Introduction

Climate change (CC) is modifying the habitat and distribution of wild fauna, causing distribution area reduction and/or altitudinal or latitudinal movements of species (^{Thuiller et al. 2004}; ^{Leach et al. 2014}) in response to increased temperatures (^{Camacho et al. 2010}) and changes in precipitation, thereby increasing their risk of extinction. Species with small populations, habitat specialists, or those that are included in a category of risk could be the most vulnerable to a reduced distribution range and to degradation or loss of their habitat in the future (^{Thuiller et al. 2004}; ^{Camacho et al. 2010}). Models of potential distribution and ecological niche (the set of ecological conditions required by a species to maintain populations within a defined region) are reliable tools for determining the factors that control the distribution of species (^{Hutchinson 1957}; ^{Peterson et al. 2002}; ^{Peterson 2003}; ^{Siqueira and Peterson 2003}; ^{Peterson et al. 2004}; ^{Thuiller et al. 2004}; ^{Soberón and Peterson 2005}; ^{Jackson et al. 2009}; ^{Armenteras and Mulligan 2010}) as well as their environmental requirements and therefore the availability and suitability of habitat (^{Peterson 2003}; ^{Siqueira and Peterson 200}3; ^{Peterson et al. 2004}; ^{Parra et al. 2005}; ^{Sánchez-Cordero et al. 2005}; ^{Levinsky et al. 2007}; ^{Elith and Leathwick 2009}; ^{Jackson et al. 2009}).

Detailed observations (occurrence data) of the species and maps of climatic and slope variables, are required as explanatory variables of species distribution (^{McCarty 2001}; ^{Peterson et al. 2004}; ^{Sánchez-Cordero et al. 2005}; ^{Soberón and Peterson 2005}). Previous studies have used climatic variables to project future distributions of species (^{Hampe 2004}; ^{Thuiller et al. 2004}; ^{Bomhard et al. 2005}; ^{Heikkinen et al. 2006}; ^{Triviño et al. 2011}).

Knowledge of the current potential distribution and future of species in danger of extinction in Mexico would allow conservation efforts to be focused on sites that present the ecological features required by these species. Therefore, it is important to determine the implications of climate change for these species, using the Tehuantepec jackrabbit Lepus flavigularis, as a case study. This species is endangered according to NOM-ECOL-059 (^{SEMARNAT 2010}), and Red List (^{IUCN 2012}), is endemic to the south of the Isthmus of Tehuantepec and has a very restricted distribution (^{Cervantes 1993}; ^{Lorenzo et al. 2000}; ^{Lorenzo et al. 2006}; ^{Carrillo-Reyes et al. 2010}). This jackrabbit inhabits savannas and introduced grasslands (^{Farías 2004}; ^{Lorenzo et al. 2006}; ^{Carrillo-Reyes et al. 2010}; ^{Sántiz et al. 2012}), and feeds mainly on grasses (66.7 %; ^{Lorenzo et al. 2011}). The major current threats to this species are loss and fragmentation of its habitat and poaching (^{Lorenzo et al. 2006}; ^{Cervantes et al. 2008}; ^{Carrillo-Reyes et al. 2010}; ^{Rioja et al. 2011}).

The effect of climate change on 87 species of lagomorphs (rabbits, jackrabbits and pikas) is predicted to be a reduction and/or modification in their distribution, with latitudinal movements towards the poles, especially in the species of rabbit and jackrabbit, while altitudinal changes are predicted for the pikas. Specifically, a 45 % decrease in the distribution of L. flavigularis is predicted, along with latitudinal changes of around 1º towards the poles and an average altitudinal increase of approximately 450 m (^{Leach et al. 2014}). That study subjectively evaluates the effect of climate change on distribution (good, medium or poor) from just eight historical and modern records. For these reasons, it is considered that the model offers a poor representation of past and future distributions of L. flavigularis, and is thus unreliable (^{Leach et al. 2014}). In the case of grasses, the global warming or the increase in CO2, may favor the distribution of some species depending on their photosynthetic pathways (^{Collatz et al. 1998}; ^{Duckworth et al. 2000}; ^{Hopkins and del Prado 2007}; ^{Osborne 2008}; ^{Lattanzi 2010}; ^{Chamaillé-Jammes and Bond 2010}).

The present study aims to: a) model the potential distribution of L. flavigularis and grasses that support it in the south of the Isthmus of Tehuantepec, Oaxaca, México, b) understand the implications of climatic variables (temperature and precipitation), on the distribution of suitable areas for the jackrabbit in the present (2014) and future (2050 and 2070), under climate change scenarios, and c) identify the specific climatic variables that determine the distribution of these species (jackrabbit and relevant grasses).

Material and Methods

Study area. The study area is located in the state of Oaxaca, México, in the south of the Isthmus of Tehuantepec (16.61º N, -95.36º W and 16.11º N, -94.24º W; 0 a 100 m; ^{Ortiz et al. 2004}; Figure 1). The climate is warm sub-humid with the occurrence of a dry period in mid-summer. Annual mean temperatures are above 26 °C and total annual precipitation is 800 mm (^{García 1988}). The vegetation is composed of scrub and mangrove with submerged vegetation in the coastal lagoons (^{Ortiz et al. 2004}). There are also extensive areas of savanna with isolated nanche (Byrsonima crassifolia (L.) Kunth (1822) and morro (Crescentia alata Kunth (1819) trees (^{Rzedowski 2006}; ^{Sántiz et al. 2012}). In the driest parts of the Isthmus there are important areas of tropical low deciduous and low spiny deciduous forests that are remnants of tropical medium sub-deciduous forest (^{Torres-Colín 2004}), and introduced grasses occur at sites from which the original vegetation has been eliminated. The Gramineae that dominate the herbaceous stratum of the savannas and introduced grasslands are of the genera Aristida, Bouteloua, Cathestecum, Cenchrus, Digitaria, Eragrostis, Panicum, Paspalum, and Schizachyrium (^{Rzedowski 2006}; ^{INEGI 2012}; ^{Sántiz et al. 2012}).

Figure 1 Location of the study area in the south of the Isthmus of Tehuantepec, Oaxaca, México. 

Potential distribution models. The models were constructed using the Maxent v. 3.3 software (^{Elith et al. 2010}) that generates probability maps of suitable or unsuitable habitat for the species of interest (^{Elith and Leathwick 2009}). Maxent only requires presence data and provides species response curves for different environmental variables, estimating the importance of each variable in terms of the distribution of the species (^{Phillips et al. 2006}). Records (205) of L. flavigularis were compiled from the GBIF (Global Biodiversity Information Facility; www.gbif.org) database of 1960 to 2000, and field records from 2001 to 2014 were also used. Additionally, 124 records of 8 species of grasses that feed the jackrabbit were collected from 1959 to 2012, and were taken from the databases of the CONABIO (^{Herrera 1997}; ^{Aragón 2001}; ^{Panero 2003}; ^{Villaseñor 2008}), data provided by ^{Rioja et al. (in review)} and field data from 2011 to 2014. The database of records was cleansed of repeated records, those from outside the study area, and those located in water bodies by errors.

Explanatory variables. The climatic variables for 2000, 2050, and 2070 conditions were taken from WorldClim (resolution: 30"; ^{Hijmans et al. 2005}). Initial selection was carried out with a test of correlation between the climatic variables, choosing those with correlation values lower than 0.7 (^{Gormley et al. 2011}). A first analysis in Maxent was then conducted with the variables chosen according to the correlation test and variables with a contribution of less than 0.5% were eliminated (^{Warren et al. 2014}) in order to produce the final models. The variables considered in the final analysis for jackrabbits were: slope (slop), annual mean temperature (BIO1), mean diurnal range (BIO2), temperature Seasonality (BIO4), mean temperature of coldest month (BIO6), temperature annual range (BIO7), mean temperature of driest quarter (BIO9), annual precipitation (BIO12), precipitation of wettest month (BIO13), precipitation seasonality (BIO15), precipitation of wettest quarter (BIO16), and precipitation of warmest quarter (BIO18). The variables used for grasses were: slope (slop), mean diurnal range (BIO2), isothermality (BIO3), temperature seasonality (BIO4), temperature annual range (BIO7), mean temperature of warmest quarter (BIO10), precipitation of wettest month (BIO13), precipitation of driest month (BIO14), precipitation seasonality (BIO15), precipitation of wettest quarter (BIO16), precipitation of warmest quarter (BIO18), precipitation of coldest quarter (BIO19).

In order to highlight the effect of climate on the distribution of the jackrabbit and to forecast trends for the years 2050 and 2070, the bioclimatic layers of the HadGEM2-ES model with scenarios 4.5 and 8.5 were chosen. These represent the less severe and most extreme conditions with high concentrations of CO2, respectively (^{Hijmans et al. 2005}; ^{Riahi et al. 2007}). All of the layers of the explanatory variables were homogenized to the raster format (pixels of 30 m), datum WGS84 and to geographic coordinates, and subsequently to the ASCII format. The models of 2014, 2050 and 2070 were constructed with 123 records for training and 82 records for test with a total of 205 records, and for grasses were developed with 124 total records, of which 93 were used in the calibration and 31 for the validation of the models. A total of 20 replicates were run, with 1000 iterations, and a total of 5 distribution maps were selected based on highest AUC (area under the curve) values. The AUC of the Receiver Operating Characteristic (ROC) graphics was used. This statistic allows evaluation of correctly classified cases for all of the prediction thresholds (from 0 to 1). The value of the AUC is between 0.5 and 1. A value of 0.5 equates to a random classification, while a value of 1 indicates a perfect fit, i.e., that all of the cases have been correctly classified (^{Fielding and Bell 1997}).

The five chosen distribution maps were converted to maps of presence-absence, fixing the thresholds of probability of species presence (^{Jiménez-Valverde and Lobo 2007}) in ArcGis 10.2 software (^{ESRI 2011}). Fixed cumulative value 10 threshold was selected because the species of interest has a limited distribution. When predicted area is limited by this factor, the false positives are also reduced (predicted distribution sites, which lack suitable conditions for species distribution), and the true positives increase (predicted distribution sites, in which suitable conditions for species distribution are present). This area could be of utility in the identification of new potentially suitable zones for the establishment of the species, possible zones of reintroduction or vital zones for establishing protection measures, and reducing human pressure (^{Peterson et al. 2011}). The final maps were obtained from multiplying the five best maps with map algebra in Spatial analyst tool of ArcGis 10.2 software (^{ESRI 2011}).

In order to know the types of vegetation in climate suitable sites where the jackrabbits and grasses occur, a map of the vegetation and land use for 2015, was developed through a multispectral satellite image of Landsat 8 Oli, and with resolution of 30 m from 2015. Maximum likelihood algorithm was used in the GRASS module for QGIS (version 2.10), to carry out a supervised classification. Classification discriminated 80% of the errors obtained to carry out a classification at random. The kappa index resulted in an accuracy of 82%.

Results

Environmental requirements of Lepus flavigularis. According to the records of the presence of the Tehuantepec jackrabbit in the study area, it is found in zones with little slope (0 - 7.34º), with mean annual temperatures of 27.15 ºC, and with total annual precipitation of 1,165 mm in 2014 (Table 1).

Table 1 Mean values of the conditions of temperature (ºC) and precipitation (mm) for 2014 in the site where Lepus flavigularis has been recorded, and the possible conditions in 2050 and 2070, according to the HadGEM2-ES model with scenarios 4.5 and 8.5, in the Isthmus of Tehuantepec, in Oaxaca, México. 

Distribution models of Lepus flavigularis and grasses. Over the total study period, distribution models estimate an increase in suitable area for the species in both scenarios (4.5 and 8.5). The climate suitable area for the presence of jackrabbits and its food resource according to the models, is 9,274 has (Figure 2), in the current scenario (2014). In the future, scenario 4.5 estimated a reduction of the available range to 7,510 ha's, by 2050 and an increase to 11,071 ha's in 2070. However, scenario 8.5 estimates increases of the suitable range to 11,649 ha's in 2050 and 10,111 ha's in 2070 (Figure 3).

Figure 2 Based on the model for 2014, the potential distribution in the south of the Isthmus of Tehuantepec of: habitat suitable for the presence (in red) of Lepus flavigularis and forage grasses; the vegetation of savannas (dark blue); and grasslands dominated by introduced species (light yellow). 

Figure 3 Using the models for 2050 and 2070, scenarios 4.5 and 8.5, potential distributions in the south of the Isthmus of Tehuantepec of: suitable habitat for the presence (in red) of Lepus flavigularis and forage grasses; the vegetation of savannas (dark blue); and grasslands dominated by introduced species (light yellow). 

Contribution of the variables. According to a Jacknife analysis (which determines the variable that most contributes to the potential niche), of the 12 explanatory variables included in the potential distribution models of suitable habitat, 10 variables explain more than 85% of the ecological niche of L. flavigularis, in the models. At the current scenario, variables BIO18 and BIO13 contributed the most to the model. At the 4.5 scenario, most important variables were BIO13 and BIO7 in 2050, and BIO13 and BIO6 for 2070. At the 8.5 scenario, variables which contributed the most were BIO13 and BIO12 in 2050, and BIO18 and BIO7 for 2070 (Table 2). For grasses, the variables with the greatest contribution to the construction of models was BIO2 for the 3 years (2014, 2050 and 2070), and the two scenarios analyzed, followed by BIO16 in 2014, and 2070 with the scenario 4.5, BIO13 in 2050 with the scenario 4.5, BIO 19 in 2050 with the scenario 8.5 and the isothermality for 2070 with the scenario 8.5 (Table 2).

Table 2 Estimated sizes of suitable grassland habitat, and variables of greatest contribution to the ecological niche models of Lepus flavigularis for each of the years 2014, 2050 and 2070. 

Discussion

The potential distribution of suitable habitat (sites climatically suitables) for the jackrabbit (and grasses) is restricted to types of vegetation as savannas and introduced grasslands that present little slope. This coincides with the information available for this species, which places it in zones dominated by open grasslands (^{Cervantes 1993}; ^{Vargas 2000}; ^{Lorenzo et al. 2006}), or savannas that are associations of natural grasslands with isolated morro trees (Crescentia alata), or isolated patches of nanche trees (Byrsonima crassifolia) typical of flat or slightly inclined areas (^{Farías 2004}; ^{Sántiz et al. 2012}). Unfortunately, the vegetation inhabited by the jackrabbit is strongly threatened by agriculture (e.g., crops of sorghum and maize) and livestock production, activities that have led to the degradation and reduction of the natural vegetation, and that cause the disappearance of the original pristine grasses and modify the plant species composition, as reported in other regions of the country (^{Anderson 2006}; ^{CONABIO 2006}; ^{Rzedowski 2006}; ^{Estrada-Castillón et al. 2010}).

Climatic variables such as Precipitation of Wettest Month (BIO13), and Precipitation of Warmest Quarter (BIO18), Min Temperature of Coldest Month (BIO6), Temperature Annual Range (BIO7) and annual precipitation (BIO12), are also factors of great importance in the delimitation of the niche of this jackrabbit. In previous studies, temperature and precipitation have been closely related to the reproduction and density of lagomorphs (rabbits and jackrabbits; ^{Pévet 1987}; ^{Portales-Betancourt et al. 2012}). The increased availability of food signals the beginning of the reproductive season (^{Pévet 1987}). In Texas, USA, the peak in reproduction of the lagomorphs (from January to April) coincided with an increase in this vegetation-type that is determined mainly by the temperature and quantity of precipitation (^{Dunn et al. 1982}). Lepus californicus in New Mexico, USA, and the Mapimi Biosphere Reserve in Durango, Mexico, presented differences in densities between years, due to variations in levels of precipitation (^{Daniel et al. 1993}; ^{Portales-Betancourt et al. 2012}). In L. flavigularis, there is a strong positive relationship between seasonality and reproduction, defined by photoperiod (day length) and ambient temperature and rainfall, since this influences testicular development (^{Rioja et al. 2011}) and, as with other jackrabbits such as L. californicus, this influences the phases of estrus and female receptivity (^{Portales-Betancour et al. 2004}; ²⁰¹²).

This study estimated an increase in the distribution of the jackrabbit of 19.3% with the scenario 4.5 and 9% with scenario 8.5, in a period of 56 years (2014 - 2070), with no changes to its latitudinal and altitudinal distribution predicted. These results differ from that estimated previously, where a 45% decrease in its distribution was predicted with latitudinal changes of around 1º towards the poles and a mean altitudinal increase of around 450 m over a longer time period (150 years; 1930 to 2080; ^{Leach et al. 2014}).

The HadGem model with the 4.5 and 8.5 scenario in 2050 and 2070 estimates increased temperature and decreased precipitation in the study area and, according to the predictions of the models, the future availability of sites with weather conditions favorable for jackrabbits and their food (grasses) may increase. However, even if there are appropriate climatic conditions for the establishment of jackrabbits and their food in the future, the outlook for L. flavigularis is not favorable. Changes in vegetal cover and land use, agriculture, livestock, and the presence of human settlements, have led to the decrease of the area of distribution and abundance of L. flavigularis (^{Lorenzo et al. 2000}, ²⁰⁰⁶, ²⁰⁰⁸, ²⁰¹⁴; ^{Vargas 2000}; ^{Farías 2004}; ^{Sántiz 2002}, ²⁰⁰⁵), and have led to the fragmentation of its habitat, causing the isolation of their populations and low genetic variability, which has aggravated their situation (^{Flux and Angermann 1990}; ^{Cervantes 1993}; ^{Vargas 2000}; ^{Rico et al. 2008}).

Moreover, changes in cover and land uses, have caused loss and modification of the vegetal cover that favors both the jackrabbit (savannas) and an increase in human settlements. In the study area, an increase of 2.1% per year has been reported in the area of human settlements since 1986. Moreover, savannas have decreased their distribution at a rate of 1.2% per year in the same period, with a loss of 6,850.6 ha's over a period of 26 years (1986-2012; unpublished data). Overhunting is also another anthropogenic factor that has caused a decrease in the density of the populations of this species, contributing to the current situation (^{Farías 2004}; ^{Lorenzo et al. 2008}). The impacts of these threats in the future to populations of jackrabbits and habitat could be greater and felt over a shorter time span than those of climate change.

Implications for the conservation of Lepus flavigularis. Conservation of L. flavigularis will present new challenges; success will depend on our capacity for working in unison with local inhabitants and decision-makers (^{Lidicker 2015}). While climate change will modify the area considered as suitable habitat for L. flavigularis, other factors already exist that threaten and modify the distribution of its populations and habitat. Land use changes for activities such as livestock production and agriculture, as well as the expansion of human settlements, are already reducing the natural vegetation (savannas) in the study area (^{Lorenzo et al. 2006}; ^{Rzedowski 2006}; ^{INEGI 2012}). For this reason, promotion of the establishment of a communal ecological reserve, administrated by the local authorities and with the technical assistance of academic personnel is one option by which to avoid the continued reduction of the habitat of the jackrabbit and to contribute to the protection of the species. Local protected areas have proven to have strengths related to dedication of the required land and its heritage of natural resources by the local communities for conservation. Such actions can be expected to also improve the quality of life for local inhabitants, since these people are currently in a socioeconomically poor situation with a lack of opportunities for work (^{Vargas 2001}). Further research is necessary to improve our understanding of the composition of and interactions among the species of grasses comprising the plant associations in which the jackrabbit lives. It will also be critically important to establish appropriate management strategies for the grassland and savanna communities in order to avoid further degradation of the remaining populations of the Tehuantepec jackrabbit.