Introduction

Tropical dry forests (TDF) are important biodiversity hotspot areas and the second-largest tropical forest in Latin America (Quijas et al. 2019). A close historical relationship between human settlements and TDF use has led to the conversion of approximately 80% of the original TDF surface into pastures for cattle ranching and agricultural activities (Balvanera et al. 2011; Dirzo 2011; Gavito et al. 2014). Growing efforts have taken place to monitor and understand the social-ecological dynamics in these forests (Mastrangelo and Laterra 2015; Quijas et al. 2019; Jara-Guerrero et al. 2019). Exemplary social-ecological dynamics in TDF include ecosystem services provided to society (Maass et al. 2005), ecosystem services contribution to human well-being (Tauro et al. 2018), management strategies that guarantee the sustainable provision of ecosystem services (Mastrangelo and Laterra 2015; Trilleras et al. 2015; Monroy-Sais et al. 2020; Sánchez-Romero et al. 2021), and the successional dynamics of the forest recovery (Jara-Guerrero et al. 2019; Gavito et al. 2021; Cortés-Calderón et al. 2021; Pérez-Cárdenas et al. 2021). However, the concept of social-ecological systems (SES; Berkes et al. 1998; Ostrom 2009) is underexplored as a means to understand the complex interactions of social-ecological systems in TDFs.

A challenge to spatially identifying social-ecological units is the nested and hierarchical nature of social and biophysical components, which underpin management decisions (Hanspach et al. 2016; Martín-López et al. 2017; Lazzari et al. 2019) in response to local people's needs (Castillo et al. 2018). Social factors, which are embedded and driven by culture, politics, economics, and governance across scales (Chapin et al. 2009; Martín-López et al. 2017), determine the conditions in which people manage the land. Therefore, identification of social-ecological units (sensu Martín-López et al. 2017) entails understanding the interactions between the social and ecological factors that affect decision-making (Cumming et al. 2005; Virapongse et al. 2016). However, scales at which ecological and social data are collected do not correspond to the scales at which people make land management decisions. Among the few studies have spatially characterized social-ecological systems in land and seascapes (e.g. Martín-López et al. 2017; Lazzari et al. 2019; Pacheco-Romero et al. 2020); efforts to do so in TDFs have already started (Monroy-Sais et al. 2020; Sánchez-Romero et al. 2021). However, incorporating the minimum local governance level and decision-making scales remains a methodological challenge.

The goal of this study was to provide a methodological approach to identify and characterize the components of social-ecological system units at three relevant decision-making scales. Specifically, we: (1) identified ecological clusters (EC) at the plot scale, (2) identified social-management clusters (SC) at the plot owner scale, and (3) spatially characterized the social-ecological system units (SESU) within the landscape at the smallest governance unit scale. We discussed how ecological-social interlinkages determine current landscape configuration in the western TDF in Mexico. We then explored how this methodological approach can contribute to the identification of opportunities to reconcile productive activities and biodiversity conservation.

Methods

Study area

The Chamela-Cuixmala region is part of the TDF biome located along the Mexican Pacific coast (Ceballos and García 2010). It is a biodiversity hotspot area where a Long Term Social-Ecological Research network has been working for almost three decades (Maass et al. 2005; Castillo et al. 2018; Balvanera et al. 2021). The region comprises the Chamela-Cuixmala Biosphere Reserve and its transition area (UNESCO 2022), located in the municipalities of La Huerta and Villa Purificación in the state of Jalisco, Mexico (Fig. 1). Topography is dominated by hills between 20 and 180 m, although some flatlands occur in floodplains and valleys along the main rivers and seasonal streams (Cotler and Ortega-Larrocea 2006). Soils on hills are poorly developed, predominated by entisols with sandy loams in texture (Cotler and Ortega-Larrocea 2006). Rainfall is seasonal with an annual mean of 800 mm, concentrated between June and October (Maass et al. 2018). The mean annual temperature is 25.6 °C (1980–2015), with a monthly minimum and maximum of 16.4 °C and 32.6 °C respectively (Maass et al. 2018).

Fig. 1
figure 1

Location of the Chamela-Cuixmala region, comprised of the Biosphere Reserve Chamela-Cuixmala (BRChC; green color) and the transition areas at the governance units (ejidos; grey color) in the adjacent area, in the Municipality of La Huerta and Villa Purificación, Jalisco, Mexico

Biosphere reserves have core, buffer, and transition areas with different protection and restriction levels that foster social-ecological sustainable activities; therefore this region is mainly covered by TDF in different successional stages and patches (Sánchez-Azofeifa et al. 2009; Flores-Casas and Ortega-Huerta 2019). Within the reserve, most of the forest is old-growth with no signs of human intervention in recent decades. In the surroundings, the region has undergone extensive land-use change in the last five decades, mainly to crop fields in flatlands and induced pastures for cattle grazing on hills. Pastures are burned to reduce woody species and foster pasture growth, which leads to nutrient depletion and long-term reduction in forage quality (Burgos and Maass 2004; Trilleras et al. 2015). These lands are sometimes left unmanaged, allowing the forest to regrow. As a result, the landscape outside the reserve is a mosaic of grassland patches, secondary forests, and old-growth forests (Sánchez-Azofeifa et al. 2009; Flores-Casas and Ortega-Huerta 2019).

The ownership regime is a critical factor that affects the landscape configuration in the transition areas. Most of the land (70–80%) is under a governance unit specific to Mexico, called ejido, a semi-communal land tenure regime that emerged from the land redistribution policies following the Mexican Revolution of the 1910s (Castillo et al. 2005; Monroy-Sais et al. 2020; Fig. 2). Local collective management arrangements have been developed in many ejidos, and are operationalized through an ejidal assembly (Toledo 1996; Agrawal 2007; Schroeder and Castillo 2013). In the Chamela-Cuixmala region, ejido formation occurred between 1950 and 1975 and was linked to a governmental program called “March to the sea” (“Marcha al mar”), designed to colonize uninhabited and isolated coasts and to promote tourism (Castillo et al. 2005; Lazos-Chavero et al. 2016). Today, the region comprises the Chamela-Cuixmala Biosphere Reserve core area, the buffer area that corresponds to some lands surrounding the southern area of the reserve (Ceballos et al. 1999), including private ecotouristic alternatives from private owners, and the transition area that includes five small towns (Careyes, Cuixmala, Zapata, Villa, Chamela), eight ejidos in the Biosphere Reserve boundaries, and three tourist developments (Costa Cuixmala, Club Med, Careyes) (Ceballos et al. 1999; Sánchez-Azofeifa et al. 2009). In this study, we focused on seven of the eight ejidos in the transition area, plus two more in the contiguous area of the Northern part where farming activities and forest land cover are highly represented (Fig. 1).

Fig. 2
figure 2

The methodological approach used to spatially identify the Social-Ecological Systems Units (SESU). PCA principal component analysis. FAMD factorial analysis of mixed data. Ejidal plots are owned by the ejidatarios, who have legal rights to inherit and sell the land, as well as vote in the Assembly for making decisions on the communal areas. Ejidos are semi-communal land tenures with private and communal lands

Land rights within ejidos in this region can take place in three different ways (Schroeder and Castillo 2013). First, ejidatarios, or the landholding members of the ejido, can inherit the land right (ejidal plots), sell it, and vote in the ejidal assembly to take communal decisions. Ejidatarios have rights over communal lands within the ejido. Second, posesionarios possess land within the ejido but cannot pass it to the following generation. Posesionarios do not hold rights over communal lands and cannot vote in the ejidal assembly. Each ejido determines the level of posesionarios participation in collective management. Finally, avecindados are those who have settled within the ejido for more than a year, and neither possesses land rights nor vote in the Assembly. Traditionally, men hold most of these three types of land rights and make land-related decisions; although there are few “ejidatarias” (women).

Ejidatarios within ejidos surrounding the Chamela-Cuixmala Biosphere Reserve are aligned with the extensive cattle ranching and silvopastoral culture (Tauro et al. 2021). Cattle ranching is strongly limited by biophysical aspects such as water availability, as well as economic aspects such as financial resources to invest in cattle maintenance (Maass et al. 2005). There have been identified three different types of ejidatarios in the area; the first one includes ejidatarios with a high financial income that own larger extensions of land and cattle. They have better means to fulfil their livelihood needs. The second group is comprised of ejidatarios with lower education and are highly dependent on cattle activities. The third group is ejidatarios, which have a diversity of productive activities and a high number of plot areas (Naime Sánchez Henkel 2016). Emigration in the area has resulted in a lack of young producers and many abandoned areas (Cohen-Salgado 2014; Torales-Ayala 2015). Traditionally in the region, a strong emphasis on biodiversity conservation has excluded the local communities and neglected their needs, particularly ejidatarios surrounding the Biosphere Reserve (deeper historical explanation in Supplementary information 1). This has led to a general rejection of the Biosphere Reserve and conservation activities (Castillo et al. 2018).

Methodological approach

Our methodology was adapted from Martín-López et al. (2017), and structured in three phases (Fig. 2) that align with our objectives (see above). Social-ecological dynamics are the result of interconnection among three decision-making scales (Supplementary information 2).

For our study site, we characterized homogeneous spatial clusters based on topography, soil information, and landscape ecological conditions. We used the ejidal plot (individual plot) as the unit of ecological analysis since it is the minimum decision-making spatial scale. Then, we characterized clusters of ejidatarios (the plot owners) based on similar social and management decisions. Finally, we identified social-ecological system units at the ejido scale (minimum governance unit) based on the governance and infrastructure connectivity context at the landscape level, where we described the existing relations between the ecological and social-management clusters. We relied on available data on ecological, social, and management decisions as well as on governance dynamics (Table 1).

Table 1 Core variables used to identify ecological clusters (ecological variables), social-management clusters (social-management variables), and social-ecological system units (social-ecological variables)

Data sources

Individual plots: We selected 63 ejidal plots (7–30 ha) for this study. The first 30 were randomly selected to cover: (i) a range of land cover and topographic landscape composition patterns; (ii) heterogeneous land use intensity; and (iii) geographic dispersion across the ejidos surrounding the reserve (Pérez-Cárdenas et al. 2021). The remaining 33 were randomly selected to represent variation in the stand age and structure of TDF across the hilly region (Mora et al. 2018). We also used soil data available for a subset of these (Supplementary information 3). The georeferenced location of each plot was used to identify the corresponding polygon reported by the Mexican Agrarian Record “Registro Agrario Nacional” (RAN 2022). For 26 plots for which polygon data were missing, a polygon with an area equal to the median area size of the plots across the region (~ 25 hectares) was simulated. Polygons for which no ecological data were available (n = 1757) were excluded. In the case of two or more points located within the same polygon (ejidal plot), we calculated the average value for each quantitative variable among all the points within the plot. The 63 plots assessed here included different successional forest stages, and represent 3.5% of the existing ejidal plots (n = 1820). For each plot, we calculated nine topographical variables using data from the Shuttle Radar Topography Mission (SRTM) (Farr 2000) in Google Earth Engine (Gorelick et al. 2017; Tables 1, S.1).

Plot owners: We focused on 67 ejidatarios who participated in previous studies and for whom management data were available (Cohen-Salgado 2014; Mora et al. 2018; Pérez-Cárdenas et al. 2021, Ramírez-Ramírez unpublished data). These ejidatarios were selected to maximize the representation of different ejidal plot successional stages and by financial resources and educational level. There are at least three ejidatarios per ejido, covering ejidatarios from the nine ejidos surrounding the reserve, which results in a high representation of the ejidatarios in the area despite the small sample size. We selected five ordinal variables and five quantitative variables (see Table 1) for which consistent information was available across ejidatarios, and had shown to be key descriptors of tropical forest (dry and wet) management intensity: (1) land extension, (2) time of management, and (3) intensity of use (Benitez-Malvido 2006; Holl 2007; Martínez-Ramos and García Orth 2007; Zermeño-Hernández et al. 2015). We developed an index of management intensity in which the qualitative intensity (1 low, 2 intermediate, 3 high) was assessed for the nine most relevant variables that explained the variance among ejidatario.Footnote 1 Higher values in the index indicated greater management intensities. We summed the ordinal assessment to obtain a management intensity index per social-management cluster. The social data were linked with the ecological data based on their belonging to the same ejido, as there was not 100% correspondence between the ejidal plots and the ejidatarios (plot owners). We considered core variables for undertaking the analyses as those with less than 15% of missing data. The rest of the variables were used to complement the description of the clusters (more details in Table S.1). Nine variables with missing data (less than 15%) were imputed using the package “missMDA” (Josse and Husson 2016).

Minimum governance units: Data were gathered on the land tenure and on the communications infrastructure for nine ejidos (1400–18,000 ha); seven are in the transition area (Schroeder and Castillo 2013). Land rights held within the ejidos (Monroy-Sais et al. 2020) and distance to human settlements and roads (Maass et al. 2005; Flores-Casas and Ortega-Huerta 2019) have an important effect on the land cover transformation in the area. Therefore, data on the type of land tenure management (collective or individual), the types of land rights holders, and the number of years under private legal schemes (Program for Certification of Ejido Rights and Titling of Urban Plots, PROCEDE) were extracted from the National Agrarian Registry “Registro Agrario Nacional” (RAN 2022). The number of roads crossing the ejidos was obtained from the cartography provided by RAN (see Table 1).

Data analysis

The general clustering procedure consisted in performing a factorial analysis on each set of data (i.e. ecological, social-management, governance). Then we applied a hierarchical clustering on the components (HCPC) to identify clusters. Only principal components weighted with eigenvalues higher than 1 were included in the clustering procedure (i.e. Kaiser Criteria; e.g. Andrews et al. 2004). To identify the suitable number of clusters (e.g. similar plot types), we followed the entropy criterion that stops aggregating clusters when dissimilarity significantly decreased (Cornillon et al. 2012).

Ecological clusters (EC)

To find ecologically homogeneous ejidal plots, we first performed a principal component analysis (PCA) on the core ecological variables to summarise main patterns (Table 2). All the variables were previously transformed with log10 (n + 0.5) to avoid biases in the analysis. Once the clusters were created we used supplementary variables from soil data to further describe their characteristics (see Table 1). For points lacking soil information, data were estimated using Ordinary Kriging, which is considered a robust technique for spatial interpolation of soil properties (e.g. Robinson and Metternicht 2006).

Table 2 Variables used for the ecological, social-management, and social-ecological characterization

Social-management clusters (SC)

To identify the social-management clusters, we performed a Factorial Analysis of Mixed Data (FAMD) using the core variables of the management carried out by the ejidatarios (Table 2; Lê et al. 2008). We described the clusters according to their management intensity.

Social-Ecological-Systems Units (SESU)

To identify Social-Ecological-Systems Units, we performed a PCA. We used the ejido as a unit of analysis and nine social-ecological variables that refer to access to land, land tenure, and governance (Table 2). Once the clusters were created, we used the percentage of plots from each ejido belonging to EC and the percentage of ejidatarios belonging to each SC as supplementary. We used these two variables to visualize how each SESU is associated with both EC and SC using a scatter plot.

We tested for significant differences among clusters at each scale (EC, SC, and SESU) by conducting ANOVA (for variables that are normally distributed) and Kruskal–Wallis tests to analyze differences in quantitative variables among clusters (p < 0.05). To test for the normal distribution of these variables, we used the Shapiro–Wilk test (Shapiro and Wilk 1965). Post hoc tests were implemented when significant differences among clusters were identified, using a Tukey and Dunn´s (with Bonferroni correction) test. To evaluate differences among clusters for qualitative data, we used the chi-square test.

All the analyses were carried out using the FactoMineR package (Lê et al. 2008) for R version 4.0.5 (R core Team 2014). We used the packages “car” (Fox and Weisberg 2019) for the ANOVAs, the package “multcomp” (Hothorn et al. 2008) for the post hoc tests, and “FSA” (Ogle et al. 2023) for the Dunn test.

Results

Ecological clusters (EC)

Ejidal plots mainly differed with respect to their land cover, carbon storage and topography (elevation, slope, aspect) (Fig.S.1). The three first principal components explained 69.5% of the variance (Table S.2). The first dimension PCA1 (34%) divided the plots along a gradient ranging from those covered by old growth forest to those with introduced grasslands. PCA2 (20%) showed a strong association between carbon storage, elevation, and slope. PCA3 (15%) grouped plots at higher altitudes covered with secondary forests (Table S.3).

The four clusters represented a gradient of land-use intensity that is embedded into the heterogeneous landscape of the Chamela-Cuixmala region (Tables 3, S.4; Fig. S.2). The first cluster, EC1 comprised Ejidal plots that had significantly more conserved forests (72%) of older ages (~ 52 years old), as well as those with the highest percentage of permanent crop cover, which are located adjacent to the Biosphere Reserve (Fig. 3). The second cluster, EC2 included the Ejidal plots found at the highest elevations (> 160 m) and in sites with steepest slopes (> 11.40°), mostly including older aged forests (~ 42 years old), and were mostly found northeast of the Biosphere Reserve (Fig. 3). The third cluster, EC3, comprised Ejidal plots with the highest values of carbon storage (27.85 mgC ha−1), and soils with the highest levels of phosphatase activity (827), were mostly found in moderate north-facing slopes with high coverage of secondary forests (23%), and north of the Biosphere Reserve. The fourth cluster, EC4, was dominated by Ejidal plots covered by introduced grasslands (47%) with the most compacted soils (with a high bulk density 1.42), northeast-oriented, and found across the Chamela-Cuixmala region (Fig. 3).

Table 3 Mean values and statistical differences for ecological variables in each Ecological cluster (EC)
Fig. 3
figure 3

Geographical representation and description of the ecological clusters (EC) at the ejidal plot level. EC1 = Dominance of conserved and old forests. Permanent crops. EC2 = Highest elevations and slope. Dominance of old forests. EC3 = Dominance of secondary forests. High carbon storage. High phosphatase activity in soil. Moderate slopes facing north. EC4 = Dominance of introduced grasslands. Compacted soil (high bulk density). Northeast oriented

Social-Management Clusters (SC)

Ejidatarios mainly differed in the way they manage their land with respect to the number of cattle owned, the number of years using their plot, the intensity of wood extraction and plot size (Fig.S.3). The first seven components of the FAMD explained 77% of the variance (Table S.5). The first dimension FAMD1 (21%) divided the owners based on the number of cattle owned. FAMD2 (15%) was associated with the number of years of using the plots. FAMD3 (11%) was related to the intensity of wood extraction and plot size. FAMD4 (8%) represented the intensity of selective slashing. FAMD5 (7%) represented the differences regarding the number of paddocks. FAMD6 (7%) was associated with cattle rotation. FAMD7 (6%) was related to the number of clearings per year (Table S.6).

Four social-management clusters (SC) of the plot owners revealed a gradient in management intensity (Figs. 4, S.4). SC1 managed their plots for a longer time (35 years on average), had no forest clearings and no or low wood extraction, lowest cattle owned (10–32), and 93% of them rotated the cattle among paddocks. SC2 had the largest plot sizes (~ 136.4 ha) and numbers of cattle owned (~ 82 cows). SC3 performed the most intense pasture management, with the highest frequency of burning (5 times) and clearing (9), and owned the smallest plots (44 ha). SC4 undertook the highest intensity of wood extraction and slashing (> 200 rods/ha, and 80 poles/ha respectively); 100% of them do clearings, and 75% do not rotate their cattle. While the nature of the management is heterogeneous within and among SC, the intensity index revealed a gradient from SC1 with the lowest overall management intensity, to SC2, SC3 and SC4, with the highest overall management intensity (Tables S.7, S.8).

Fig. 4
figure 4

Dendrogram of the social-management clusters (SC) at the ejidatarios level and their description

We observed an association between plot size and management intensity; larger plots tended to have the lowest management intensity (SC1), while plots with the most intensive management were smaller (SC3 and SC4) (Table 4). A gradient was also observed in cattle rotation since larger plots represented by SC1 were the ones where 100% of the ejidatarios rotated cattle, while smaller plots are related to less cattle rotation (SC3) (Table 4). In addition, those plots represented by SC4, where there is the least cattle rotation, are the ones with the highest intensity of wood extraction (Table 4).

Table 4 Mean values and statistical differences for social-management variables in each Social cluster (SC)

Social-ecological systems units (SESU)

The variance among ejidos was mainly explained by the percentage of total ejidal surface allocated to individuals, the number of years under private tenure (registration in PROCEDE), and the average individual ejidal plot extension (Fig. S.5). The first three components of the PCA presented eigenvalues greater than 1 and explained 87% of the variance (Table S.9). PCA1 (41% of the variance) divided the ejidos based on the percentage of ejidal surface allocated to common lands versus those allocated to individuals. There was a positive relation between the number of ejidatarios per ejido and the percentage of common lands, and between the number of avecindados and the percentage of land allocated to individuals. PCA2 (26%) showed a strong relation with the number of years under the private tenure of PROCEDE. PCA3 (20%) grouped ejidos according to the average size of the individual plots (Table S.10).

The four social-ecological systems units (SESU) differed with respect to communal or individual governance and tenure rights. SESU1 and SESU4 were the most dissimilar, representing a gradient characterised by the percentage of land allocated to individuals, the percentage of avecindados, and the percentage of ejidatarios (Figs. 5, S.6; Table 5). The duration under private tenure regulated by PROCEDE explained the differences between SESU2 and SESU3 (Tables 5, S.11).

Fig. 5
figure 5

Left: Dendrogram of the Social-Ecological Systems Units (SESUs) at the Ejidataro level. Right: SESUs shown in the map of Ejidos surrounding the Chamela-Cuixmala Biosphere Reserve

Table 5 Comparison of mean values for governance variables at the ejido level per each Social-Ecological System Unit (SESU)

SESU1 comprised the ejidos Nacastillo and José María Morelos in the eastern part of the Biosphere reserve (Fig. 5). These ejidos do not have land allocated to individuals and had the highest percentage of ejidatarios (97%). SESU2 included the ejidos Los Ranchitos and Juan Gil Preciado at the North of the Biosphere Reserve, and had the higher number of years under the private schemes of PROCEDE (26 years) and the highest average size of individual (ejidal) plots (27 ha). SESU3 comprised two ejidos at the north and south of the Biosphere, i.e. Santa Cruz de Otates and Ley General de Reforma Agraria. The ejidos of SESU3 were the last ones to join the private scheme of PROCEDE (18 years under PROCEDE). SESU4 included the ejidos Emiliano Zapata, La Fortuna, and San Mateo, which are at the north and south of the Reserve, closest to the coastline and with the highest percentage of avecindados (72%), and the lowest percentage of ejidatarios (24%). Finally, SESU3 and SESU4 presented the highest percentage of surface allocated to individuals (79%).

While topography (EC) and plot owner individual resources (SC) underpin land cover transformations and management intensity, they are also modulated by communal governance (Fig. 6, SESU). SESU4, the most distinct one (Fig. 6), was ecologically characterised by flatter lands at lower elevations (less EC2) and secondary forests with high phosphatase (as represented by EC3) (Tables 3, 5). At the same time, SESU4 was dominated by ejidatarios who frequently undertake burnings and clearings (SC3) (Table 4). It presented a highest percentage of avecindados. By contrast, SESU1 was characterised by the high % of ejidatarios, and dominated by mature forests, including the oldest groves (EC1), and those at the highest elevations (EC2); socially it was dominated by ejidatarios with the largest plot size (SC2), moderate cattle management intensity with no rotation and the highest wood extraction most frequent forest management (SC4) (Table 4). In between SESU1 and SESU4, SESU2 and SESU3 presented an intermediate land-cover transformation (Fig. 6), but differed with respect to the number of years under private tenure of PROCEDE (Table 5). In addition to the longest period under PROCEDE, SESU2 is also characterized by a majority of ejidatarios with the lowest management intensity (SC1). SESU3 instead, had the lowest number of years under private tenure of PROCEDE and was dominated by plots at the highest elevations (EC2), where ejidatarios had the largest plot size (SC2) (Tables 3, 4).

Fig. 6
figure 6

SESU association regarding the highest percentage of Ecological clusters (EC) and Social-management clusters (SC). Right boxes represent the SESU description regarding social-ecological variables

Discussion

Land-use intensity and trade-offs between nature’s contributions to people: the relevance of co-production

Topography was a major driver of land use change within individual ejidal plots. Areas with rugged topography tended to maintain more forest cover while flatter areas have been more drastically transformed into pastures. This supports other studies focused on the role of topography in land cover change (Martín-López et al. 2017; Flores-Casas and Ortega-Huerta 2019; Aik et al. 2021). The prevalence of secondary forest in some of the plots results in a combination of productive activities that suggested that biodiversity conservation and livelihoods can be reconciled under certain conditions, similar to as found elsewhere (Pérez-Cárdenas et al. 2021; Balvanera et al. 2021). From our results, we found that important nature contributions to people, such as regulation of soil quality (represented by high phosphatase activityFootnote 2 and less soil compaction) and regulation of climate change (measured with the variable carbon storage) are provided in these forests (EC3; Table 3). By contrast, in the introduced grasslands (EC4; Table 3), phosphatase is usually low and soil compaction is high, indicating the low provision of soil quality. It is also in the introduced grasslands where the most widespread land use is intensive cattle farming (SC4; Table 4), which suggested a relation between intensive rangeland use and soil degradation, something previously reported at the plot scale (Jaramillo et al. 2003; Trilleras et al. 2015; Ayala-Orozco et al. 2018).

The size of the plots owned by ejidatarios underpins their decisions about management. For example, most ejidatarios do not use their entire plot for cattle due to factors like the high cost to transform forests into grasslands in areas with high slopes (such as SESU 1 and SESU 2, Fig. 6). Conversely, ejidatarios of small-sized plots use the greatest amount of available resources, removing the forest area and intensifying the management of the land (SC3 and SC4; Table 4).

Decisions on how to manage land in the Chamela-Cuixmala Region are based on adaptive management and learning processes, as well as access to anthropogenic capitals (Sánchez-Romero et al. 2021). Ejidatarios' motivations to burn or not to burn their land depend on their benefit–cost knowledge (i.e. human capital) (Ramírez-Ramírez et al. in review), which supports that nature’s contributions also require inputs from humans, a process known as “co-production” (Díaz et al. 2015; Palomo et al. 2016). Recent empirical research has shown that the type of anthropogenic capital involved in the co-production determines the level of land-use intensity and leads to trade-offs and synergies among nature’s contributions (Torralba et al. 2018; Lavorel et al. 2020; Bruley et al. 2021). García-Llorente et al. (2015) found that while high use of inorganic pesticides, fertilizers, and technology (manufactured and financial capitals) was strongly used in intensively managed greenhouses in the lowlands of Sierra Nevada Mountains, small-scale farming systems at higher altitudes were mainly supported by collective action of irrigation communities (i.e. social capital). Studies on how spatial configuration of the use of anthropogenic capitals lead to trade-offs between nature’s contributions to people and maintenance of multi-functional landscapes contribute to reconciling biodiversity conservation and productivity activities (e.g. Schermer et al. 2016; Pachoud et al. 2020; Grosinger et al. 2021).

The examples portrayed above highlight the relevance of operationalizing social-ecological system units (SESU), only possible if there is long-term interdisciplinary research, and collaborative efforts to create place-based social and ecological datasets (Haberl et al. 2006; Collins et al. 2010; Maass et al. 2016). In this study, the characterization of a large number of plots (N = 67) based on a wide range of ecological variables (N = 20) was only possible because of the collaborative nature of the long-term explorations of the social-ecological dynamics in the Chamela-Cuixmala region (Maass et al. 2005, 2016; Balvanera et al. 2021). Likewise, the identification of social-management clusters relied on a rich database of in-depth interviews to ejidatarios (N = 63) conducted over time.

Although there have been some efforts to collect social and ecological data in long-term research programs (e.g. Fischer et al. 2010; Bretagnolle et al. 2019), challenges remain around mismatches between ecological, social, and management data. Although our methodological approach was limited by our sample size that might not have fully represented existing ecological conditions within ejidal plots (Cohen-Salgado 2014), the spatial representation of ejidal plots and ejidos was critical to explore social-ecological dynamics for the Chamela-Cuixmala region.

Land management decisions: the relevance of governance systems across scales

Decisions of plot owners (ejidatarios) were bounded by the topographical characteristics of their plot, and by the governance system in which their decision-making is embedded. Historical privatization trends and level of communal management (see the introduction and Supplementary information 1) have had a clear impact on the social-ecological dynamics in the Chamela-Cuixmala region. For example, the ejidos that were the last in applying the private tenure fostered by PROCEDE (i.e. Santa Cruz de Otates and Ley General de Reforma Agraria—SESU3) are the ejidos with the highest percentage of surface allocated to individuals, with a high number of ejidatarios and large plot sizes (Table 5). This governance of land tenure has led to a moderate management intensity that allows the co-existence of forest preservation and productive activities. Including the governance level in our approach made visible similarities between ejidos that are very distinct in ecological and social-management conditions (i.e. SESU3). The Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) explicitly recognized this central role of governance in its conceptual framework and stated that governance systems determine, to various degrees, the access to, and the control, allocation, and distribution of components of nature and anthropogenic assets and their benefits to people” (Díaz et al. 2015, p. 6).

There are social-ecological dynamics mediated by land management decisions that were not covered in this study and should be considered in future research. For example, (Vallet et al. 2019) found that the co-production of nature’s contributions to people were subject to inequalities in access to different types of capital. In addition, Martín-López et al. (2019) found that power relations were exerted across governance scales, where institutions and stakeholders at larger scales often shape the decisions of local actors. Therefore, future social-ecological research in the Chamela-Cuixmala region needs to explore the role played by external stakeholders, such as external land buyers who recently arrived in the region, or landless inhabitants that can rent or work the land, power dynamics that shape the distribution of access to land, and the anthropogenic capitals underpinning co-production (Felipe-Lucia et al. 2015; Berbés-Blázquez et al. 2017; Vallet et al. 2019; Martín-López et al. 2019).

Moreover, this study did not evaluate the optimum number of clusters produced by different clustering statistical methods, and did not validate the SESU characterizations with the ejidatarios and other relevant stakeholders of the Chamela-Cuixmala region. Future applications of this methodology should consider a broad spectrum of available biophysical, social, and governance data, test for different clustering methods, and validate the results with relevant stakeholders. Here, it is important to point out that the resulting SESU maps are statistical constructs and might differ from the maps constructed by different stakeholders. Yet, the SESU maps obtained through the suggested methodological approach can create spaces for dialogues with different stakeholders about sustainable management options.

Conclusions

Place-based research on social-ecological systems has immensely advanced in the last decade by deepening the understanding of human-nature interactions across scales (Epstein et al. 2015; Folke et al. 2021; Norström et al. 2022). This body of research demonstrates that although social-ecological interactions are of relevance at larger scales than locally, it is usually at the local scale where diverse and innovative solutions emerged to reconcile productive activities and biodiversity conservation (Norström et al. 2022). This study provided a multi-scale methodological approach to identify spatially explicit social-ecological units across three decision-making scales. This approach helped address scale mismatches between ecological, social, and governance data, and navigate the inherent complexities of the interactions between people and nature.