Deforestation affects biogeographical regionalization: a case study contrasting potential and extant distributions of Mexican terrestrial mammals

We used ecological niche modelling projected as species' potential (based on the original vegetation map) and extant (based on the 2000 land use and vegetation map) distributions to analyse changes on patterns of endemism of terrestrial mammals occurring in Mexico. Based on the biogeographic method of Parsimony Analysis of Endemicity, we obtained cladograms under scenarios of species' potential distribution (t1) and extant distributions (t2). We found that the resolution of consensus cladogram in t2 was poorer, while there were more geographic synapomorphies in t1, and more autapomorphies in t2 due to a reduction of species' distributions as a consequence of deforestation. We defined a hierarchical regionalization with two regions with the cladogram of t1; a transitional zone, two subregions, five dominions, and 15 provinces. Conversely, the consensus cladogram of t2 had a basal trichotomy, and the position of the Sierra Madre Occidental changed compared with t1. In t1 and t2, the Yucatán Peninsula+Chiapas+Isthmus of Tehuantepec clade was maintained, although in t2 it was separated from the remaining areas of the country. The impact of deforestation on species distributions strongly affected the biogeographic regionalization of terrestrial mammals in Mexico.


Introduction
Biogeographic regionalization consists of hierarchical systems for categorizing biodiversity into realms, regions, dominions, provinces, and districts (Cabrera & Willink 1973;Brown & Lomolino 1998). Typical biogeographic schemes rely on geographic, palaeontological, faunistic or floristic criteria, but few studies attempt to build natural regionalization systems (Espinosa-Organista et al. 2001;Morrone et al. 2002). Recent approaches based on patterns of successively nested endemism propose hierarchical schemes based on a historical perspective, in which the biotic or biogeographic province, the smallest endemism area identified, is the basic unit (Espinosa-Organista et al. 2001). Patterns of successively nested endemism can be identified using Parsimony Analysis of Endemicity (PAE) (Rosen 1988; Rosen & Smith 1988;Morrone 1994). Briefly, in a PAE cladogram, groups of areas sharing unique endemic taxa are identified assuming a common biogeographic history.
Several biogeographic regionalizations, based on global similarity of faunas and habitats (vegetation, climate), have been developed for Mexico, some of them using predominantly terrestrial mammals (e.g. Ramírez-Pulido & Castro-Campillo 1990, 1993Ramírez-Pulido et al. 1994;Arriaga et al. 1997; see Morrone 2005). Morrone (2005) synthesized the main Mexican biogeographical schemes developed in the 20th century, emphasizing those employing patterns of distributions of vegetation, animals and habitat characteristics. An underlying assumption of these approaches is that biotic similarity does not necessarily imply naturalness; consequently, recent methods for regionalization in Mexico are based on identifying biogeographic homology (Morrone et al. 1999Espinosa Organista et al. 2000;Morrone 2001Morrone , 2005Morrone & Márquez 2001, 2003. A potential shortcoming of previous biogeographic regionalizations efforts based on species' point occurrence data from museum specimens involves their taxonomic and geographic biases . To improve the distributional maps of species obtained from museum data, we can use novel approaches of species' ecological niche modelling (Stockwell & Peters 1999). Ecological niche modelling projected as species' potential distributions using point occurrence data, environmental layers and GIS platform provides a robust predictor for species presence and absence in a given region (Peterson et al. 2000Illoldi-Rangel et al. 2004). Species potential distribution models seem to improve significantly identification of endemism areas (Escalante et al. 2003, in press;Rojas-Soto et al. 2003), and consequently biogeographical regionalization, which holds promise for further research (Guisan & Thuiller 2005).
The geographical classification of areas is also relevant for conservation planning (Brown et al. 1993;Márquez et al. 2001). The transformation of natural landscapes into agrosystems and urban settlements is a major threat for biodiversity conservation. Several investigations have analysed the implications of climatic change on species distributions and their conservation status (Peterson et al. , 2002bWalther et al. 2002;Hannah et al. 2005;Thuiller et al. 2005), whereas others have investigated the impact of landscape fragmentation on species distribution, richness, density, and abundance (Bethke & Nudds 1995;Kolozsvary & Swihart 1999;Debinski & Holt 2000;Sánchez-Cordero et al. 2005a, 2005b. High deforestation rates usually result in reduction of species distributions , 2005a, 2005bFuller et al. 2007). Since deforestation impacts species distributions, we suspect that distributional patterns derived from individual species distributions change accordingly; examples depicting species distributional changes under t1 and t2 scenarios are shown in Figure 1.
Natural patterns of endemism are the bases of hierarchical biogeographic regionalizations, but those can be modified by transformation of natural habitats. To our knowledge, this is the first study addressing the impact of transformation of natural habitats on patterns of endemism and regionalization. Understanding the biological implications of land use changes is relevant in predicting patterns of biodiversity (Peterson & Navarro-Sigü enza 2006;. Our aim was to determine the impact of deforestation on patterns of endemism by contrasting potential and extant distributions using terrestrial mammals in Mexico as a case study. Specifically, we show changes in distributional patterns of terrestrial mammals under two scenarios: by modelling species' ecological niche projected as potential distributions (t1), using the original natural vegetation map, and projected as extant distributions (t2), using the 2000 land use and vegetation map (Figure 1). Figure 1. Ecological niche modelling projected as potential (based on the natural vegetation, t1) and extant (based on land use and natural vegetation, t2) distribution of two terrestrial mammals. Black areas depict part of distributions of (a) Chaetodipus spinatus in the Baja California Peninsula, and (b) Cabassous centralis in Chiapas.

Point occurrence data
We used a database of 19,058 museum specimen records in unique localities of 429 terrestrial mammals occurring in Mexico, following taxonomic nomenclature of Villa and Cervantes (2003). Point occurrence data for each species were compiled from national and international scientific collections (see Acknowledgements) and from the Comisión Nacional para el Conocimiento y Uso de la Biodiversidad (Conabio, www.conabio.gob.mx), and were georeferenced to the nearest 0.01u of longitude and latitude for each locality using 1:250,000 topographic maps (Conabio 1998;www.conabio.gob.mx). The database had 6574 unique localities for Mexico. The average of unique localities per species was 42.5; most species (.200) had less than 15 unique localities, whereas one species (Liomys irroratus) showed 660 unique localities.

Ecological niche modelling and species' potential and extant distributions
To overcome geographic biases on delimiting species distributions, we generated models of ecological niche for each species, using the Genetic Algorithm for Rule set Prediction (GARP; Stockwell & Peters 1999; available for download at www.lifemapper.org/ desktopgarp). We characterized ecological niches using environmental data layers (0.0460.04 pixel resolution), including one dataset with elevation, slope, and aspect (from the US Geological Survey's Hydro-1K data set; www.usgs.gov); climatic parameters, including mean annual precipitation, mean daily precipitation, maximum daily precipitation, minimum and maximum daily temperature, and mean annual temperature (Conabio, www.conabio.gob.mx); and a map (t1) with the original natural vegetation (see Rzedowski 1990), producing species' potential distributions. We also used another dataset with a map (t2) including the 2000 land use (transformed areas) and natural vegetation (see SEMARNAT 2001) and other environmental layers. Using the module of GARP for the projection, we produced species' extant distributions (t2).
For each species (both datasets), we ran 100 replicate models, retaining the 20 models with lowest omission error. We retained the 10 models with moderate commission error (i.e. we discarded the 10 models with areas predicted as 'present' showing greatest deviations from the overall median areas predicted as present across all models). This ''best subset'' of models was summed producing species' distributional maps (Anderson et al. 2003). We further refined these distribution models by delimiting presumed distributional over-predictions with the physiographic provinces (Cervantes-Zamora et al. 1990) where species have been collected. Finally, models of species' potential and extant distributions were overlapped to a grid of 1u latitude61u longitude.

Biogeographical regionalization
Two matrices (matrix t1, using species distributions t1; and matrix t2, using species distributions t2) for PAE were performed for 248 quadrats. If a taxon was present in a quadrat, the entry was assigned '1', and if it was absent, the entry was assigned '0'. A hypothetical area coded with '0' for all columns was added to root the cladogram. Both PAE (t1 and t2) were performed in PAUP 4.0b10 (Swofford 2002), using ACCTRAN optimization (Geraads 1998) and k50 (Goloboff 1993;Luna-Vega et al. 2000;Escalante et al. 2007). We employed consistency (CI) and retention (RI) indices to choose important taxa: CI51.0 denotes geographic synapomorphies (endemic taxa) or geographic autapomorphies (characteristic taxa), whereas CI50.50 may indicate possibly endemic taxa (only when they are geographic synapomorphies with a posterior reversal, indicating a probable extinction) . Occasionally, CI50.33 may as well indicate possibly endemic taxa. For t1 and t2, characteristic and endemic species were mapped with Arc View 3.2 (ESRI 1999), and these species and cladograms were used for biogeographic regionalizations. Province boundaries were taken from the digital maps of Ramírez-Pulido and Castro-Campillo (1990) and Arriaga et al. (1997).

Results
The matrix t1 had 39 uninformative (autapomorphic) species and resulted in 1140 cladograms of 4505 steps (CI50.09 and RI50.79). The strict consensus cladogram had 4532 steps (CI50.09 and RI50.79) and 52 geographic synapomorphies (including all CI.0.33). The matrix t2 had 44 uninformative species and resulted in 3100 cladograms (which is the top number of cladograms able to be retained by our computer) of 4267 steps (CI50.08 and RI50.80). The strict consensus cladogram had 4271 steps (CI50.08 and RI50.80), with 49 geographic synapomorphies (including all with CI.0.33). Synapomorphic and autapomorphic species and their status, and the simplified consensus cladograms under t1 and t2 scenarios are shown in Table I and Figure 2, respectively. By using the potential distribution maps of synapomorphic and autapomorphic species, we obtained a mammalian biogeographic regionalization of Mexico (Table II). In some cases, we estimated the covered area of the species' potential (t1) and extant (t2) distributions, respectively, and calculated the percentage of transformed area.
We found important differences between cladograms produced using species potential and extant distributions. The cladogram resolution in t2 was poorer than t1, because the consensus cladogram had more polytomies and CI was lower. The number of cladograms for t1 was lower than t2, because we were not able to obtain all possible cladograms for t2; however, all cladograms and the consensus cladogram obtained in t2 were shorter than t1 (this had more steps). From 52 geographic synapomorphies in t1, only 49 were present in t2, whereas 10 species were added as synapomorphies and three species as autapomorphies in t2 (Table I). Eleven synapomorphic species in t1 were not identified in t2. There were three autapomorphies in t2 that appeared in t1 as possible synapomorphies (CI.0.33): Orthogeomys cuniculus, Perognathus longimembris and Thyroptera tricolor; these species were present in two or three quadrats in t1, but were reduced to one quadrat in t2. There were three geographic synapomorphies in t1 (being only in two quadrats) that changed to autapomorphies in t2, due to their reduction of their areas of distribution: Corynorhinus mexicanus, Neotoma bunkeri, and Macrophyllum macrophyllum. Five species changed their CI, from 0.5 in t1 to 1.0 in t2: Mimon crenulatum, Molossus bondae, Peromyscus guardia, Scapanus latimanus, and Sorex sclateri. Four species changed from synapomorphies in t1 CI51.0 to CI50.5 in t2: Microtus californicus, Phyllostomus stenops, Sorex ornatus and Tylomys bullaris. In each case (change of CI50.5 to CI51.0 and vice versa), the species' distributions were not remarkably reduced, but the changes were due to a different cladogram topology.
The consensus cladogram of t2 had a basal trichotomy, whereas in t1 it was absent (Figure 2a, b). In t1, the northwest (Sonora+Baja California Peninsula+Sierra Madre Occidental) was initially separated from the rest of country, which was not observed in t2. Moreover, the Sierra Madre Occidental was not linked in t2 to Sonora and the Baja  California Peninsula. In both t1 and t2, the Yucatán Peninsula+Chiapas+Isthmus of Tehuantepec clade was maintained, although in t2 it was separated from the remaining areas of the country.

Mammalian biogeographic regionalization
Based on the consensus cladogram t1, we built a biogeographical regionalization of the terrestrial mammals in Mexico, contrasting differences with t2 ( Figure 3; Table II).
Regions. Traditionally, Mexico has been divided into the Nearctic and Neotropical regions ( Figure 3; Table II), with a transitional zone referred to as the Mexican Transition Zone (Halffter 1962(Halffter , 1965(Halffter , 1972(Halffter , 1974(Halffter , 1976(Halffter , 1978(Halffter , 1987Escalante et al. 2004;Morrone 2005 (Figure 2a, b). The Sonora+Baja California Peninsula+Sierra Madre Occidental clade is separated from the rest of the country in t1. It was justified by Lasiurus xanthinus, but this species is distributed mostly in the Nearctic region in Mexico. For the other clade (rest of the country) of t1, the species that justified this grouping were Philander opossum and Sigmodon hispidus, both occurring throughout the Neotropical region in Mexico. The species restricted to the Neotropical region, located in the Yucatán Peninsula+Chiapas+Isthmus of Tehuantepec clade, were Alouatta pigra, Artibeus jamaicensis, Ateles geoffroyi, Lontra longicaudis, Mazama americana, Odocoileus virginianus, and Saccopteryx bilineata (Table II).

Campeche, Yucatán and Quintana Roo
The consistency (CI) and retention (RI) indices have been used to identify geographic synapomorphies (similar to cladistic analysis), which should be overlapped to delineate areas of endemism. Only CI and RI.0.5 (or even 0.33) can be used as possible synapomorphies . However, on our cladograms, there are major clades justified by species with poor CI, but better RI. Because RI had advantages over CI (Farris 1989), we still maintained some species with low CI in major clades (marked with *). All the other species had CI and RI.0.5. In t2, the species that supported the Neotropical region of t1 had more steps (Philander opossum: L510; Sigmodon hispidus: L529), and were in a different position on the cladogram than in t1. The third clade was defined by Baiomys taylori, Mephitis mephitis, Myotis californica, and Sylvilagus floridanus. Baiomys taylori is distributed nationwide, except for the Baja California and Yucatán Peninsulas; Mephitis mephitis is restricted to the north; Myotis californica is restricted to the north and Baja California Peninsula; and Sylvilagus floridanus is distributed nationwide, except Baja California Peninsula.
Subregions. We followed the classification of Morrone et al. (1999) and Morrone (2001), who described the North American Pacific and Caribbean subregions, and added a Mexican Transition Zone.
Dominions. We propose the following dominions under t1:  Escalante et al. (2004) suggested that the Chiapas Highlands should be included in the Mexican Transition Zone. This zone was recognized by the presence of panbiogeographic nodes of mammals (Escalante et al. 2004), but we did not find any synapomorphy, except Corynorhinus mexicanus (synapomorphy in t1) is distributed in the Sierra Madre Oriental and Transmexican Volcanic Belt, whereas Megadontomys nelsoni (autapomorphy) is distributed in the boundary of these provinces. 4. Mesoamerican dominion: Morrone and Márquez (2003) defined this dominion including all the tropical areas of Mexico, mainly the centre-southern coasts, Isthmus of Tehuantepec and Chiapas, except the Yucatán Peninsula. We did not find species distributed in all the dominion, since all Neotropical species identified occurred also in the Yucatán Peninsula. Thus, this domain requires additional analyses with the inclusion of more taxa. 5. Antillean dominion: it includes the Yucatán Peninsula. Its endemic species are the same as the Yucatán Peninsula province (see below).
In t1 and t2 cladograms, this province coincides with the state of Sonora, with the endemics Neotoma varia (insular), Peromyscus crinitus, P. pembertoni (insular) and Sorex arizonae. 4. Mexican Plateau: has been divided in two different provinces: Northern and Southern Plateau (Arriaga et al. 1997;Escalante et al. in press), but we failed to find this separation. It was defined by species that may extend their distributions to the rest of North America: Ammospermophilus interpres, Bison bison, Cratogeomys castanops, Dipodomys compactus, Geomys personatus, Microtus pennsylvanicus, Myotis planiceps, Onychomys arenicola, Peromyscus hooperi, and P. nasutus. Microtus pennsylvanicus was not an endemic under t2 due to 19% reduction of its distribution. 5. Tamaulipas: we failed to identify important species (endemic and autapomorphic), although a detailed analysis of species of clades of the rest of the country allowed identification of two species, mostly distributed in this province: Nycticeius humeralis and Onychomys leucogaster. They reduced their distributions in t2 by 35% and 18%, respectively. 6. Sierra Madre Occidental: located on the homonymous physiographic province. Escalante et al. (2004) did not find panbiogeographic nodes, and our analysis under t1 showed that the Sierra Madre Occidental province is more closely related to the Nearctic than to the Neotropical provinces. This pattern suggests that this province perhaps does belong to the Nearctic region. Conversely, Marshall and Liebherr (2000) undertook a cladistic biogeographic analysis, where they considered that the northern portion of the Sierra Madre Occidental and the Mexican Plateau are a unique area of endemism, whereas a small area in the south differs from the rest of the province. We found few endemics: Sciurus aberti inhabits the entire province, but was only recognized as a synapomorphy under t2, because it also occurs in a small part of the Mexican Plateau. Euderma maculatum occurs in a small area in the central part, and Neotoma palatina is located only in the south. Surprisingly, under t2, this area was relocated to the clade of the rest of the country, far from the Nearctic provinces of Sonora and Baja California Peninsula, and close to the northern Mexican Plateau (Figure 2b) (Fa 1989;Fa & Morales 1991;Escalante et al. 2002;Monroy-Vilchis et al. 1999). In our analyses, Nelsonia goldmani, Neotoma nelsoni, Pappogeomys alcorni, Sorex emarginatus, and Zygogeomys trichopus were autapomorphic for this province under t1 and t2.
Corynorhinus mexicanus was originally distributed in the Sierra Madre Oriental and the Transmexican Volcanic Belt in t1, but in t2 it was lost from the Transmexican Volcanic Belt. Megadontomys nelsoni was on the Transmexican Volcanic Belt in t1, and it was restricted to the Sierra Madre Oriental province in t2.
9. Balsas Basin: the Balsas Basin is located in the lowlands south to the Transmexican Volcanic Belt. We were not able to identify endemic species for this province. 10. Sierra Madre del Sur: Liomys salvini, Myotis ciliolabrum and Sylvilagus insonus are characteristic of this province under t1 and t2. 11. Mexican Pacific: its endemic species are Cratogeomys fumosus and Xenomys nelsoni, found as possible geographic synapomorphies under t1 and t2, whereas Rhogeessa mira was autapomorphic. Megatondomys thomasi, Peromyscus melanocarpus and P. melanophrys occur in this province, but expand their distribution to the Balsas Basin and adjacent provinces. Peromyscus melanocarpus and P. melanophrys were synapomorphies in t2. 12. Mexican Gulf: Geomys tropicalis occurred in the northern portion of this province as a synapomorphy under t1 and t2, but with an evident reduction of its distributions of 50%. Orthogeomys lanius was autapomorphic under both t1 and t2. 13. Isthmus of Tehuantepec: the Oaxaca province was synonymized by Morrone (2001) with the Sierra Madre del Sur province. Ramírez-Pulido and Castro-Campillo (1990) defined this province as Oaxaco-Tehuacanense; we followed their limits for this province, although it requires further refinement including other taxa. Escalante et al. (2004) found a generalized track in this zone, suggesting a shared history of its biota. The endemics under t1 were Centronycteris maximiliani, Enchistenes hartii, Habromys lepturus, Megadontomys cryophilus, Microtus oaxacensis, M. umbrosus, Molossops greenhalli, Orthogeomys cuniculus and Rheomys mexicanus. Orthogeomys cuniculus changed its status of synapomorphy in t1 to apomorphy in t2. Rheomys mexicanus was not identified in t1, but in t2 was present as a synapomorphy. 14. Chiapas: our analyses coincided with the Chiapaneca province of Ramírez-Pulido and Castro-Campillo (1990), which is larger than the province of Morrone (2001), who added the Soconusco province of Arriaga et al. (1997). It is likely that the lowlands of Chiapas do not belong to this province. Escalante et al. (2003) found an important area of endemism for mammals in the highlands of Chiapas, and Escalante et al. (2004) identified two panbiogeographic nodes in this province (nodes 8 and 9). Further, Escalante (2003) prioritized Altos de Chiapas as an important area for mammal conservation. We found the following species (some with their distributions extended to Central America): Cabassous centralis (Figure 1a) Morrone's (2001) provinces, and Escalante et al. (2003) found a similar endemism area. Under t1 and t2, this province was identified, but five species changed their status in the cladograms between both scenarios.

Changes in biogeographical regionalization
By combining ecological niche modelling and PAE, we produced two biogeographic regionalizations under an historical scenario of no deforestation (scenario t1), and including deforestation (transformed areas into agrosystems and urban settlements; scenario t2). Presumably, species show ecological niche conservatism, so transformed areas constitute unsuitable niches for population persistence. Consequently, deforestation can have profound effects on species distributions (Sánchez-Cordero et al. 2005b;Fuller et al. 2007), and consequently on biogeographic regionalization. Relevant changes between t1 and t2 cladograms were observed, resulting in dramatic loss of biogeographical patterns at major scales. In the t2 cladogram, the boundary between Nearctic and Neotropical virtually disappears. In addition, the Sierra Madre Occidental province was not identified. Those changes probably are due to modification of individual distributional areas of species when a change in ecological conditions occurs. Species distributional areas may change by (1) reducing distribution areas including only remnant suitable habitats, (2) modifying their distributional area using transformed habitats, and/or (3) moving to other remnant suitable habitats (Holt 1990). As species can respond differentially to drastic habitat transformation, distributions are consequently modified, and therefore, biogeographic patterns are lost.
The Nearctic-Neotropic regions and the transitional zone are biogeographic patterns that were conformed over evolutionary times. Each region has a distinctive biota, with particular geological and ecological conditions, and a shared history. Nevertheless, these patterns can be lost due to intensive transformation and degradation of natural habitats (see Challenger 1998). Similarly, the Sierra Madre Occidental is a relevant montainous chain in Mexico, harbouring part of the richest pine-oak forest in the world, but only a few large patches of forest remain capable of maintaining its rich biota (Ricketts et al. 1999). This province has been identified by species of different taxa (Morrone 2005); so, it is likely that loss of biogeographic patterns for mammals may reflect equivalent loss of patterns for other taxa.

Conclusions
Changes on biogeographic regionalizations should be expected over evolutionary time. Changes in distributional patterns of biota result from unique geologic, geographic, and climatic changes at a global scale. In ecological time scales, climate change (Peterson et al. 2002a(Peterson et al. , 2002b and deforestation (Sánchez-Cordero et al. 2005b) can presumably modify biogeographic regionalizations as well. We showed that changes in species' distributions attributable to deforestation (scenario t2) resulted in significant changes in biogeographical regionalization using terrestrial mammals as a case study. If further deforestation continues, it is likely that some natural patterns of species distributions may be lost. Studies incorporating areas of endemism and transition zones can help identify conservation priorities.