It is argued that forest fragmentation has negative effects on biodiversity at the short and long term; however, these effects might be dependent on the specific vegetation of the study area and its intrinsic characteristics. The processes leading to fragmentation are very diverse and many of them have anthropogenic causes as logging actions and clearings for agricultural fields. Furthermore, it is thought that scale plays an important role in the expected effects of fragmentation on biodiversity. In this study the effect of forest fragmentation and its impact on the woody plants species, richness and diversity are analysed considering three vegetation types in a poorly studied and difficult access biodiversity hotspot in northern Mexico. The results show that the effects of fragmentation are dependent on the vegetation type and that these are not strongly related to the species richness, and diversity in a microscale (100 m2). Fragmentation effects on biodiversity must be analysed in a broad scale, considering the fragment as a whole. Furthermore, conservation priority should be given to the larger fragments, which could potentially maintain a higher portion of biodiversity. Management should also be focused on increasing the connectivity between these big and medium size forest patches.
The processes generating forest loss and habitat fragmentation are recognized as a principal cause of biodiversity decline and are the greatest ecological problem in tropical developing countries [
It is argued that larger fragments might contain more species than small fragments and that when the patch’s area decreases, it has higher implications for core area species than for the edge ones. However, also contradictory results have been obtained (i.e., [
It has also been proposed that the number of individuals per species in a given “forest patch” is conditioned not only by the area available but also by the dispersal capacity of each species [
The scale at which the study is carried out is one of the most important factors when analysing species, richness and diversity changes as the variables impacting the species in a given area at a given scale might not be as important when analysed at a different scale [
This study analyses the impact that forest fragmentation, as a function of the forest patch characteristics, has on the woody plants species, richness and diversity of a poorly studied mountainous ecosystem in northern Mexico. The fragmentation impacts on biodiversity are studied in the complete landscape as an entity. Then, this is further divided by vegetation types (coniferous, nonconiferous, and scattered vegetation) as a surrogate for spatial scale in order to investigate how fragmentation affects the different vegetation types separately.
The study area is located in a poorly studied mixed coniferous-nonconiferous forest ecosystem in the Sierra Madre Occidental, in the northern state of Chihuahua, Mexico (Figure
Study area location within the Sierra Tarahumara in the Sierra Madre Occidental, Chihuahua Mexico, and sampling sites.
One cloud free Landsat ETM+ datasets was downloaded from the Global Land Cover Facility database (
The Landsat image was classified using a combined pixel-based and object-based method for satellite image classification [
Description of the Land Use/Land Cover classes used during the classification analysis.
Land cover class | Description |
---|---|
(1) Coniferous forest | Forest patches dominated by coniferous plant species, mostly |
(2) Scattered vegetation | Mixed vegetation with a scattered distribution—shrubs and small trees |
(3) Water | Water bodies such as lakes and reservoirs |
(4) Nonconiferous forest | Forest patches dominated by nonconiferous plant species, as |
(5) Bare soil | Areas without vegetation, not water and not urban |
(6) Agriculture | Crop fields, pastures |
(7) Urban | Built-up land for residential or commercial purposes |
Maps of the classified study area for 1999 and 2006 where the observed increase in bare soil and agriculture classes and decrease in forested land are presented.
The forest patches of the complete study area were classified by size as small, medium, and large. For the patch size classification threshold, the 33rd and 66th percentile of the polygons area values were used. The small patches had sizes up to 0.005 km2, medium patches had sizes above 0.005 km2 but equal to or below 0.014 km2, and large fragments were those with areas above 0.014 km2. The three forest classes obtained from the image classification process coniferous, nonconiferous and scattered vegetation, were used as the sampling sites and are used for the subsequent analysis.
The species sampling was carried out in 100 forest fragments chosen by a stratified random sampling procedure using Hawth’s analysis tools [
Each selected forest fragment was surveyed using one plot of 10 × 10 meters, as suggested by Zuñiga [
The fragment variables used during the analysis to describe the fragments characteristics or fragmentation status of the forest class (Table
Patch variables used for the analysis (fragmentation variables from [
Variable | Explanation | Range |
---|---|---|
Area (km2) | Equals the patch’s area | 0.0009–10.13 |
CONTIG | Patch’s contiguity | 0–0.92 |
Core area (km2) | Equals the core area | 0.0000–9.40 |
Core FD | Core area fractal dimension | 0–1.86 |
Core PAR | Cores’ area perimeter-area ratio | 0–0.78 |
Core perimeter (m) | Core area perimeter | 0–73363 |
Core SI | Shape index of the core area | 0–6.74 |
ECON* (%) | Patch’s edge contrast | 0-1 |
Elevation (masl.) | Patch’s elevation | 2500–2800 |
ENN (m2) | Patch’s Euclidean nearest neighbour | 60–942 |
FD | Patch’s fractal dimension | 1.31–1.55 |
Latitude (m) | Geographic location | 3084590–3094000 |
Longitude (m) | Geographic location | 262754–272468 |
PAR | Perimeter-area ratio | 0.007–0.20 |
Perimeter (m) | Patch’s perimeter (m) | 120–71276 |
PROX | Patch’s proximity index | 0.001–6610.48 |
SI | Shape index of the patch | 1.12–6.31 |
Slope (degrees) | Patch’s slope | 0°–30° |
The fragmentation indices were obtained with the Patch Analyst extension v.4 [
Preliminary analysis of the explanatory variables showed that four of the sampled polygons had extreme values for the shape index variable (SI > 2.62) in comparison to the other polygons’ values. Consequently these fragments were not used for further analyses as they could mask the real pattern of the data [
A principal component analysis (PCA) was carried out in order to concentrate the information of the fragmentation variables into a small number of components. The association between all pairs of explanatory variables was evaluated by means of Spearman correlation coefficients. The differences between forests types with respect to the patch variables analysed were assessed with the Kruskal-Wallis test. These analyses were carried out using “R” statistical software v.2.11.1 [
Subsequently, multiple regression analyses with a backward selection procedure were used to determine how the Shannon diversity index (
The PCA analysis resulted in the extraction of four components considered for further analysis. Following the Kaiser-Guttman criterion [
Loadings for the four components extracted from the PCA analysis and the % of variation explained by each one.
Axes extracted | PCA 1 (50.96%) | PCA 2 (16.74%) | PCA 3 (11.65%) | PCA 4 (7.66%) |
---|---|---|---|---|
Variable | ||||
Area (m2) | 0.98 | 0.09 | −0.03 | 0.06 |
CONTIG | 0.34 | −0.22 | 0.29 | 0.68 |
Core area (m2) | 0.97 | −0.03 | −0.09 | −0.15 |
Core FD | −0.49 | 0.48 | 0.11 | −0.48 |
Core PAR | −0.94 | 0.28 | 0.13 | 0.04 |
Core perimeter | 0.94 | 0.12 | −0.06 | −0.22 |
Core SI | 0.60 | 0.74 | 0.09 | −0.06 |
ECON | −0.02 | −0.24 | 0.59 | −0.10 |
ENN | −0.45 | 0.17 | −0.57 | 0.42 |
FD | −0.57 | 0.71 | 0.30 | 0.20 |
PAR | −0.96 | 0.21 | 0.14 | 0.01 |
Perimeter (m) | 0.94 | 0.28 | 0.04 | 0.11 |
PROX | 0.13 | −0.34 | 0.80 | 0.01 |
SI | 0.56 | 0.74 | 0.23 | 0.21 |
The Spearman rank correlations were carried out between the four PCA factors extracted, slope, elevation, longitude, and latitude variables.
Fragment variable | Slope | Elevation | Longitude | Latitude | PCA 1 | PCA 2 | PCA 3 | PCA 4 |
---|---|---|---|---|---|---|---|---|
Slope | ||||||||
Correlation coeff. | 1 |
|
0.15 |
|
−0.03 | −0.04 | 0.02 | −0.10 |
|
. | 0.02 | 0.14 | 0.00 | 0.74 | 0.71 | 0.84 | 0.31 |
Elevation | ||||||||
Correlation coeff. | 1 | 0.01 |
|
|
0.07 | −0.06 | −0.07 | |
|
. | 0.89 | 0.00 | 0.04 | 0.50 | 0.57 | 0.52 | |
Longitude | ||||||||
Correlation coeff. | 1 | 0.10 | −0.11 | −0.05 |
|
0.06 | ||
|
. | 0.32 | 0.29 | 0.60 | 0.00 | 0.56 | ||
Latitude | ||||||||
Correlation coeff. | 1 | −0.02 | 0.01 | −0.05 | −0.07 | |||
|
. | 0.88 | 0.91 | 0.64 | 0.48 | |||
PCA factor 1 (area) | ||||||||
Correlation coeff. | 1 | 0.07 | 0.05 |
|
||||
|
. | 0.47 | 0.62 | 0.00 | ||||
PCA factor 2 (shape) | ||||||||
Correlation coeff. | 1 | −0.03 | 0.05 | |||||
|
. | 0.74 | 0.63 | |||||
PCA factor 3 (proximity) | ||||||||
Correlation coeff. | 1 | −0.05 | ||||||
|
. | 0.60 | ||||||
PCA factor 4 (contiguity) | ||||||||
Correlation coeff. | 1 | |||||||
|
. |
Significant values are highlighted.
Difference in mean values of the patch’s variables used as input data for the linear regression analysis. The data is analysed between forest types. Information on geographic location of the patches is not included in the table. PCA 1: area; PCA 2: shape; PCA 3: proximity; PCA 4: contiguity.
Forest class | PCA 1 mean ± 1 SD | PCA 2 mean ± 1 SD | PCA 3 mean ± 1 SD | PCA 4 mean ± 1 SD | Slope mean ± 1 SD | Elevation mean ± 1 SD |
---|---|---|---|---|---|---|
Coniferous |
|
|
|
|
|
|
Nonconiferous |
|
|
|
|
|
|
Scattered veg. |
|
|
|
|
|
|
Kruskal-Wallis H | 0.501 | 11.07 | 35.08 | 0.58 | 4.54 | 1.95 |
|
0.77 | 0.00 |
|
0.74 | 0.10 | 0.37 |
(a) The 1st and 2nd PCA factors extracted; these are mainly composed of variables related to area-perimeter and shape index. (b) The 3rd and 4th PCA factors loaded by variables related to the patch’s proximity and contiguity.
In total, we recoded 2491 woody plants. Data from 139 individuals was discarded as they belonged to the 4 eliminated plots. The final analyses were carried out on the basis of data from 2352 plant individuals belonging to 17 species within 7 plant families, which presented two main seed dispersal mechanisms—anemochory and zoochory (Table S3) A significant difference in species richness and diversity among forest classes, where the coniferous forest showed the highest values, followed by the non-coniferous forest and the scattered vegetation class, was found (Table
Comparison of the mean values of woody plant species’ richness and species’ diversity between forest classes.
Forest class | Species |
Species |
---|---|---|
Coniferous |
|
|
Nonconiferous |
|
|
Scattered veg. |
|
|
Kruskal-Wallis H | 10.34 | 13.25 |
|
0.005 | 0.001 |
The results of the multiple regression models (96 plots) showed that only the “Proximity” between forest fragments had a significant effect on the woody plant species, richness but none on the species diversity (Table
Results of the multiple regression analyses using (a) the complete plots dataset (96 plots) and (b) the dataset divided by forest type.
Analysis | Response variable |
|
|
|
Explanat. variable | Standardized coefficient |
|
---|---|---|---|---|---|---|---|
(a) Complete dataset |
|
0.16 | 5.64 | 0.00 | Slope | 0.34 | 0.00 |
Elevation | 0.25 | 0.02 | |||||
|
0.13 | 8.38 | 0.00 | Slope | 0.32 | 0.00 | |
Proximity | 0.20 | 0.03 | |||||
(b) Forest types | |||||||
Coniferous |
|
— | — | — | — | — | — |
|
0.13 | 5.17 | 0.030 | Longitude | −0.40 | 0.03 | |
Nonconiferous |
|
0.17 | 7.84 | 0.00 | Slope | 0.44 | 0.00 |
|
— | — | — | — | — | — | |
Scattered veg. |
|
0.19 | 3.66 | 0.02 | Latitude | −0.38 | 0.02 |
|
0.20 | 5.30 | 0.01 | Slope | 0.35 | 0.03 |
—: not significant.
The coniferous forest species’ richness was negatively related to the longitudinal location but did not present any relation with the fragmentation factors analysed. Furthermore, none of the variables included in the regression analysis interacted with species, diversity of this forest type. For the non-coniferous forest a positive correlation between diversity and the slope of the forest fragments was obtained. In this forest class the steeper sites showed higher species, diversity. Species, richness was not significantly related to any of the fragments variables studied.
The regression models for the scattered vegetation showed no effect of fragment characteristics on the woody plants species, richness and diversity (Table
Fragmentation effects on the plant community strongly differed according to forest class (Table
The results showed that greater proximity between patches (reducing isolation) within forest class increases species’ richness and diversity. Similar results have been found in the highlands of Chiapas, Mexico, by Ochoa-Gaona et al. [
In this study it was observed that the area-perimeter of the patches did not have an effect on the species, richness and diversity. This finding concurs with those of Ochoa-Gaona et al. [
Species’ diversity showed a negative relation with the increment in latitude in the case of the scattered vegetation forest class. The scattered vegetation forest class is the one with the lowest values for the species’ richness and diversity, which might be explained by three reasons: (1) the scattered vegetation class comprises mostly shrub species spread out in the patch area, (2) this forest class has more bare soil area available than the coniferous and non-coniferous forest classes, and (3) as observed in the field data collection, these scattered vegetation areas emerge more often after the logging processes, which mostly select
It is shown in this study that fragmentation effects on the woody plant species are not strong and that these depend on the forest type studied. Furthermore, it is shown that environmental factors, as slope characteristics, play a major role in determining species, richness and diversity at the plot scale (100 m2 sampling plot). In our study it is apparent that different forest classes respond in a different way and have different resilience to habitat fragmentation effects. It is further shown the contrasting effects of environmental characteristics and fragmentation depend on the vegetation studied and render more insight into the discussion on the effects of habitat fragmentation on plants, richness and diversity. The results of this study suggest that fragmentation must be analyzed in a broader scale, for the fragment as a whole, as the community assembly at a small sample scale area (100 m2) seems to be more influenced by ecological filters related to environmental variables limiting plant establishment and biotic filters related to competition than to factors limiting species colonization (e.g., isolation). Essentially, conservation actions should be guided by the particular vegetation class analysed and by the specific fragmentation/environmental variable(s) impacting the forest type under study, as the different vegetation types responses to the fragmentation effects have been widely shown. It is essential to investigate the forest fragmentation effects on biodiversity at the landscape level, followed by an in-depth analysis at the forest type level (finer scale) in order to obtain more detailed information about the effect of fragmentation in the species’ richness and diversity on different vegetation classes.
In this particular area, conservation efforts should be directed to reduce isolation between forest fragments, as this has a significant impact on the species’ richness and diversity. This highlights the importance that seed dispersal plays for the plant communities’ in this particular study area, as colonization and genetic exchange might be facilitated in forest fragments that are contiguous or closer to each other. There was no evidence showing an impact of fragment size on the woody plant species’
The author declares that there is no conflict of interests regarding the publication of this paper.
This research was funded by the Netherlands Universities Foundation for International Cooperation (Nuffic), a grant from the “Amsterdamse Universiteits Vereniging (AUV)”, and a “STUNT” grant from the University of Amsterdam. Misión Indígena Bawinokachi and Maria Pontes are thanked for their logistical support. Luz Maria Stanek and Jorge Alberto Pérez de la Rosa (University of Guadalajara at CUCBA) are thanked for their help with the species identification.