Multitemporal Analysis of Tree Cover, Fragmentation, Connectivity, and Climate in Coastal Watersheds of Oaxaca, Mexico

Abstract

The synergistic interaction between landscape fragmentation and climate change poses a critical threat to tropical forests. However, the long-term dynamics of these coupled pressures have been little explored. This study analyzes half a century (1979–2023) of changes in landscape structure and climate across five coastal watersheds in Oaxaca, Mexico a region of high biological and socio-economic diversity. Using multitemporal satellite imagery (Corona, Orthophotos, RapidEye and Planet), we quantified the trajectories of tree cover, fragmentation (Largest Patch Index, LPI; Simpson’s Diversity Index, SIDI), and connectivity (Probability of Connectivity Index, PC); and contrasted these with temperature and precipitation trends. Our results reveal that during the period 1979–2010, there was a slight increase in tree cover accompanied by positive landscape metrics, whereas in the period 2010–2023 a loss of tree cover was observed. Nonetheless, overall, between 1979 and 2023, the analysis indicates a net gain of 59,725 ha of tree cover, a reduction in fragmentation (LPI increased by 26.33% and SIDI decreased by 0.23), and an improvement in connectivity (PC increased by 0.35). During the same period, the average annual temperature increased by 2.3 °C, and precipitation decreased by 219 mm annually. The study concludes that the system is undergoing a transition from a spatial configuration limitation to a climate-induced habitat quality limitation.

1. Introduction

In recent decades, forests worldwide have undergone intense deforestation processes [1]. The consequences resulting from the decline in tree cover and the recent effects of global climate change do not act in isolation; rather, their synergy degrades the integrity of ecosystems and landscapes [2,3,4,5]. On one hand, fragmentation creates smaller and more isolated forest patches, which decreases connectivity, alters microclimates, and drastically limits species’ ability to migrate in response to new conditions [6]. On the other hand, increasing temperatures and altered precipitation patterns exacerbate this vulnerability by directly affecting vegetation composition and distribution [7]. These combined effects have more persistent and damaging consequences in smaller fragments, where the impacts are magnified over time and compromise their long-term sustainability [2]. Taken together, landscape structure, defined by patch configuration and connectivity, acts as a response mechanism by ecosystems to environmental pressures from fragmentation and climate change [8]. The vast majority of changes in landscape structure are aimed at improving human well-being and economic development. However, excessive and irrational growth has come at the cost of the unsustainable exploitation of natural resources [9]. The drastic changes observed in ecosystems are associated with both external factors (climate change), and primarily with anthropocentric internal factors (population growth, economic expansion, industrial development, the expansion of agricultural and livestock activities, and the increase in tourism) [10]. They have influenced the landscape trajectories of ecosystems, leading mainly to soil degradation, loss of natural vegetation and forests, and overexploitation of groundwater, resulting in a reduction in ecosystem services (ES) and decreased human well-being [11,12,13].

However, landscape trajectories do not always follow a linear path toward degradation. Based on forest transition theory [14], various socioeconomic factors such as the abandonment of agricultural lands [15], shifts in land use policies [16] or the implementation of conservation programs like Payments for Environmental Services (PES) [17], can reverse deforestation and initiate a process of structural forest recovery. This phenomenon can lead to patch coalescence and increased connectivity, theoretically improving landscape resilience, despite the forcing of climate variables [5,18].

The relationships between landscape metrics and climate are key to understanding how ecosystems in these watersheds respond to environmental pressures. Previous studies show that increasing temperatures can affect the composition and distribution of vegetation, which modifies tree cover area, as well as intensifying drought periods and wildfires [7]. Studies have demonstrated that human activities and climate change can both contribute to increases in tree cover. For example, between 1984 and 2020, these two variables accounted for approximately 85.68% and 14.32%, respectively, of the dynamics of vegetation change in the Jilin Momoge National Nature Reserve in northeastern China. In areas experiencing vegetation degradation, human activities represented a significant portion of the change, with contributions exceeding 60% in many regions, indicating that anthropogenic effects have been the dominant factor influencing vegetation dynamics in the area [19]. Similarly, other research has shown that rising temperatures promote vegetation growth, particularly in spring, while decreasing precipitation limits vegetation cover growth, with conditions varying by region. Nonetheless, it has been reported that climatic variability accounts for between 38% and 64% of vegetation changes [20]. This phenomenon is especially critical in tropical and subtropical regions, such as southern Mexico, where the increased frequency of wildfires and droughts has contributed to greater landscape fragmentation [21].

This creates a knowledge gap. What happens when the structural recovery of a landscape coincides with sustained climatic forcing? Is the improvement in spatial configuration (less fragmentation, greater connectivity) sufficient to increase ecosystem resilience, or can the intensification of climatic stress (rising temperatures, droughts) negate these benefits, creating structurally intact but functionally vulnerable landscapes? How do reforestation and land management efforts influence the improvement of landscape structure, curb the causes of deforestation, and mitigate adverse climatic forcing? Answering these questions is crucial for the management of tropical socio-ecosystems.

This study addresses a critical knowledge gap by analyzing the trajectories of landscape structure and climate over half a century in the southern region of Oaxaca State, Mexico. In this region, communal and ejidal land tenure (forms of social land tenure established through agrarian reforms in Mexico, where land is collectively managed by a community or ejido) systems predominate [22,23]. These tenure regimes directly influence local strategies for territorial management as well as the implementation of conservation and landscape restoration policies [22]. For this reason, the area represents an ideal natural laboratory, as it is a mosaic of communal land tenure and features conservation initiatives such as PES [16] and the coexistence of development pressures and increasingly pronounced climate change [22]. Therefore, the objective of this study was to analyze the dynamics of tree cover and landscape metrics (fragmentation and connectivity), and to examine their relationship with temperature and precipitation trends in five watersheds along the coast of Oaxaca at the municipal level during the period from 1979 to 2023. The central hypothesis was that, despite a possible recovery of landscape structure driven by forest transition dynamics in the earlier decades, intensified climatic forcing in the more recent period would become the primary modulator of the system, limiting or reversing gains in tree cover and thereby compromising long-term ecosystem resilience.

2. Materials and Methods

2.1. Study Area

The study area is located in southern Mexico, within the state of Oaxaca, and corresponds to a portion of Hydrological Region 21, known as the ‘Costa de Oaxaca’. Hydrological regions in Mexico are delineated by the National Water Commission (CONAGUA, by its acronym in Spanish) according to watershed boundaries and drainage patterns, serving as planning units for water resource management. Although this hydrological region spans approximately 1,006,786 ha [24], the present study focuses on five representative coastal watersheds within it, which together span 281,744 ha. The area covers 23 municipalities and 665 localities, and is characterized by its high biological, climatic, and cultural diversity [25]. As shown in Figure 1, the area includes five hydrological watersheds: (i) Rio Copalita 1 (from their headwaters to the La Hamaca hydrometric station), (ii) Rio Copalita 2 (from La Hamaca to their mouth at the Pacific Ocean), (iii) Rio Coyula (from the headwaters of the Coyula and Cuajinicuil rivers and the Súchil stream to their mouths), (iv) Rio Zimatan 1 (from their headwaters to the Zimatan hydrometric station), y (v) Rio Zimatan 2 (from the Zimatan hydrometric station to their mouth at the Pacific Ocean).

The coastal watersheds cover the following territorial areas: Copalita 1, 132,792 ha, Copalita 2, 19,259 ha, Coyula 64,322 ha, Zimatan 1, 37,490 ha, and Zimatan 2, 27,881 ha [26]. In addition to their ecological value, these watersheds supply water to important tourist areas such as Huatulco Bays, La Crucecita, and Santa Maria Huatulco, highlighting their environmental as well as socioeconomic significance. The main vegetation types of the most representative ecosystems in the watersheds are medium deciduous forest (29.0%), pine forest (17.4%), pine-oak forest (17.0%), cloud forest (11.8%), medium sub-evergreen forest (8.1%), medium sub-deciduous forest (2.7%), oak-pine forest (0.7%), oak forest (0.3%), gallery forest (0.3%) and oyamel forest (0.1%) [27,28,29].

To analyze the change processes, considering biophysical and territorial factors that influence landscape configuration.

Table 1. Components, variables, and indicators for the multitemporal analysis of tree cover, fragmentation, connectivity, and climate in coastal watersheds in Oaxaca, Mexico, 1979–2023.

Component	Variable	Indicator	Description
Land use and vegetation	Tree cover	Area with tree cover (ha)	Dynamics of changes in tree and non-tree cover over time
Fragmentation	Structural fragmentation	Largest patch index (LPI)	Quantification of the percentage of the total landscape area represented by the largest patch
	Landscape heterogeneity	Simpson’s Diversity Index (SIDI) [30]	Dominance of one species over another
Connectivity	Functional connectivity of the landscape	Probability of Connectivity Index (PC) [31]	Landscape structure that determines the ease or difficulty of species movement, as well as ecological processes between patches
Climate	Annual average temperature	Temperature (°C)	Average value of the daily mean temperatures recorded throughout the entire year
	Total annual precipitation	Accumulated precipitation (mm/annual)	Total amount of rain, snow, or hail falls throughout the year

presents the ecological and climatic components with their respective variables and indicators used to analyze landscape transformations in five hydrological watersheds in southern Oaxaca, Mexico, over the past 44 years.

2.2. Land Use and Vegetation

Land use and vegetation maps are visual representations that present the distribution, extent, and types of cover present in each area. To analyze cover change from 1979 to 2023, maps of tree (deciduous forest, pine forest, pine-oak forest, cloud forest, medium sub-evergreen forest, medium sub-deciduous forest, oak-pine forest, oak forest, gallery forest and oyamel forest), and non-tree cover (agriculture, pasture, urban areas, water bodies, and other non-vegetated surfaces), were obtained for five hydrological watersheds for the years 1979, 1993, 2010, 2018 and 2023.

For 1979, images from the Corona satellite were used, with panchromatic spectral resolution and a spatial resolution of 1 m [32]. Using an unsupervised approach, eight land cover classes were defined with the Iso Cluster tool in ArcMap 10.8. Each class was visually identified as belonging to either the tree cover or non-tree cover group. For 1993, orthophotos from the National Institute of Statistics and Geography of Mexico (INEGI, by its acronym in Spanish) were used, with panchromatic spectral resolution and a spatial resolution of 2 m [33]. The tree and non-tree cover classes for this year were generated in a similar manner to the 1979 classification.

For 2010, RapidEye satellite images were used, with spectral resolution of 5 bands (red, green, blue, red edge, and near-infrared) and a spatial resolution of 5 m [34]. Because information was required for all forest species and not for any specific one [35]. To maximize contrast in dry deciduous areas, the image selection considered the phenological pattern of the vegetation, prioritizing scenes from October (rainy season) and April (dry season). Using the object-based Random Forest classifier in the software eCognition Developer 9 [36]. Two land cover classes were defined: areas with tree cover (pine forest, pine-oak forest, cloud forest, fir forest, gallery forest, medium dry deciduous and subdeciduous forest, and evergreen seasonal forest), and areas without tree cover (human settlements, agriculture, grasslands, and water bodies). One hundred control points were used (50 for classification and 50 for validation), evenly distributed across classes, determined based on fieldwork and analysis in Google Earth. For 2018 and 2023, Planet satellite images were used, with spectral resolution of 4 bands (red, green, blue, and near-infrared) and a spatial resolution of 4.78 m [37]. The classification was performed similarly to that of the RapidEye satellite images. The maps with their respective classifications from 2010, 2018, and 2023 were evaluated using the confusion matrix and Kappa coefficient [38] and were validated through fieldwork or reference data. Subsequently, the maps corresponding to the five years were compared to determine the tree cover dynamics over the period [39].

2.3. Land Use and Vegetation Changes

Land use and vegetation change was analyzed using the deforestation rate for the periods 1979–1993, 1993–2010, 2010–2018, 2018–2023, and 1979–2023, applying Formula (1) for the compound interest rate (r) [40]:

$$\mathrm{r}={\displaystyle \frac{1}{{\mathrm{t}}_2-{\mathrm{t}}_1}}\ln {\displaystyle \frac{\mathrm{A}2}{\mathrm{A}1}}$$

(1)

where A1 represents the tree cover area at ${\mathrm{t}}_1$ (initial area); A2 refers to the tree cover area at ${\mathrm{t}}_2$ (final area); t is the time difference between ${\mathrm{t}}_1$ and ${\mathrm{t}}_2$ in years.

2.4. Landscape Fragmentation

Landscape fragmentation occurs when habitat or vegetation cover is divided into small, isolated patches, mainly resulting from anthropogenic activities [30]. With the help of the software Fragstats 4.2 [30,41], Fragmentation was evaluated using the LPI and SIDI, two landscape-level metrics frequently used in recent research [42,43]. The first index estimates the annual fragmentation rate, while the second index considers the number of landscape elements and the dominance of one species over another. A high value suggests a greater likelihood of dominance by one species and, therefore, lower diversity, while a low value indicates higher diversity. Both indices were used to analyze changes in landscape structure and heterogeneity at the municipal level, focusing on the five selected years for comparison (1979, 1993, 2010, 2018, 2023).

2.5. Connectivity

Landscape connectivity refers to the ability of ecosystems to allow the movement of organisms or matter between fragments or patches, and is assessed using geoprocessing tools, network algorithms, and dispersal models. It also reveals the structure and functionality of ecosystems [31]. The first studies the continuity between ecosystems, while the second focuses on the movement of species and processes within ecosystems. One of the most used metrics to study functional landscape connectivity, which was used in this study, is the PC. The index evaluates the importance of each patch and allows for the identification of ecological sources and network connectivity [44,45]. Based on Formula (2) [31], and the software Conefor 2.6 [46] (Appendix A, Figure A4). The PC was also estimated annually at the municipal level for the five selected years, and to obtain nodes and distances, the ‘ID Within Distance Parameters’ extension of the ArcMap 10.8 software was used.

$$\mathrm{P}\mathrm{C}={\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{a}}_{\mathrm{i}}{\mathrm{a}}_{\mathrm{j}}{{\mathrm{p}}^{\ast }}_{\mathrm{i}\mathrm{j}}}{{{\mathrm{A}}_{\mathrm{L}}}^2}}$$

(2)

where a_i and a_j represent the areas of the habitat patches i and j; A_L denotes the total area of the study region; p_ij it is the maximum probability of the product of ij of all possible paths between the patches i and j [31]. The PC value ranges between 0 and 1. A value of 0 corresponds to the point where there are no habitat patches within the study area. A value of 1 corresponds to when no patches are present.

2.6. Temperature and Precipitation

Climate refers to the average atmospheric conditions that characterize a place over a prolonged period, typically more than 30 years. Climate indicators such as temperature, precipitation, relative humidity, atmospheric pressure, solar radiation, wind speed and direction are obtained through weather stations, radars, or satellites [47]. In recent decades, climate change studies have frequently analyzed the effects of temperature and precipitation on natural and social ecosystems. For the design and implementation of territorial and environmental strategies [48,49]. For this reason, this study analyzes the behavior of temperature and precipitation over a long period of time to observe patterns and examine their possible relationship with tree cover in the study area.

During the collection of climatological data for the period 1979–2023, a total of 34 weather stations were identified, irregularly distributed within and outside the study area [50]. Nineteen inactive stations provided insight into climate behavior during the 1970s and 1980s, while the remaining fifteen active stations offered greater data density in more recent years.

Daily average temperature data were collected, and subsequently, the mean temperature for each month was calculated. Even with Appendix A from the National Water Commission (CONAGUA, by its acronym in Spanish) through the national transparency portal, none of the stations had 100% complete climate data. For the months with insufficient data, a linear regression was performed to predict the daily and monthly average temperature. Next, the data were spatialized across the study area using the Empirical Bayesian Kriging (EBK) interpolator available in ArcGIS Pro [51,52,53]. Finally, the 12 months of the year were averaged to obtain an annual average temperature map by municipality for the years 1979, 1993, 2010, 2018, and 2023 (Appendix A, Figure A5).

For precipitation, the amount of water that fell daily as rain or hail was summed for each month. The interpolation of the data across the entire study area was then carried out using the EBK method. Subsequently, the total amount of water from the 12 months was summed to obtain an annual accumulated precipitation map by municipality for each year analyzed (Appendix A, Figure A6). Figure 2 summarizes the methodological procedure of this study.

3. Results

3.1. Land Use and Vegetation

The land use and vegetation maps generated for the years 2010, 2018 and 2023 show high levels of accuracy, ensuring their reliability for multitemporal analysis of tree cover along the coast of Oaxaca. The overall accuracy was 96%, 98%, and 99%, respectively, while the corresponding Kappa coefficients were 0.92, 0.96, and 0.98, indicating a strong agreement between the classification results and the observed reality (Appendix A,

Table A1. Confusion matrix for the supervised classification 2010.

	Tree Cover	Non-Tree Cover	Total	User’s Accuracy (%)	Commission Error (%)
Tree cover	50	-	50	100.00%	0.00%
Non-tree cover	4	46	50	92.00%	8.00%
Total	54	46	100	-	-
User’s accuracy (%)	92.59%	100.00%	-	-	-
Omission error (%)	99.07%	0.00%	-	-	-
Overall accuracy	96.00%	-	-	kappa	0.92

,

Table A2. Confusion matrix for the supervised classification 2018.

	Tree Cover	Non-Tree Cover	Total	User’s Accuracy (%)	Commission Error (%)
Tree cover	50	-	50	100.00%	0.00%
Non-tree cover	2	48	50	96.00%	4.00%
Total	52	48	100	-	-
User’s accuracy (%)	96.15%	100.00%	-	-	-
Omission error (%)	99.04%	0.00%
Overall accuracy	98.00%	-	-	kappa	0.96

and

Table A3. Confusion matrix for the supervised classification 2023.

	Tree Cover	Non-Tree Cover	Total	User’s Accuracy (%)	Commission Error (%)
Tree cover	50	-	50	100.00%	0.00%
Non-tree cover	1	49	50	98.00%	2.00%
Total	51	49	100	-	-
User’s accuracy (%)	98.04%	100.00%	-	-	-
Omission error (%)	99.02%	0.00%	-	-	-
Overall accuracy	99.00%	-	-	kappa	0.98

). In addition, both the user’s and producer’s accuracies were above 92% in all instances, which supports the quality of the classifications by minimizing both omission and commission errors. Two aspects explain the high accuracy obtained in our results. First, the classification scheme used was relatively simple, distinguishing only between areas with tree cover and those without it. This reduced the complexity of the analysis and lowered the likelihood of confusion between classes. Second, the strong spectral contrast between vegetated and non-vegetated surfaces made the separation easier, since woody vegetation tends to show distinctive reflectance patterns that can be identified in remote sensing imagery. These results support the conclusion that the generated maps offer a reliable depiction of changes in tree and non-tree cover over the study period.

Table 2. Tree cover (ha) and annual tree cover change rate (%) in coastal watersheds of Oaxaca, 1979–2023.

Period	Initial Tree Cover (A1)	Final Tree Cover (A2)	Change in Tree Cover (ha)	Years	% Gain/Loss	Annual Change in Tree Cover Rate (%)
1979–1993	189,528.08	227,147.75	37,619.67	14	19.85	1.29
1993–2010	227,147.75	275,274.66	48,126.91	17	21.19	1.13
2010–2018	275,274.66	257,112.05	−18,162.62	8	−6.60	−0.85
2018–2023	257,112.05	249,253.92	−7858.13	5	−3.06	−0.62
1979–2023	189,528.08	249,253.92	59,725.84	44	31.51	0.62

reveals the dynamic trajectory of tree cover change over the past 44 years. Between 1979 and 2010, a sustained expansion of tree cover is observed, with net gains of 37,619.67 ha in the first period (1979–1993) and 48,126.91 ha in the second period (1993–2010). This growth is reflected in positive annual rates of 1.29% and 1.13%, respectively. However, from 2010 onwards this trend reverses. The period 2010–2018 shows a loss of 18,162.62 ha, and the period 2018–2023 records an additional loss of 7858.13 ha, corresponding to negative annual rates of −0.85% and −0.62%, respectively. Figure 3, despite these recent setbacks, the overall balance for the period 1979–2023 is positive, with a net increase of 59,725.84 ha at an average annual rate of change of 0.62% (upper figure). Therefore, the tree cover in the study area exhibits non-linear trends, characterized by phases of both forest recovery and loss.

3.2. Fragmentation

The analysis of landscape indicators shows a general trend toward a decrease in fragmentation in the study region between 1979 and 2023 (Appendix A,

Table A4. Annual fragmentation indicators by municipality in the coastal watersheds of Oaxaca 1979–2023. Rio Copalita 1 (RC1), Rio Copalita 2 (RC2), Rio Zimatan 1 (RZ1), Rio Zimatan 2 (RZ2) and Rio Coyula (RC).

Watershed	Municipality	LPI (%)					SIDI
		1979	1993	2010	2018	2023	1979	1993	2010	2018	2023
RC1	Candelaria Loxicha	86.43	84.09	97.62	92.03	94.92	0.22	0.24	0.02	0.13	0.08
RC1-RC	Pluma Hidalgo	81.41	92.93	97.68	92.97	83.37	0.28	0.13	0.05	0.11	0.19
RC1	San Agustin Loxicha	51.59	60.05	96.43	96.41	94.79	0.49	0.46	0.07	0.07	0.10
RZ1-RZ2	San Carlos Yautepec	61.23	83.18	96.13	85.48	86.61	0.44	0.27	0.07	0.24	0.22
RC1	San Francisco Ozolotepec	47.97	55.11	39.33	76.54	56.68	0.50	0.48	0.46	0.34	0.39
RC1	San Juan Mixtepec	61.15	91.17	98.39	84.01	83.64	0.47	0.15	0.03	0.23	0.20
RC1	San Juan Ozolotepec	29.49	74.28	81.02	56.51	84.88	0.49	0.36	0.28	0.27	0.24
RC1	San Marcial Ozolotepec	29.49	74.28	81.02	56.51	84.88	0.49	0.36	0.28	0.27	0.24
RC1-RC	San Mateo Piñas	57.47	80.72	98.45	96.03	71.53	0.47	0.30	0.03	0.08	0.23
RC1	San Mateo Rio Hondo	50.58	71.77	97.16	96.64	95.60	0.48	0.39	0.05	0.06	0.08
RC1-RC2-RZ1-RZ2	San Miguel del Puerto	61.14	86.20	95.13	68.18	71.06	0.40	0.23	0.09	0.15	0.14
RC1	San Miguel Suchixtepec	37.65	67.15	97.76	95.67	92.91	0.50	0.42	0.04	0.08	0.13
RC1	San Pedro el Alto	34.92	40.58	95.68	90.12	87.97	0.49	0.48	0.08	0.18	0.21
RZ2	San Pedro Huamelula	87.10	83.99	95.16	79.77	78.32	0.22	0.26	0.09	0.18	0.19
RC1	San Pedro Mixtepec	70.00	86.45	91.02	91.89	89.71	0.40	0.21	0.16	0.15	0.18
RC	San Pedro Pochutla	41.28	38.41	89.51	45.92	43.66	0.42	0.39	0.18	0.20	0.24
RC1	San Sebastian Rio Hondo	36.84	52.98	84.42	82.47	81.14	0.50	0.47	0.16	0.20	0.22
RC1-RC2-RC	Santa Maria Huatulco	81.78	82.91	87.72	83.12	74.99	0.29	0.26	0.18	0.24	0.28
RC1	Santa Maria Ozolotepec	36.81	71.26	93.89	90.67	67.24	0.50	0.39	0.11	0.17	0.31
RC1-RZ1	Santiago Xanica	33.56	77.22	95.28	92.60	65.56	0.50	0.34	0.09	0.14	0.17
RC1	Santo Domingo Ozolotepec	42.45	73.86	95.82	83.73	83.83	0.50	0.37	0.08	0.19	0.18
Average		53.35	72.79	90.70	82.73	79.68	0.43	0.33	0.12	0.17	0.20

). The annual average of the LPI increased from 53.35% to 79.68%, indicating a greater dominance of the largest patch in each municipality and, consequently, a reduction in structural fragmentation. Complementarily, the SIDI decreased from 0.43 to 0.20, indicating an increase in landscape heterogeneity and, therefore, greater diversity, which is also associated with reduced functional fragmentation. Collectively, these indicators demonstrate that the landscape has become increasingly compact and less fragmented. At the municipal level, notable cases include San Miguel Suchixtepec, San Mateo Rio Hondo, and San Agustin Loxicha, where the LPI exceeds 90% in 2023. Conversely, municipalities such as San Pedro Pochutla, Santa Maria Ozolotepec, and Santiago Xanica exhibit relatively low LPI values alongside high SIDI values (Appendix A, Figure A1 and Figure A2).

3.3. Connectivity

The annual average PC exhibited an increasing trend between 1979 and 2010, followed by a slight decrease in the most recent decade (Appendix A,

Table A5. Annual Probability of Connectivity Index (PC) by municipality in the coastal watersheds of Oaxaca 1979–2023. Rio Copalita 1 (RC1), Rio Copalita 2 (RC2), Rio Zimatan 1 (RZ1), Rio Zimatan 2 (RZ2) and Rio Coyula (RC).

Watershed	Municipality	PC
		1979	1993	2010	2018	2023
RC1	Candelaria Loxicha	0.77	0.74	0.97	0.87	0.92
RC1-RC	Pluma Hidalgo	0.69	0.87	0.95	0.89	0.79
RC1	San Agustin Loxicha	0.33	0.41	0.93	0.93	0.90
RZ1-RZ2	San Carlos Yautepec	0.46	0.71	0.93	0.74	0.77
RC1	San Francisco Ozolotepec	0.21	0.36	0.40	0.62	0.53
RC1	San Juan Mixtepec	0.39	0.84	0.97	0.75	0.78
RC1	San Juan Ozolotepec	0.47	0.58	0.69	0.69	0.74
RC1	San Marcial Ozolotepec	0.31	0.62	0.96	0.95	0.86
RC1-RC	San Mateo Piñas	0.39	0.67	0.97	0.92	0.74
RC1	San Mateo Rio Hondo	0.35	0.54	0.94	0.93	0.91
RC1-RC2-RZ1-RZ2	San Miguel del Puerto	0.74	0.75	0.91	0.82	0.84
RC1	San Miguel Suchixtepec	0.22	0.49	0.96	0.92	0.86
RC1	San Pedro el Alto	0.19	0.37	0.92	0.81	0.78
RZ2	San Pedro Huamelula	0.76	0.71	0.91	0.81	0.79
RC1	San Pedro Mixtepec	0.53	0.78	0.83	0.84	0.81
RC	San Pedro Pochutla	0.48	0.54	0.81	0.78	0.72
RC1	San Sebastian Rio Hondo	0.25	0.39	0.82	0.78	0.75
RC1-RC2-RC	Santa Maria Huatulco	0.68	0.71	0.81	0.73	0.69
RC1	Santa Maria Ozolotepec	0.25	0.53	0.89	0.83	0.65
RC1-RZ1	Santiago Xanica	0.23	0.61	0.91	0.86	0.82
RC1	Santo Domingo Ozolotepec	0.29	0.57	0.92	0.80	0.81
Average		0.43	0.61	0.88	0.82	0.78

). The annual average PC exhibited an increasing trend between 1979 and 2010, rising from 0.43 in 1979 to 0.61 in 1993, and reaching its maximum value of 0.88 in 2010, followed by a slight decrease in the most recent decade. After this year, a slight decline in connectivity is observed, potentially indicating disturbances or threats to the structural integrity of the landscape. By 2018, the annual average was 0.82, decreasing to 0.78 in 2023. Despite this recent decrease, the current value remains above the levels of previous decades. This suggests that a process of landscape conservation and regeneration has occurred, facilitating the reconnection of patches or fragments. At the municipal level, significant differences are observed. Municipalities such as San Marcial Ozolotepec, San Mateo Piñas, San Miguel Suchixtepec, and Candelaria Loxicha have consistently exhibited high values in recent years. The less affected municipalities experience fewer disturbances and better tree cover conservation, which favors connectivity. In contrast, other municipalities such as San Francisco Ozolotepec, San Pedro el Alto, and Santa Maria Ozolotepec exhibit lower or fluctuating values (Appendix A, Figure A3 and Figure A4). These municipalities have more fragmented landscapes with patches that are more isolated, resulting in lower connectivity.

3.4. Temperature and Precipitation

The climatic averages recorded throughout the period 1979–2023 indicate a distinct trend toward warmer and drier conditions within the analyzed municipalities. During this period, the average annual temperature rose from 18.68 °C to 20.98 °C, while the average annual precipitation decreased from 1272.75 mm in 1979 to 1053.60 mm in 2023. The temperature increase was more pronounced in municipalities such as Santa Maria Huatulco, San Miguel del Puerto, and San Pedro Huamelula, which exceeded 23 °C in 2023, while localities such as San Miguel Suchixtepec and San Mateo Rio Hondo maintained temperatures below 18 °C (Appendix A,

Table A6. Annual mean temperature and total annual precipitation by municipality in the coastal watersheds of Oaxaca 1979–2023. Rio Copalita 1 (RC1), Rio Copalita 2 (RC2), Rio Zimatan 1 (RZ1), Rio Zimatan 2 (RZ2) and Rio Coyula (RC).

Watershed	Municipality	Annual Average Temperature (°C)					Total Annual Precipitation (mm)
		1979	1993	2010	2018	2023	1979	1993	2010	2018	2023
RC1	Candelaria Loxicha	18.76	20.40	19.66	21.18	20.56	1247.41	1620.81	1391.18	1391.30	1684.12
RC1-RC	Pluma Hidalgo	18.46	22.02	21.71	22.39	22.64	1310.23	1699.23	1405.71	1405.73	1499.88
RC1	San Agustin Loxicha	19.04	19.92	19.28	19.99	20.22	1235.15	1595.21	1401.23	1400.28	2209.15
RZ1-RZ2	San Carlos Yautepec	23.97	24.28	23.77	23.61	22.64	1166.17	1342.99	1341.94	1338.04	749.35
RC1	San Francisco Ozolotepec	16.39	21.86	21.37	20.82	20.77	1339.98	1540.68	1302.66	1299.93	801.31
RC1	San Juan Mixtepec	15.74	19.25	18.54	17.20	18.88	1092.73	1457.55	1182.77	1182.90	754.96
RC1	San Juan Ozolotepec	16.82	21.58	21.05	20.45	20.56	1295.97	1527.76	1286.04	1284.37	794.95
RC1	San Marcial Ozolotepec	15.21	19.44	18.73	18.52	19.63	1256.69	1625.64	1324.36	1324.98	1153.05
RC1-RC	San Mateo Piñas	17.38	21.68	21.36	21.72	21.87	1354.49	1662.15	1384.74	1385.06	1046.05
RC1	San Mateo Rio Hondo	15.83	17.69	16.75	16.35	17.30	1219.63	1642.82	1273.30	1272.99	1302.77
RC1-RC2-RZ1-RZ2	San Miguel del Puerto	25.28	24.60	25.07	26.48	23.95	1375.82	1508.61	1389.07	1385.98	704.47
RC1	San Miguel Suchixtepec	14.75	17.86	16.83	16.55	17.63	1233.14	1565.77	1281.99	1283.03	1236.47
RC1	San Pedro el Alto	16.67	19.63	18.73	19.24	19.63	1243.47	1628.01	1362.48	1362.92	1433.62
RZ2	San Pedro Huamelula	28.56	25.58	26.52	28.37	25.58	1244.16	1303.46	1396.71	1393.27	537.43
RC1	San Pedro Mixtepec	17.42	21.56	20.56	19.66	20.04	1185.99	1449.13	1264.55	1261.79	755.69
RC	San Pedro Pochutla	22.28	24.00	23.91	26.27	25.10	1469.03	1522.73	1443.82	1442.99	1109.86
RC1	San Sebastian Rio Hondo	15.10	17.65	16.75	16.07	17.30	1159.27	1593.52	1195.04	1195.08	879.69
RC1-RC2-RC	Santa Maria Huatulco	25.50	25.34	25.57	28.43	26.18	1483.59	1467.96	1462.77	1458.54	892.35
RC1	Santa Maria Ozolotepec	14.72	19.25	18.68	17.99	19.28	1249.24	1578.21	1264.63	1265.06	948.18
RC1-RZ1	Santiago Xanica	19.19	22.72	22.59	23.15	22.01	1417.75	1615.77	1362.53	1361.32	827.36
RC1	Santo Domingo Ozolotepec	15.19	19.27	18.63	17.44	18.90	1147.81	1499.70	1204.74	1204.63	804.85
Average		18.68	21.22	20.76	21.04	20.98	1272.75	1545.13	1329.63	1328.58	1053.60

and Figure A5).

In the same year, at the municipal level, the reduction in precipitation was more pronounced in localities such as San Pedro Huamelula, San Miguel del Puerto, San Juan Mixtepec, San Pedro Mixtepec, San Carlos Yautepec, and San Juan Ozolotepec, where values fell below 600 mm. In contrast, municipalities such as Santa Maria Ozolotepec and San Marcial Ozolotepec recorded precipitation levels above 900 mm (Appendix A, Table A6 and Figure A6).

Overall, the results of this study show a slight increase in tree cover in the coastal watersheds of Oaxaca during the period 1979–2023. This trend reflects a complex landscape dynamic, characterized by a first period (1979–2010) of tree cover expansion, reduced fragmentation, increased species diversity, and improved functional connectivity across watersheds—despite the rise in average annual temperature and the decline in average annual precipitation. This was followed by a second period (2010–2023) marked by a loss of tree cover, a slight increase in structural landscape fragmentation, and a modest decline in both diversity and connectivity within the study area—accompanied by a continued, albeit gradual, increase in temperature and a decrease in precipitation.

However, human actions can alter the climate and, in certain cases, neutralize the climatic benefits provided by vegetation. For example, vegetation-related climate models at a global scale show that vegetation dynamics could cool the climate by approximately 0.22 K by the year 2300, due to increased carbon sequestration and tree cover, which results in lower atmospheric CO₂ concentrations [54]. Nevertheless, vegetation expansion does not follow the same pattern everywhere and, in some cases, may generate counterproductive effects. The conversion of savannas into grasslands has been associated, in some studies, with a reduction in precipitation of around 10% and an increase in surface temperature (~0.5 °C). This creates a feedback cycle in which vegetation degradation amplifies drought conditions [55]. At regional scales, vegetation–climate feedback has been observed to intensify extreme events such as heat waves and droughts, thereby increasing the spatial heterogeneity of the climate [56]. It is important to recognize that the capacity of vegetation to mitigate climate change is not immutable, but rather vulnerable to alterations driven by anthropogenic climate trends.

4. Discussion

4.1. Decoupled Trajectories of Landscape and Climate

This study quantifies a paradigmatic divergence in the trajectories of change within a tropical socio-ecological system. Our results reveal a critical decoupling between landscape structural dynamics and regional climate forcing over the past half-century. Specifically, the period between 1979 and 2010 was characterized by a marked structural recovery, evidenced by a net increase in tree cover exceeding 85,000 has, a significant rise in connectivity (average PC from 0.43 to a maximum of 0.88), and consolidation of the forest matrix, reflected in reduced fragmentation (average LPI increased from 53.35% to 90.70% and SIDI decreased from 0.43 to 0.12), which aligns with global trends [57]. However, this phase of apparent improvement in landscape integrity coincided with a sustained degradation of climatic variables, manifested by an increase of 2.3 °C in average temperature and a water deficit of approximately 219 mm. This temperature increase is consistent with the trends reported by the IPCC [7], which show accelerated warming caused by the increase in greenhouse gases. In the state of Oaxaca, the Mixteca region exhibits the same trends, where it is projected that over the next 49 years the temperature will increase by more than 3 °C and precipitation will decrease by slightly more than 140 mm [58]. The reversal of the tree cover trend post-2010 suggests that the system may be approaching an ecological threshold, where the effects of climate forcing begin to outweigh the benefits of structural reconfiguration.

4.2. Underlying Processes of Forest Transition and Its Reversal

The forest recovery dynamics observed up to 2010 are consistent with ‘forest transition’ theories, where socioeconomic factors such as agricultural land abandonment and the intensification of conservation programs (e.g., PES) act as primary drivers of change [48,49]. The drastic increase in LPI and the decrease in SIDI indicate a process of patch coalescence and landscape homogenization [59], where the dominant matrix transitions from agricultural to forest use. This process has direct implications for biodiversity, potentially favoring interior species at the expense of edge species or those from open habitats [60].

The subsequent net loss of nearly 26,000 ha of tree cover between 2010 and 2023 indicates a destabilization of this trajectory. This reversal can be attributed to two mechanisms that are not mutually exclusive. First, the intensification of direct anthropogenic pressures, such as urban and tourism expansion in the coastal area [61], may be generating new deforestation frontiers. Second, accumulated climate stress may be inducing increased tree mortality rates through droughts and fires, which have become more frequent in southern Mexico [21], a phenomenon that would not be captured solely by cover change analysis but results in a net loss of biomass and, eventually, forest area.

Furthermore, this reversal in the forest transition between 2010 and 2023 reflects that reforestation strategies, environmental degradation mitigation, and conservation efforts through programs such as PES, which have been implemented in recent decades in these watersheds by international foundations [17] and the federal government [62] have not had the desired impact. This may be due to natural deforestation processes driven by climate stress and anthropogenic factors exceeding the implemented efforts, or because such reforestation and conservation efforts have not been efficient in recovering tree cover.

Additionally, the results show that connectivity increased significantly, and fragmentation decreased. This is even though the Huatulco Bays area, one of the coastal zones, has undergone restructuring due to increased tourism and population growth [61]. However, this area includes five protected natural areas (PNAs) covering 12,459 ha, which have been established with the aim of mitigating the effects of climate change and reducing biodiversity loss. These PNAs have been designated by the federal government from 1998 to 2024 [63,64,65,66,67].

The increase in tree cover found in this study could be attributed to the type of land tenure in the study area, as well as initiatives to mitigate threats related to deforestation and environmental degradation, in which international foundations have been involved in funding restoration and biodiversity conservation campaigns [17]. First, Oaxaca is the state with the largest communal agrarian land area in the entire country. Communal agrarian land accounts for 71.3%, followed by ejidal land with 25.8%, and the remainder consists of private possessors and settlers [68]. Furthermore, the organization of their authorities is mostly based on internal normative systems, which means that the regulation of their assets relies on the establishment of their own rules, traditions, and customs to govern and resolve their internal issues. Secondly, intervention actions have focused on reforesting with native species, achieving the integration and participation of 18 municipalities from the upper, middle, and lower parts of the watersheds [69]. This is reflected in the recovery of areas with tree cover, which has been more evident in the upper part, where agricultural production systems and pastures have been transformed into pine plantations and reforestation areas [27]. Similarly, the lower parts of the region have also experienced recovery, suggesting that the restoration programs have had positive impacts across different areas of the landscape.

These results, at least for two watersheds—Copalita 1 and 2—are contradictory. On one hand, various authors report that tree cover has decreased, associated with degradation processes in most ecosystems and the conversion to other types of land use [27,29,70]. While, on the other hand, it is consistent with what is reported in the land use and vegetation maps from INEGI [71,72].

Table 3. Tree Cover Area (ha) of the Coastal Watersheds Rio Copalita 1 and 2. The results shown in this Table correspond only to the Rio Copalita 1 and 2 watersheds, as the studies found on land use and vegetation in the study area are limited exclusively to the tree cover analysis in these two watersheds.

Reference	Years of Evaluation
	1979	1993	1995	2000	2010	2015	2018	2020	2023
[29]	-	-	-	149,727	-	-	-	136,575	-
[27]	-	-	130,386	-	-	125,929	-	-	-
[72,74]	-	129,117	-	-	-	-	131,471	-	-
In this study	90,796	117,804	-	-	142,257	-	140,426	-	136,193

indicates a slight growth trend in tree cover from 1993 to 2018. At the global level, our results are consistent with long-term studies that show similar trends [57,73], which implies the emergence of positive scenarios.

4.3. Functional Implications of the Structure–Climate Mismatch

From the perspective of metapopulation theory, the landscape configuration in 2010–2023, with greater structural connectivity (PC) and dominance of a large patch (LPI), should, in principle, confer greater resilience to the system. A more permeable matrix with larger patches facilitates gene flow, dispersal, and rescue effects, mitigating demographic and environmental stochasticity [75]. However, the increase in temperature and the reduction in precipitation act as an environmental filter at the landscape scale, potentially decreasing the intrinsic quality of all habitat patches, regardless of their size or spatial configuration [21]. Therefore, the landscape could be undergoing a transition from fragmentation-limited to climate-induced habitat quality-limited conditions. Structural connectivity, although high, could become functionally irrelevant if the destination patches are climatically unsustainable [76].

Furthermore, vegetation is related to changes in precipitation and temperature due to the carbon sequestration provided by plant cover, which mitigates the effects of global warming [77]. Otherwise, when deforestation occurs, the adverse effects of lost carbon sequestration can be even worse than those emitted by fossil fuels [78]. This, in turn, leads to greater greenhouse gas emissions into the atmosphere, causing alterations in climate patterns.

4.4. Study Limitations and Future Directions

We acknowledge that the nature of this study does not allow for definitive causal attribution between socioeconomic drivers, conservation policies, and the observed ecological outcomes. The spatialization of climatological data from a heterogeneous network of meteorological stations introduces inherent uncertainty that must be considered when interpreting the results. Likewise, the landscape metrics used quantify connectivity (PC) and structural fragmentation (LPI, SIDI), but not necessarily functional connectivity, which depends on the biological characteristics of the species.

Furthermore, while tree cover change is discussed in terms of spatial extent, it is important to note that we were unable to assess the functional ecological implications of these changes over time. This is mainly due to the lack of historical environmental data and official records documenting where and how deforestation occurred in previous decades, particularly in regions with complex land tenure such as this one [6,21]. This gap makes it difficult to evaluate how forest composition, structure, and associated ecosystem functions—such as carbon storage, habitat quality, or microclimate regulation—have evolved alongside land cover transitions [1,2,18]. Future studies integrating field data, forest inventories, and ecological indicators will be necessary to overcome this limitation and complement remote sensing approaches [5,14].

While this study includes a fundamental analysis of long-term temperature and precipitation trends, both global and local influences are likely to shape the patterns we observed [2]. On the one hand, regional warming and reduced rainfall align with broader anthropogenic climate change trends in Mexico [7,13]. On the other hand, changes in tree cover and structure, such as fragmentation or deforestation, can also lead to localized shifts in temperature and moisture [1,2,20]. These microclimatic effects may not be as noticeable, but they interact with broader climate pressures and could worsen them. Although our analysis does not separate the weight of each factor, we believe both play a role. Further research is needed to understand better how these drivers combine, especially using finer spatial and temporal data.

To advance the understanding of these systems, future research should focus on three areas: (1) The use of ecological niche models and dynamic vegetation models to project future climatic suitability of habitats within the topography of the watersheds. (2) The quantification of the impact of specific programs (e.g., PES) through quasi-experimental research designs. (3) Modeling functional connectivity for key species guilds (e.g., seed dispersers, pollinators) that may be particularly sensitive to both landscape structure and climatic thresholds.

Finally, although our results reveal contrasting trends, an expansion of tree cover during the three decades (1979–2010), followed by a decline in the most recent decade, we did not attempt to project future forest dynamics. Anticipating such trajectories would require specific modeling approaches (e.g., land use change scenarios or spatially explicit simulations), which were beyond the scope of this study. Nevertheless, understanding these potential pathways is essential for land use planning and conservation strategies, and future research should incorporate socioeconomic variables and policy dynamics to provide a more comprehensive basis for decision-making.

5. Conclusions

The analysis of the coastal watersheds of Oaxaca reveals that, over more than four decades, landscape structure has tended to improve, with increases in tree cover, reduced fragmentation, and greater connectivity. However, these advances have occurred in parallel with sustained climatic deterioration, rising temperatures and declining precipitation, highlighting a decoupling between structural landscape recovery and environmental conditions. In the tropics, where biodiversity and ecosystem services largely depend on the continuity of tree cover, fragmentation represents one of the main threats. The loss of connectivity disrupts key ecological flows such as seed dispersal, wildlife mobility, and the hydrological cycle. In contrast, a more integrated matrix enhances resilience by allowing species to find refugia and migration routes in the face of climate change.

Finally, this study provides empirical evidence of a fundamental dilemma in the management of tropical landscapes under climate change: ignoring climatic trajectories when planning landscape restoration risks creating ecosystems that are structurally intact but functionally failed. In this sense, the research results contribute to the field of landscape ecology by highlighting that ecosystem resilience is an emergent property of the interaction between its spatial configuration, environmental conditions, and climatic trajectories—not an attribute of structure alone [79]. They could also aid in the design and implementation of land use planning and sustainable development strategies that minimize environmental degradation and promote the resilience of the coastal watersheds of Oaxaca. For this, conservation and land management policies are needed that go beyond the paradigm focused solely on reforestation and fragmentation mitigation [79]. It is imperative to adopt a ‘climatically explicit’ approach that, in addition to reforestation, integrates habitat suitability and climate vulnerability into spatial planning. This implies that ecosystem restoration strategies should prioritize not only structural connectivity but also the creation of altitudinal and microclimatic corridors that serve as refuges and escape routes for biota. Additionally, the selection of species for reforestation should be based on their endemism profile, drought tolerance, and thermal stress resistance.