Impacts of land cover change on the plant resources of an endangered pollinator

One of the key drivers of pollinator declines is land cover change. We documented for the first time the impacts of over three decades of land cover change in Mexico on the plant resources of an endangered migratory pollinator, the Mexican long-nosed bat, Leptonycteris nivalis. This species is considered endangered under national and international criteria due to population declines over 50% in the past 10 years. Pregnant females of this bat species migrate every year following the blooms of Agave spp. from central Mexico to the southern United States; moving pollen over its 1,200 km long migratory corridor and pollinating distant populations of Agave spp. Increases in human populations density and agricultural expansion may be reducing agave habitat over time. The objective of our study is to understand the land cover change trends in the northern range of the bat and identify potential fragmentation patterns in the region. We analyzed changes that occurred in three vegetation types where agaves are found in five time periods 1985, 1993, 2002, 2007 and 2011. The area of the three vegetation types selected was reduced by using only the overlap with potential agave habitat created with ecological niche modeling algorithms to obtain the available agave habitat. We then calculated fragmentation metrics for each period. We found a significant portion of habitat lost mainly due to expansion in agriculture. The total number of patches increased after 1985. Only 9% of the available agave habitat in 2011 is inside the limits of protected areas. We recommend restoring agave populations in depleted areas to help prevent soil erosion and provide multiple socio-economic benefits for the region in the short term, and, in the long-term maintaining foraging resources for nectar-feeding bats.


INTRODUCTION
Human impacts on planetary biodiversity have increased greatly over the past century, prompting some conservationists to propose that we have entered a defining new period, the Anthropocene (Lewis & Maslin, 2015). Declines in biodiversity vary across taxa and are driven by multiple threats, including habitat loss and fragmentation, climate change, and overconsumption (Lacher & Roach, 2018). All these factors combine to varying colonies in the northern range (Hensley & Wilkins, 1988;Moreno-Valdez, Honeycutt & Grant, 2004;Adams, 2016). Researchers have suggested that pregnant females give birth in northern Mexico prior to their arrival in Emory cave at Big Bend National Park in Texas (Easterla, 1972). Therefore, it is a conservation priority to maintain foraging resources in the northern range of L. nivalis.
Among terrestrial habitats, arid and semi-arid lands are under particular threat due to degradation and fragmentation associated with grazing and agriculture (Galvin et al., 2008). This has also been a recent trend in northern Mexico, much of this driven by urbanization and agriculture (Pérez Miranda et al., 2012). The objective of this analysis is to assess the degree of land-cover change in the northern range of L. nivalis, and to quantify trends in the distribution of dry forest and desert scrub habitats where agaves occur. We hypothesize that increasing human population density and associated agricultural and livestock expansion have reduced available agave habitat, impacting both the availability of resources for L. nivalis and traditional human livelihoods in the region. In this study we used geographic information system tools to analyze the effects of three decades of land cover change (LUCC) on the potential area of distribution of Agave species occurring in the northern Mexican portion of the bat's range. We identified fragmentation trends and suggest conservation opportunities to help maintain the endangered L. nivalis-Agave pollination corridor.

MATERIALS & METHODS
The study area encompassed the potential distribution of the Agave species that form the L. nivalis-Agave pollination corridor (created in Gómez-Ruiz & Lacher (2017)). The northern portion of the corridor is where L. nivalis females give birth and where agaves are the main food source for the species (US Fish and Wildlife Service, 2018). We focused this analysis on the northern portion of the corridor within Mexico (Fig. 1). We focused on paniculate Agaves (subgenus Agave). There were 11 possible species in the region used in the analysis of the pollination corridor (Gómez-Ruiz & Lacher, 2017) which covered a much larger area, and nine were included in the analyses. The region selected for the analysis in this paper has at least eight confirmed species. The entire region used in this study is contained within the known points of occurrence and distribution of L. nivalis (Gómez-Ruiz & Lacher, 2017).
We used Land Use/Land Cover (LC) maps, scale 1:250 000, created by Mexico's National Institute of Geography and Statistics (INEGI). To date, INEGI has produced LC maps at 1:250,000 scale for the following time periods: Series I 1985, Series II 1993, Series III 2002, Series IV 2007, and Series V 2011. These maps are available for download from INEGI's website (http://www.inegi.org.mx) in vector format and are the best available information on land use and land cover for Mexico at a regional scale. We used the software ArcGIS 10.2 R to convert the LC maps to raster format with a pixel size of 100 m and projected them to North America Albers Equal Area Conic to estimate areas correctly. The Land Use/Land Cover classes used by INEGI in each period varies due to adjustments in the methodologies and the use of new types of interpretable inputs (aerial photography, satellite imagery, etc.) therefore we reclassified into only nine classes of interest (Table 1). We used the Land Change Modeler for ArcGIS extension (2.0) (Clark Labs R ) and the software ENVI 5.2 R to obtain land cover change statistics and transition matrices. In addition, we produced change maps of the following vegetation classes where agaves occur: desert scrub, conifer and deciduous forest. We calculated percent change for each class. Percent change is the area for a class divided by the area of a class in the later cover image and multiplied by 100.
To identify potential fragmentation trends on the agave corridor, we created maps of available agave habitat for each time period, identifying the overlap of the potential agave habitat created in previous studies (Gómez-Ruiz & Lacher, 2017), with the areas of the generalized land cover classes where wild agaves occur. Other studies have used ecological niche models to evaluate the impacts of land cover change on species (Sánchez-Cordero et al., 2005;Peterson et al., 2006;Ríos-Muñoz & Navarro-Sigüenza, 2009;López-Arévalo et al., 2011;Yañez Arenas et al., 2012). First, we extracted from the LC maps the areas of the three generalized land cover classes where wild agaves occur (desert scrub, deciduous forest, conifer forest), and created a binary raster where the value of 1 indicated presence of the targeted classes. Next, we used the potential distribution maps of agaves and combined them into one binary raster where 1 indicated presence of at least one agave species, and 0 indicated absence of all species. Then we summed the binary raster obtained from the LC map and the binary raster obtained from the agave distribution models. We reclassified the output raster to create a binary raster where 1 indicated available agave habitat. This raster had a pixel resolution of 1 km. Finally, using the raster of available agave habitat, we calculated landscape metrics commonly used as indicators of fragmentation: total number of patches, mean patch size, and mean nearest neighbor distance. These metrics were obtained using Patch Analyst extension for ArcGIS (Rempel, Kaukinen & Carr, 2012). Additionally, we estimated the amount of remaining agave habitat in the 2011 land cover map that is located within protected areas designated by Mexico's national commission on protected areas (CONANP). In this study, we focused only on human-driven land-use changes and did not include other factors, such as climate change, that could affect the distribution of agaves.

RESULTS
Of the three generalized land cover classes where the agaves are found, desert scrub had the largest cover area in all five LC maps used ( Table 2). The land cover change analysis indicates that desert scrub was the class with the largest negative net change from 1985 to 2011, and most of this change occurred between 1985 and 2002 (Fig. 2). The annual net loss in desert scrub area has decreased from the early time intervals (1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993) to the latest (2007-2011) ( Table 3). The analysis of contributions to the net change in desert scrub shows that most of the area transitioned to agriculture (Fig. 3).  The desert scrub area loss between 1985 and 1993 occurred in areas located in the central eastern portion of Coahuila adjacent to Nuevo Leon; and the area loss between 1993 and 2002 occurred in areas located in the northern portion Coahuila within the study area.
Our analysis indicates that conifer forest and deciduous forest increased mainly between 1993 and 2002 (Fig. 2, Table 3). Deciduous forest increased primarily in the northeastern portion of the state of Coahuila, within the study area; and the increase in conifer forest is   stable after 1993. The mean nearest neighbor distance between patches remained similar in all the time periods analyzed (Table 4).

DISCUSSION
The vegetation of northeastern Mexico ranges from low elevation semi arid scrib (Tamaulipan Biotic Province) and arid high elevation Chihuahuan desert (Lacher, 1999) to a complex of woody formations including mid-elevation Mexican shrublands and high elevation oak and pine forests (Valiente-Banuet & Verdú, 2020). There is exceptional diversity in the region due to the complex interaction of these vegetation formations, and regions like the Mexican shrublands are among the most poorly studied of Mexico's ecosystems (Valiente-Banuet & Verdú, 2020). These regions also show high levels of fragmentation in a recent mapping exercise (Haddad et al., 2015) resulting in a high likelihood of future local species loss through extinction debt (Halley et al., 2016). This fragmentation can have severe negative consequences on ecosystem function through the loss of habitat specialists and the increase in generalists once critical thresholds in available habitat are reached (Pardini, Nichols & Püttker, 2018).
Pollination is an important ecosystem function known to be sensitive to habitat loss and high levels of fragmentation (Kolb, 2008;Hadley et al., 2014). We centered our analysis on the changes in the three classes where agaves occur in the study region: desert scrub, deciduous forest, and conifer forest. Within our study area, the LUCC analysis indicates that loss in desert scrub occurred primarily between 1985 and 1993 and occurred in the  central eastern portion of Coahuila adjacent to Nuevo Leon, probably due to the expansion of the main urban areas in the region (Monterrey, Saltillo, and Monclova) along with agricultural activities around the cities. This may indicate that at the regional level the cause of the LUCC may vary, depending on the land use dynamics of each region. Our findings on the degradation of desert scrub are consistent with other LUCC studies for the study region (Pérez Miranda et al., 2012;Trucios Caciano et al., 2012). Desert scrub is the most common vegetation type in Mexico, occupying 26.2% of the territory (CONABIO, 2009). By 2011, 10% (5.8 million hectares) of the desert scrub area was transformed, and the main cause of this transformation is the expansion of agricultural areas (SEMARNAT, 2013). These results were also observed, albeit in a much broader scale analysis, by Bonilla- Moheno & Aide (2020). They note agricultural expansion in the Chihuahuan ecoregion, precisely where we also observed these changes on a finer scale.
Our results also indicate an increase in deciduous and conifer forest replacing areas previously occupied by desert scrub, a trend also noted in another recent analysis (Bonilla-Moheno & Aide, 2020). This finding is expected considering the possibility of natural forest succession over the time lapse of our analysis. In the desert scrub category, we included scrublands, which are found adjacent to forest in the foothills of mountains, and it is possible that forest extended into those areas. Other land cover change studies in Mexico report similar trends, showing an increase in forest areas (Moreno-Talamantes & Garcia-Aranda, 2012;Trucios Caciano et al., 2012). The gain in deciduous forest between 1993 and 2002 in areas located in northern Coahuila could have been the result of the implementation of conservation activities in the area after the creation of the federal protected area Maderas del Carmen in 1994 (CONANP, 2008).
The increase in number of available agave habitat patches since 1985 suggests a fragmentation trend. There were three land-cover classes in the fragmentation analysis: conifer forest, deciduous forest, and desert scrub. The fragmentation statistics show a nearly constant nearest neighbor distance over time and a fairly stable number of landscape patches of habitats supporting agaves. This was likely the result of increases in conifer and deciduous forest land cover and patches from 1985 to 2002 with large declines in desert scrub from 1985 to 2002, so agave habitat patches were replacing each other. From 2002 to 2011 numbers were similar except for a large jump in the loss of desert scrub habitat in 2007-2011. As desert scrub is the most representative habitat for agaves in the study region, fragmentation statistics of agave habitat classes alone do not reflect the degree of loss of agaves in the region.
The increase in human settlements after 1985 in the region was likely the result of migration of people seeking employment opportunities with the growing industries (e.g., mining, energy, manufacturing, construction) in the main cities of the region: Monterrey, Saltillo, and Monclova. Since the 1970s, Mexico has seen an urbanization trend, with approximately 77% of the total population currently living in cities. The abandonment of rural areas reduces the local pressure on the land use and has been related to natural vegetation recovery in several countries, including Mexico (Bonilla-Moheno, Aide & Clark, 2012). The abandonment of rural areas in the study region could have contributed to the increase in forest areas.
Human productive activities (agriculture, livestock, and urban development) are still the most important contributor to habitat loss and subsequent fragmentation. Agriculture and cattle ranching are the main driving forces of the degradation of ecosystems in arid regions, such as our study region (SEMARNAT, 2019). Still, there are other human activities that are threatening this region, such as mining (especially coal mining, shale gas extraction, and material banks). Mining has been an important economic activity since the early 1900s, and significant portions of the landscape have been directly and indirectly affected by this activity, by degradation of the vegetation and pollution of the soil (Marroquín-Castillo et al., 2017). Mining activities are indirectly included in the land cover maps used in this study, particularly the areas affected by open pit mining which are classified as bare ground and mining facilities as human settlement.
Moreover, significant portions of the region have been affected by severe drought and fire. Between 1998 and 2017, the state of Coahuila had the second largest proportion of area affected by fires in the country, with more than 598 000 hectares burned (SEMARNAT, 2019). Most of this area is located within our study area. The effect of fire on the agave species present in our study area has not been studied, however, agave traits such as succulent leaves and a thick cuticle decrease flammability and facilitate fire resistance (Rodríguez-Trejo, Pausas & Miranda-Moreno, 2019). Studies in other regions indicate low mortality of agave after fire. Slauson (2002) reports 3.3% mortality after fire for Agave palmeri. Rodríguez-Trejo, Pausas & Miranda-Moreno (2019) reports 10% mortality for Agave potatorum.
During our field surveys in 2012 and 2013, we observed the effects of the severe drought that occurred in 2011 affecting large areas in northern Mexico. The drought affected extensive cattle ranching activities because there was not enough forage available for the cattle and, as an alternative, the cows foraged on any vegetation available, including agaves, increasing the level of degradation.
The recovery of depleted vegetation in arid lands is slow. For instance, agaves are slow-growing plants that bloom only at the end of their life cycle at 15-20 or more years (Gentry, 1982). The migration of the endangered L. nivalis relies on the agave blooming events (Gómez-Ruiz & Lacher, 2017). To have blooming agave every year, there needs to be agaves in late life stages. Even if depleted areas are restored with agaves, there will be a time lag for blooming events to occur meaning that restoration needs to happen as soon as possible to account for this lag. The fragmentation and disturbance on the available agave habitat can result in the lack of mature agave to produce flowers, and this would disrupt L. nivalis migration. Climate change is another factor that would affect the distribution of agave habitat (Gómez-Ruiz & Lacher Jr, 2019), however for this study we focused on direct human-driven land-use changes.
An additional but unrelated threat to the migratory L. nivalis in the study region is the development of wind farms (Hernández-Escobedo et al., 2014). Currently, wind turbines are operating in areas dominated by desert scrub in Coahuila, Nuevo Leon and Tamaulipas, sites of high potential for wind energy development (Wood et al., 2012) within the potential distribution of the migratory corridor (Moreno-Talamantes & Tovar, 2018). Large numbers of bats are killed at wind energy facilities in North America (Hammerson et al., 2017;Frick, Kingston & Flanders, 2019), but this impact can possibly be mitigated (Arnett et al., 2016), particularly during the spring and summer months when the endangered L. nivalis migrates to the region (Gómez-Ruiz & Lacher, 2017;Burke et al., 2019).

CONCLUSIONS
The consequences of the large-scale modification of terrestrial ecosystems are well documented and tropical and subtropical dry forests have been particularly heavily impacted (Miles et al., 2006;Watson et al., 2016). This makes the attainment of conservation goals and targets particularly challenging (Watson et al., 2016), all this complicated by the additional synergistic impacts of climate change and habitat loss (Mantyka-Pringle, Martin & Rhodes, 2012), in particular the documented impacts of climate-change on the L. nivalis-Agave complex in this region (Gómez-Ruiz & Lacher Jr, 2019). Overall, our results highlight the need for the implementation of conservation strategies to mitigate fragmentation and degradation in the Agave-L. nivalis corridor. The pressures on the land cover change are a result of actions performed by various actors (government, private sector, local communities). Conservation strategies should be developed with participation of all actors. Recent efforts to engage local actors for the protection of bats and their habitat in our study region have demonstrated the value of engaging regional NGOs and local communities (Gómez-Ruiz et al., 2015).
We recommend implementing restoration activities of agave populations in depleted areas. This would prevent soil erosion in the short term and would provide foraging resources for the endangered nectar-feeding bats in the long term and maintain the Agave-L. nivalis pollination corridor.