Keywords: Biogeography, biodiversity, fragmentation, connectivity, matrix
Quantifying primary forest
fragmentation and matrix quality on
Mauritius
Author: Felix J.M. van Leuven
Supervisors: Dr. K.F. Rijsdijk & S.J. Norder MSc University of Amsterdam
Bachelor project Earth Science Bachelor Thesis 30-06-2014 Word count: 5892 Contact details UvA Studentnumber: 10264035 Adres: Oleanderstraat 29, 1031GP Amsterdam Email: felixvanleuven@hotmail.com Tel.: 0031 681 036 731
Darwin (1835), a first reference to an awakening interest in evolution: “When I see these Islands in sight of each other, & possessed of but a scanty stock of animals, tenanted by these birds, but slightly differing in structure & filling the same place in Nature, I must suspect they are only varieties… If there is the slightest foundation for these remarks the Zoology of Archipelagoes will be well worth examination; for such facts would undermine the stability of species.”
2
Abstract
This thesis is about the fragmentation of primary forest with the incorporation of the matrix and its potential effect on endemic biodiversity on the island of Mauritius. Species-area theories indicate that the number of species increases with area, consequently suggesting that smaller areas can sustain a lower biodiversity. However, the matrix theory suggests that the area surrounding a habitat can also to some extent support endemic biodiversity, effectively increasing the habitat’s area. The natural habitat on Mauritius has decreased by 98%, while the endemic biodiversity on Mauritius has not suffered such a reduction. Therefore, the question is raised: How can the endemic biodiversity of Mauritius still be so high, while the primary forest area has decreased to 2% of its initial area? Previous work has failed to quantitatively combine habitat fragmentation with the matrix’ effect on habitat connectivity. This research introduces a new method for the incorporation of the matrix in fragmentation studies and furthermore a tool for the quantification of fragmentation is given. The results suggest that the matrix adjacent to the primary forest can dramatically decrease the effective fragmentation of natural habitats, in this case, the primary forest supporting endemic arthropods. This can help conservationists efficiently manage matrix areas critical to a species survival and furthermore increases awareness on the importance of the matrix.
- Figure 1 Title picture: Mauritian mountains in the middle of a fragmented landscape (Wikipedia, 2012)
3
Contents
Abstract ... 2
1. Introduction ... 4
1.1 Context ... 5
1.2 Relevance ... 6
1.3 Research questions ... 7
1.4 Hypothesis ... 7
2. Methods & Materials ... 8
2.1 Digitalizing the land-use of Mauritius ... 8
2.2 Creating buffers around primary forest patches ... 9
2.3 Creating the fragmentation index ... 11
3. Results ... 12
3.1 NoBuffer data ... 12
3.2 RoadBlock (impermeable roads) Buffer data ... 12
3.3 NoRoads (permeable roads) Buffer data ... 14
3.4 Fragmentation index ... 14
4. Discussion ... 15
4.1 Discussion of Results ... 15
4.2 Discussion of Methods ... 17
4.3 Implications of Results ... 18
5. Conclusion ... 18
Acknowledgements ... 19
References ... 19
Appendix ... 22
Figures ... 22
Tables ... 30
4
1. Introduction
Mauritius is an island of the Mascarene archipelago, located in the Indian Ocean to the east of Madagascar (see Figure 2 and Figure 3). It has a surface area of 1850 km2 and a population density of 673 per km2 (Norder, et al., 2012). This archipelago belongs to the top three archipelagos with the richest plant biodiversity, of which 79,5% are single-island endemics (Caujapé-Castells, et al., 2010). Furthermore, Mauritius is home to 1032 beetle species, of which 46% are endemics (Motala, et al., 2007). Demand for land for sugarcane plantations and urbanization (Norder, et al., 2012) has led to an increasingly fragmented native forest on Mauritius, of which only 2% remains in 2000 (Kalla, 2006). Fragmentation and habitat loss can lead to an isolated and unstable population, vulnerable to disturbances like fire etc. (Caujapé-Castells, et al., 2010).
‘Fragmentation’ is defined as: “a large expanse of habitat transformed into a number of smaller patches of smaller total area, isolated from each other by a matrix of habitats unlike the original” (Wilcove, et al., 1986). This definition leads to the following effects of fragmentation on habitat patterns: 1) reduction in habitat amount, 2) increase in number of patches, 3) decrease in size of habitat patches, 4) increase in isolation of patches (Fahrig, 2003) (see Figure 5).
However, while the native forest on Mauritius has decreased to only 2% of its initial area, biodiversity on Mauritius is still high and thus did not follow the trend as would be expected when relating to the species-area theories by Arrhenius (1921) and MacArthur & Wilson (1963).
Figure 3 Location of Mauritius relative to Africa (Google Inc., 2014)
5
The aim of this research is approaching this peculiarity from a habitat fragmentation and matrix perspective. The organisms of interest are endemic Mauritian beetles (Coleoptera) as the method was based on a research on beetles by Meijer, et al. (2011). The main research question is: How can the endemic Coleoptera biodiversity of Mauritius still be so high (46%) while the native forest area has decreased by 98%?It is hypothesised that the matrix surrounding the primary forest can act as an extension of primary forest and therefore increase the effective habitat area and subsequently increase habitat connectivity, reducing the vulnerability of endemic populations in primary forest patches.
To answer the research question, an ArcGIS study was performed in which the size and abundance of primary forest patches will be assessed. Subsequently, the quality of the matrix surrounding these patches will be assessed. The analysis will be performed on a manually digitalized land-use map of Mauritius. For this research a new method was designed to assess the critical matrix areas supporting endemic biodiversity. And furthermore, a tool to quantify forest fragmentation was created.
In this paper first the context and relevance will be explained, followed by the research questions and hypothesis. Secondly, the used method is described and the results and a discussion of the results are given.
1.1 Context
The ecological species-area law by (Arrhenius, 1921) states that “the number of species increases continuously as the area increases”. Furthermore, MacArthur & Wilson (1963) argued in their ‘Equilibrium theory of insular zoogeography’ that there exists an equilibrium between extinction and immigration of species on islands. This theory gives a general view on the biodiversity of islands; with increasing distance from islands, the immigration will decrease and with decreasing surface area of islands, the extinction rate will increase. Therefore (generally speaking), a large island near a migration source (e.g. a continent) would have the highest biodiversity (see Figure 4).
The same theory might be applied to fragmented patches of nature (i.e. fragmented primary forest). The bigger the patch, the more species the patch can sustain. Small and isolated habitat fragments may be more vulnerable to disturbances such as fires, inbreeding and genetic variation, and edge effects may reduce the effective area for endemic species to live on. Furthermore, habitat fragmentation can help invasive species replace endemics, reduces endemic population sizes and increases the isolation of these populations (Caujapé-Castells, et al., 2010) (Lienert, 2004). Therefore, effects of fragmentation are especially felt by endemic populations (Meijer, et al., 2011).
6
However, fragmentation does not bydefinition lead to a negative situation. Meijer, et al. (2011) state that “natural habitat patches are not necessarily islands in a sea of uninhabitable matrix; the landscape surrounding reserves is an important factor in supporting biodiversity.” The degree to which the matrix can benefit species in natural habitats is dependent on its quality. Here, the matrix is defined as: “the non-habitat surrounding the native non-habitat patches of interest” (Ricketts, 2001). A ‘patch’ is defined as a piece of land covered by a different land-use than the surrounding area. For instance, when looking at primary
forest, a high-quality matrix would be secondary forest, allowing interfragment migration and therefore reducing extinction risks. In case of a low quality matrix, such as intensive agriculture, inevitable (and very natural) local extinctions cannot be balanced by regional populations due to the absence of interfragment migration; in this way local extinctions can ultimately lead to regional extinctions (Vandermeer & Perfecto, 2007).
1.2 Relevance
Islands are an enormously important source of information and an unparalleled testing ground for various scientific theories. Studies on insular biogeography have contributed a major part of evolutionary theory. Clusters of islands can be seen as simpler microcosms, as opposed to continental and oceanic biogeography, of which resident populations can be identified. The very multiplicity and variation of islands and their ecology provide the necessary replications in natural experiments (MacArthur & Wilson, 1967).
Figure 5 Forest fragmentation in Warwickshire, England, from 400-1960 A.D. (black is forest) (Wilcove, et al., 1986)
Figure 4 Equilibrium theory: Rate of extinction/immigration on y-axis and number of species on x-y-axis (MacArthur & Wilson, 1963)
7
Furthermore, islands are of great importance for global plant diversity; about one quarter of all known extant vascular plant species are endemic to islands. The Mascarenes belong to the top three archipelagos with the richest plant biodiversity, of which more than 60% of the endemics are single-island endemics (79,5% in the Mascarenes). These single-single-islands endemics have a very limited distribution and are therefore at higher risk of extinction than widespread endemics. Furthermore, rare and endemic plant species are more common in undisturbed (and often small) patches of primary forest than in secondary vegetation (Caujapé-Castells, et al., 2010). The same goes for many other endemic species (Meijer, et al., 2011).In the near future, human pressure on ecosystems will likely increase more markedly on islands than on continents. Additionally, small (and therefore vulnerable) islands have a significantly higher number of endemics per unit area than large islands. However, one major problem here is that the percentage of highly threatened taxa on these islands is also much higher than on large islands (Caujapé-Castells, et al., 2010).
1.3 Research questions
- How can the endemic Coleoptera biodiversity of Mauritius still be so high (46%) while the primary forest area has decreased to 2% of its initial area?
o What role does the matrix surrounding primary forest play in supporting endemic biodiversity, if any?
o What functions, if any, of the original habitat can the matrix adopt for specific species?
o To what extent do roads affect habitat fragmentation?
1.4 Hypothesis
Literature on the matrix and its effect on biodiversity suggest that the matrix, when sufficient in quality, can significantly increase the effective habitat area of certain species, especially in a highly fragmented landscape. This will allow those species to balance local extinctions by migrating and thus lessening their vulnerability. Furthermore, the matrix is said to provide additional foraging and breeding area for certain species. Therefore, it could be that while the native forest of Mauritius has significantly decreased in size, the surrounding matrix (partially) compensates for this decrease. This hypothesis will be tested by using endemic ground-foraging arthropods as model species.
8
2. Methods & Materials
The following method contained three major steps: First, a land-use map along with a roadmap and a forest-quality map were digitalized and combined. Secondly, the combined land-use map was used to create buffers around primary forest patches. And lastly, these maps were used to create a Forest Fragmentation Index indicating the degree of habitat fragmentation on Mauritius. This procedure was realised using ArcMap 10.0 (ESRI, 2014) to create the maps, MATLAB R2012b (MathWorks, 2014) was used to create the metrics and Microsoft Excel 2010 (Microsoft, 2014) was used to organize the data.
2.1 Digitalizing the land-use of Mauritius
The land-use map used in this research was provided by the ‘Resource Atlas for Mauritius and Rodrigues’ (Kalla, 2006) (see Appendix, Figure 15) and contained seven land-use categories: reservoirs/water bodies, built-up areas, forest plantation, tea, forests, savannah and scrubs, and sugarcane. This map was georeferenced as it initially did not have any spatial reference. After georeferencing, a shapefile was created to be able to create polygons around the different land-use types. The polygons were cut at a 1:6000 scale. After creating all polygons, specific Id’s were ascribed to each polygon: 1 = Water, 2 = Roads, 3 = Forest plantations, 4 = Tea, 5 = Primary forest, 6 = Savannah & Scrubs, 7 = Sugarcane, 8 = Urban areas, 999 = Unknown.
After identifying all polygons, a road map of Mauritius was created using the ‘Streets’ basemap provided by ArcGIS (ESRI, 2014) and integrated into the digitalized land-use map. The road polygons were cut at a 1:4000 scale and enclosed the whole road (with lines on the left and right side of the road and the top and bottom, making the roads polygons instead of lines). Areas with a high road density not identified on the land-use map as urban areas were enclosed along the edges of the roads and assigned to Urban areas. The road polygons were merged to one polygon and then exploded (using the Explode Multi-part Feature) to make an entity of all roads, only separated by Urban areas and Water bodies. Furthermore, the ‘Streets’ basemap was used to create more accurate water body polygons on places that were already identified as water bodies by the land-use map.
As the Kalla (2006) land-use map only contained one area identified as primary forest (Forest), a forest cover map indicating forest quality on Mauritius in 1995 (Page & d’Argent, 1997) (see Appendix, Figure 16) was integrated into the digital land-use map to expand the primary forest area. This map indicates six different forest qualities of which three are different native forest qualities. For this research only these three were used. These were combined to one category, Primary forest, as the Kalla (2006) land-use map only identifies one quality native forest. The new primary forest
9
polygons could overlap any existing land-use polygons except for water bodies, roads and urban areas.2.2 Creating buffers around primary forest patches
The next step in this research was to create buffers around primary forest patches. First, the coordinate system had to be changed so that area calculations could be performed. The coordinate system was changed to Sphere Sinusoidal, GCS Sphere. Data provided by Meijer, et al. (2011) on decreasing abundance and number of single-island endemic (SIE) species of Arthropods with distance from native forests was used to determine the size of these buffers. Buffers were used in this research as a representation of the quality of the matrix adjacent to primary forest and its functions (such as increased connectivity) from the perspective of an endemic arthropod. These buffers indicated the maximum distance an endemic arthropod travels outside of its natural habitat (primary forest). The size of this buffer was dependent on the land-use type adjacent to the native forest. For this a few assumptions were made: 1) Forest plantations and Savannah & Scrubs both counted as secondary forest, while 2) Tea and sugarcane plantations both counted as intensive agriculture. Furthermore, 3) Buffers could not overlap urban areas, water bodies and roads.
The maximum distance an Arthropod SIE travels outside of primary forest was 4 kilometres when the adjacent use was secondary forest. The maximum distance was lower when the adjacent land-use was intensive agriculture. As Meijer, et al. (2011) did not provide any data on the distance a SIE travels into intensive agriculture, the assumption was made that this distance is 1 kilometre. This distance was calculated using Meijer, et al. (2011) Table 2 (see Table 5) indicating the number of single-island endemic species per land-use
type. The number of endemics in intensive pastures was roughly 25% of the number of endemics in exotic forests. This percentage was used to calculate the distance for intensive agriculture: 25% of 4 km = 1 km.
Buffers of both sizes were created individually around the primary forests, giving two shapefiles. These shapefiles (one with a 1 km buffer and one with a 4 km buffer) were intersected with intensive
10
polygons respectively. The output shapefiles contained the intensive agriculture and secondary forest area’s overlapped by a 1 km buffer and a 4 km buffer respectively. The polygons in these shapefiles were then combined with the primary forest polygons to create a shapefile containing the ‘new’ primary forest.The new primary forest patches were analysed on any oddities. For this analysis a few assumptions were made: 1) Patches overlapped by a buffer had to be directly connected (or via another buffer) to primary forest to be qualified as an extension of primary forest (patches that were not connected in this way were removed). However, 2) roads between buffers did not hamper this connection, so it was possible to have the following sequence: primary forest – buffer – road – buffer (see Figure 6). This second assumption was, however, dependent on the analysis:
1. Roads are impermeable: Roads were absolute barriers and separated buffer areas from primary forest. If a buffer area was not directly connected to a primary forest, the buffer was removed. As such, the buffer to the left of the road, found in the sequence in Figure 6, would be removed. This analysis was based on research concerning the movement pattern of arthropods in areas adjacent to agricultural roads. The main outcome of this research was that roads hampered or even stopped the dispersal of ground-foraging arthropods and reduced the immigration rate (Mader, et al., 1990).
2. Roads are permeable: The secondary forest and intensive agriculture area covered by the buffer was seen as an extension of primary forest and was therefore classified as such. Furthermore, the roads running through the primary forest (and buffer) were in this case merged with the new primary forest polygons. This analysis did not apply to arthropods, but acted more as a means to compare and to see how roads could contribute to forest fragmentation.
Lastly, these shapefiles (containing primary forest and buffers) were integrated into the land-use map without buffers and the area of each polygon was calculated to form the maps that were used to answer the research questions. Integrating the shapefiles was done by first converting the shapefiles (polygons) (the ones with buffer data and the one without a buffer) to rasters (using the ‘Polygon to Raster’ tool). The ‘Cellsize’ was set to 1, for an optimal resolution. The rasters (buffer and no buffer) were then Mosaicked to form a new raster. Lastly, the new raster was converted back to a polygon shapefile (using the ‘Raster to Polygon’ tool) also with a ‘Cellsize’ of 1. The new shapefiles had a large amount of oddities (1x1 polygons), as the conversion from and to rasters is not one-on-one. These oddities were removed using the ‘Integrate’ tool, also with the ‘Cellsize’ set to 1. This left a clean shapefile, ready for analysis. However, due to the small changes made as a result of the conversion to and from rasters, the areas were a little different from the NoBuffer map; though the difference in the total area of Mauritius was as small as 0,0003%.
11
2.3 Creating the fragmentation index
The data obtained from ArcMap was used to create metrics in MATLAB. The MATLAB output was further analysed and organized in Microsoft Excel.
The fragmentation index was based on the main effects of fragmentation on habitat patterns as stated by Fahrig (2003): 1) reduction in habitat amount, 2) increase in number of patches, 3) decrease in size of habitat patches, 4) increase in isolation of patches. However, isolation was not used as there was no data on isolation. To create this index there was a need for metrics that quantify these three points. All the metrics can be found in the Tables section of the Appendix (Table 2, Table 3 and Table 4). The metrics used for the fragmentation index are:
1. H = Change in habitat amount (ranging from 0 – 1) , PA = Effective primary forest area (km2), TA = The total area of Mauritius (km2). A higher H means a larger primary forest area:
Eq. 1
2. P = Patchiness of primary forest (ranging from 0 – 1), PP = Effective number of primary forest patches. A higher P means a lower patchiness. Dividing it by 37 was based on the fact that this is the maximum amount of primary forest patches:
Eq. 2
3. PS = Change in habitat patch size (ranging from 0 – 1), ME = Mean effective primary forest patch area (km2), MA = Mean patch area (km2). A higher PS means a larger primary forest patch size.
Eq. 3
After calculating these 3 values for the ‘NoBuffer’, ‘RoadBlock’ and ‘NoRoads’ data, the Fragmentation index was calculated using equation 4:
Eq. 4
The FFI has a range from 0 – 1 (0 = no fragmentation, 1 = extremely fragmented). The actual creation of the fragmentation index by combining these metrics was based on the methods used by Butler, et al. (2004), who stated that a solution to choosing among multiple fragmentation metrics is to
12
combine them into a single index. Furthermore, this index allows the degree of fragmentation to be compared within a study area.3. Results
3.1 NoBuffer data
The NoBuffer data (see Appendix, Table 2) was obtained from the digitalized land-use map (see Figure 9) without a buffer. This data was used for comparison with land-use maps that did include a buffer zone.
The area of Mauritius taken up by secondary forest when including primary forest was almost 28% (secondary forest was also seen as a possible habitat). This area was distributed over little more than 25% of the total amount of individual landscape patches found on Mauritius (205 of the 813 patches), with a mean patch area of 2,5 km2 (and a variance of 44 km2). Of the total habitat area (secondary forest and primary forest), 38% was covered by primary forest. The primary forest was divided over 37 distinct patches, with a mean of 5,3 km2 and a high variance of 161 km2. Primary forest accounted for 10,6% of the total Mauritian surface area.
3.2 RoadBlock (impermeable roads) Buffer data
The RoadBlock data (see Appendix, Table 3) was obtained from an altered version of the NoBuffer land-use map (see Figure 10 and Figure 11). The RoadBlock map contained a buffer area around the
NoBuffer RoadBlock NoRoads Mean effective PF patch area (km2) 5 27 117 Total (effective) PF area (km2) 197 544 584 0 100 200 300 400 500 600 700 0 20 40 60 80 100 120 140
NoBuffer RoadBlock NoRoads
Total patches 813 720 678 Effective PF patches 37 20 5 0 5 10 15 20 25 30 35 40 600 650 700 750 800 850
13
primary forest overlappingsecondary forest and intensive agriculture. Furthermore, roads were seen as absolute barriers in this scenario.
There was a clear decrease in effective primary forest patchiness (20 patches instead of 37) when considering the matrix adjacent to primary forest (the buffer) as an extension of primary forest area (the overall patchiness had also decreased from 813 to 720 patches) (see Figure 7, right). Consequently, the mean effective primary forest patch area was more than 5 times larger when including the matrix. Furthermore, there was a dramatic increase in effective total primary forest area of 176% (see Figure 7, left). There was a notable difference in number of buffer patches when comparing the scenario with impermeable roads to the permeable roads scenario (118 patches and 71 patches respectively). Furthermore, the mean area of primary forest plus buffer was nearly half of the mean found in the permeable roads data. Part IA in Buffer of total area IA Part SF in Buffer of total area SF RoadBlock 12% 70% NoRoads 11% 76% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%
Buffer area (km2) Actual PF area (km2)
RoadBlock 347 197 NoRoads 387 197 0 50 100 150 200 250 300 350 400 450 Mean actual PF
area (km2) Actual PF patches
NoRoads 8,6 23 RoadBlock 5,3 37 0,0 5,0 10,0 15,0 20,0 25,0 30,0 35,0 40,0 0,0 5,0 10,0 15,0 20,0 25,0
Figure 8 Top: Part of intensive agriculture area (IA) covered by buffer of total IA area & Part secondary forest (SF) covered by buffer of total SF area; Middle: Buffer area & Actual PF area; Bottom: Mean actual PF area & Actual PF patches
14
3.3 NoRoads (permeable roads) Buffer data
The NoRoads data (see Appendix, Table 4) was obtained from a map (see Figure 12 and Figure 13) very similar to the RoadBlock map, with the important difference that roads were seen as permeable and therefore served as no barrier whatsoever.
The total effective area of Primary forest was almost 200% larger when roads were seen as permeable than in the NoBuffer map and covered more than 30% of Mauritius (see Figure 7, left). Furthermore, a large part (about 76%) of the secondary forest was used as an extension of primary forest. Whereas this was only 11% for intensive agriculture (see Figure 8, top). However, the area of intensive agriculture supporting endemic biodiversity was only half of the secondary forest area with the same function. This was probably due to the fact that 64% of Mauritius was covered by intensive agriculture, while this was only 18% for secondary forest, making the intensive agriculture area a larger potential for being a buffer. The primary forest was distributed over only 5 patches, with a mean of 117 km2 (see Figure 7) and a range between 6 and 307 km2. The total patchiness compared to the RoadBlock data had decreased by only 5,8%, which was due to the merging of effective primary forest area with the roads running through this area (see Figure 7, right).
It was notable that about two-thirds of the new primary forest area was found in the buffer zone. Interestingly, when buffer zones were counted as equal to primary forest, the patchiness of primary forest decreased from 23 patches to 5 patches. Furthermore, the buffer zone contained about 3 times the amount of primary forest patches (Figure 8, middle).
3.4 Fragmentation index
Table 1 shows the Forest Fragmentation Index (FFI) and the values from which it was calculated. While the changes in habitat amount and habitat size were only slightly higher when comparing the RoadBlock data with the NoRoads data, the NoBuffer data was noticeably lower. This was different for the P values due to the big changes in primary forest patchiness (decreasing from 37 to 20 to 5). When studying the FFI values and comparing the NoBuffer data to the RoadBlock and NoRoads data, there was a decrease of 42% in fragmentation for the RoadBlock data and an additional decrease of 37% in fragmentation for the NoRoads data.
H
P
PS
FFI
NoBuffer
0,106
0,000
0,569
0,775
RoadBlock
0,292
0,459
0,905
0,448
NoRoads
0,313
0,865
0,976
0,282
Table 1 The Forest Fragmentation Index (FFI) and the metrics from which it’s calculated: H = change in habitat amount, P = patchiness of primary forest, PS = change in habitat size
15
4. Discussion
4.1 Discussion of Results
The fragmentation of primary forest on Mauritius is dependent on the approach taken to quantify fragmentation. When assuming that habitat patches are islands in an uninhabitable matrix one could argue that the primary forest on Mauritius is highly fragmented with a FFI of 0,775, a mean primary forest patch area of 5,3 km2 and a total of 37 isolated (either by roads or matrix) patches. What is more, when relating the fragmentation of primary forest to the theories on ‘species and area’ such as those by Arrhenius (1921) and MacArthur & Wilson (1963), one could further argue that the endemic biodiversity on Mauritius is highly threatened.
However, data on biodiversity shows that the Mauritian endemicity is still very high and very well developed given its size, age and isolation. The current endemism among beetle species is as high as 46% (Motala, et al., 2007) which is odd considering the drastic deforestation Mauritius underwent. If those species-area theories were applied to the fragmented primary forest of Mauritius, an endemicity as high as found would be impossible as endemic populations would be extremely isolated, decreased in size and vulnerable to disturbances. So the question arises: Is there anything apart from primary forest positively affecting the existence of these endemics? Several studies have found that populations in forest remnants can be positively influenced by the surrounding matrix due to increasing foraging or breeding area and increasing immigration through the matrix (allowing interfragment migration). Some species can even compensate for habitat loss by moving to other habitat types (Kupfer, et al., 2006).
The research by Meijer, et al. (2011) revealed how endemic Arthropods were affected by different matrix types and qualities, with special interest on endemic dispersal through the matrix. Secondary forests were significantly better at supporting endemics as opposed to intensive pastures. This relates to the research by Kupfer, et al. (2006) who argue that species are able to make better use of the matrix when its structure is closer to the natural habitat of that species (e.g. primary forest is more similar to secondary forest than it is to intensive pastures). However, the effects of the matrix are highly species specific (Prevedello & Vieira, 2010). Though, it is suggested that species avoiding the matrix tend to decline or disappear in fragments (Kupfer, et al., 2006). Which in turn would suggest that all species are affected by the matrix, either positively or negatively. Then again, the degree to which species are affected would vary greatly.
Now, the results of this research indicate that the matrix can be of great importance to the livelihood of endemic Arthropod species by drastically increasing the connectivity between native habitats. The effective fragmentation of the primary forest would decrease by 42% due to the ability of endemic beetle species to use the matrix as a sort of extension of their natural habitats. Moreover, the total
16
effective habitat area of an endemic arthropod increases by 175% due to its use of the matrix adjacent to primary forest. These results can be related to research by Ricketts’ (2001) who found that the type of intervening matrix can significantly influence the effective isolation of habitat patches.Interestingly, while more than 65% of Mauritius is covered by intensive agriculture and therefore qualifies as a low quality-matrix, the effective area adjacent to primary forest which could be used to support endemic biodiversity is only two-fifth smaller than the high-quality matrix area supporting biodiversity. However, in a scenario in which all area surrounding primary forest is qualified as a high-quality matrix, the total potential area able to sustain endemic biodiversity is almost 5 times larger than in case of a low-quality matrix (see Figure 14). This strengthens Vandermeer & Perfecto’s (2007) claim that “an improvement in matrix quality can outweigh the negative effects of habitat loss”. Nonetheless, these improvements are still very species specific and might depend on the degree to which a species is a habitat specialist or generalist, with the latter being more resilient and adaptable. Then again, when only considering primary forest as a natural habitat to endemic species, and the improvements follow the idea by Kupfer, et al. (2006), increasing the matrix’ structural similarity to primary forest would benefit all endemic species.
As shown in this research, the matrix could play a critical role in supporting endemic biodiversity especially by increasing habitat connectivity and expanding overall effective habitat area. However, for some species, and especially the epigeic arthropods in question, roads are still an absolute barrier according to Mader, et al. (1990) and serve as an important factor in habitat fragmentation. The results from this study show the extent to which roads influence this. For the sake of habitats, roads (especially paved) should be removed from native forests and the adjacent matrix when possible. As for these species the matrix is of no use if a road needs to be crossed to reach into the matrix, however high its quality. To illustrate this, the removal of roads would for epigeic arthropods further increase habitat connectivity by 37% as opposed to the scenario in which roads are barriers (and the effective primary forest patchiness would decrease from 20 to 5 patches). This is mainly due to the fact that roads hamper direct connectivity between habitat patches and therefore only allow indirect connection through the matrix.
The main research question can hereby be partially answered. The matrix surrounding primary forest can dramatically increase the effective habitat area and connectivity for endemic Arthropods. Whereas the primary forest has decreased to 2% of its initial area, the matrix can increase the effective habitat area to around 29% of the total area of Mauritius. Furthermore, increased connectivity between primary forest patches can help interfragment migration to stabilize endemic populations and protect them from local extinctions. Admittedly, there is a possibility that the remaining endemics are for the most part habitat generalists, suggesting that endemic extinctions in
17
the past were especially felt by specialists. Although, if that is the case, it would not be of any concern. It is however recommendable to research the degree to which species are specialists or generalists as this could influence the conservationist’s tactics. If a patch would consist primarily of specialist species, it is of great importance to keep the natural habitat as intact as possible. Whereas, in case of generalist species, it can be very beneficial to improve the quality of the matrix, as this might drastically increase the habitat area and connectivity for these species.Despite the fact that this research shows a decreased fragmentation of primary forest due to the matrix, the endemics (especially specialized endemics with low mobility) in the remaining forest are still facing many extinction pressures by, for example: the invasion of exotic plants, and increased predation and competition (by and with exotic species) (Kupfer, et al., 2006).
4.2 Discussion of Methods
There are many possible improvements on the methods used in this research. One of the most important improvements includes the use of more detailed maps. As for the roadmap, only the main roads, as shown in the ArcGIS ‘Streets’ basemap, were digitalized. Smaller roads were simply left out, while these can still contribute to habitat fragmentation (Mader, et al., 1990). Furthermore, the Kalla (2006) land-use map that was used was highly generalized; e.g. many urbanized areas were labelled as sugarcane plantations. Moreover, the Page & d’Argent (1997) forest-quality map was digitalized at a very low resolution, and for this research only the 3 types of native forests were digitalized and combined to form primary forest, which was done to correspond with the Kalla (2006) forest area. Using the generalized maps had for the most part to do with a lack of time. Increasing the resolution and accuracy of these maps will give better insight in the matrix and its quality and, as for roads, what kind of road you are dealing with (paved, sandy etc.). Moreover, the Page & d’Argent (1997) forest-quality map dates back to 1995 and the Kalla (2006) land-use map date back to 2000; both maps are of different years and are very likely not representative of the actual situation in Mauritius.
Further research could focus on more species with varying mobility to assess how they respond to fragmentation and habitat loss and to what extent they compensate for this by using the matrix. Furthermore, the fragmentation index could be tweaked to contain matrix effects. For example a metric could be created describing how resistant an island’s matrix is to animal movement, based on research such as the one by Ricketts (2001) about the effective isolation of habitat patches in fragmented landscapes.
18
4.3 Implications of Results
This research presented a new method that can be used to analyse species specific fragmentation, with the incorporation of matrix effects on connectivity (with the help of research on species specific matrix impacts, such as by Meijer, et al. (2011)). Furthermore, it can point out areas in the matrix that are critical to the conservation of endemic species and therefore help to efficiently manage these areas, either by improving them (making these matrix areas more similar to the native habitat in terms of structure) or stopping them from further degradation (stopping the conversion to a lower-quality matrix; especially urban areas). The results show just how important the matrix may be and can therefore increase public awareness on the correct use of the matrix.
It is, however, recommended to also conserve natural habitats as best as possible, as not all endemic species (specialists) can benefit from an increased matrix-quality. In addition, a tool was created (the Forest Fragmentation Index) for the analysis of the severity of fragmentation and as a means to compare different fragmentation scenarios (for different species). The combination of the method and tool can help landscape architecture in thinking of ecologically responsible ways of shaping the landscape, as different scenarios can be compared.
5. Conclusion
Islands can give crucial insight in how ecosystems behave and organize themselves. They are home to many unique species and contribute greatly to global biodiversity. However, many islands are under human pressures such as habitat destruction and fragmentation, and their biodiversity is at high risk. This is especially the case for endemic species that prefer their natural habitat.
When looking at the primary forest patches as ‘islands in an uninhabitable matrix’, one could say that Mauritius is highly fragmented (with a FFI value of 0,775) and that the primary forest is very susceptible to perturbations and its biodiversity thus at high risk. However, ecologists nowadays are more and more focussing on the matrix as a possible medium for intra-island connectivity, connecting forest patches and serving as an habitat extension rather than a barrier to species dispersal. Therefore, an approach to fragmentation with the incorporation of the matrix as an extension of primary forest is more accurate. Hence, when viewing the primary forest from a matrix perspective, fragmentation is reduced to 0,448 (FFI) due to an increased habitat area and patch size, and decreased patchiness. Moreover, it is shown that roads contribute to 37% of the fragmentation in primary forest and its adjacent matrix on Mauritius.
19
The decrease in native habitat area has not translated well into the observed decrease in endemic biodiversity, especially when only considering species-area theories. It is argued that the endemic arthropod biodiversity is probably partially resorting to the matrix for additional foraging and breeding area, and uses it as a means for interfragment migration. The degree to which species are able to effectively use the matrix is, however, very species specific. But it can be considered a fact that the matrix has an influence on species, albeit negatively in case of a very low-quality matrix (e.g. built-up areas) isolating populations, or positively due to a high-quality matrix, effectively expanding a species’ habitat and connecting populations.The tools presented in this research can be used to accurately point out areas critical to the conservation of a species. This can increase the efficiency of habitat conservation and increase the local awareness about the possible effects of converting the matrix surrounding the natural habitat. Therefore, not only the habitat itself but also the matrix can be managed correctly.
Acknowledgements
I want to thank Kenneth Rijsdijk for being such an enthusiastic and critical supervisor. And Sietze Norder for providing a reality check after a wave of Rijsdijk’s enthusiasm.
Furthermore, I’d like to thank my parents for sheltering me during two extremely busy weeks and catering to my every need (often late at night).
And, last but not least, I’d like to thank my little brother for the times he forced me to cut the crap and go to work.
References
1. Arrhenius, O., 1921. Species and Area. Journal of Ecology, 9(1), pp. 95-99.
2. Butler, B. J., Swenson, J. J. & Alig, R. J., 2004. Forest fragmentation in the Pacific
Northwest: quantification and correlations. Forest Ecology and Management,
Volume 189, pp. 363-373.
3. Caujapé-Castells, J. et al., 2010. Conservation of oceanic island floras: Present and
future global challenges. Perspectives in Plant Ecology, Evolution and Systematics,
Volume 12, pp. 107-129.
20
4. Diamond, J. M., 1975. The island dilemma: Lessons of modern biogeographic studies
for the design of natural reserves. Biol. Conserv., Volume 7, pp. 129-146.
5. ESRI, 2014. ArcGIS 10.0. [Online]
Available at: http://www.esri.com/software/arcgis
6. Fahrig, L., 2003. Effects of Habitat Fragmentation on Biodiversity. Annual Review of
Ecology, Evolution, and Systematics, Volume 34, pp. 487-515.
7. Google Inc., 2013. Google Earth. [Online]
Available at: http://www.google.nl/earth/download/ge/agree.html
8. Google Inc., 2014. Google Maps. [Online]
Available at: https://www.google.nl/maps/@o.2136714,16.98485,4z?hl=en
[Accessed 10 May 2014].
9. Kalla, C., 2006. ELP/Longman Resource Atlas for Mauritius and Rodrigues.
10. Kupfer, J. A., Malanson, G. P. & Franklin, S. B., 2006. Not seeing the ocean for the
islands: the mediating influence of matrix-based processes on forest fragmentation
effects. Global Ecology and Biogeography, Volume 15, pp. 8-20.
11. Lienert, J., 2004. Habitat fragmentation effects on fitness of plant populations - a
review.. Journal of Nature Conservation, Volume 12, pp. 53-72.
12. MacArthur, R. H. & Wilson, E. O., 1963. An Equilibrium Theory of Insular
Zoogeography. Evolution, 17(4), pp. 373-387.
13. MacArthur, R. H. & Wilson, E. O., 1967. The Theory of Island Biogeography.
Princeton, New Jersey: Princeton University Press.
14. Mader, H. J., Schell, C. & Kornacker, P., 1990. Linear Barriers to Arthropod
Movements in the Landscape. Biological Conservation, Volume 54, pp. 209-222.
15. MathWorks, 2014. MATLAB R2012b. [Online]
21
16. Meijer, S. S., Whittaker, R. J. & Borges, P. A., 2011. The effects of land-use change on
arthropod richness and abundance on Santa Maria Island (Azores): unmanaged
plantations favour endemic beetles. J Insect Conserv., Volume 15, pp. 505-522.
17. Microsoft, 2014. Microsoft Excel 2010. [Online]
Available at: http://office.microsoft.com/en-us/excel/
18. Motala, S. M., Krell, F. T., Mungroo, Y. & Donovan, S. E., 2007. The terrestrial
arthropods of Mauritius: a neglected conservation target. Biodivers. Conserv.,
Volume 16, pp. 2867-2881.
19. Norder, S., 2012. Land use classification and spatial pattern analysis of forest cover in
two Mascarene Islands (Mauritius and Reunion). pp. 1-14.
20. Norder, S., Rijsdijk, K., Dietz, T. & Rughooputh, S., 2012. Mauritius: Ruimtegebrek en
suikerriet. Geografie, Issue February, pp. 25-27.
21. Prevedello, J. A. & Vieira, M. V., 2010. Does the type of matrix matter? A quantitative
review of the evidence. Biodivers. Conserv., Volume 19, pp. 1205-1223.
22. Ricketts, T. H., 2001. The Matrix Matters: Effective Isolation in Fragmented
Landscapes. The American Naturalist, 158(1), pp. 87-99.
23. Vandermeer, J. & Perfecto, I., 2007. The Agricultural Matrix and a Future Paradigm
for Conservation. Conservation Biology, Volume 21, pp. 274-277.
24. Wikipedia, 2012. Wikipedia. [Online]
Available at:
http://commons.wikimedia.org/wiki/File:Sugarcane_plantation_in_Mauritius.jpg
[Accessed 28 April 2014].
25. Wilcove, D. S., McLellan, C. H. & Dobson, A. P., 1986. Habitat fragmentation in the
temperate zone. Conservation biology. The science of scarcity and diversity, pp.
237-256.
22
Appendix
Figures
23
Figure 10 Digitalized land-use map of Mauritius with the buffer area merged with primary forest. Here the roads are impermeable (RoadBlock)
24
25
Figure 12 Digitalized land-use map of Mauritius with the buffer area merged with primary forest. Here the roads are permeable (also merged with primary forest) (NoRoads)
26
27
28
29
30
Tables
Total area 1865,523 km2
Total area PF 196,854 km2
Total area SFP* 513,713 km2
% Area PF OF Total area 10,55% area
% Area SFP OF Total area 27,54% area
% Area PF OF SFP 38,32% area
Variance area PF 160,835 km2
Variance area SFP 43,955 km2
Mean area PF 5,320 km2
Mean area SFP 2,506 km2
Total patches 813 patches
Patches PF 37 patches
Patches SFP 205 patches
% Patches PF OF Total patches 4,55% patches
% Patches SFP OF Total patches 25,22% patches
*SFP = Savannah & Scrubs + Forest plantations + Primary forest; PF = Primary forest
NoBuffer - Only land-use
Table 2 NoBuffer metrics
Total area 1865,528 km2 Total area PF + buffer 543,988 km2
Total area PF 543,968 km2 Total area buffer 347,135 km2
Total area SF 103,723 km2 Total area PF 196,854 km2
Total area IA 1072,844 km2 Total area PF + buffer OF Total area 29,16% area
Total area SF in buffer 238,112 km2 Part PF OF Total area 10,55% area
Total area SF + buffer 341,834 km2 Part PF OF PF + buffer area 36,19% area
Total area IA in buffer 146,948 km2 Mean area PF + buffer 3,510 km2
Total area IA + buffer 1219,792 km2 Mean area PF 5,320 km2
Part PF OF Total area 29,16% area Variance area PF + buffer 55,431 km2
Part IA buffer OF Total area IA 12,05% area Variance area PF 160,835 km2
Part SF buffer OF Total area SF 69,66% area Patches PF + buffer 155 patches
Mean area PF 27,198 km2 Patches buffer 118 patches
Variance area PF 1122,148 km2 Patches PF 37 patches
Total patches 720 patches Part buffer OF Total patches 16,39% patches
PF patches 20 patches Part PF OF Total patches 5,14% patches
Part PF OF Total patches 2,78% patches Part PF OF PF + buffer patches 23,87% patches
*PF = Pri ma ry fores t; SF = Seconda ry fores t; IA = Intens i ve a gri cul ture
RoadBlock (roads are IMPERMEABLE) RoadBlock (roads are IMPERMEABLE)
31
Total area 1865,528 km2 Total area PF + buffer 584,162 km2
Total area PF 584,160 km2 Total area buffer 387,054 km2
Total area SF 78,758 km2 Total area PF 197,108 km2
Total area IA 1059,888 km2 Total area PF + buffer OF Total area 31,31% area
Total area SF in buffer 255,885 km2 Part PF OF Total area 10,57% area
Total area SF + buffer 334,644 km2 Part PF OF PF + buffer area 33,74% area
Total area IA in buffer 131,168 km2 Mean area PF + buffer 6,214 km2
Total area IA + buffer 1191,056 km2 Mean area PF 8,570 km2
Part PF OF Total area 31,31% area Variance area PF + buffer 293,234 km2
Part IA buffer OF Total area IA 11,01% area Variance area PF 787,024 km2
Part SF buffer OF Total area SF 76,47% area Patches PF + buffer 94 patches
Mean area PF 116,832 km2 Patches buffer 71 patches
Variance area PF 13721,787 km2 Patches PF 23 patches
Total patches 678 patches Part buffer OF Total patches 10,47% patches
PF patches 5 patches Part PF OF Total patches 3,39% patches
Part PF OF Total patches 0,74% patches Part PF OF PF + buffer patches 24,48% patches
*PF = Pri ma ry fores t; SF = Seconda ry fores t; IA = Intens i ve a gri cul ture
NoRoads (roads are PERMEABLE) NoRoads (roads are PERMEABLE)
Table 4 NoRoads metrics
Table 5 The number of single-island endemic species and the percentage single-island endemics out of the total number of Azorean endemic species in the four habitat types of Santa Maria (Meijer, et al., 2011)
