also analysed grasshopper range shifts relative to four functional groups that represent interactions between mobility and habitat specificity: (a) less mobile generalists; (b) highly mobile generalists; (c) less mobile specialists; and (d) highly mobile specialists. According to previous research, the extent of range shifts among insect species depends on both habitat specificity and mobility (cf. Buse & Griebeler, 2011;Hill et al., 2002;Steck et al., 2007). Although we hypothesize that grasshopper specialists have suffered greater range losses than generalists owing to their higher sensitivity to change in land use, we expect that highly mobile grasshopper species might cope better with environmental changes than less mobile grasshopper species owing to their higher dispersal ability.

| Study area
The study area consisted of Germany (centre: 51°9′48″ N, 10°26′52″ E), covering more than one-third of Central Europe (Ellenberg & Leuschner, 2010; Figure 1). Germany is divided into seven biogeographical regions: (a) the north-western and (b) north-eastern lowlands (hereafter considered as North German Plain), which pass southwards to (c) the western, (d) the eastern and (e) the south-western low mountain ranges (hereafter considered as Central German Uplands), extending to (f) the foothills of the Alps, which then adjoin (g) the Northern Limestone Alps (hereafter considered as Alps and Alpine Foreland) (cf. Maas et al., 2002). Elevation ranges from slightly below sea level in the F I G U R E 1 The (a) historical range shifts (pre-1990 vs. 1990-1999) and (b) recent range shifts (1990-1999 vs. 2000-2017) of Calliptamus italicus in Germany are intended to illustrate the modelling approach, which is based on spatial point pattern analysis (see Section 2.3.4). Local range-trend values of the species were calculated as the differences between the "local relative raster frequencies" of two time intervals that are taken into account within the respective time period (historical or recent) (see Section 2.3.4). To represent the overall distribution trend (range shift) of a species across its German range, the mean of all local range-trend values was calculated. In the case of C. italicus, its range changed on average by −15.06 in the historical period (a) and by 11.54 in the recent period (b) (cf. Table 1) TA B L E 1 Historical (pre-1990 vs. 1990-1999) and recent (1990-1999 vs. 2000-2017) range shifts among all studied grasshopper species classified by habitat specificity (HS; S = habitat specialist; G = habitat generalist) and mobility (M; H = high mobility; L = low mobility)

| Change in land use
For centuries, Central Europe was characterized by low-intensity land use, resulting in open and heterogeneous cultural landscapes (Ellenberg & Leuschner, 2010). However, since the beginning of the industrial era, both intensification of land use and land abandonment have led to a severe loss of semi-natural habitats, habitat fragmentation and homogenization of the landscape (Burns et al., 2016;Stoate et al., 2009;WallisDeVries et al., 2002). These changes were especially severe during the period from 1950 to 1990 (Carvalheiro et al., 2013;Stoate et al., 2009;van Strien et al., 2019). Since then, the rate of landscape change has slowed, and the remaining habitats have been increasingly protected [e.g., by the Habitats Directive (EC, 1992)] and managed to maintain their biodiversity (Carvalheiro et al., 2013;Löffler, Poniatowski, & Fartmann, 2020). Nevertheless, biodiversity in these remnants still suffers frequent degradation of habitat quality and fragmentation Fischer & Lindenmayer, 2007;Helbing et al., 2017).

| Climate change
Since the end of the 19th century, the mean annual temperature in Germany has increased by 1.3 °C (Brasseur, Jacob, & Schluck-Zöller, 2017). However, global warming became especially evident in the late 1980s and has now reached an unprecedented level (IPCC, 2014). Accordingly, temperature anomalies increased significantly in Germany (across all biogeographical regions). As a result, recent annual temperatures usually exceeded the long-term mean over the last three decades (Supporting Information Figure S1). Given that actions against climate change are still insufficient, it is expected that temperatures will increase by at least a further 2.6 °C, and possibly up to 3.7 °C by the end of the 21st century (Brasseur et al., 2017;IPCC, 2014).
Although annual precipitation has increased slightly, drought periods during summer have become more frequent owing to shifts in precipitation from summer to winter and higher evapotranspiration rates from increased summer temperatures (Brasseur et al., 2017).
TA B L E 1 (Continued)

| Model organisms
Currently, 79 grasshopper species are regarded as indigenous to Germany (Fischer et al., 2020). Of these, 58 species with sufficient availability of distribution data were analysed here (Table 1; Supporting   Information Table S1). To ensure statistical reliability, only species with a frequency of > 10 occupied grid cells in at least one time interval per reference time period (see Section 2.3.2) were included in analyses.
Synanthropic, data-deficient, non-resident and cryptic species that have been inadequately sampled were not considered (cf. Reinhardt et al., 2005;van Strien et al., 2013). Owing to possible sampling deficits in high mountain areas (cf. Gedeon et al., 2014), we additionally excluded predominantly alpine species within their German range.

| Distribution data
Most of the data used were originally compiled for a comprehensive atlas project on German grasshoppers (Maas et al., 2002).
Additionally, more recent data (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) were provided by the federal state authorities and regional grasshopper societies in the German federal states (see Acknowledgments). Grasshopper distribution data were based on a grid-map system with a spatial resolution of 10 km × 11 km covering the whole of Germany (in total, Germany comprises 3,004 of these grid cells, known in German as "Messtischblatt"; cf. Reinhardt et al., 2005). Grid cells were classified as occupied when at least one observation in the considered time interval was available (cf. Termaat et al., 2019;van Strien et al., 2019). Overall, the data set included 109,351 aggregated observational records (redundant information for the respective grid cells were excluded before the analyses; see also Section 2.3.2).

| Reference time periods and intervals
Grasshopper range shifts were analysed for two reference time For each of the time intervals, the amount of available data was high and similar (pre-1990: 28,648 observations; 1990-1999: 46,009 observations; 2000-2017: 34,694 observations).
The two reference time periods show clear differences in the intensity of change in land use and climate. The historical period (pre-1990 vs. 1990-1999) represents an era of severe change in land use, but only weak effects of climate change (see Sections 2.1.1 and 2.1.2). In contrast, the recent period (1990-1999 vs. 2000-2017) was characterized by rapid climatic change, whereas habitat loss and deterioration have slowed (see Sections 2.1.1 and 2.1.2).

| Reference grid cells
As the basis for spatial analyses, we defined well-surveyed reference grid cells (see Section 2.3.4). A grid cell was considered well surveyed if all national and regional grasshopper reference species were recorded during the respective time interval (Supporting Information Table S2). The national reference species represent common species that are assumed to be present throughout Germany. Owing to physiogeographical differences within the study area, we additionally defined regional reference species for each of the seven biogeographical regions of Germany

| Spatial analyses
To detect species range shifts using presence-only distribution data, we chose a modelling approach derived from case-control studies in spatial epidemiology and based on kernel smoothing; a method from spatial point pattern analysis (SPPA). The general idea behind our approach is to model the relative probability of occurrence (i.e., local relative raster frequency) of a species based on sizeable distribution data (i.e., occupied grid cells) and local sampling intensity (i.e., the local density of well-surveyed grid cells around occupied grid cells) (cf. Figure 1). Differences in "local relative raster frequencies" in the time intervals before and after the threshold years were calculated to quantify range shifts during the historical and recent period, respectively (see Section 2.3.2). The main advantage of this method for distribution modelling is that it reduces potential biases attributable to differences in sampling effort across space and time (cf. Bivand et al., 2013;van Strien et al., 2013). A detailed description of the spatial analyses is available in the Supporting Information (Supplementary methods).

| Functional groups
We classified grasshopper species by their mobility (less mobile versus highly mobile species) and habitat specificity (habitat generalists versus specialists). These traits crucially affect the response of grasshoppers to environmental change (e.g., Fartmann et al., 2012;Löffler et al., 2019;Reinhardt et al., 2005). The classification underlying our analyses was mainly derived from Reinhardt et al. (2005).
However, we revised this classification for some species by considering more recent information on habitat selection and species mobility (cf. Löffler et al., 2019;Poniatowski, Münsch, Helbing, & Fartmann, 2018; Table 1). Afterwards, we differentiated four functional groups representing interactions of mobility and habitat specificity: (a) less mobile generalists; (b) highly mobile generalists; (c) less mobile specialists; and (d) highly mobile specialists (Figure 2).

| Global change indices
We calculated the species temperature index (STI) and the species farmland index (SFI) for all grasshopper species included in our analyses. The STI is defined as the mean temperature within the distribution range of a species (cf. Devictor et al., 2012). It should be noted that the STI is a relative measure usually based on geographically restricted distribution data. Thus, STI values might differ from the "true thermal niche" of a species across its entire range.

| Statistical analyses
In order to detect the most important drivers behind species range shifts at the national scale, we calculated multivariate linear models (LMs), for both the historical and the recent time period (Table 2). Species traits (i.e., mobility and habitat specificity; see also Section 2.4) were used as categorical predictors in these models.

Furthermore, they included the STI and SFI as continuous measures
of species sensitivity to changes in climate and land use (see also Section 2.5). The STI and SFI were transformed before the analyses [e x and ln(x), respectively; Figure 3; Table 2]. In order to determine the impact and relative variable importance (RI) of the predictors in the two time periods, we applied model averaging based on an information-theoretic approach (Grueber, Nakagawa, Laws, & Jamieson, 2011). Model averaging was conducted using the "dredge" function (R package MuMIn; Bartón, 2020) and included only top-ranked models within ∆AIC C < 3 (cf. Grueber et al., 2011).
We additionally calculated univariate LMs, in order to illustrate the role of mobility and habitat specificity responsible for grasshopper range shifts in Germany (Figure 2). The relationships between STI/SFI and historical or recent range shifts were additionally proved at a regional scale using univariate LMs (Supporting Information   Table S3). Statistical differences in these models were assessed using likelihood ratio tests (type III test). (pre-1990 vs. 1990-1999) and (b) recent range shifts (1990-1999 vs. 2000-2017) (mean ± SD) of habitat generalists and habitat specialists with low (light blue) and high (dark blue) mobility. Differences between the groups were tested using univariate linear models (LMs; see Section 2.6). Statistical significances are indicated as follows: n.s. p > .05; *p < .05; **p < .01; ***p < .001. Range shifts were calculated as described in Figure 1 (see also Section 2.3.4) **

Generalists Specialists
Range shifts

| Historical range shifts
During the historical period (pre-1990 vs. 1990-1999), 77% of species showed distinct range retractions, whereas only 23% of species expanded their ranges (Table 1; Figure 2a). Habitat specificity has been identified as the driver with the highest relative variable importance (RI = 1.00) considering historical range shifts (Table 2a).
Except for highly mobile habitat generalists, which extended their ranges, all other functional groups suffered from severe range reductions during the historical period (Figure 2a). Considering highly mobile species, range shifts of habitat generalists were significantly different from those of habitat specialists. Although both less mobile habitat generalists and specialists have undergone range retractions, habitat specialists experienced significantly greater range losses than habitat generalists. In addition, range shifts of highly mobile habitat generalists and specialists differed significantly from those of the less mobile ones.
The SFI exhibited a significantly negative correlation with historical range shifts (RI = 0.64; Figure 3c; Table 2a). This means that species with a higher SFI underwent more severe range losses during the historical period. Furthermore, we found a positive relationship between the STI and historical range shifts (RI = 0.56; Figure 3a; Table 2a), which indicates that thermophilic species experienced smaller range losses or even expanded their ranges. The findings of the national analyses largely correspond to the results at the regional scale. Although we found a negative relationship between historical range shifts and SFI across all biogeographical regions, the positive impact of STI on historical distribution trends was restricted to the North German Plain (Supporting Information Table S3a).

| Recent range shifts
During the recent period (1990-1999 vs. 2000-2017), 60% of the species expanded their ranges, whereas the remaining 40% of species still experienced range retractions (Table 1). Although recent range shifts were significantly driven by the mobility of the species (RI = 1.00), they were no longer affected by habitat specificity (Table 2b). Recent range expansions were mostly observed in highly mobile species, including both habitat generalists and specialists (Figure 2b). In contrast, ranges of less mobile species on average remained stable in the study area.
Furthermore, recent range shifts exhibited a significantly positive relationship to STI (RI = 1.00; Figure 3b). However, there was no relationship between SFI and recent grasshopper range shifts ( Figure 3d). This was confirmed by the regional analyses, which revealed that SFI did not affect recent range shifts accross the biogeographical regions, whereas theVy were significantly driven by STI in the North German Plain and Central German Uplands (Supporting Information Table S3b). Note: Influence of species temperature index (STI), species farmland index (SFI) (see Section 2.5), mobility and habitat specificity (see Section 2.4) on (a) historical (pre-1990 vs. 1990-1999) and (b) recent (1990-1999 vs. 2000-2017) range shifts of the 58 grasshopper species (Table 1). Model-averaged coefficients (conditional average) were derived from the top-ranked linear model (ΔAIC C < 3; RI = relative parameter importance; see Section 2.6). The R 2 values represent the range of model accuracies within the top-ranked models. Statistical significances are indicated as follows: n.s. p ≥ .05; *p < .05; **p < .01; ***p < .001.

TA B L E 2
Results of multivariate linear model analyses (model averaging)

| D ISCUSS I ON
Most of the grasshopper species examined here have experienced large-scale range shifts over the last few decades. However, the patterns differed clearly between the two reference time periods (historical and recent). During the historical period (pre-1990 vs. 1990-1999), most species experienced severe range losses.
Furthermore, habitat specialists declined to a muVch greater extent than generalists with equal mobility. These range retractions were related to species associated with HNV farmland (i.e., species with high SFI values). In contrast, during the recent period (1990-1999 vs. 2000-2017) the distribution of less mobile species generally remained stable, and highly mobile species even expanded their ranges, irrespective of their habitat specificity. Also, range expansions were correlated with thermophilic species, with high STI values.

| Historical range shifts (pre-1990 vs. 1990-1999)
Change in land use contributed to a severe loss of semi-natural habitats across Europe before 1990 (Carvalheiro et al., 2013;Stoate et al., 2009;van Strien et al., 2006). Grasshoppers are also highly sensitive to changes in land use Marini, Fontana, Battisti, & Gaston, 2009;Torma, Gallé, & Boszó, 2014), especially in the case of habitat specialists Poniatowski & Fartmann, 2010). Our findings are in accordance with these, with range retractions of habitat specialist grasshoppers during the historical study period generally being greater than those of habitat generalists. Irrespective of the spatial scale (i.e., national or regional), the grasshopper species with the highest SFI values generally showed the greatest declines. This species group includes HNV farmland species (e.g., Decticus verrucivorus, Omocestus rufipes and Psophus stridulus), depending on the maintenance of low-intensity land use (Maas et al., 2002), and additionally, species inhabiting dynamic, alpine river systems (e.g., Bryodemella tuberculata, Chorthippus pullus and Tetrix tuerki). In Central Europe, the few remaining habitats of the latter species group are embedded in the traditionally managed alpine landscape (Helbing, Blaeser, Löffler, & Fartmann, 2014;Fischer et al., 2020). Their SFI values are correspondingly high.
Climate change was becoming evident in the late 1980s (IPCC, 2014; see Section 2.1.2). Accordingly, it had not yet been seen as playing any major role in the observed range shifts during the historical period (cf. Pöyry et al., 2009). However, species with high STI values had already experienced small range losses or, alternatively, had already expanded their ranges slightly in the 1990s across Germany. However, regional analyses revealed that this trend was mainly evident in the North German Plain, were temperatures might already have become favourable for some thermophilic grasshoppers during F I G U R E 3 Relationship between (a,c) historical range shifts (pre-1990 vs. 1990-1999) or (b,d) recent range shifts (1990-1999 vs. 2000-2017) (58 species) and species temperature index (STI; a,b) or species farmland index (SFI; c,d). Regression slopes and significant differences were fitted based on the results of multivariate linear models (LMs; Table 2; see Section 2.6). RI, relative parameter importance, see Section 2.6. Statistical significances are indicated as follows: n.s. p ≥ .05; *p < .05 the historical period (cf. Supporting Information Figure S1b). With the exception of a few highly mobile specialists, such as Oecanthus pellucens and Phaneroptera falcata, historical range shifts were especially true for highly mobile habitat generalists. These findings are in accordance with previous studies, where only a few highly mobile and predominatly generalist species benefited from increased temperatures during the initial stage of elevated anthropogenic global warming (e.g., Beckmann et al., 2015;Hill et al., 2002;Warren et al., 2001 (Holusa, 2012;Maas et al., 2002;Reinhardt et al., 2005).
Especially in the North German Plain and in the Central German Uplands, numerous species were able to colonize new areas. Given that range expansions were most obvious for species with high STI values, it can be assumed that these positive shifts were driven mainly by climate change. This is in accordance with previous studies, which provided evidence that climate warming has led to range expansions of thermophilic insects in general (Devictor et al., 2012;Hickling et al., 2006;Termaat et al., 2019) and grasshoppers in particular (Fumy et al., 2020;Löffler et al., 2019). However, in contrast to previous studies that detected range expansions attributable to climate warming largely restricted to habitat generalists (Beckmann et al., 2015;Hill et al., 2002;Warren et al., 2001), we found that both generalists and specialists were able to track global warming in the recent period, as long as they were of high mobility. For example, dry grassland specialists, such as Calliptamus italicus, Oedipoda caerulescens, Platycleis albopunctata and Sphingonotus caerulans, which all experienced severe range losses pre-1990 (Maas et al., 2002;Reinhardt et al., 2005), rapidly expanded their ranges over the last two decades. High summer temperatures are known to increase insect flight activity (Kuussaari, Rytteri, Heikkinen, Heliölä, & von Bagh, 2016) and favour colonization of areas where the climate was previously unsuitable (Hickling et al., 2006;Hill et al., 2002;Parmesan et al., 1999). Although the biodiversity of wet habitats is generally assumed to be threatened by climate change (Streitberger et al., 2016), even highly mobile wet grassland specialists, such as

Mecostethus parapleurus, Ruspolia nitidula and Stethophyma grossum,
have recently expanded their range, as we show here. These species require sufficient habitat moisture only during winter (i.e., the high-precipitation season; see Section 2.1.2) for successful egg development (Poniatowski, Münsch, et al., 2018). The situation is different for other hygrophilous species, such as Omocestus viridulus, Pseudochorthippus montanus and Tetrix undulata. They are among the few species with range retractions during the recent period, which is probably attributable to their high sensitivity to summer drought (cf. Gardiner, 2010;Poniatowski, Münsch, et al., 2018).

| Outlook on future range shifts
Although many insects, especially habitat specialists, across different taxa have experienced severe range losses during the second half of the 20th century (Maes & Van Dyck, 2001;Reinhardt et al., 2005;van Swaay et al., 2006;Warren et al., 2001), we found that many grasshopper species have recently expanded their ranges across Central Europe. However, it is uncertain whether these species will continue to benefit from future climate change. Steck et al. (2007) predicted that climate change will foster grasshopper species richness in Central Europe. In contrast, Wessely et al. (2017) suggested that recent conservation strategies might not compensate for climate-induced range losses in the future. Although that study focused mainly on alpine or rare species, further habitat loss and deterioration might be so great that many grasshopper species will be unable to react to further increased global warming by movement (Buse & Griebeler, 2011;Hill et al., 2002;Steck et al., 2007). If that is the case, we expect that future climate change will be particularly threatening to hygrophilous and alpine species unable to make use of suitable refuge habitats in intensive agricultural landscapes.
Many less mobile habitat specialists among grasshoppers have become extinct in large parts of Central Europe as a result of past habitat destruction (Holusa, 2012;Maas et al., 2002;Reinhardt et al., 2005). This development has also affected numerous other groups of insects, such as bees, butterflies and carabid beetles (Brooks et al., 2012;Goulson, Lye, & Darvill, 2008;van Strien et al., 2019;Wagner, 2020). Given that the intensive agricultural landscape in Central Europe is largely unsuitable for most insect species, their remaining populations are currently dependent on a few isolated remnants of semi-natural habitats (Helbing et al., 2017;Maes & Van Dyck, 2001;Öckinger & Smith, 2007). However, habitat quality within these remnants continues to decrease, which is likely to lead to time-delayed local species extinctions Löffler et al., 2020).
This means that conservation measures should focus on adaptation to future climate changes by building ecosystem resilience and

B I OS K E TCH
In our recent studies, we focus on the effects of changes in land use and climate on biodiversity patterns (global change ecology). Furthermore, we investigate how species richness of ecosystems is influenced by disruptive events, such as windthrow, fire and floods (disturbance ecology). In the field of restoration ecology, we evaluate habitat management and measures that aim to restore threatened ecosystems. We use a wide range of model organisms.