Habitat fragmentation causes immediate and time-delayed biodiversity loss at different trophic levels

Intensification or abandonment of agricultural land use has led to a severe decline of semi-natural habitats across Europe. This can cause immediate loss of species but also time-delayed extinctions, known as the extinction debt. In a pan-European study of 147 fragmented grassland remnants, we found differences in the extinction debt of species from different trophic levels. Present-day species richness of long-lived vascular plant specialists was better explained by past than current landscape patterns, indicating an extinction debt. In contrast, short-lived butterfly specialists showed no evidence for an extinction debt at a time scale of c. 40 years. Our results indicate that management strategies maintaining the status quo of fragmented habitats are insufficient, as time-delayed extinctions and associated co-extinctions will lead to further biodiversity loss in the future.

the study year of 2000, 17 of the sites were managed as pastures, whereas 13 of the sites had been abandoned from grazing 2-35 years prior to the study year and were subject to gradual overgrowth. However, only sites not covered by forest were included in the study. The current patch and landscape areas were obtained by interpretation of digitalized and orthorectified color aerial photographs (scale: 1:30000) taken between 1999 and 2005. The past patch and landscape areas were obtained by interpretation of digitalized and orthorectified black and white aerial photographs (scale 1:60000, scanning resolution 15 µm) taken in the year 1963 (except one area that was photographed in 1965).
In SW Finland (covering all the Finnish study areas) decline in the occurrence of open semi-natural habitats started already around 1880, and according to agricultural land-use statistics the sharpest decline was passed in these areas already by 1920 (Pöyry et al. 2004).
Following this early decline, the amount of semi-natural habitats has continued to decrease steadily until these days (Luoto et al. 2003) although the relative rate of decline might have been lower during the recent decades than it was in the late 1800s and early 1900s.
Germany: In Germany, in the vicinity of the city of Göttingen in Lower Saxony, 31 calcareous grassland patches were surveyed (Krauss et al. 2003;2004). The current patch and landscape areas were achieved by interpretation of digitalized and orthorectified color aerial photographs. The current aerial photographs from 2004 and 2005 had a resolution of 20 -40 cm on a scale of 1:12000 and were acquired from "Behörde für Geoinformation, Landesentwicklung und Liegenschaften Northeim", "Hessisches Landesamt für Bodenmanagment und Geoinformation" and "Landesamt für Vermessung und Geoinformation Freistaat Thüringen". The vast majority of the historical black and white photos is from 1962, had a resolution of 50 cm on a scale of 1:18000 and was achieved from "Amt für Geoinformationswesen der Bundeswehr". Further photographs were acquired from "Landesvermessungs und Geobasisinformation Niedersachsen" from the years 1967 -1968 on a scale of 1:12000 or 1:13000, from "Hessisches Landesamt für Bodenmanagment und Geoinformation" from the year 1959 and from "Landesvermessung und Geobasisinformation Brandenburg" from the year 1971. Approximately 90% of the 31 patches were interpreted for 1962 and 2005, which is a time period of 43 years between current and historic landscape variables.
The maximum distribution of calcareous grasslands in the region is assumed to be in the 19th century with the most essential losses in area being in the 20th century (Von Drachenfels personal communication). In southern Germany there is evidence for a severe habitat loss of calcareous grassland after 1960, with a reduction of more than 50 % until 1990. Between 1900 and 1960 the habitat loss in this region was only 25 % (WallisDeVries et al. 2002).

Spain:
In Spain, 30 calcareous submediterranean pastures dominated by hemicryptophytes and chamaephytes (Al. Aphyllanthion monspeliensis) were selected in the calcareous massifs in Southern Catalonia (NE of Spain, 41ºN, 0º 30' E). These habitat patches were scattered across 100 km, distributed in 4 dry mountain platforms sited at more than 800 m a.s.l. and showing a mountain Mediterranean climate with cold winters and mild summers (average temperature, 14.1ºC, average annual rainfall 650 mm). Patches summarize a variety of sizes and intensities of change (from stable to high regression), in contrastingly forested landscapes. The historical (1956) aerial photographs were obtained from archives of Spanish Army. Scale of original photos was approximately 1:30000, and pixel size of the resulting orthophoto-maps was 1m. Present-day (2004) images corresponded to orthophoto-maps assembled and orthorectified by the Cartographic Institute of Catalonia (www.icc.cat). Thus, there is a time period of 48 years between current and historical landscape information. Both historical and present-day grassland patches of the study areas were digitized by the Spanish team, by on-screen photo-interpretation of the respective orthophotomaps.
There are no general data for the study sites about the onset and maximum rate of pasture loss, but in general it is assumed that it started in the decade of the 1940, in accordance with the observed in nearby regions such as Southern France (Debussche et al. 1999). Historical data, only available for some municipalities in the study region, confirm this trend (Lloret et al. 2002). The number of semi-natural grasslands has decreased drastically during the last century.
Today only 10% of the semi-natural grassland and 2% of the hay meadows remain. The largest decrease took place before 1950 and during the last decades the decrease has slowed down mostly due to restoration of former semi-natural grasslands (Eriksson et al. 2002).

Biodiversity surveys in different countries. Plant species richness in Estonia was recorded
in July 2001 with one intensive survey in a 30 m radius circle area per patch (see details in Helm et al. 2006). Such sampling covers the total number of specialist species from the whole habitat patch due to the homogeneity of alvar grasslands, confirmed by additional transect walks over remaining areas. In Finland one survey was conducted between June and August 2000 using one 0.25 ha plot and 15 x 1 m² plots, plus a searching time adjusted to habitat area to complement the species list (for details see Raatikainen et al. 2007). In Germany one survey in May/June plus one in August 2000 were conducted. Depending on patch area 1 -3 plots with 25 m² area were mapped and additional plant transects per patch in the same year to complete species richness data (details see Krauss et al. 2004). In Spain and Sweden plant records were conducted in 2007. One survey in March and one in May with 25 x 0.5 m² plots plus additional transects were conducted in Spanish grasslands. In Sweden one survey in July with 10 x 1 m² plus additional transects per patch were carried out.

Butterfly transects (including burnet moths) were carried out in 2000 in Finland and
Germany, and in 2007 in Estonia, Spain and Sweden. All transects were adjusted to patch area and conducted between end of April and end of August to cover the whole flying period of butterfly species. In Estonia four transects, in Finland seven, in Germany five, in Spain three and in Sweden four transects were conducted under conditions suitable for butterfly activity (Pollard 1977; for details on Finnish and German transects see Krauss et al. 2003;Pöyry et al. 2009).

Additional analyses
In addition to the extinction debt analyses, we tested whether (1) patch area loss and (2) landscape area loss of each study site (N = 147) explains current species richness of plants and butterflies. Species richness of specialized vascular plants (patch area loss: slope -7.61, p = 0.007; landscape area loss: slope -8.30, p = 0.03) and butterflies (patch area loss: slope -3.23, p = 0.005; landscape area loss: slope -4.81, p = 0.003) decreased significantly with increasing area loss. Country was included as a random intercept in these general linear mixed effects models.
To validate the findings for the habitat specialists, we tested whether habitat generalists also showed an indication for an extinction debt. We assumed they should not, as their distribution is not restricted to semi-natural grasslands. The calculations indeed showed no indication for an extinction debt, as slopes for full models, as shown for specialists in   Figure S1. Effects of past and current patch area and past and current landscape area on specialized vascular plant species richness in five European countries (regression lines are only shown, when significantly positive (p < 0.05); best country model in box, see Table  S6a). shown, when significantly positive (p < 0.05); best country model in box, see Table S6b).