Multi‐criteria systematic conservation planning for terrestrial vertebrates in a biodiversity hotspot in Mexico

Anthropogenic loss of biodiversity continues to increase worldwide, and existing conservation area networks (CANs) are inadequate for its adequate representation and persistence. To identify a set of new nominal conservation areas in Oaxaca, a Mesoamerican biodiversity hotspot in Mexico, for terrestrial vertebrate species, we used a multi‐criteria systematic conservation planning approach. Besides minimizing the area incorporated into the nominal CAN, we incorporated 25 socioeconomic variables using multi‐attribute value theory. We constructed a portfolio of nominal CAN solutions for four different scenarios all of which satisfied a 10% representation target for the modeled suitable habitat of each vertebrate species: (1) existing protected area‐based (PA) solution; (2) voluntary conservation area‐based (VCA) solution; (3) PA‐VCA solution; and (4) R‐C solution (rarity‐complementary algorithm). The PA‐VCA and PA solutions were the most expensive in terms of area that had to be included in the nominal CANs (13,352 km2 and 12,587 km2, respectively). In all our multi‐criteria analyses, highest costs were associated with maximizing the number of airports, amount of tourism, and length of available highways in a nominal CAN. We have thus established a portfolio of multi‐criteria solutions to the problem of creating an adequate CAN for the representation of terrestrial vertebrate species.

The purely ecological problem of the prioritization of CANs includes the selection of biodiversity features or "constituents" (Sarkar, 2014) for representation within the network, establishing targets and goals for the coverage of these constituents or their surrogates, selecting priority areas for the optimal satisfaction of these targets, as well as optimizing the spatial configuration (i.e., size, shape, dispersion, connectivity, alignment, and replication) of selected priority areas (Franco et al., 2009;Fuller et al., 2006;Kullberg et al., 2019;Peralvo et al., 2007;Sarkar et al., 2009;Triviño et al., 2018).
Major threats to biodiversity loss are increasingly related to human activities including habitat loss due to land use-land cover (LULC) change including urbanization as well as legal and illegal resource extraction (including hunting), often resulting in pollution (Brockington et al., 2008;Margules & Sarkar, 2007;Powers & Jetz, 2019).In most contexts, protected areas constitute important national and international instruments for conservation and management of biodiversity and natural resources (Kullberg et al., 2019;Naughton-Treves & Holland, 2019).This is particularly true in some developing countries that contain exceptionally high levels of biodiversity because of which protected areas become increasingly important for conservation (Brockington et al., 2008;Naughton-Treves et al., 2005).However, strictly protected areas alone are known to be insufficient for adequate representation of biodiversity globally or locally and additional priority areas for conservation must be selected to establish CANs (Brockington et al., 2008;Esmail & Geneletti, 2018;Margules et al., 2023;Margules & Pressey, 2000).These additional areas can also mitigate the risk from climate change-induced biological range shifts (Hole et al., 2011).
In the identification of these additional priority areas, biodiversity conservation strategies can avail themselves of the opportunity to encourage sustainable management of natural resources while benefiting local stakeholders such as residents who use these resources (Sarkar & Montoya, 2011;Spenceley et al., 2017).For decades, it has been explicitly recognized that these strategies can help embed prioritizing areas for biodiversity conservation goals within broader programs for social and economic development (Brown, McAlpine, et al., 2019;Heiner et al., 2019;Lessman et al., 2019;Margules & Sarkar, 2007;Moffett et al., 2005;Sarkar, 2012;Tali et al., 2019;Williams et al., 2003).Because these strategies must try to satisfy multiple goals simultaneously, at the technical level they involve tradeoffs and require multi-criteria approaches to area prioritization (Esmail & Geneletti, 2018) using decision-theoretic tools of multicriteria analysis (MCA) (Moffett et al., 2005;Moffett & Sarkar, 2006;Sarkar et al., 2017).There is a wide range of methods for approaching multi-criteria problems, and the appropriate method depends on the context of analysis (Figuera et al., 2005;Liu et al., 2019;Margules & Sarkar, 2007;Moffett & Sarkar, 2006;Sarkar et al., 2006).
Despite extensive transformation of natural ecosystems due to different types of LULC change (Ellis & Ramankutty, 2008;Haberl et al., 2007;Sanderson, 2002), many rural landscapes still provide important suitable habitat conditions for the persistence of many species (Iverson et al., 2019).The complementary ability of transformed and untransformed landscapes to harbor biodiversity jointly depends fundamentally on the size and spatial configuration of remnant natural habitat as well as the type of human activities in these landscapes and their relationship to various land uses (Franco et al., 2009;Fuller et al., 2006;Geldmann et al., 2019;Geneletti, 2007;Harlio et al., 2019;Iverson et al., 2019;Peralvo et al., 2007;Sarkar et al., 2009;Spenceley et al., 2017).
Such a complex conservation strategy is the one that is most appropriate for the state of Oaxaca in southwest Mexico which is recognized as a Mesoamerican biodiversity hotspot because of its high species richness and endemism (García-Mendoza et al., 2004).This is particularly true for well-studied faunistic groups in the region such as terrestrial vertebrates, of which 113 species are amphibians (Casas-Andreu et al., 2004), 245 are reptiles (Casas-Andreu et al., 2004), 736 are birds (Berlanga et al., 2019;Navarro et al., 2004), and 194 are mammals (Alfaro et al., 2007;Botello et al., 2007;Briones-Salas & S anchez-Cordero, 2004;Ramírez-Pulido et al., 2014).Unfortunately, the region is undergoing increasing deforestation that has resulted in a high loss of vegetation cover due to extensive cattle ranching, cropland expansion, and human settlement, all involving a disproportionate exploitation of forests (Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).
At the same time, Oaxaca also exemplifies a wide range of successful conservation strategies involving local communities that have combined conservation of species and natural habitats with traditional knowledge of the use and management of natural resources (García-Mendoza et al., 2004;Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).In addition, conservation areas that have emerged from federal governmental programs, including the provision of funds for biodiversity conservation, ecosystem services, and reforestation, have all mitigated the loss of remnant natural habitats (Illoldi-Rangel et al., 2008;Londoño-Murcia et al., 2010;Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).Previous studies in Oaxaca proposing priority areas for biodiversity conservation analyzed both single and multiple terrestrial vertebrate taxonomic groups as biodiversity constituents (Illoldi-Rangel et al., 2008;Monroy-Gamboa et al., 2019).A few studies have incorporated non-biological criteria like land use criteria (Fuller et al., 2006;Tali et al., 2019).Other studies have attempted to optimize the spatial configuration of nominal additional terrestrial conservation areas (Fuller et al., 2006;Kullberg et al., 2019;Lessman et al., 2019;Prieto-Torres et al., 2021;Tali et al., 2019).
Few studies on prioritizing conservation areas in Mexico integrate socioeconomic and LULC change information with biological data in multi-criteria analyses at a regional scale, despite the urgent need to produce conservation scenarios that optimize social opportunities for sustainable resource use while preventing biodiversity loss (Lessman et al., 2019;Liu et al., 2019;Mendoza-Ponce et al., 2020;Prieto-Torres et al., 2021;Rodríguez-Romero et al., 2018).In this study, we used a systematic conservation planning protocol implementing MCA for prioritizing protected areas in Oaxaca to represent all terrestrial vertebrate species in a minimal set of conservation areas while satisfying other goals.There are many methodologies for MCA (Moffett & Sarkar, 2006); following Moffett et al. (2006) and Sarkar et al. (2017), the methodology used here was multi-attribute value theory (MAVT) (Moffett et al., 2005;Sarkar et al., 2017).
We prioritized sets of sites in CANs with the goal of representing biological constituents (terrestrial vertebrates) adequately (i.e., up to a specified target of representation) while incorporating social and land use goals.The allocation of CANs considered land use patterns in transformed landscapes, and multiple opportunities for the involvement of rural communities in conservation actions.Specifically, we constructed CAN solutions for four different scenarios.All these solutions used an algorithm that selected new areas by optimizing the inclusion of species as yet inadequately represented in the solution, that is, using rarity and complementarity (Margules & Sarkar, 2007).However, they differed in how the process was initialized: one initialized the selection process with existing protected areas (CAN solution PA); one initialized the selection process with community-sponsored or voluntary conservation areas (CAN solution VCA; see Monroy-Gamboa et al., 2019); one initialized the selection process with both protected areas and voluntary conservation areas, (CAN solution PA-VCA), and one did not initialize the selection process with any existing set but used the rarity-complementary algorithm to prioritize areas for selection (CAN solution R-C).

| Study area
The state of Oaxaca ranks fifth in land area nationwide in Mexico (15 39 0 to 18 42 0 S, and from 93 52 0 to 98 32 0 W) and has an area of approximately 93,757 km 2 .Its altitudinal profile ranges from 0 to 3750 m above sea level.Oaxaca has a complex physiography and high environmental and climatic diversity resulting in a complex mosaic of vegetation types (García-Mendoza et al., 2004).There are 12 ecoregions (of the 51 present nationwide for Mexico) represented in Oaxaca based on climatic parameters, topography, landscapes, and vegetation types (Figure 1) (CONABIO, 2009;Urquiza-Haas et al., 2011).LULC change resulting in habitat fragmentation as well as hunting continue to negatively affect the biodiversity of the region (Monroy-Gamboa et al., 2019).

| Fundamental objectives and objectives hierarchy (OH)
Following a previously published protocol for MCA (Sarkar et al., 2017), we interpreted the prioritization of a nominal CAN as a decision scenario in which each possible CAN is a feasible alternative available for selection.We set a hard constraint (i.e., one that cannot be violated) on this set of feasible alternatives: each had to satisfy a biodiversity representation target of at least 10% of the modeled habitat of each vertebrate species.Thus, this target can be interpreted as the minimum threshold for the representation of biodiversity; though it has no firm biological basis it has been widely used in the absence of a better alternative (Margules et al., 2023;Margules & Sarkar, 2007).
The best alternatives were the CANs with the smallest area that simultaneously tried to satisfy all criteria relevant to the decision.Inclusion of multiple criteria was achieved through an OH-based MCA: i.We constructed an OH which consists of multiple hierarchical trees, each rooted at a fundamental objective with its sub-objectives as nodes.Fundamental objectives are those held to be the most important goals to be achieved through a decision.
This means that decision makers felt that no further justification were required for having these objectives; the question, "Why is x a fundamental objective?"did not lead to deeper reasons (Keeney, 1992).
A typical fundamental objective in a conservation decision is the adequate representation of biodiversity.To construct an OH, fundamental objectives are iteratively decomposed into sub-objectives at each level of the tree until measurable criteria (or attributes) are reached.Decomposition required that subobjectives must be complete, non-redundant, concise, specific, understandable, and most importantly, measurable criteria for the decision process (Keeney, 1992;Sarkar et al., 2017).ii.The OH was constructed through systematic consultation of stakeholders comprising local communities throughout Oaxaca potentially affected by the prioritization of areas as described here (i.e., communities which would be affected if any of the CANs constructed here are implemented).Fifteen consultations of stakeholder groups were carried out by the first author between February and March 2012, under an anthropological framework of polls (Montes, 2000;Sautu et al., 2005).The poll consisted of the presentation of the socioeconomic variables with each interviewed indicating how important a criterion was for that subject.The raw elicitations used a (comparative) ratio scale; final values were normalized between 0 and 1.This process was carried out within the ConsNet software package (see Section 2.5).
There were three fundamental objectives (see Table 1): (i) landscape preservation, that is, local communities living inside protected areas were interested in the preservation of their natural resources for subsistence; (ii) human development, that is, for many local communities lack of typically urban services and other such resources constituted a perceived problem and misuse of natural resources; and (iii) employment and subsistence, that is, local communities desired job opportunities so as to avoid recourse to deforestation, hunting, traffic in native species, or displacement.For each fundamental objective, we constructed an OH relating the fundamental objectives to sub-objectives and eventually to a set of quantitative attributes that can be easily measured (see Table 1) (Keeney, 1992;Mattsson et al., 2019;Sarkar et al., 2017).The sub-objectives were chosen in such a way that each of them was connected to a measurable quantitative attribute at the bottom of the hierarchy.
Values for these quantitative attributes were obtained from the 2010 survey conducted by the National Institute of Geography, Statistics and Informatics (Instituto Nacional de Estadística, Geografía e Inform atica; INEGI) for each municipality in Oaxaca.The information provided by the INEGI were obtained from an extensive set of interviews of adult individuals from each municipality nationwide in Mexico (INEGI, 2013(INEGI, , 2014(INEGI, , 2016)).Initially, we compiled a comprehensive list of 100 attributes for which data were available.These data were tested for normality (Kolmogorov-Smirnov test) to reject attributes that did not satisfy this requirement.Auto-correlated attributes (Spearman correlation, SPSS 10.01) were discarded to avoid redundancy though this is not strictly required for MCA (Sarkar et al., 2017).The remaining attributes were then further reduced by consultation with stakeholders to ensure that they were relevant and could be credibly measured; the remaining 25 quantitative attributes were finally included in the final OH set (Table 1).
T A B L E 1 Construction of an objective hierarchy (OH) with the identification of fundamental objectives, meeting the criteria of being essential for conservation, and controllable for assessing relative satisfaction by alternatives.

Fundamental objectives Sub-objectives Attributes
Maximize landscape protection (0.33) Note: The fundamental objectives were decomposed into sub-objectives including specific characteristics and being measurable in Attributes.The weights assigned for each objective is given in parentheses; the sum of the weights at each level of the OH must add up to.See Section 2 for details.
We then assigned weights to all objectives (more important, less important, or equal) and these weights were elicited from stakeholders so that they represented their perception of the value of each objective for the community.These consultations were also used to test whether the attributes satisfied preference independence and difference independence criteria.Preference independence incorporates the following requirement: Suppose that two alternatives that are evaluated based on n criteria differ in their weight only in the case of m of these n criteria.Preference independence is satisfied if preferences between these two alternatives are determined only by the m criteria that have different weights.Difference independence means that the difference between two alternatives differing in weight with respect to some subset of criteria does not depend on the shared weights of the other criteria.Satisfaction of these condition is required for the use of MAVT which was the MCA methodology used for this analysis (Moffett et al., 2005;Sarkar et al., 2017).

| Species distribution models (SDMs)
Suitable habitats were assigned to each species using SDMs.A total of 1063 species of terrestrial vertebrates were included in our analyses.SDMs were constructed using the Maxent software package using species presence records and environmental layers as predictive variables.These models have been previously published; their construction and validation are described in Monroy-Gamboa et al. (2019) (Data S1).

| Socioeconomic, land use, and risk variables
LULC data were obtained from the National Forestry Inventory (Inventario Nacional Forestal III series; see INEGI, 2000) in order to identify cells with evidence of human-induced vegetation changes (introduced grasslands, human settlements, areas without apparent vegetation, agriculture and livestock activity zones, introduced palm areas and urban areas).We excluded these areas from further analyses.Socioeconomic data were obtained from the 2010 survey by INEGI of each municipality, including forested and reforested areas, total population, quantity of garbage produced, populations of indigenous communities, poverty, illiteracy, communication options available (roads and airports), agricultural area, availability of government economic support (e.g., PROCAMPO, a direct support program for local communities), number of livestock (cows, sheep, goats, horses, pigs, poultry, and rabbits), forestry management strategies, extent of tourism (tourist facilities and volume), GDP (at local scale, is the income from the economic activities of the communities), manufacture, immigrants, and mining (Table 1).These variables are directly involved with the socioeconomic activities, and the values given by every municipality to each variable were assigned cell by cell to the grid covering all of Oaxaca.All cells were used in the selection process for the remaining variables.

| Area selection
For each of the four CAN solutions (PA, VCA-VCA, and R-C) independent searches or processes were carried out to incorporate MCA in the cell prioritization process.All CAN solutions met the hard constraint of satisfying the 10% representation target for each of the 1063 species of terrestrial vertebrates.Each search attempted to minimize the total area of the solution and the perimeter-area ratio during the MCA.The ConsNet software package (Ciarleglio et al., 2009(Ciarleglio et al., , 2010) ) was used for the searches.
ConsNet uses a modified analytic hierarchy process (AHP) (Moffett et al., 2006) to create a general multicriteria objective function.The modification makes the solutions obtained identical to those produced by MAVT (Moffett et al., 2006) and, thus, consistent with classical economic analysis.The results are different from what would be obtained using the AHP, but the transparent elicitation process of the AHP can still be used.This elicitation process requires decision-makers to evaluate the importance of criteria to each other on a ration scale of how many times one criterion is more important than another; the raw scores obtained are then normalized between 0 and 1.
ConsNet's objective function is a weighted linear combination of all the multi-criteria variables, and optimization requires either maximization or minimization of each variable (Ciarleglio, 2008;Ciarleglio et al., 2010).Performance is measured by the value of this function for a solution (i.e., a set of selected areas) which thus integrates all criteria from biodiversity to socioeconomic variables (Ciarleglio et al., 2009).In contrast, the cost of a solution is simply its area which represents how much of the region must be subject to conservation measure.

| RESULTS
Oaxaca had high socioeconomic complexity with 570 municipalities with a total population approximately four million people by 2015 and a 1% of annual population growth rate.The unemployment rate was 6.34% in 2010, a high of 21.6% in the immigration rate was recorded in 2009; for the per capita GDP growth, a high of 36.25% was recorded in 2008.More than 40% of the population is involved in activities using extensive natural resources such as cropping, fishing, and forestry (INEGI, 2013(INEGI, , 2014(INEGI, , 2016)).

| Protected area performance
The best CAN solutions for each scenario (PA, VCA, PA-VCA, and R-C) were the ones in which the most species were represented in the least area possible, that is, the ones with the least area.These solutions were obtained using the general MCA methodology incorporated in ConsNet.
F I G U R E 2 Solutions showing selected areas in the four different conservation area network scenarios (CAN).(a) Using existing official protected areas (PA solution, blue), (b) using community-sponsored or "voluntary" conservation areas (VCA solution, green), (c) using both these sets of areas (PA-VCA solution, pink), (d) not including these existing conservation areas a priori and used a rarity-complementary algorithm for area selection (R-C Solution, red), and (e) the overlap of the four given solutions (all colors).We used a different color for the selected areas for prioritization in each solution; the green areas represent the protected areas (PA) and the yellow areas are the voluntary conservation areas (VCA).
Selected areas meeting the 10% terrestrial vertebrate target representation were more costly in area in PA-VCA and PA solutions (13,352 km 2 and 12,587 km 2 , respectively) than in the R-C and VCA solutions (11,781 km 2 and 11,805 km 2 , respectively) (Figure 2).In the MCA, variables imposing the highest cost were the presence of airports, tourism (tourist facilities and volume), and existence of highways.The quantity of trash produced was a variable with a medium cost (measured by solution area) in all CAN solutions; maximizing the total population was only costly (in terms of area) in the VCA and R-C solutions (Table 2).The PA-VCA and PA solutions included the largest areas represented from all ecoregions, with over 11,200 km 2 ; the VCA and R-C solutions showed lower areas from some regions with slightly less than 10,000 km 2 in each case.The ecoregion Sierra Madre del Sur de Guerrero and Oaxaca pine-oak and mixed forests were best represented, and the Gulf of Mexico coastal plain moist evergreen forest was the least represented in all CAN solutions, respectively (Table 3).

| CAN
To fulfill the 10% target of terrestrial vertebrate representation not achieved by the protected areas, additional conservation areas needed to be included.The PA and PA-VCA solutions optimized more socioeconomic variables (16 variables in both cases) than VCA and R-C solutions (14 variables, respectively) but the former was more costly in terms of the area of the selected cells than the latter.The costs of the socioeconomic variables for the PA and PA-VCA solutions and the VCA and R-C solutions (16 and 14 variables each, respectively; Table 2) Note: Columns: PA solution, using existing official protected areas; VCA solution, using community-sponsored conservation areas; PA-VCA solution, using both these sets of areas; RCA solution, without initialization by existing set.See text for more detail.Abbreviation: CAN, conservation area network.
were remarkably similar.The variable of total population was costly in the VCA and R-C solutions, given that the size of the selected areas was too small.In the PA solution, the area of the additional selected cells was 68.8% of the total cells of Oaxaca (78,168).For the VCA solution, this number was 90.9%; for the PA-VCA solution it was 62.5%; for the R-C solution, it was 99.4%.Also, we overlap all the solutions in order to have a better view of the overlapped areas and the missing ones (Figure 2a-e).

| DISCUSSION
Our study illustrates how systematic conservation planning can be used to construct nominal CANs that incorporate socioeconomic criteria without compromising adequate terrestrial vertebrate representation in these CANs.Regional biodiversity conservation planning can further use this strategy for promoting sustainability, and for regulating social demands for resources by allocating areas to diverse types of land use activities to maximize net benefits to rural communities and local landowners, where two or more ecosystem and environmental services can be achieved in a single area (Faith & Walker, 2002).The four CAN solutions showed different trends across variables, reflecting the complex interactions and trade-offs between the socioeconomic context, patterns of land use, and biodiversity conservation (Geist & Lambin, 2002;Lambin et al., 2001;Luck, 2007).Increasing our understanding of the conservation-development tension should help us to assign more realistic threshold values and weights in our analyses, thus improving the proposed CAN solutions (Esmail & Geneletti, 2018;Figueroa et al., 2009).The integration of biodiversity conservation and human development still remains a challenge due to its complexity, as successful conservation and development demands active mutual engagement from stakeholders, environmental professionals, governmental institutions, nongovernmental organizations, and local communities, among others (Figueroa & S anchez-Cordero, 2008;Figueroa et al., 2009;Lavariega et al., 2017).
Conservation actions typically must ultimately be implemented at a regional scale.Our study provides an example for regional systematic conservation planning by identifying potential sites for different types of land use management and resource allocation at the local resolution of municipalities while simultaneously identifying priority areas for terrestrial vertebrate conservation (Figure 2).Conservation and development goals can be established using our still coarse-grained framework at a local scale, although detailed analyses relating socioeconomic, land use, and biodiversity conservation contexts for each particular site are necessary at finer spatial resolutions.For example, Figueroa et al. (2009) analyzed socioeconomic conditions in Mexican biosphere reserves, adjacent areas, and the ambient ecoregions, including the impacts of land use activities such as agriculture, livestock grazing, and forestry.They concluded that it was critical to include socioeconomic conditions in conservation area management and rural development planning for improving strategies leading to the confluence of conservation and development goals (Figueroa et al., 2009).
It would be interesting to expand Figueroa et al.'s (2009) study for other protected areas in Mexico and linking these analyses with our study in the case of Oaxaca.We expect that such analyses would contribute to integrating biodiversity conservation and development goals so as to benefit rural communities and local stakeholders and landowners.
Globally, existing protected areas only conserve a small fraction of biodiversity (Brown, Lockwood, et al., 2019;Geldmann et al., 2019), and this is also the case for Oaxaca at a more regional scale (Briones-Salas et al., 2016;Illoldi-Rangel et al., 2008;Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).Consequently, new CANs must be incorporated into conservation and management planning for adequate protection of biodiversity (Cunningham et al., 2008;Geneletti, 2007;Kullberg et al., 2019;Margules & Sarkar, 2007).Meanwhile the goals of sustainability and development must also be achieved across the same landscapes.We identified several high priority areas for both biodiversity conservation and for allocating resources to achieving sustainable development by including a multi-criteria regional perspective for area prioritization (Figure 2).
For example, selected areas in the terrestrial ecoregion of Sierra Madre del Sur of Guerrero and Oaxaca pine-oak and mixed forests (which are part of the Sierra Madre de Oaxaca sub-province) provided by the VCA solution are important for the conservation of microendemic species such as the mouse Habromys chinanteco.The terrestrial ecoregions of Tehuantepec canyon and plain dry forest and thorn forest and the moist evergreen forest hills are also very important for the conservation of vertebrate species because this region represents the limit of the distribution of many of the species occurring in Oaxaca (García-Mendoza et al., 2004).This region is also lacking in conservation initiatives and thus needs more conservation action plans (Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).
Selected areas in the VCA and R-C solutions include the ecoregions of Sierra Madre del Sur of Guerrero and Oaxaca, pine-oak and mixed forests, the Balsas depression dry forest and xeric scrub, Oaxaca and Puebla valleys, and depression dry forests and xeric scrub are remarkable as they hold the lowest records of terrestrial vertebrates in Oaxaca (Table 3).There is clearly a high need to conduct faunistic surveys to assess their importance for conserving biodiversity.It is interesting the fact that in the R-C solution no selected area included the National Park of Huatulco of the ecoregion of Mexican south Pacific hills and foothills thorn forests.These areas represent potential opportunities for the involvement of local stakeholder and landowners in conservation initiatives with economic benefits along with improving habitat quality.
Although a high proportion of the population in Oaxaca has been moving to urban settlements, poverty remains concentrated in rural areas.Rural poverty is often coincident with areas holding high biodiversity (Naughton-Treves et al., 2005).Programs for providing payment for environmental services or for reducing emissions from deforestation and degradation focus frequently in some of these regions, such as the terrestrial ecoregion Sierra Madre del Sur of Guerrero and Oaxaca pine-oak and mixed forests located in the Sierra Madre de Oaxaca, but ignore other potential areas (Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).Our study can help stakeholders to identify regions in which such schemes can be optimally established while it simultaneously provides guidelines for establishing new protected areas.
Protected areas are major conservation instruments for deterring human-induced habitat transformation (Geldmann et al., 2019;Naughton-Treves & Holland, 2019).Conservation actions across the landscape matrix including protected areas, complementary selected priority areas, agrosystems, areas that can be targeted for restoration, and rural and urban settlements must be integrated to protected areas for optimal conservation outcomes (Margules & Sarkar, 2007).In the context of expanding the CAN of a region, for these efforts to be normatively justifiable and also to be successful at the practical level, local and regional conservation programs should aim for providing benefits for local stakeholders (Margules & Sarkar, 2007;Naughton-Treves & Holland, 2019;Naughton-Treves et al., 2005).Globally, a large increase in the number of protected areas has been seen in the last 20 years (Margules & Sarkar, 2007;Naughton-Treves & Holland, 2019;Naughton-Treves et al., 2005) Nonetheless, our results show the low efficiency of protected areas in representing biodiversity in Oaxaca.More studies involving systematic conservation planning for developing CAN solutions are needed to establish protected areas in regions holding high biodiversity representation.
Oaxaca ranks third nationwide holding high levels of social inequality, unemployment rate, and extreme poverty (INEGI, 2013(INEGI, , 2014(INEGI, , 2016)).Equally important is the ethnic and linguistic diversity, with 16 of the 56 ethnic groups of Mexico, representing 60% of the total population of Oaxaca (García-Mendoza et al., 2004).The socioeconomic complexity challenges traditional static conservation strategies as depending solely on protected areas, which are insufficient to conserve the exceptional biodiversity of Oaxaca (Illoldi-Rangel et al., 2008;Monroy-Gamboa et al., 2019;Monroy-Gamboa et al., 2015).To propose CAN solutions including protected areas, additional conservation instruments, remnant natural habitat, and agrosystems in rural landscapes provides a more plausible alternative (Harvey et al., 2008;Margules & Sarkar, 2007;Sarkar et al., 2009).The present proposal and the results of this analyses were given to the statal conservation authorities.Conservation actions based on sustainable use of natural resources can contribute to reducing poverty while supporting biodiversity conservation in a systematic conservation planning context (Margules & Sarkar, 2007;Roe, 2008).

ACKNOWLEDGMENTS
A. G. Monroy-Gamboa thanks the Posgrado de Doctorado en Ciencias Biomédicas of Universidad Nacional onoma de México, Consejo Nacional de Ciencia y Tecnología (doctoral scholarship 200469) and Consejo Nacional de Humanidades, Ciencias y Tecnologías (postdoctoral fellowship CVU 206047 at Centro de Investigaciones Biol ogicas del Noroeste, S.C.).We thank the Instituto de Biología, Universidad Nacional Aut onoma de México for logistical and financial support.We thank the reviewers for their useful comments that improved this work.

F
I G U R E 1 Map of Oaxaca in Mexico.Terrestrial ecoregions include Sierra Madre Centroamericana pine-oak and mixed forests (SMC), Sierra Madre del Sur de Guerrero and Oaxaca pineoak and mixed forests (SMS), Tehuantepec Canyon and Plain dry forest and thorn forest (TCP), Balsas Depression dry forest and xeric scrub (BAD), Moist Evergreen forest hills (MEF), Mexican South Pacific hills and foothills thorn forest (MSP), Gulf of Mexico coastal plain moist evergreen forest (GME), Soconusco Coastal Plains and Hills moist evergreen forest (SCP), and Oaxaca and Puebla Valleys and Depression dry forest and xeric scrub (OPV).
Percentage contribution values for each criterion in the four ConsNet CAN solutions (see Section 2 for details).
T A B L E 3 Number of selected cells (measured in km 2 ) for the four CAN solutions in the terrestrial ecoregions of Oaxaca, Mexico.Ecoregions: Sierra Madre Centroamericana pine-oak and mixed forests (SMC), Sierra Madre del Sur de Guerrero and Oaxaca pine-oak and mixed forests (SMS), Tehuantepec Canyon and Plain dry forest and thorn forest (TCP), Balsas Depression dry forest and xeric scrub (BAD), Moist Evergreen forest hills (MEF), Mexican South Pacific hills and foothills thorn forest (MSP), Gulf of Mexico coastal plain moist evergreen forest, (GME), Soconusco Coastal Plains and Hills moist evergreen forest (SCP), and Oaxaca and Puebla Valleys and Depression dry forest and xeric scrub (OPV).Solutions: using existing official protected areas (PA solution), using community-sponsored or "voluntary" conservation areas (VCA solution), using both these sets of areas (PA-VCA solution), and not include these existing conservation areas a priori and used a rarity-complementary algorithm for area selection (R-C solution).Abbreviation: CAN, conservation area network.