Ecological corridors in Costa Rica: An evaluation applying landscape structure, fragmentation‐connectivity process, and climate adaptation

In recent years, ecological corridors have been proposed on a global scale as a response to the accelerated process of natural ecosystem fragmentation, mainly as a result of human impact. In accordance with this trend, Costa Rica has undergone a process of implementing ecological corridors as to promote ecological connectivity since the 1990s, with the establishment of 44 ecological corridors covering 38% of Costa Rica's territory. Nevertheless, there is no research evaluating these corridors on a national scale that takes into account their functions as conduits, barriers, and habitats. Thus, the objective of this research was to describe the process of biological corridor formation in Costa Rica, and to evaluate the potential effectiveness of corridors by considering aspects of landscape structure and ecological processes related to connectivity and fragmentation. We used the National Program of Ecological Corridors database along with coverage analysis from Landsat images from 2000 and 2015.The composition of the biological corridors was determined at the landscape scale and related to potential to maintain a specific population of wild mammals weighing more than 10 kg. The composition of the ecological corridors was highly variable in terms of total area, proportion of natural habitat, and fragmentation process. Most biological corridors are capable of maintaining viable populations of Pecari tajacu and Tapir bairdii, while none could maintain populations of Panthera onca and Tayassu pecari. Only 50% of the biological corridors had improved in their connectivity. Therefore, public policies, such as master plans focusing on ecosystem restoration must be established. In addition, only two biological corridors incorporate the majority of elevation ranges (Life Zones) present in the country, which reduces the potential of the corridor system as a tool for climate change adaptation.


| INTRODUCTION
The increase of natural ecosystem fragmentation has been one of the most important ecological issues in the past century. According to Fahrig (2003), fragmentation is a landscape-scale process involving the loss and breaking apart of habitat independent from habitat loss. Actually the fragmentation process activities that have been exacerbated by climate change as a result of natural and anthropogenic (Fung et al., 2017). The creation of habitat edge through the fragmentation process affects many species with smaller home ranges and provokes different species responses where members of the community may be using resources across heterogeneous types of landscapes (Fletcher Jr et al., 2018). The maintenance and enhancement of natural ecosystem connectivity is a challenging issue for biodiversity conservation and landscape planning (Ersoy, Jorgensen, & Warren, 2018).
One response to this ecological condition is the establishment of ecological corridors (ECs), which has become a common strategy for biodiversity conservation (Hilty et al., 2006). ECs are spaces that allow for movement of individuals of species and related ecological processes like energy movement, gene flow, and nutrient cycles (Palmeri, Barausse, & Jorgensen, 2017). Theoretical and empirical observations demonstrate that increased interchange of individuals among populations, particularly small and isolated ones, may improve persistence and increase of local and regional populations (Rosenberg, Noon, & Meslow, 1997). According to Moreno and Guerrero-Jiménez (2019), ECs have many variations depending on their objectives, scale, and contextualization. A related concept is green corridors, which is usually applied to connectivity in urban areas that have fewer species present (Zhang, Meerow, Newell, & Lindquist, 2019). In Costa Rica, the biological corridor concept is the most common and is synonymous of ecological corridors.
Although there is limited knowledge regarding the ecological effects of ECs (Haddad et al., 2014), investigations of the last century dealing with climate change and biodiversity conditions have pointed out that protected areas are isolated islands of wilderness that need to be connected (MacArthur & Wilson, 2001;Worboys, Francis, & Lockwood, 2010). The main conclusion of the applications biogeography theory of conservation is that protected areas should be the largest possible single area (Simberloff & Abele, 1976). Acceptance of ECs is increasing as land managers seek ways to ensure that species can shift their ranges through fragmented landscapes in response to climate change (Beier et al., 2012;Krosby, Tewksbury, Haddad, & Hoekstra, 2010). As a result, ECs across different scales have been established covering diverse landscape structures, socio-economic conditions, and natural cover characteristics.
In Costa Rica, the first EC to be established during the 1990s was the Talamanca-Caribe EC which has since been used as a model to define the Mesoamerican Corridor on the Caribbean coast of Central America (Solís, Madrigal, Cruz, & Fonseca, 2002). Currently, Costa Rica has 44 corridors covering 38% of the country's surface. Several publications assessing the ecological connectivity of ECs have been carried out (Acuña, Molina, & Rodríguez, 2017;Alarc on et al., 2003;Bosselmann, 2012;Chinchilla, 2015;Villate, Canet-Desanti, & Chassot, 2009) although most studies evaluating ECs are carried out at the individual corridor level, with only one study assessing them on a national scale (Calvo, 2009). In addition, a research was carried out to identify potential sub-corridor at the actual establish EC's in the Norwest Costa Rica (Moran, Monroe, & Stallcup, 2019).
ECs are not considered an official category of protected areas and most of their lands are privately owned. However, the government established the National Ecological Corridor Program (NECP) through Executive Order 33106 of the Ministry of Environment, Energy and Telecommunications (MINAET) in 2006 (Villate et al., 2009), which also gave priority to the Payment for Environmental Services Program within corridors. According to national regulations, the requirement to be recognized formally as an EC include establishment of a local council, a technical profile, and a strategic plan. Only 30 of Costa Rica's ECs meet these conditions. Due to the lack of studies evaluating EC's connectivity with protected areas (intact habitats) on a national level, the objective of this article is to describe the ecological corridor system and model their potential effectiveness by considering landscape-level characteristics and patterns of habitat patches within corridors such as: area, number, mean size, and SD size, followed by the analysis of fragmentation/connectivity processes and elevation range distribution in Costa Rica.

| Study area
Costa Rica is located on the Central American isthmus, which connects the Neotropical region (South America) with the Nearctic region (North America) and has historically allowed for migration of flora and fauna in both directions (Kappelle, 2016). This bridging function, together with a wide range of elevations, diverse climate, geomorphology, and soils, has produced a region with high ecological variation presently comprising around 5% of the world's biodiversity, despite its relatively small area (51,100 km 2 ), (Criado-Hern andez & Marín-Cabrera, 2008). The country has numerous distinct ecosystems that include: paramo, tropical rainforest, tropical dry forest, mangroves, savanna, flooded forest, and wetlands (Morera & Sandoval, 2020). The Protected Area System (SINAC) was established in 1994 with the goals of preserving biodiversity and conservation management. Categories of protected areas include biological reserves, national parks, wildlife refuges, water protection zones, forest reserves, and wetlands corresponding to officially protected areas (Morera & Nel-lo, 2017) (Figure 1). In addition, the country has 44 ECs covering 38% of its continental surface, 25%).

| Field information and database
The first step in describing an EC is to analyze its landscape features (Pierik, Dell'Acqua, Confalonieri, Bocchi, & Gomarasca, 2016). The present research used a GIS database administered by the NECP in order to define the boundaries of each EC, including the application of a polygon boundary correction. Land use coverage surveys for the years 2000 (January, February, and March) and 2015 (February, March, and April) were developed using six satellite images (Landsat7 and Landsat8) with a spatial resolution of 30 m, with both sets of images having the same data composition. In order to validate the 2015 land use map with the use of ERDAS software, 70 samples were taken for each of the six images, resulting in 420 control points for which the supervised reclassification showed an accuracy of 93%. Each land use category was converted from raster to vector and integrated into a single vector layer to establish the land use cover mosaic for the whole country. Consequently, the mosaic was cleaned by using a minimum cartographical unit of 3,600 m 2 . The following categories of land-use cover were determined for each year: primary forest, secondary forest, mangroves, flooded vegetation, paramo, grassland, agriculture, and urban areas. The first five categories (considered to be natural covers) were used in the present research for the purpose of EC evaluation.
The GIS program Arc-Map 10.5 and the Patch Analysis extension were applied to analyze the landscape structure indices of ECs as: area, number, mean size, SD, and F I G U R E 1 Protected areas and ecological corridors, Costa Rica nearest neighbors for natural patches. Costa Rica has 44 ECs ( Figure 1) which were declared by the NECP in 2006, despite the fact that in 2018, 14 of them (Aguirre, Cordillera a Cordillera, Fila Langusiana, Fila Nambiral, Fuente de vida La Amistad, Las Camelias, Miravalles-Rinc on de la Vieja, Moín-Tortuguero, Osa, OSREO, Pirris, Playa Hermosa, Rinc on Cacao, and Rinc on Rain Forest) did not meet the government requirements and therefore were not used in the present study. In addition, during recent years, SINAC has been establishing green corridors that enhance both structural and functions connectivity in urban areas (Zhang et al., 2019), such as: Torres, María Aguilar, Cobric Surac, Par a, Garcimuñoz, and Achiote. Nevertheless, these are not included in the present research because their establishment was based on different criteria.
A theoretical framework was developed to estimate the capacity of ECs as a potential habitat for four wild mammals: jaguar (Panthera onca), tapir (Tapirus bairdii), whitelipped peccary (Tayassu pecari), and collared-peccary (Pecari tajacu). Distribution of these native species throughout the country was analyzed, considering their presence in corridors and home range areas (Zeller, McGarigal, & Whiteley, 2012). To analyze the EC's importance to wildlife use, we obtained records of the GBIF Database (GBIF.org, September 28, 2020). We downloaded data presence for tapir flag species from 2000 to 2020 that are recorded based on citizen science (Silvertown, 2009). The presence data records were associated with two conditions: (a) In ECs (1) and outside ECs (0); (b) and with natural cover (1) and without natural cover (0). Previous publications about home range area for the tapir flag species were reviewed in order to determine the suitable habitat area (natural cover) present in the ECs.
Based on the conceptual framework of the Protected Area Design by Primack, Rozzi, Feinsinger, and Dirzo (2001), we determined protected area size required to preserve wildlife populations, considering home range and number of individuals, in order to show the required habitat conditions. Using body size and home range to explain fauna mobility (McCauley et al., 2015), a habitat matrix of cover categories was designed (Caelen, 2017).
The Bayesian approach has been used in studies regarding ecosystem analysis (Gonzalez-Redin, Luque, Poggio, Smith, & Gimona, 2016) specializing in multi-dimensional data. Naives Bayes is one of mostly used models in the Bayesian approach (Fytilis & Rizzo, 2013) and it was used (Smith, Howes, Price, & McAlpine, 2007) to evaluate the occurrences of the tapir flag species in the ECs. A buffer influence (BIA) was delimited based on the average home range according to social and environmental variables related with tapir occurrences. The use of road density is very common (Bennett, 2017) as a social variable associated with ecosystem degradation and disturbed wildlife behavior. Likewise, the water source and forest ecosystem are specialized conditions required for suitable habitats (Naranjo, 2009). Response variables were standardized and the model was trained with 25% of the data. The Gaussian Naives Bayes approach was applied as a classifier in order to evaluate the efficiency of the model, for which the confusion matrix, precision, recall, and f1 scores were calculated. Its metrics have been defined in terms of true and false positives as well as true and false negatives. In this research, the Naives Bayesian model was implemented to consider true and false corridors and true and false not corridors. Therefore, a true positive is when the actual class is positive as is the estimated class. A false positive is when the actual class is negative, but the estimated class is positive. Additionally, we adjusted the receiver operating characteristic curve as a differing tool from binary classifiers.
In order to calculate the Fragmentation/Connectivity Index (FCI) for assemblages of patches of natural cover, Vargas' (2008) equation was adapted to compare natural cover dynamics between 2000 and 2015. In this sense, when the value of FCI is closer to 0, the fragmentation process is higher, whilst when the value is farther from 0, there is more connectivity.
where: FCI: Fragmentation/Connectivity Index; SPTA: surface of the ecological corridor; Nm: number of patches with natural cover; Sm: surface of patches with natural cover; Dm: average distance between two patches with natural cover calculated from the center of each one. To evaluate the ECs location according to elevation range, a digital elevation model (DEM) was built using vector shape curves scaled at 1:500000 (IGN, sf). An interpolation process was subsequently carried out to determine the seven intervals of each 500 m level based on the life zone range (Holdridge, 1967).
We estimated the size effect with a 95% confidence interval of mean differences (Spiegelhalter, 2019) between the area of natural cover, patch number and index fragmentation for 2000 and 2015 in order to compare both periods. Python 3.7 and the ANACONDA platform (2020 Anaconda Inc.) were applied for statistical analysis.

| Landscape and fragmentation process
The probability size distribution shows high variability; around 75% of those ECs were less than 500 km 2 , thus the range of area was 95% CI 313.6-656.0 km 2 (Figure 2). Similarly, there were large differences in natural cover (mean = 246.5 km 2 , CI 95% = 160.9-332.1) and the majority were dominated by tropical old-growth forest and secondary forest (<25 years recovery) in 2015 (Figure 2b). Between 2000 and 2015, the total natural cover area increased (mean = 222.4 km 2 ), but this was not statistically significant (Diff CI 95% = À92.6 to 140.8 km 2 ), while the number of natural cover patches have decreased from 2000 (mean = 501.1) to 2015-year (mean = 343.2), although not statistically significant (Diff CI 95% = À329.3 to 13.7). This possibly indicates a trend toward lower levels of isolation within corridor landscapes (Figure 3). The mean fragmentation index (Table A2) was also significantly higher in the year 2000 than in 2015 (Diff IC 95% = 1.2-5.6), representing empirical evidence of a decrease in forest fragmentation within ECs (Figure 4).

| Home range and habitat use
Home ranges for species vary from small to large areas. For example, herds of collared peccaries were 0.06-0.11 km 2 , while in the case of jaguars, we found 3.4-26.3 km 2 according to previous research (Table A1). We found that no ECs have the ability to support a large population (<1,000 individuals) of target species (Figure 5). In order to filter and debug the database for tapir flag species, 102 presence data from the performance model were obtained. The confusion matrix predicted 14 positive values and four negative ones correctly and two positive and four negative ones incorrectly. The metrics obtained for precision, recall and f1-scores are presented in Table 1.

| Climate change adaptation
Costa Rica has a large ecological gradient divided into seven zones that are frequently used for research regarding species distribution. However, only two ECs include most of the elevation (6) ranges present in Costa Rica; Santos and Volcanica-Central. Four (33%) ECs are located in two elevation ranges between 0 and 1,000 m while four (33%) ECs present three elevation ranges and four (33%) present five ranges. Six (20%) ECs had four elevations range and three ECs (10%) located in one sole range.

| DISCUSSION
According to Farina (2000), connectivity, which is the degree of physical connection of patches of natural cover, is essential. Most (80%) ECs have undergone sufficient F I G U R E 4 Fragmentation/ Connectivity Index (FCI) 2000-2015 for ecological corridors, Costa Rica: The vertical bar shows the FCI for 2000 and 2015 as well as change during this period for each ecological corridor F I G U R E 5 Natural cover of ecological corridors and its relationship with tree indicator's species (individual number and home range). Red points for jaguar (Panthera onca), blue points for tapir (Tapirus bairdii) and brown points for collared peccary (Pecari tajacu). Circles represent each relation between natural cover and total area by corridor (gradient green color) recuperation of this type of cover. Nevertheless, when other aspects such as patch size and distance between neighboring patches are considered, the results were quite different. After their establishment, only 50% of the ECs improved their connectivity conditions according to the FCI. For this reason, it is important to develop land use policies, such as county-level master plans, that can support conservation efforts appropriate for ecological restoration strategies at local level.

| Fragmentation-connectivity process
The quality of the ecosystems that lie within ECs is a key aspect determining whether those areas are spaces that facilitate wildlife movement. Bennett (1999) states that for ECs to allow for animal transit and function as dispersal corridors (Forman, 1997), the must have continuous or almost continuous connection of habitats in mosaics of inhospitable uses. Alarc on et al. (2003) demonstrated that the intense human use of forests in the Osa Peninsula of Costa Rica has affected the functionality of ECs.
The present research found that many ECs have been improving in structural connectivity with recovering natural vegetation, although it is necessary to also improve other dimensions such as the functional connectivity, which refers the extent to which wildlife move through the area. The quality of the ecosystems within the ECs is a key aspect in determining if those areas are spaces where wildlife actually moves.
After establishment of ECs, natural cover has generally improved throughout the country. A factor behind this trend, is the Payment for Environmental Services Program, which gave priority to ECs during the 2011-2015 period and allocated 52% of its resources incentivizing forest retention and restoration to these areas (Morera, Sandoval, & Loría, 2015), that. However, land zoning and spatial planning do not prioritize factors critical to natural cover restoration, such as size and the evaluation of distance between natural patches (FCI index).

| ECs size and potential habitat
ECs can be studied over broad spatial and temporal scales because they are distinctive spaces where fauna and flora interchange can take place. In consequence, area is a key element for a corridor, and it is an essential requirement to determine their adequate size (Hilty, Keeley, Merenlender, & Lidicker Jr, 2019). According to Forman (1997), corridor width as an indicator has been very well studied (ex. wind-breaks) though more research is still required in order to comprehend their dynamics. On the other hand, the increase of the EC's edge increases the human use and reduces ecological connectivity. Animal movement is necessary to predict trends in species distribution and community assemblage, as well as in conservation planning (Khazan, 2014). An adequate size of EC is needed for true habitat connectivity; however, none of Costa Rica's ECs are of adequate size. Based on 30 years of research experience in the Amazon, the size of a nature reserve should be large (surpassing 10 4 km 2 ) in order to protect suitable habitat (Laurance et al., 2018). This is relevant to Costa Rica's conditions due to on the similar biogeographical conditions, even though scale conditions are quite different. In addition, suitable habitat increases according with area, with larger areas increasing water, food, shelter, and reproduction probability (Sinclair, Fryxell, & Caughley, 2006).
In spite of the smaller surface area of Costa Rica, some ECs with large areas occur along national borders and highlight the importance for trans-border cooperation. One such example is the San Juan-La Selva EC, which preserves green macaw habitat (Chassot & Monge-Arias, 2012). The adequate size of an EC is determined by the species that use such an area, which in some cases signifies migratory behavior across national borders. Considering this assumption, few of the ECs researched have an optimal size for the P. onca (Portugal, Morato, Ferraz, Rodrigues, & Jacobi, 2020). Some of them are adequate for T. pecari, while the majority of ECs are suitable for T. bairdii and P. tajacu. On the other hand, while P. onca is a species that requires wide ranges of habitat in order to develop its daily activities as it consumes large prey and searches for proper refuge (Azevedo & Murray, 2007), it has the capacity to cross forest gaps (Cavalcanti & Gese, 2009). The minimum range obtained for this species was 19-23 km 2 per jaguar (Table A1), individuals. In addition, T. pecari form large groups (10 to over 300 individuals) with a gregarious behavior, which is why it requires extensive areas of suitable habitat to meet its dietary needs (Altrichter, Carrillo, S aenz, & Fuller, 2001; Moreira-Ramírez, L opez, García-Anleu, C ordova, & Dub on, 2015) and migratory movements (Biondo, Keuroghlian, Gongora, & Miyaki, 2011;Carrillo, Saenz, & Fuller, 2002). In the case of T. bairdii, even though it has a smaller habitat area than other species, it requires a larger expanses of extension of habitat to conserve one or more viable populations due to its nongregarious behavior (Cove et al., 2013) and high diversity and vegetation biomass to supplies the variety of items diet (Tobler, Naranjo, & Lira-Torres, 2006), for example to a population of 1,000 individuals, the range area must be 1,200-1,400 km 2 (using minimum home range in Table A1), assuming a suitable habitat conditions for all area. Regarding species such as P. tajacu, group size associated with smaller habitat areas represents less habitat demand required to sustain a viable population.
Measures such as habitat restoration and remnant forest retention within ECs are essential to conserve species of smaller home ranges with gregarious behavior and less demanding feeding habits. Management actions seeking greater connectivity within and between ECs (Araya-Gamboa & Salom-Pérez, 2016) are oriented toward species with greater spatial demands, such as large carnivore species with migratory movements as well their dietary diversity and food requirements (Moran et al., 2019).
The heterogeneous size of ECs in Costa Rica is a result of the government not being obliged to purchase the land where they are established, and perhaps because they are also expected to meet social goals that contribute to sustainable development (Newcomer, 2002). This relatively small country has implemented different scales of ECs where size has not been considered as an important element. In addition, ECs borders can be modified through proposals by their local committee and approval by the Conservation Area Board without rigorous scientific studies.

| Tapir presence
Even though the model shows a better adjustment based on the metrics (Table 1) and predictions presented properly in the confusion matrix, it offered better precision to evaluate tapir presence outside of the corridor rather than inside. Habitat variables defined the presence of tapir in ecological corridor with more occurrences in major density river (Figure 6), forest density (Figure 7), but also in grassland (Figure 8). Therefore, the tapir species flag has specific habitat conditions including well preserved forests. Street density is considered an important threat to wildlife movement and tapir presence; however, this indicator does not show such response. Conversely, a greater density of water sources is a key factor for tapir presence. Grassland within an ecological corridor could be proximity of forest patch than out. Tapir is a strong species indicator for assessing connectivity regarding landscape conservation targets due to its habitat demands (Naranjo, 2009).

| Climate change adaptation
Increased connectivity is often proposed as a strategy for biodiversity adaptation to climate change (Heller & Zavaleta, 2009;Krosby et al., 2010). Regarding to the relationship between climate change and connectivity, it is imperative to consider the area and quality of the landscape in addition to the strategic location. The location of ECs, in relation to elevation ranges, acts as a climate refuge, and is a key aspect for climate change adaptation which, as observed, has not been an important criterion in the definition of these areas in Costa Rica. Most ECs are located over four ranges and only 7% (2 ECs) consider seven ranges. Climate change has not been a key element considered in the establishment of EC's. However, it is necessary to take climate change into account and prioritize this within the corridor system. Even though EC's are a well-established practice for conservation connectivity in Costa Rica, there is a lack of research and management attention to their ecological functions as well the human use in these areas. For example, the habitat requirements of the species need to be evaluated in order to determine the optimal size of each EC. The main outcome of an EC is to connect landscape habitat allowing for small populations using its functional connectivity to become a viable population. In addition, elements such as restoration and remnant forest management need to be incorporated in order to increase the connectivity of natural ecosystems. ECs need to be a realistic space for species movement, and this needs to be considered in land use planning.