Journal of Mammalogy, 89(6):1502–1511, 2008

LAND-COVER CHANGE AND THE FUTURE OF THE APENNINE BROWN BEAR: A PERSPECTIVE FROM THE PAST ALESSANDRA FALCUCCI,* LUIGI MAIORANO, PAOLO CIUCCI, EDWARD O. GARTON,

AND

LUIGI BOITANI

Department of Animal and Human Biology, Sapienza Universita` di Roma, Viale dell’Universita` 32, 00185 Rome, Italy (AF, LM, PC, LB) Department of Fish and Wildlife Resources, University of Idaho, Moscow, ID 83843, USA (AF, LM, EOG)

The Apennine brown bear (Ursus arctos marsicanus) is an endangered subspecies endemic to Italy, where a small population, estimated at 40–50 bears, inhabits a human-dominated landscape. Although little is known of the ecology of this population, habitat loss and fragmentation often has been considered one of the main threats for small and endangered populations. To assess habitat availability at the landscape scale, we used a distribution model to compare historical, present, and future land-cover suitability for the Apennine brown bear population in central Italy. The 4 models are based on 3 existing land-cover maps (1960, 1990, and 2000) and 1 simulated map for 2020, obtained from a cellular-automata Markov-chain land-transition model. We also compared changes in human population density as a surrogate for human pressures on bear habitat, and we measured the contribution of protected areas to the bear’s conservation. Our results show that, at the landscape level and assuming that current human population trends continue in the future, land-cover suitability does not seem to be an issue or priority. The current negative trend of this population, despite opposite trends in land-cover suitability, suggests that conservation efforts should focus more on direct actions aimed at reducing human-caused mortality and enhancing population expansion into suitable unoccupied areas. Key words: Abruzzo, bear conservation, geographic information system, habitat modeling, Italy, landscape dynamics, Ursus arctos marsicanus

The brown bear (Ursus arctos) was widespread in the entire holarctic region until the beginning of 1800s when, facing persecution and habitat destruction, its population decreased markedly (Swenson et al. 2000). Nowadays, in Europe, it is abundant only in the eastern and northern parts (Zedrosser et al. 2001), whereas in western Europe it is facing high risks of extinction and is restricted to small, isolated, and endangered populations (Swenson et al. 2000; Taberlet et al. 1995). In Italy, the brown bear has seen a progressive reduction in its range since the 1600s due to legal hunting (before 1938), poaching, and habitat destruction (Fabbri et al. 1983; Febbo and Pellegrini 1990). Today it is restricted to 2 populations in the Alps and 1 in the Apennines (Boitani et al. 2003; Linnell et al. 2007). The central-Apennine population is a subspecies endemic to Italy (U. a. marsicanus, hereafter, the Apennine brown bear—Loy et al. 2008; Randi et al. 1994; Vigna-

* Correspondent: [email protected]

! 2008 American Society of Mammalogists www.mammalogy.org 1502

Taglianti 2003) and it is restricted to roughly 5,000–8,000 km2 (Fig. 1). However, the most densely and consistently occupied area (1 bear/50–80 km2—Lorenzini and Posillico 2000) is only 1,500–2,500 km2 (Fig. 1), mainly in Abruzzo-Lazio-Molise National Park (hereafter, Abruzzo National Park) and surrounding areas (Bologna and Vigna-Taglianti 1992; Boscagli 1999; Meriggi et al. 2001; Posillico et al. 2004). In other parts of the central Apennines, the presence of the species is not stable and densities are very low (Ciucci and Boitani 2004; Posillico et al. 2004). The subspecies U. a. marsicanus is protected by law and Abruzzo National Park, along with other adjacent protected areas (PAs), has been created to secure the bear’s conservation. The Apennine brown bear is considered endangered by the Italian World Wildlife Fund red list (Bulgarini et al. 1998), and critically endangered under Criterion D by the International Union for Conservation of Nature European Mammal Assessment (International Union for Conservation of Nature 2007). Moreover, Swenson et al. (2000) considered the subspecies as highly threatened; Convention on International Trade in Endangered Species of Fauna and Flora includes the Apennine brown bear in its Appendix II, the European Community listed

December 2008

FALCUCCI ET AL.—LAND-COVER CHANGE AND APENNINE BEARS

FIG. 1.—A) Distribution of brown bears (Ursus arctos) in Italy and in the surrounding countries (modified from Large Carnivore Initiative for Europe 2007; http://www.lcie.org/). B) Distribution of brown bears in the study area and existing protected areas.

the bear as endangered in the Habitat Directive in 1992, and the international community classified the bear as strictly protected in the Bern convention in 1979. The central-Apennine population has been completely isolated from other bear populations for 400–600 years (Lorenzini et al. 2004; Randi et al. 1994; but see Loy et al. [2008], who suggested a postglacial separation), and current population size has been estimated at 43 individuals (95% confidence interval: 35–67—Gervasi et al. 2008), which may correspond to an effective population size of 4–10 adult females (Ciucci and Boitani 2008), below the number that is required for a viable population (Wiegand et al. 1998; Wielgus 2002; but see Sæther et al. 1998). Nevertheless, Lorenzini et al. (2004) did not reveal any indications of inbreeding depression among 30 individually genotyped bears. Moreover, the Apennine brown bear lives in a human-dominated landscape, suffers consistent humancaused mortality (Posillico et al. 2002; L. Gentile, Veterinary Service, Abruzzo-Lazio-Molise National Park, pers. comm.), and lacks a coordinated conservation strategy (Swenson et al. 2000) to overcome the existing political and administrative fragmentation (i.e., the existence of more than 1 administrative entity issuing laws and regulations on the same bear population). Such a strategy is difficult to achieve in the absence of reliable knowledge on the status, ecology, and threats to the population. Thus, the Apennine brown bear faces serious problems and needs significant efforts to achieve successful conservation.

1503

Habitat loss and degradation have been considered 2 of the main reasons for the Apennine brown bear’s decrease (Bologna and Vigna-Taglianti 1992; Fabbri et al. 1983; Febbo and Pellegrini 1990), and are considered among the major longterm threats for its survival (Swenson et al. 2000). Nonetheless, no formal evaluation of the status and trends of the bear habitat has been undertaken in the region. Simply assuming that habitat loss is (and has been) the main cause of bear population decrease might diminish the potential role of other threats (e.g., human-induced mortality and low recruitment rates). Thus, conducting a large-scale habitat evaluation is important because any conservation strategy for demographic recovery must consider space and habitat availability, possibly with explicit reference to habitat changes over time. Our primary goal is to test the hypotheses that land-cover suitability for the Apennine brown bear in central Italy is declining and that the current system of PAs is insufficient to ensure the persistence of the subspecies. In particular, we assessed historic (1960s and 1990) to current (2000) changes in land-cover suitability for Apennine brown bears and used these changes to project likely effects of future (2020) land-cover changes on this population. We also evaluated changes in human population density for the same time frames as a surrogate for human encroachment into the bear’s habitat. In addition, we investigated distribution and location of suitable areas in central Italy that are still unoccupied by bears and where individuals from the core population might eventually expand in the near future. Our aim is to facilitate conservation strategies based on metapopulation dynamics, providing spatially explicit references for intensive management actions (e.g., conflict resolutions and population augmentation), and habitat restoration. This further allows us to measure the potential contribution of existing and proposed PAs to the goal of large-scale and long-term conservation for the Apennine brown bear.

MATERIALS AND METHODS Our study area (22,000 km2) corresponds to the approximate range of the Apennine brown bear in 1800 (Boitani et al. 2003; Fig. 1). We built a distribution model for the Apennine brown bear using 4 geographic information system layers: a digital elevation model (cell size ¼ 75 m), and 3 land-cover maps, including the Land-Cover 1960 (LC1960), the CORINE LandCover 1990 (CLC1990), and the CORINE Land-Cover 2000 (CLC2000). The LC1960 (scale 1:200,000; 22-class legend) was produced from 1956 to 1968 by the National Research Council, Rome, Italy. The CLC maps (scale 1:100,000; 44class legend) were produced in 1990 and in 2000 by the Italian Ministry for the Environment (Rome, Italy) for the European Community (Bruxelles, Belgium). To obtain 3 thematically homogeneous geographic information system layers, we reclassified the 2 CLC maps and the LC1960 map (Table 1), slightly modifying the legend proposed by Falcucci et al. (2007) to include more thematic details for agricultural areas and forests. To obtain 3 spatially homogeneous layers, we transformed the land-cover maps from

1504

Vol. 89, No. 6

JOURNAL OF MAMMALOGY

vector to raster, assigning to each map the same origin, extent, and cell size (200 " 200 m). Cell size was chosen to be consistent with the coarser resolution of the LC1960. The digital elevation model was resampled to match the resolution and the extent of the land-cover layers. For the period 1960–2000, we used spatially explicit data on the human population based on national censuses (Italian Institute of Statistics; www.istat.it). Existing PAs and the NATURA2000 network (a network of conservation areas proposed by the European Community for the conservation of species and habitat listed under the Habitat Directive) were obtained from Maiorano et al. (2006, 2007). The Forestry Service (Ufficio Territoriale per la Biodiversita` di Castel di Sangro, Castel di Sangro, L’Aquila, Italy) provided 407 locations of Apennine brown bears, including hair collected systematically through baited traps, scats collected opportunistically along trails, specimens collected occasionally, and intensive surveys within patches of Rhamnus alpinus, that is, seasonal aggregation sites for bears (Randi et al. 2004, 2005). The data were collected from 2000 to 2003 within a European Union LIFE project following standard protocols (Woods et al. 1999), inside the park and within its buffer zone, and met guidelines approved by the American Society of Mammalogists (Gannon et al. 2007). Laboratory protocols were implemented to obtain reliable individual genotypes (42 individual genotypes were identified—Randi et al. 2005) and molecular sex determination (19 females and 23 males). We retained only 304 locations (those with .95% probability of being bears on a genetic basis), of which 33% occurred between May and August (spring to early hyperphagia, as defined by Swenson et al. [2000]), and 67% between September and December (late hyperphagia). Land cover in 2020.— To project land cover in 2020, we used the 2 CLC maps in a combined cellular-automata Markov-chain multicriteria–multiobjective land allocation land-cover prediction procedure (CA_MARKOV procedure in IDRISI3.2—Eastman 2001). The CA_MARKOV algorithm is based on the assumption that future land-cover changes can be predicted using past land-cover changes (Eastman 2001). We used CLC1990 and CLC2000 to obtain a transition-area file, quantifying changes from each land-cover category in 1990 to each other category in 2000. The suitability of each pixel for each land-cover type is determined on the basis of a set of land-cover–suitability maps, 1 for each land-cover type. We built each land-cover–suitability map using Mahalanobis distance statistics (De Maesschalck et al. 2000) based on topographic and anthropogenic layers (aspect, elevation, slope, main and secondary road densities, and distance to main and secondary roads). We used a contiguity filter to downweight the land-cover suitability of pixels far from existing areas of each land-cover class, thus giving preference to contiguous suitable areas (Eastman 2001). Using CLC2000 as the starting point for the land-cover change projection in 2020, we ran the entire procedure 100 times and we assigned the value that occurred most often for each pixel to the final 2020 land cover. Moreover, to evaluate the stability of the land-cover projection, we measured the

TABLE 1.—Scores assigned to each land-cover class for the distribution model and for the possible alternative models (0 ¼ land-cover class that does not support the presence of the bear; 1 ¼ land-cover class that fulfils partial resource requirements in terms of food, cover, and water; 2 ¼ land-cover class that fulfils resources requirements at a suboptimal level; 3 ¼ land-cover class that fulfils optimal resources requirements). A dash (—) indicates that no alternative score has been considered. Land-cover class

Model

Alternative score

Artificial Non-irrigated arable lands Irrigated arable lands Vineyards Wooden cultivations Olive groves Pastures Complex agricultural areas Agricultural areas with natural vegetation Broadleaf forests Coniferous forests Mixed forests Natural prairies Moors Sclerophyllous vegetation Forest#shrubs transitional areas Beach and dunes Scarcely vegetated areas Marshes Bare rocks Rivers Lakes

0 1 0 0 2 0 1 1 2 3 2 3 1 2 2 3 0 1 0 1 0 0

— 0 — — 1 — — 0 — — 1 2 — — 1 2 — — — — — —

percentage of the study area classified in the same land-cover class in all the 100 runs of the CA_MARKOV procedure. Further details on the CA_MARKOV algorithms can be found in Eastman (2001). Land-cover–suitability models.— To model the land-cover suitability for the Apennine brown bear, we used information available on the ecology of the species (Boitani et al. 2003; Maiorano et al. 2006; Posillico et al. 2004; Swenson et al. 2000) to build 1 deductive model (DM, sensu Corsi et al. [2000], i.e., expert and literature based) for each time frame (i.e., 1 DM for 1960, 1 for 1990, 1 for 2000, and 1 for 2020). We chose a deductive approach because it allows for the generalization of species–habitat relationships over large areas, providing a synthesis of the available knowledge (Johnson and Gillingham 2004), and because points of presence were available only for 2000. Following Clevenger et al. (2002) and Maiorano et al. (2006), land cover and elevation were used as surrogates of bear habitat qualifiers. The 2 layers can be easily related to the existing knowledge on the ecology of the population, and, finally, digital maps of these variables exist (or can be obtained through simulations) for all of the 4 time frames that we considered. For each time frame, our modeling approach involved 2 steps: reclassification of the 2 environmental layers based on the species’ land-cover and elevation preferences, and combination of the 2 reclassified layers to obtain a final suitability score for each 200 " 200-m cell in the study area. In the 1st

December 2008

FALCUCCI ET AL.—LAND-COVER CHANGE AND APENNINE BEARS

TABLE 2.—Combinations of elevation and land-cover scores to obtain the suitability scores for the deductive models. Elevation scores based on expert opinion and literature data: 0 ¼ out of the usual elevation range of presence (,500 m); 1 ¼ inside the usual elevation range of presence (from 500 m to the maximum elevation possible in the study area); 2 ¼ inside the ‘‘core’’ elevation range (from 800 m to 1,800 m). See Table 1 for land-cover scores. Elevation Land cover 0 1 2 3

0

1

2

Unsuitable Unsuitable Unsuitable Unsuitable

Unsuitable Low suitability Low suitability Medium suitability

Unsuitable Low suitability Medium suitability High suitability

step, the land-cover map was reclassified into 4 categories (Table 1), whereas the elevation map was reclassified into 3 categories (Table 2). In the 2nd step, the scores available for land-cover classes and elevation were integrated to obtain the final suitability scores for the DM (Table 2). Land-cover suitability and human population density.— We investigated the relationship between land-cover suitability and human population changes. For each administrative unit (n ¼ 592) we calculated changes in human population density from 1960 to 2000. Administrative units were classified into 2 groups: those where human population increased and those where human population decreased. For each unit we calculated the percentage of area occupied by each suitability class and tested the significance of the differences using a median test. Land-cover suitability and PAs.— We measured land-cover suitability inside PAs and inside NATURA2000 and tested the difference in suitability with the rest of the study area using a median test. In particular, we drew 1,000 random points in PAs and 3,500 outside (both corresponding to 0.2 points/km2). For each point, we built a circular buffer with a 500-m radius and measured the area occupied by the different suitability classes. The same procedure was used considering PAs plus the NATURA2000 network. Large-scale connectivity and landscape indices.— To measure large-scale connectivity between the area of stable presence (Fig. 1) and the rest of the study area, we ran a least-cost–path analysis (Walker and Craighead 1997) using suitability scores as proxies for movement costs (with lower suitability implying higher costs). To monitor the changes occurring in the landscape structure and to numerically quantify the least-cost–path analysis, we measured the following landscape indexes for each DM (over the entire study area, over the PAs, and over the non-PAs): percentage of the landscape, number of patches, largest patch index, euclidean nearest neighbor median distance, and normalized landscape shape index (McGarigal and Marks 1995; Turner et al. 2001). Validation and sensitivity analyses.— To evaluate the DM developed for 2000, we calculated the Boyce index (Boyce et al. 2002; Hirzel et al. 2006) using the 304 available points of presence. The index goes from #1 to 1, with negative values indicating a model that predicts poor-quality areas where presence is more frequent, with positive values indicating a

1505

model whose predictions are consistent with the data set, and with values close to 0 indicating that the model is not different from a chance model. The index is built calculating the ratio of predicted frequency versus the expected frequency for each suitability class. The predicted frequency for a given suitability class is calculated as the number of evaluation points predicted by the model to fall in that suitability class; the expected frequency is calculated as the percentage of evaluation points expected in the suitability class on the assumption of a random distribution across the study area. To calculate the predicted frequency, we built for each validation point a 500-m circular buffer and we assigned to each point the suitability class with the highest share inside the buffer. Alternatively, if 2 or more suitability classes had the same share, the highest suitability was chosen. To calculate the expected suitability we used 20,000 random locations (corresponding to roughly 0.9 points/km2, a density similar to that measured for the validation points), and we assigned a suitability class to each location following the same procedure outlined for the points of presence. No validation was possible for the DMs of time frames other than 2000, because no presence data sets were available. However, we performed a sensitivity analysis (Ray and Burgman 2006) to evaluate the extent of the deviations in the results by using different suitability scores (see Table 1) and changing the elevation ranges (original ranges 6 100 m). We ran a total of 639 different models and measured the correlation existing among them and the final DM using Cramer’s V (significance tested with a chi-square statistic—Ott et al. 1983).

RESULTS Land cover in 2020.— The 2020 land-cover change projection was performed 100 times. With just 5 simulations, more than 97.7% of the study area was always classified in the same land-cover class, and after 25 simulations the percentage did not change markedly (96.4%, 96.1%, and 96.05% after 25, 50, and 100 simulations, respectively). The remaining areas of uncertain assignment (roughly 4% of the study areas) were small (median size ¼ 1 ha) and located at the boundaries between land-cover classes. Land-cover suitability models.— Most of the high- and medium-suitability areas for the Apennine brown bear are in the internal, mountainous parts of the study area (Fig. 2), whereas most of the low-suitability and unsuitable areas are in lowlands, roughly corresponding to areas with high human population densities. From 1960 to 1990 the high-suitability classes increased (Figs. 2 and 3). In particular, unsuitable areas covered 24% of the study area both in 1960 and in 1990. However, low-suitability areas occupied almost 50% of the study area in 1960 and 29% in 1990. The opposite was true for higher suitability classes that increased in 1990. The pattern of suitability did not change markedly for 1990, 2000, and 2020 (Figs. 2 and 3). Land-cover suitability and human population density.— Median human population density decreased from 1960 to 2000 (Fig. 4). In particular, human population decreased in 487

1506

JOURNAL OF MAMMALOGY

FIG. 2.—Deductive models for the Apennine brown bear (Ursus arctos marsicanus) in 1960, 1990, 2000, and 2020.

(82.3%) administrative units from 1960 (median ¼ 0.65 inhabitants/ha) to 1990 (0.45 inhabitants/ha) and in 398 (67.2%) units from 1990 to 2000 (0.44 inhabitants/ha). Administrative units where human population decreased from 1960 to 1990 had a higher percentage of medium- and highsuitability areas when compared with administrative units where human population increased (P , 0.0001). The same was true for 1990–2000.

Vol. 89, No. 6

Land-cover suitability and PAs.— Protected areas covered .23% of the study area (5,230 km2). Including also the NATURA2000 network, the percentage of PAs exceeded 38% of the study area (851,530 km2). Compared to the rest of the study area, PAs hosted a higher percentage of high- and low-suitability areas, and a lower percentage of unsuitable and mediumsuitability areas (P , 0.0001; Fig. 5), and the same results were obtained combining PAs and the NATURA2000 network. Large-scale connectivity and landscape indices.— According to our least-cost–path analysis, in 1960 the areas with a stable presence of the bears appeared to be fairly well connected to the southern portion of the study area, whereas northward connections were limited. In 1990, the entire landscape was much more homogeneous, and large-scale connectivity increased widely, with barriers to movement of the animals existing only in correspondence to intensively cultivated areas. In 2000 and in 2020, the connectivity within the landscape was similar to that estimated in 1990 (Fig. 6) but with a clear trend toward increasing levels of connectivity. The landscape indices are consistent with the least-cost–path analysis, suggesting decreasing fragmentation of habitat for brown bears. Across time, most landscape indices show different trends for the unsuitable and the low-suitability areas versus medium- and high-suitability areas (Fig. 7). In 1960, unsuitable areas were highly contiguous (low number of patches, high percent of the landscape, and low normalized landscape shape index), but in 1990, the number of patches for unsuitable areas increased and the largest patch index decreased, indicating an increasing fragmentation. Projections from 2000 to 2020 suggest a tendency toward increasing fragmentation of unsuitable areas, with normalized landscape shape index, percent of the landscape, and largest patch index that should remain almost unchanged, whereas number of

FIG. 3.—Percentage of the landscape (PLAND) of our study area in central Italy in the 4 suitability classes from 1960 to 2020.

December 2008

FALCUCCI ET AL.—LAND-COVER CHANGE AND APENNINE BEARS

FIG. 4.—Median human population density per administrative unit from 1951 to 2001 (data source: Italian Institute of Statistics; http:// www.istat.it).

patches should decrease and euclidean nearest neighbor distance increase (Fig. 7). We revealed a similar pattern for low-suitability areas, which in 1960 represented the prevalent class (36% of the landscape), but appeared increasingly fragmented in 1990 (,30% of the landscape, largest patch index , 5%, and both number of patches and normalized landscape shape index increased with respect to 1960). Projections from 2000 to 2020 suggest that patterns of low-suitability areas should stabilize or show a tendency toward higher fragmentation (Fig. 7). Medium- and high-suitability areas showed different trends. In 1960, the 2 classes represented only a minor portion of the study area and were relatively fragmented (Fig. 7). In 1990, percent of the landscape and largest patch index increased,

1507

whereas number of patches, normalized landscape shape index, and euclidean nearest neighbor distance decreased, indicating a lower fragmentation. We revealed the same tendency for medium- and high-suitability areas for both 2000 and 2020. Overall, by contrasting PAs and the rest of the study area (results not shown), we revealed a greater fragmentation for unsuitable areas and a lower fragmentation of high-suitability areas inside PAs. However, no temporal trends of the landscape metrics were detected between PAs and the rest of the study area. Validation and sensitivity analyses.— The Boyce index calculated for the 2000 DM was 0.8, indicating that the DM is sufficiently consistent with the evaluation data set (bear ! 6 SD; presence data points). As indicated by Cramer’s V (X 1960: 0.78 6 0.11; 1990: 0.85 6 0.06; 2000: 0.84 6 0.07; 2020: 0.82 6 0.07; all P , 0.001) all 639 alternative DMs used for sensitivity analysis were not significantly different from the final model.

DISCUSSION Although we checked for validity and stability of all the models, we had no way of validating the 2020 land-cover projection, which is based on the assumption that the pattern of land-use change observed during 1990–2000 will remain constant up to 2020. Although indications provided by the 2020 projection should be interpreted cautiously, the resulting output was very stable after 20 simulations, which indicates, at least, that if our assumptions hold, then our results are consistent. We recognize that development projects (e.g., tourist and ski resorts, roads, and wind farms) can drastically change the landscape. Nevertheless, our 2020 landscape is based on layers that represent proxies for the probability of

FIG. 5.—Percentage of protected areas (PAs) and nonprotected areas (No PAs) in central Italy occupied by the different suitability classes in 1960, 1990, 2000, and 2020.

1508

JOURNAL OF MAMMALOGY

FIG. 6.—Estimated movement cost for the Apennine brown bear (Ursus arctos marsicanus) from the area of stable presence to the rest of the study area from 1960 to 2020.

future developments, thus minimizing the possibility for unforeseen events. Moreover, most of the projects currently considered will be implemented in areas of low suitability for the bears (Piano d’azione per la tutela dell’orso marsicano;

Vol. 89, No. 6

www.regione.abruzzo.it), and therefore should not alter significantly the pattern of land-use change in high-suitability areas. We did not take into account the effects of climate change on habitat availability for the bears. However, both scale and time frames are too detailed to build realistic models on climate change and our data set does not allow building a reliable climatic envelope. The DM for 2000 was validated using field data. The concordance between true presence locations and suitability predictions indicates that our deductive approach provides a reliable synthesis of the species distribution in the study area (but consider that no evaluation of the specificity of our DM was possible, given that no true absence point was available). No validation was possible for the DM of the other time frames, but sensitivity analyses showed extremely low departures to changes in suitability scores, further supporting the reliability of our DM. Overall, our results suggest that large-scale availability of areas suitable for the bears is not a major issue for their conservation, provided that recent and current land-use trends are maintained and that our DM is valid. Although increasing habitat destruction is continuing today in many developing countries (e.g., Sodhi et al. 2004), different trends have been observed in many parts of Europe (Debussche et al. 1999; Falcucci et al. 2007; Olsson et al. 2000). During the 20th century, traditional agriculture, grazing, and forestry activities, following rural depopulation, became increasingly economically nonviable (Cohen 2003; Olsson et al. 2000). As a consequence, vegetation succession is progressing toward forest reestablishment and spread (Falcucci et al. 2007; Laiolo et al. 2004), and, at the same time, large carnivores are increasing their ranges, at least in some areas of Europe (http:// www.lcie.org/).

FIG. 7.—Landscape indexes (NP ¼ number of patches; LPI ¼ largest patch index; ENNMD ¼ euclidean nearest neighbor median distance; NLSI ¼ normalized landscape shape index [dimensionless]) calculated for the entire study area in 1960, 1990, 2000, and 2020.

December 2008

FALCUCCI ET AL.—LAND-COVER CHANGE AND APENNINE BEARS

Considering land-cover only, we showed that habitat availability also has increased for Apennine brown bears, especially from 1960 to 1990. After 1990, we measured limited changes in habitat availability, probably because most of the remaining unsuitable areas are in lowlands, where agriculture and urbanization are the dominant features, and human population did not, and most probably will not, decrease. The least-cost–path analysis indicated that natural recolonization by bears of the northern part of the study area is possible, at least at the landscape scale and according to the variables we modeled. Once more, it is important to underline that our results for 2020 do not account for future developments of roads and highways, considered among the most important barriers to movements by bears (Swenson et al. 2000). We have no direct validation of the least-cost–path analysis, but our results are confirmed by bear sign (obtained by cameratraps, sightings, and scats) collected over the last 15 years in the northern portion of the study area about 140 km from the park (P. Ciucci, L. Carotenuto, P. Morini, and P. Forconi, pers. comm.), where the Apennine brown bear was considered extinct by the 1930s (Boitani et al. 2003). It is therefore evident that natural recolonization of the study area is possible. In summary, habitat availability (as we defined it) will not be an issue for bear conservation in the foreseeable future on the landscape scale. This indication is important in shaping and prioritizing renewed conservation efforts to ensure the survival of the species in central Italy. Our results are valid for a landscape-scale approach; land cover and topography do not represent all the ecological parameters that affect habitat suitability for bears. The focus of our approach is to analyze and project major trends in suitability throughout the bear range, and to evaluate large-scale connections among the different ‘‘core’’ portions of the study area. Therefore, our approach should be considered complementary to fine-grained, inductive modeling that integrates occupancy and mortality risks, allowing for a better understanding of the distribution of sink habitats and of their dynamics through time. Such a modeling approach is planned and the collection of new, better data sets is ongoing (Ciucci and Boitani 2006). According to our results, the existing PAs provide a reasonable coverage of the area occupied by Apennine brown bears, with limited possibilities of additional PAs being established (Maiorano et al. 2007). In addition, our least-cost–path analyses suggest that bears can move from one PA to another, using areas with high- to medium-suitability outside PAs. These potential corridors for the bears should be considered as a fundamental component of any conservation strategy. However, habitat protection alone is not a sufficient solution for bear conservation. This is indicated by the number of bears accidentally killed and poached inside the park and its buffer area (Ciucci and Boitani 2008). Seventy-four bears have been found dead in the last 30 years, and in at least 21 cases poaching was the cause of death. Conservation of Apennine brown bears must focus most of its effort on issues other than establishing additional PAs and promoting further habitat restoration. The amount, configuration, and connectivity of suitable habitat at the landscape scale

1509

seem appropriate to host a larger bear population and to allow for the natural recolonization of its former range. Despite the positive trends in suitability for the past 40 years, the bear population apparently continues to decrease (Ciucci and Boitani 2008; International Union for Conservation of Nature 2007; Linnell et al. 2007; Wilson and Castellucci 2006). At the landscape scale, the level of connectivity indicated by our model should be preserved from potential developments or habitat alteration. As in the Greater Yellowstone Ecosystem (Gunther et al. 2004), direct conservation actions aimed at reducing human-caused mortality appear extremely urgent.

ACKNOWLEDGMENTS The Abruzzo-Lazio-Molise National Park administration, through its Scientific and Surveillance Services, and the Forestry Service (Ufficio Territoriale per la Biodiversita`; Castel di Sangro) collaborated in the collection of the data on bear presence used for model evaluation. Funding for this project was partially provided by a Wildlife Conservation Society grant and by a Ministry of the Environment (Directorate for Nature Conservation) grant to the Department of Human and Animal Biology (Sapienza Universita` di Roma). The Institute of Applied Ecology provided logistical and technical support for the analyses. R. A. Powell, R. Harris, and 2 anonymous referees provided useful comments on an earlier version of the manuscript.

LITERATURE CITED BOITANI, L., S. LOVARI, AND A. VIGNA-TAGLIANTI (EDS.). 2003. Fauna d’Italia. Mammalia III: Carnivora–Artiodactyla. Calderini Editrice, Bologna, Italy. BOLOGNA, M. A., AND A. VIGNA-TAGLIANTI. 1992. Osservazioni sull’areale dell’orso marsicano con particolare riferimento al Gran Sasso e ai Monti della Laga. Hystrix—Italian Journal of Mammalogy 5:75–80. BOSCAGLI, G. 1999. Status and management of the brown bear in central Italy (Abruzzo). Pp. 81–84 in Bears: status, survey and conservation action plan (C. Servheen, S. Herrero, and B. Peyton, eds.). International Union for Conservation of Nature Species Survival Commission, Gland, Switzerland. BOYCE, M. S., P. R. VERNIER, S. E. NIELSEN, AND F. K. A. SCHMIEGELOW. 2002. Evaluating resource selection functions. Ecological Modelling 157:281–300. BULGARINI, F., E. CALVARIO, F. FRATICELLI, F. PETRETTI, AND S. SARROCCO. 1998. Libro rosso degli animali d’Italia, Vertebrati. World Wildlife Fund Italia, Roma, Italy. CIUCCI, P., AND L. BOITANI. 2004. Protocollo per il campionamento genetico non invasivo. PNALM, Pescasseroli, Italy. CIUCCI, P., AND L. BOITANI. 2006. Conservation of large carnivores in Abruzzo. A research project integrating, species, habitat and human dimension. 2006 Annual Report. Wildlife Conservation Society, New York. CIUCCI, P., AND L. BOITANI. 2008. The Apennine brown bear: a critical review of its status and conservation problems. Ursus. CLEVENGER, A. P., J. WIERZCHOWSKI, B. CHRUSZCZ, AND K. GUNSON. 2002. GIS-generated, expert-based models for identifying wildlife habitat linkages and planning mitigation passages. Conservation Biology 16:503–514. COHEN, J. E. 2003. Human population: the next half century. Science 302:1172–1175. CORSI, F., J. DE LEEUW, AND A. SKIDMORE. 2000. Modeling species distribution with GIS. Pp. 389–434 in Research techniques in

1510

JOURNAL OF MAMMALOGY

animal ecology: controversies and consequences (L. Boitani and T. K. Fuller, eds.). Columbia University Press, New York. DEBUSSCHE, M., J. LEPART, AND A. DERVIEUX. 1999. Mediterranean landscape changes: evidence from old postcards. Global Ecology and Biogeography 8:3–15. DE MAESSCHALCK, R., D. JOUAN-RIMBAUD, AND D. L. MASSART. 2000. The Mahalanobis distance. Chemometrics and Intelligent Laboratory Systems 50:1–18. EASTMAN, J. R. 2001. IDRISI3.2 release 2. Guide to GIS and image processing. Clark Labs, Clark University, Worcester, Massachusetts. FABBRI, M., G. BOSCAGLI, AND S. LOVARI. 1983. The brown bear population of Abruzzo. Acta Zoologica Fennica 174:163–164. FALCUCCI, A., L. MAIORANO, AND L. BOITANI. 2007. Changes in landcover patterns in Italy and their implications for biodiversity conservation. Landscape Ecology 22:617–631. FEBBO, D., AND M. PELLEGRINI. 1990. The historical presence of the brown bear on the Apennines. Aquilo 27:85–88. GANNON, W. L., R. S. SIKES, AND THE ANIMAL CARE AND USE COMMITTEE OF THE AMERICAN SOCIETY OF MAMMALOGISTS. 2007. Guidelines of the American Society of Mammalogists for the use of wild mammals in research. Journal of Mammalogy 88:809–823. GERVASI, V., ET AL. 2008. A preliminary estimate of the Apennine brown bear population size based on hair-snag sampling and multiple data-source mark–recapture Huggins model. Ursus. GUNTHER, K. A., M. A. HAROLDSON, K. FREY, S. L. CAIN, J. COPELAND, AND C. C. SCHWARTS. 2004. Grizzly bear–human conflicts in the Greater Yellowstone ecosystems, 1992–2000. Ursus 15:10–22. HIRZEL, A. H., G. LE LAY, V. HELFER, C. RANDIN, AND A. GUISAN. 2006. Evaluating the ability of habitat suitability models to predict species presences. Ecological Modelling 199:142–152. INTERNATIONAL UNION FOR CONSERVATION OF NATURE. 2007. Ursus arctos marsicanus. In European mammal assessment. http:// ec.europa.eu/nature/conservation/species/ema/. Accessed 5 November 2007. JOHNSON, C. J., AND M. P. GILLINGHAM. 2004. Mapping uncertainty: sensitivity of wildlife habitat ratings to expert opinion. Journal of Applied Ecology 41:1032–1041. LAIOLO, P., F. DONDERO, E. CILIENTO, AND A. ROLANDO. 2004. Consequences of pastoral abandonment for the structure and diversity of the alpine avifauna. Journal of Applied Ecology 41:294–304. LINNELL J., V. SALVATORI, AND L. BOITANI. 2007. Guidelines for population level management plans for large carnivores in Europe. Final draft May 2007. A Large Carnivore Initiative for Europe report prepared for the European Commission, Rome, Italy. LORENZINI, R., AND M. POSILLICO. 2000. DNA fingerprinting and brown bear conservation in Abruzzo—central Italy. Italian Journal of Zoology 67:326–327. LORENZINI, R., M. POSILLICO, S. LOVARI, AND A. PETRELLA. 2004. Noninvasive genotyping of the endangered Apennine brown bear: a case study not to let one’s hair down. Animal Conservation 7:199–209. LOY, A., P. GENOVESI, M. GALFO, M. G. JACOBONE, AND A. VIGNATAGLIANTI. 2008. Cranial morphometrics of the Apennine brown bear (Ursus arctos marsicanus) and preliminary notes on the relationships with other southern European populations. Italian Journal of Zoology 75:67–75. MAIORANO, L., A. FALCUCCI, AND L. BOITANI. 2006. Gap analysis of terrestrial vertebrates in Italy: priorities for conservation planning in a human dominated landscape. Biological Conservation 133:455–473. MAIORANO, L., A. FALCUCCI, AND L. BOITANI. 2007. Contribution of the Natura 2000 network to biodiversity conservation in Italy. Conservation Biology 21:1433–1444.

Vol. 89, No. 6

MCGARIGAL, K., AND B. J. MARKS. 1995. FRAGSTAT. Spatial analysis program for quantifying landscape structure. United States Department of Agriculture Forest Service General Technical Report PNW-GRT-351:1–122. MERIGGI, A., O. SACCHI, U. ZILIANI, AND M. POSILLICO. 2001. Definizione dell’areale potenziale di cervo sardo, muflone e orso bruno. Pp. 117–235 in Progetto di monitoraggio dello stato di conservazione di alcuni mammiferi particolarmente a rischio della fauna italiana (S. Lovari and A. Sforzi, eds.). Ministry of the Environment, Department for Nature Conservation, Rome, Italy. OLSSON, E. G. A., G. AUSTRHEIM, AND S. N. GRENNE. 2000. Landscape change patterns in mountains, land use and environmental diversity, Mid-Norway 1960–1993. Landscape Ecology 15:155–170. OTT, L., R. F. LARSON, AND W. MENDENHALL. 1983. Statistics: a tool for the social sciences. Duxbury Press, Massachusetts. POSILLICO, M., A. MERIGGI, E. PAGNIN, S. LOVARI, AND L. RUSSO. 2004. A habitat model for brown bear conservation and land use planning in the central Apennines. Biological Conservation 118:141–150. POSILLICO, M., A. PETRELLA, AND L. SAMMARONE. 2002. Piano preliminare di conservazione dell’orso bruno (Ursus arctos L. 1758). Ministero delle Politiche Agricole e Forestali—Commissione Europea, Progetto LIFENAT99/IT/006244:1–48. RANDI, E., ET AL. 2005. Non-invasive genetic monitoring of the brown bear population in the central Apennines, Italy. Pp. 1–27 in 16th International Conference on Bear Research and Management, September 27th–October 1st 2005. Riva del Garda, Trentino, Italy. RANDI, E., L. GENTILE, G. BOSCAGLI, D. HUBER, AND U. U. ROTH. 1994. Mitochondrial DNA sequenze divergence among some west European brown bear (Ursus arctos L.) populations. Lessons for conservation. Heredity 73:480–489. RANDI, E., ET AL. 2004. Relazione finale sul conteggio della popolazione, sullo status genetico e demografia/dinamica della popolazione. Ministero delle Politiche Agricole e Forestali, Commissione Europea, Roma, Prodotto identificabile del progetto LIFENAT99/IT/006244:1–48. RAY, N., AND M. A. BURGMAN. 2006. Subjective uncertainties in habitat suitability maps. Ecological Modelling 195:172–186. SÆTHER, B.-E., S. ENGEN, J. E. SWENSON, Ø. BAKKE, AND F. SANDEGREN. 1998. Assessing the viability of Scandinavian brown bear, Ursus arctos, populations: the effects of uncertain parameter estimates. Oikos 83:403–416. SODHI, N. S., L. P. KOH, B. W. BROOK, AND P. K. L. NG. 2004. Southeast Asian biodiversity: an impending disaster. Trends in Ecology and Evolution 19:654–660. SWENSON, J. E., N. GERSTL, B. DAHLE, AND A. ZEDROSSER. 2000. Action plan for the conservation of the brown bear (Ursus arctos) in Europe. Nature and Environment 144:1–69. TABERLET, P., J. E. SWENSON, F. SANDEGREN, AND A. BJA¨RVALL. 1995. Localization of a contact zone between two highly divergent mitochondrial DNA lineages of the brown bear Ursus arctos in Scandinavia. Conservation Biology 9:1255–1261. TURNER, M. G., R. H. GARDNER, AND R. V. O’NEILL. 2001. Landscape ecology in theory and practice. Pattern and process. Springer– Verlag, New York. VIGNA-TAGLIANTI, A. 2003. Ursus arctos. Note di sistematica. Pp. 87– 92 in Fauna d’Italia. Mammalia III: Carnivora–Artiodactyla (L. Boitani, S. Lovari, and A. Vigna-Taglianti, eds.). Calderini Editrice, Bologna, Italy. W ALKER , R., AND L. C RAIGHEAD . 1997. Analyzing wildlife movement corridors in Montana using GIS. http://gis.esri.com/ library/userconf/proc97/proc97/to150/pap116/p116.htm. Accessed September 2008.

December 2008

FALCUCCI ET AL.—LAND-COVER CHANGE AND APENNINE BEARS

WIEGAND, T., J. NAVES, T. STEPHAN, AND A. FERNANDEZ. 1998. Assessing the risk of extinction for the brown bear in the Cordillera Cantabrica, Spain. Ecological Monographs 68:539–570. WIELGUS, R. B. 2002. Minimum viable population and reserve sizes for naturally regulated grizzly bears in British Columbia. Biological Conservation 106:381–388. WILSON, C. J., AND C. CASTELLUCCI. 2006. The Apennine brown bear and the problem of large mammals in small populations. Ecos 27:75–81.

1511

WOODS, J. G., D. PATKAU, D. LEWIS, B. N. MCLELLAN, M. PROCTOR, AND C. STROBECK. 1999. Genetic tagging of free-ranging black and brown bears. Wildlife Society Bulletin 27:616–627. ZEDROSSER, A., B. DAHLE, J. E. SWENSON, AND N. GERSTL. 2001. Status and management of the brown bear in Europe. Ursus 12:9–20.

Submitted 1 August 2007. Accepted 8 May 2008. Associate Editor was Roger A. Powell.