Geodiversity of Mauritius Developing a geodiversity index for Mauritius to predict native vascular plant diversity hotspots

(1)

Geodiversity of Mauritius

Developing a geodiversity index for Mauritius to

predict native vascular plant diversity hotspots

Raphael Reinegger Student Nr. 10193731 Supervisor: Kenneth Rijsdijk

5/29/2015

vascular plant diversity hotpots. Only a few attempts to create such indices for other islands and regions have been made. This paper tries to tackle some of the problems with the methodological design of previous geodiversity indices. Areas that contained the remaining good quality native forests were used as hotspots for native vascular plant diversity, because they contain over 50% of the native vascular plant cover in Mauritius. However, the index failed to show a positive

correlation between geodiversity and native vascular plant diversity. The choice of parameters could have resulted in a model that is not suited for estimating geodiversity in order to explain native vascular plant diversity patterns. The distinctive geographic isolation of the good quality forests due to deforestation on Mauritius could also be an explanation for the fact that no correlation between geodiversity and native vascular plant diversity was found.

(2)

Acknowledgements

I wish to express my gratitude to those people without whom it would not have been possible to complete this work. In no particular order they are: Kenneth Rijsdijk, for his knowledge of islands and biodiversity and of course his guidance and enthusiasm, Harry Seijmonsbergen, who provided me with great amounts of information about geodiversity, Sietze Norder, for helping me with finding datasets, Rachel Chambers, who I never got to meet in person, but whose master thesis served as the foundation of this work, and last but not least, Pieter Zitman, for helping me out and giving me advice.

(3)

Introduction

Theoretical framework

Geodiversity can be described as the variety of abiotic factors within an environment (Chambers, 2014). Murray Gray tried to define and thoroughly discuss and explain the concept of geodiversity (Gray, 2004). This has resulted in a widely accepted definition of geodiversity:

‘The natural range (diversity) of geological (rocks, minerals, fossils), geomorphological (land form, processes) and soil features. It includes their assemblages, relationships, properties, interpretations and systems (Gray, 2004).’

The diversity of these components creates environmental heterogeneity and a varied landscape. It includes many of the environmental processes that are considered to be drivers of biodiversity (Parks & Mulligan, 2009). According to Burnett et al. (1998) species richness, diversity and dominance are all related to spatial heterogeneity of abiotic properties. According to Gray (2008) the complexity within the non-living world (geodiversity) leads to complexity in the living world. ‘A varied landscape

consisting of diverse abiotic habitats and structural organisms will lead to a broader available niche space available for species to fill’ (Dufour et al., 2006). High geodiversity is therefore related to high biodiversity.

This has resulted in the use of geodiversity as a measure for explaining biodiversity patterns. According to Hjort et al. (2012) the predictive performance of their models for species richness was higher when measures of geodiversity were included in the set of explanatory variables. Their results also suggested that models with only measures of geodiversity showed a better predictive

performance than models that included other more commonly used abiotic factors. Chambers (2014) created a geodiversity index for Tenerife that could ultimately be related to a potential biodiversity (vegetation) index. A statistically significant relationship was found between the hotspots of the two indices. However, measuring geodiversity remains a challenge and only a few attempts at applying a methodology to measure geodiversity have been made (Chambers, 2014).

(5)

Study Area

Mauritius is an island in the Indian Ocean about 2000 kilometers off the coast of south-east Africa. Together with Rodrigues and Réunion it forms the Mascarene Islands. These volcanic islands share a common geologic origin, because they were all created by the Réunion hotspot beneath the

Mascarene plateau. These islands have developed a unique flora and fauna that display a high degree of endemism. Mauritius in particular has a biodiversity that is usually not found in such a small area (CBD, 2007). Ever since the discovery of Mauritius, its tropical forests have been subjected to degradation. The exploitation of the forests and introduction of alien species has resulted in a total loss of about 98% of the tropical rainforests. ‘The area of good quality native forest is estimated to cover less than 2% of the island (CBD, 2007).’ This forest once stretched from the mountain tops of the central plateau to the shore, but is now concentrated in the Black River Gorges National Park in the south-west, the Bamboo Mountain Range in the south-east, the Moka-Port Louis Ranges in the north-west and some isolated mountains (figure 1.1).

(6)

Fig. 1.2: Forest cover in Mauritius (Page & d’Argent, 1997)

As can be seen in figure 1.2, the largest part of the Mauritian forests is made up of invaded native forests and exotic plantation forests. The remaining 2% of good quality native forest is marked with red in figure 1.2. Despite of the fact that these good quality native forests are confined to small patches, Mauritius still has 671 species of indigenous flowering plants. About 310 of these plants are endemic (CBD, 2007). Mauritius has therefore been classified as a centre of (vascular) plant diversity by the IUCN (WWF/IUCN, 1994). Remarkably, over 50% of the native vascular plant cover is located within the patches of remaining good quality native forest (CBD, 2007). This means that these patches (red polygons) can be considered as hotspots of native vascular plant diversity.

(7)

Aim

The aim of this research was to create a geodiversity index for Mauritius that could be used to predict native plant diversity hotpots. Only a few attempts to create such indices for other islands and regions have been made. Two noteworthy problems with the methodological design were the identification of the elements to be included in geodiversity estimation and the selection of grid resolution (Pereira et al., 2013; Chambers, 2014). This paper tried to tackle the first problem with the help of a paper by Parks & Mulligan (2008) in which a way of developing a compound index for geodiversity that can model broad scale biodiversity patterns is proposed. The second problem was tackled by considering some analytical and empirical rules for selecting a suitable grid resolution from a paper by Hengl (2006). The model that was developed by Chambers (2014) for Tenerife in ArcGIS served as a foundation for this model. It is a model that can be applied to other volcanic islands and allows the addition and removal of different parameters.

For metrics of biodiversity this work focuses on native vascular plant richness, because geodiversity is considered to be more of a driver of plant diversity and richness than a driver of animal diversity (Araujo et al. 2001). Furthermore, the remaining good quality native forest patches indicated in figure 1.1 contain 50% of the native vascular plant cover (CBD, 2007). Because of the lack of data on native vascular plant richness, these patches will serve as vascular plant diversity hotspots.

Relevance

Classifying and quantifying geodiversity in ArcGIS is more efficient than simple documentation of observations on paper because GIS databases can be shared and updated globally. Bringing together multiple sources of information and combining them into a map can make it more understandable to people without a background in earth science and therefore makes its utilization more efficient in land-use planning and environmental management (Pereira et al. 2013; Chambers, 2014).

Furthermore, gathering data on species richness and community composition requires field work, which requires a lot of time and money. A model that can predict the location of biodiversity hotspots and explain biodiversity patterns with geodiversity patterns will save a lot of time and money,

(8)

Research question and hypothesis

Main question:

- Do geodiversity hotspots, that are located with a geodiversity index (GDI) that is based on the index that was developed by Chambers (2014), fall within the boundaries of the patches of good quality native forest that will be used as hotspots of vascular plant richness in Mauritius?

Sub question:

- What parameters should be included or removed in the model that was developed by Chambers (2014) to calculate geodiversity hotspots?

- What is the composition of the geodiversity calculated with a geodiversity index (GDI) that is based on the model that was developed by Chambers (2014) within good quality forest patches that serve as hotspots of vascular plant diversity?

Hypothesis:

Geodiversity hotspots calculated by the GDI will most likely fall within the boundaries of the good quality native forest patches (vascular plant diversity hotspots), because high geodiversity creates more environmental heterogeneity that leads to a higher complexity within the living world (Gray, 2004), creating a broader available niche space for species to fill (Dufour et al., 2006). Therefore, geodiversity hotspots will likely overlap vascular plant diversity hotspots.

(9)

Methodology

The model that was developed by Chambers (2014) was created in ArcGIS. Various datasets were used as input for the geodiversity map. A grid was created, which was used as an overlay for the different maps. For each cell the variety of the different components was calculated. Based on various models, formula 1.1 was used:

- Formula 1.1: GDI = 2(Gm) + Gs + P + 2(Sr) + Ss

Various components of geodiversity are included:

- Geology: Expressed as variety of geologic mega- and subunits (Gm and Gs). She had two geologic maps at her disposal, of which one contained mega- units (Gm) and the other sub-units (Gs). She emphasized the importance of the mega-sub-units compared to sub-sub-units by weighting it *2 in the formula. This component was included because on a volcanic island geology is inherently linked with geomorphology (Chambers, 2014). Furthermore, the arrangement and nature of rocks are often fundamental in the development of landforms (English Nature, 2004).

- Pedology: Expressed as soil variety (P). The data were derived from a soil map. Soils controls nutrient and water availability and therefore greatly affect plant growth. Furthermore, different soil types support a different variety of plants.

- Topography: Expressed as slope angle variety (Sr and Ss). It controls solar radiation received at the surface and adds complexity to the landscape. The variety is divided into slope angle standard deviation and range, which were derived from a DEM. Slope angle effects plant species richness and plant cover (Nadal-Romero et al., 2014). Furthermore, according to Hjort and Luoto (2010), the range and standard deviation of slope angle were the most correlated land surface parameters to geodiversity. The weighting of *2 is because the range is more important at a coarse scale (Chambers, 2014).

The parameters that are included in this formula are dependent of the available data. Therefore, the formula is fairly flexible and can be easily adjusted or expanded. It is important to note that the formula does not contain certain components of geodiversity that are considered to be drivers of plant diversity in the paper by Parks & Mulligan (2009). These components are climate and hydrology and are not included in formula 1.1 that was used for the model developed by Chambers (2014). Figure 2.1 shows what factors should be considered as fundamental components of geodiversity.

(10)

Fig. 2.1: Relationships between the fundamental components of geodiversity and the measure of geodiversity, biodiversity is considered in terms of tree species richness (Parks & Mulligan, 2009).

The term geodiversity in the paper by Parks and Mulligan (2009) is viewed as a measure of environmental resource availability. They deviate from the generally accepted definition of geodiversity by including climate. To avoid confusion, climate was not included in this particular geodiversity index.

The spatial variety of hydrology, topography and geology provide a measure of complexity within the non-living world (Parks & Mulligan, 2009). The formula that was used by Chambers (2014) does not contain a measure of the spatial variety of hydrology, such as drainage density or an overview of the stream network. These can be derived from a DEM and provide information about stream density and water availability within an area. Including a measure of hydrology resulted in formula 1.2:

- Formula 1.2: GDI = G + P + Sr + Ss + D (‘D’, representing stream density)

Because of the limited attempts at estimating geodiversity and limited available data, the formula has remained as simple as possible. The maps that were used to obtain the parameters were a digitized 1:100,000 soil map from a paper by Parish & Feillafe (1965) for the Mauritius Sugar Industry Research Institute (MSIRI), a digitized 1:200,000 geology map of ORSTOM and MSIRI, a 100 meter resolution DEM and a shapefile containing the outline of Mauritius. Before any of the parameters could be obtained in ArcGIS a grid had to be created with the fishnet tool. Grid resolution was determined by using a method developed by Hengl (2006).

(11)

Based on the minimum legible delineation (MLD), which is the smallest size area that is mapped (0.25 cm² on the map), and the maximum location accuracy (MLA, 0.25mm - 0.1mm on the map), the coarsest and finest grid resolution can be determined. The scale number (SN) refers to the scale of the map that serves as underlay for the grid, which is the scale of the Mauritius outline map in this case (1:400,000). This scale number is then multiplied by the MLD to get the coarsest grid resolution and multiplied by MLA to get the finest grid resolution.

- Formula 2.1: SN * 0.0025 (MLD) for maximum legible resolution (coarsest) - Formula 2.2: SN * 0.00025 (MLA) minimum legible resolution (finest)

The paper also provides a universal rule of thumb to relate scale number to grid resolution: SN * 0.0005. However, the scale of the map that was used as underlay for the grid was 1:400,000. By using this as input for SN in the rule of thumb formula and the formula for the finest grid resolution, a grid overlay is created where the individual grid cells can be barely made out. Therefore, the midpoint between the calculated values of the first two formulas was calculated. Then the midpoint was calculated again between that value and the calculated value of the second formula to get a slightly finer resolution, which was close to the value calculated with the universal rule of thumb. As a result the visibility of the individual grid cells is better. The resulting resolution is 325 * 325 meters.

Geodiversity Index

The grid was used as an overlay for the different maps in ArcGIS. The grid was clipped by using ‘selection by location’ and clipping the grid within the boundaries of the Mauritius outline. In this way, cells that contained parts of the sea were left out. Parameters P and G for each quadrant were calculated by first converting both the soil and geology maps to a raster. Then the variety was calculated with the zonal statistics tool and the results were stored in two different attribute tables. The two tables were then joined with the attribute table of the clipped grid. To obtain Sr and Ss, a slope angle map was derived from the DEM. With the zonal statistics tool an attribute table was created that contained all statistics, including standard deviation and range. These needed to be reclassified in order to normalize the data. Natural breaks classification (Jenks) was used to divide the standard deviation into 6 and range into 7 classes. The extra range class was added because of high outliers. These classes were then reclassified with the field calculator and given a value from 1 till 7. The python scripts are derived from the scripts that were created by Chambers (2014) and can be found in the appendix. These reclassified attributes were then added to the attribute table of the clipped grid. Furthermore, parameter D was derived from the DEM by creating a flow accumulation map. This map was then used as the input for the raster calculator. A threshold value needed to be determined in order to identify what cells are classified as a stream. This is a trial and error process. The resulting stream network should resemble the streams and rivers visible on satellite imagery. With the zonal statistics tool the sum of stream cells within a quadrant was calculated. Because of the visible green belts along the rivers on satellite imagery, a high amount of stream cells within a

quadrant were given a higher value than quadrants with few or no stream cells. This statistic was then too reclassified as sum category with the field calculator. This attribute was also added to the

attribute table of the clipped grid. Finally, an extra attribute named GDI was added to the clipped grid attribute table. With the field calculator, the sum of the soil and geologic variety, the slope angle classes and the stream network sum was calculated for GDI. This resulted in 5 GDI classes (Jenks classification): very low, low, medium, high and very high.

(12)

Native plant diversity

The forest quality map by Page & d’Argent (1997) was georeferenced in ArcGIS. With the editor tool, polygons were drawn around the patches of good quality forest and clipped from the map. Over 50% of the native plant cover is located within these patches (CBD, 2007). Therefore, they served as hotspots of native vascular plant diversity. The outline of the polygons was used to tell whether cells with high geodiversity fell within the boundaries of the patches. The GDI layer was then clipped within the polygons to calculate statistics and the composition of the different geodiversity

components within the polygons. Figure 3.1 shows the GDI map with the black polygons that indicate the patches of good quality forest and serve as hotspots for native vascular plant richness.

(13)

Results

Fig. 3.1: Geodiversity index for Mauritius. Black polygons indicate native vascular plant diversity hotspots.

As can be seen on the map, the hotspots of geodiversity (red quadrants) are primarily concentrated within the Black River Gorges National Park in the south- west and the mountain ranges to the north-west and south-east. Other geodiversity hotspots are concentrated around isolated mountains like Trois Mamelles, Corps de Garde and Le Morne Brabant (fig. 1.1). Furthermore, the stream network

(14)

occurrence of hotspots in the mountain ranges is due to the high variety in slope angle standard deviation and slope angle range. Most of the hotspots have very high values for slope angle standard deviation and range. However, a few hotspots lie outside of the mountainous areas (indicated by blue boxes in figure 3.1). All of the different components of the geodiversity hotspots in the centre blue box have ‘medium’ values, except for soil diversity, which is remarkably higher than in other areas. The occurrence of the hotspot in the lower blue box is due to a very high stream density in this area. The areas that contain geodiversity hotspots seem to be concentrated in and around the black polygons that indicate native vascular plant diversity (figure 3.1). Quadrants with a geodiversity value ranging from 13 to 18 were classified as ‘Very High’ by using Jenks classification and served as geodiversity hotspots. However, only 4.86% of the total geodiversity hotspot area falls within the boundary of the polygons. Moreover, only 10% of the total polygon area is covered with geodiversity hotspots.

The slope angle standard deviation class that was measured most often within the polygons was 3 (figure 3.2), which means medium slope angle variance within a quadrant. The most frequently occurring slope angle range classes were 3 and 4. These are both considered medium ranges.

Furthermore, the dominant soil type that is found within the boundaries of the polygons is lithosol. It is a poorly developed soil that is typically found on steep slopes. The steepness often causes flora to be sparse. This would mean that this area is not very suitable for the development of tropical forests and not very capable of supporting a diverse native flora.

Figure 3.2: Frequency distributions and statistics of standard deviation (top) and range (bottom) of slope angles within areas of high native vascular plant diversity (black polygons in figure 3.1).

The other components of the quadrants that fall within the boundaries of the native vascular plant diversity hotspots (black polygons) have remarkably low values.

(15)

Figure 3.3: Frequency distributions and statistics of soil (top) and geologic (bottom) variety within the areas of high native vascular plant diversity (black polygons in figure 3.1)

The soil and geologic variety are generally very low, with little (2 different types) to no variety (only 1 type).

Figure 3.4: Frequency distributions and statistics of stream density within the areas of high native vascular plant diversity

Most areas with high native vascular plant diversity do not even overlap with areas with high stream density (0 stream density). In the few quadrants that do contain stream cells the stream density is very small (only 1 or 2 stream cells). All these results contradict the hypothesis that the geodiversity hotspots calculated with the GDI based on the model that was developed by Chambers (2014) overlap with native vascular plant diversity hotspots.

(16)

Discussion

Methodological design

The hypothesis that geodiversity hotspots calculated with the GDI will most likely fall within the boundaries of the hotspots of native vascular plant diversity could not be confirmed. The geodiversity hotspots calculated with this model are not located within the boundaries of the remaining patches of good quality native forest. These patches of forest contain over 50% of the islands native plant diversity (CBD, 2007) and are considered hotspots of native vascular plant diversity. This has serious implications for the model. It could mean that the model fails to capture geodiversity, because the most frequently occurring parameter values within the polygons range from low to medium and do not indicate a geodiversity hotspot. However, the location of the polygons on the map could deviate from their actual location. The map that was used to create the polygons was georeferenced in ArcGIS. As can be seen on the geodiversity map, a small correction to the south-east would increase the amount of geodiversity hotspots that fall within the boundaries of the polygons significantly. Based on the available input data, the data seems to capture geodiversity nicely. The map shows what areas contain a high diversity of the different components that were included in the measure for geodiversity. The geodiversity is visibly higher in the mountain ranges to the east, south-west and north-south-west and the isolated mountains that are indicated in figure 3.3. This is because of the high variety of slope angles in these areas, which creates a complex landscape. The geodiversity around streams is also higher than the surrounding area, because of the higher streamline density. It can be argued that the some parameters should not be weighted equally in the formula. Firstly, geology and geomorphology are inherently linked on a volcanic island (Chambers, 2014), which means that the development of landforms like mountains, hills and their slopes is likely to be influenced by geology. Secondly, slope angle range and deviation within an area both affect the development of azonal soils. Putting emphasis on these two parameters might result in a better estimate of geodiversity that can be used to find native plant diversity hotspots. For a better measure of geodiversity that can be used to explain biodiversity patterns as proposed by Parks & Mulligan (2009), the addition of climatic factors like solar radiation and precipitation should be considered. The problem is that this conflicts with the definition by Gray (2004). However, the amount of solar radiation that the earth’s surface receives is affected by slope aspect and it could be included indirectly by adding slope aspect to the formula. Nonetheless, the effects of precipitation on plant diversity should be studied separately.

The fact that the geodiversity hotspots and good quality native forest polygons do not overlap and therefore fail to show a positive correlation between native plant diversity and geodiversity could also imply that the plant diversity hotspots are unrelated to geodiversity.

(17)

Geographic Isolation

The tropical native forests of Mauritius have been rapidly declining since the 16th_century.

Fig. 4: Forest cover loss since 1773 (Page & d’Argent, 1997)

There is little data available on native plant diversity in areas that have been deforested. Therefore, it is hard to tell whether these areas contained more, less or just as much plant diversity as the

remaining patches of good quality native forest. Except for one big patch in the left bottom corner, all patches are located at high altitude. These areas are hard to access for people, and might have been avoided for that reason. The reason that these very diverse forests still remain is because they remained unaffected by humans. These patches may have become geographically isolated at a very early stage which has caused speciation in the different patches. This could be an explanation for the high amount of native plant diversity within the patches. Less than 100 years might seem like a small amount of time for speciation, but the first steps in speciation with some plant species takes about 100 generations or even less (Hendry et al., 2007).

However, geodiversity and native plant diversity do not necessarily have to be unrelated. The remaining good quality native forests might not have been affected by human activities, because of their geographic location. Areas that have already been deforested might have had even higher native vascular plant diversity than the current good quality native forest. Therefore, areas with high

geodiversity that were calculated with the GDI that was based on the model developed by Chambers (2014) might have the potential to support an even greater diversity of native plants. Compared to these areas, the now remaining patches of good quality native forest (figure 3.1) might have had very low or intermediate native vascular plant diversity.

Furthermore, the effects of climatic factors like rainfall and temperature also need to be studied. In the lowlands of Mauritius, the rainfall varies from 890 mm on the leeward side of the island, to 1905 mm on the southeast coast. In the uplands, the rainfall varies from 2540 mm to 4445 mm per year. The rainfall is sufficient to permit the development of tropical moist forest on the windward side of the island and tropical dry forest on the leeward side (Schipper, 2015). This gradient is likely to influence the distribution of some of the native plant diversity.

(18)

Conclusion

The hypothesis that geodiversity hotspots calculated with the GDI will most likely fall within the boundaries of the good quality native forest polygons could not be confirmed because the model has failed to show a positive correlation between native plant diversity and geodiversity. This could have numerous underlying causes. Formula 1.2 (GDI = G + P + Sr + Ss + D) was used to calculate the geodiversity index and has remained rather simple and provided an acceptable measure of geodiversity (Parks & Mulligan 2009; Chambers, 2014). Putting emphasis on different parameters might provide a better estimation of geodiversity. With the addition of slope aspect the model might be able to capture variation in solar radiation, which will provide a better measure of geodiversity. However, the native plant diversity could be unrelated to geodiversity, because of the high degree of deforestation and limited data on native plant diversity. The remaining patches of good quality forest might have been geographically isolated at an early stage, which could have resulted in speciation. The fact that the good quality native forests have been preserved at high altitude may also mean that they were too difficult for humans to access. In the past some areas, that now no longer contain good quality native forest, may have had an even higher diversity of native plants. Certain areas with higher geodiversity could have potentially supported an even higher diversity of native vascular plants than the now remaining areas covered by the remaining good quality native forests.

(19)

References

- Bruneau, P. M. C., Gordon, J. E., Rees, S., (2011), ‘Ecosystem sensitivity and responses to

change: understanding the links between geodiversity and biodiversity at the landscape scale’, JNCC report, 450.

- Burnett, M. R., August P. V., Brown J. H. Jr., Killingbeck K. T., (1998), ‘The influence of

geomorphological heterogeneity on biodiversity’, Conservation Biology, 12,

363- 370.

- CBD (Convention on Biodiversity), (2007), ‘Strategy and Action plan – Mauritius’ , retrieved from http://npcs.govmu.org/English/Documents/Chapter%202-4.pdf on 16-05-2015.

- Chambers, R., (2014), ‘Development of a geodiversity index on a volcanic island: Tenerife,

Canary Islands’ (Unpublished master thesis), University of Amsterdam, Amsterdam, The

Netherlands.

- Cormier, Z., (2012), ‘Follow the rain: Tree biodiversity greatest in areas of high rainfall’, Center for International Forestry Research, retrieved from

http://blog.cifor.org/10887/follow-the-rain-tree-biodiversity-greatest-in-areas-of-high-rainfall#.VVu7ivntmkp on 16-05-2015.

- Dufour A., Gadallah F., Wagner H. H., Guisan A, Buttler A., (2006), ‘Plant species richness and

environmental heterogeneity in a mountain landscape: effects of variability and spatial configuration’, Ecography, 29, 573–584.

- External Relations Team English Nature, (2004), ‘Linking geology and biodiversity’ (Report no. 562), Peterborough, Status Design & Advertising, retrieved from

https://www.cbd.int/doc/pa/tools/Linking%20Geology%20and%20Biodiversity%20(part

%201).pdf retrieved on 16-05-2015.

- Franklin, J., Ripplinger, J., Freid, E., Marcano-Vega, H. and Steadman, D.W., in press, ‘Regional

variation in Caribbean dry forest tree species composition’, Plant Ecology, 216, 873-886.

- Gordon, J., (2014), ‘Linking geodiversity and biodiversity in protected area management:

developing a more integrated approach’, EGU General assembly 2014, Vienna.

- Gray M., (2004), ’Geodiversity: Valuing and Conserving Abiotic Nature’, Chichester, John Wiley & Sons Ltd.

(20)

- Hendry, A. P., Nosil, P., Rieseberg, L. H., (2007), ‘The speed of ecological speciation’,

Functional Ecology, 21, 455-464.

- Hjort J., Heikkinen R. K., Luoto M., (2012), ‘Inclusion of explicit measures of geodiversity

improve biodiversity models in a boreal landscape’, Biodiversity Conservation, 21,

3487–3506.

- IUCN/WWF , 1994, ‘Centres of plant diversity: A guide and strategy for their conservation’,

368, Cambridge, IUCN Publications Unit.

- Nadal-Romero, E., Petrlic, K., Verachtert, E., Bochet, E., Poesen, J., 2014, ‘Effect of slope angle

and aspect on plant cover and species richness in a humid Mediterranean badland’, Earth

Surface, Processes and Landforms, 39, 1705-1716.

- Norder, S., Rijsdijk, K. F.,Seijmonsbergen, H., Van Loon, E., Rughooputh, S., Kamminga, T.,

(2012), ‘Vruchtbare bodems van Mauritius verdwijnen in zee’, Geografie, 21, 34-36.

- Page, W., d’Argent, G., (1997), ‘A vegetation survey of Mauritius to identify priority rainforest

areas for conservation management’, Mauritian Wildlife Foundation.

- Parish, D. H., Feillafé, S. M., (1965), ‘Notes on the 1 : 100,000 soil map of Mauritius’, MSIRI

- Parks, K. E., Mulligan, M., (2009), ‘On the relationship of a resource based measure of

geodiversity and broad scale biodiversity patterns’, Biodiversity Conservation, 19, 2751-2766.

- Pereira D. I., Pereira P., Brilha J. and Santos L. (2013), ‘Geodiversity Assessment of Paran a ́

State (Brazil): An Innovative Approach’, Environmental Management, 1-12.

- Schipper, J., ’Islands of Réunion, Mauritius, East of Madagascar’, WWF website, retrieved from https://www.worldwildlife.org/ecoregions/at0120 on 03-05-2015 on 16-05-2015. - Serrano E., Ruiz-Fla no P. and Arroyo P., (2009), ̃ ‘Geodiversity assessment in a rural landscape:

Tiermes-Caracena area (Soria, Spain)’, Mem.Descr.Carta Geol.d’, 87, 173-180.

- Staub C. G., Stevens F. R., Waylen P. R., (2014), ‘The geography of rainfall in Mauritius:

Modelling the relationship between annual and monthly rainfall and landscape characteristics on a small volcanic island’, Applied Geography, 54, 222-234.

- Strasberg D., Rouget M., Richardson D. M., Baret S., Dupont J., Cowling R. M., (2005), ‘An

assessment of habitat diversity and transformation on la Réunion island (Mascarene islands, Indian ocean) as a basis for identifying broad-scale conservation projects’, Biodiversity and

Conservation, 14, 3015–3032.

(21)

Appendix

1.a – Python script for reclassifying slope angle standard deviation (Ss) def reclass(STD): if (STD < 0.98209851): return 1 elif (STD < 2.33248396): return 2 elif (STD < 3.98977519): return 3 elif (STD < 6.07673452): return 4 elif (STD < 8.9616489): return 5 elif (STD < 15.6521949): return 6 else: return 0 reclass(!STD!)

1.b – Python script for reclassifying slope angle range (Sr) def reclass(RANGE):

if (RANGE < 2.5796359): return 1

elif (RANGE < 6.26483002): return 2

(22)

elif (RANGE < 10.8713227): return 3 elif (RANGE < 15.8463348): return 4 elif (RANGE < 22.1111648): return 5 elif (RANGE < 30.0343322): return 6 elif (RANGE < 46.9862252): return 7 else: return 0 reclass(!RANGE!)

1.c – Python script for reclassifying stream network sum (D) def reclass(SUM):

if (SUM < 1.97647059 and SUM != 0): return 1

elif (SUM < 3.9882351 and SUM != 0): return 2

elif (SUM < 9.1 and SUM != 0): return 3

else: return 0 reclass(!SUM!)

(23)

[fine_soiltable_VARIETY] + [fine_geotable_VARIETY] + [fine_slopetable_STD_CATEGORY] + [fine_slopetable_RANGE_CATEGORY] + [drain_table_SUM_CATEGORY]

(24)

(25)

(26)

Geodiversity of Mauritius Developing a geodiversity index for Mauritius to predict native vascular plant diversity hotspots