Modeling the effects of sea level rise and
anthropogenic activities on landscape connectivity
in the Philippines
Ranès Rioza
10675655
Amsterdam
03 – 07 - 2017
Supervisor:
Kenneth Rijsdijk
Abstract
Since the last glacial maximum (LGM) the global sea level rose 120 meter (Rijsdijk et al.,
2014). This caused a substantial decrease in land mass on the paleo islands of the
Philippines. One of the effects of the rising sea level was the partitioning of the island
‘Paleo Luzon’ in to Luzon and Mindanao. The partitioning and change of coastline
resulted in a decrease in landscape connectivity. In present time, major threats to
landscape connectivity originate in anthropogenic activities, such as land use change and
expanding infrastructure. For conservationists it is interesting to what extend the effects
of anthropogenic activities is similar to the effects of sea level rise since the LGM.
The changes in quality of connectivity are therefore modelled using a circuit theory
model (McRae, & Kavanagh, 2011). The resulting cost weight distance/euclidean
distance ratio (CWD/ED) and a cost weight distance/least cost path ratio (CWD/LCP),
were used as indication for quality in connectivity. After statistical analysis of the results
it appeared that both sea level rise and anthropogenic activities caused a significant
decrease in landscape connectivity in the research area.
The differences in impact between the two factors were compared in percentages,
and differed per island. The effect of anthropogenic activities on Luzon was similar with
the effect that was caused by sea level rise. This similarity may be supportive to a
continuous discussion concerning the selection of priority areas. The effects of
anthropogenic activities on Mindanao however, were at least 3.5 times stronger than sea
level rise. This finding may support the discussion about human pressure on landscape
connectivity.
Introduction
The Philippine archipelago is a region of special interest to biogeographers, concerning the extraordinary biodiversity patterns (Diamond & Gilpin 1983; Brown et al. 2013). This archipelago was according to paleogeographic records connected to the Sunda shelf during the Pleistocene when sea level used to be 100 – 120 meter lower than the current sea level (Diamond & Gilpin 1983; Rijsdijk et al. 2014), thereafter the terrestrial connection disappeared with the rising sea level, which allowed the development of high biodiversity (Diamond & Gilpin 1983). This disconnection that is currently known as the Huxley line, separates the eastern Philippines from Wallacean (Brown et al. 2013).
Sea level rise isolated the Philippine islands from the continents on the one hand and altered the landscapes, island size and shape on the other. The latter controls the landscape
connectivity, which affects the distribution, range and gene flow of land animals. A study to the prolonged process of alterations in the landscape is therefore relevant to multiple social and academic disciplines.
For example a major challenge for biogeographers is finding a relation between phylogenies and paleogeographic change. A phylogeographic approach that rigorously incorporates the geographic component, and connectivity of habitat patches may help bio geographers to develop comprehensive explanations of organismal distributions (Chan, Brown & Yoder 2011). Regarding this topic it has to be mentioned that this research may be relevant to explain the distribution of land vertebrates, but it is focused on the role of landscape change and does not include an assessment of how connectivity affects specific land vertebrates.
Although landscape connectivity is exposed to growing human influence (Correa Ayram et al. 2016) that fragmentizes ecological suitable habitats (Imong et al. 2014), threats are also formed by natural hazards like fires and floods (Matisziw & Murray 2009). Multiple researches have already been conducted to the potential threat of anthropogenic activities to landscape connectivity, and in some cases natural hazards are integrated as well. However only a few researches focused on the comparison between the natural hazard formed by sea level rise and man-made threats (Leonard et al. 2017). A study that compares changes in landscape connectivity as a result of sea level rise and anthropogenic activities may provide new beneficiary insights for conservation processes with the aim to protect mobility and gene flow of endangered species in fragmented landscapes (McRae & Beier 2007; Correa Ayram et al. 2016).
Research Area
During the last 120ky there have been multiple stable sea levels (figure 1) (Rijsdijk et al. 2014). Period D during the last glacial maximum (LGM) and period F, the situation with the present day sea level are of particular interest. These are respectively the lowest recorded sea level within the time frame of the last 120ky, and interesting for the current purposes of conservationists. To investigate these periods the research area is restricted to the islands of Paleo Luzon; a major landmass during the LGM, and it’s two major remnants the known Philippine islands Luzon and Mindanao (figure 2).
The Philippine regions below 600 meter are threatened by land use changes in commercial agriculture and by urban development. The lowland forests in these regions belong to the
Figure 1: Sea level fluctuations during the last 120 thousand years.
Source: Rijsdijk et al., (2014)
Figure 2: Research Areas of Paleo Luzon, Luzon and Mindanao. Regions over 600 meter are no part of the research area.
most threatened habitats in the Philippines (Heaney et al. 2009). Additionally these regions are the only terrestrial connection between the biodiversity hotspots on higher altitudes described in Heaney et al. (2016). In combination with the fact that sea level rise always threats the lower situated areas the first, this study will focus exclusively on regions below 600 meter.
Definition of connectivity
Over the last century scientists already developed multiple definitions for connectivity. The most universal definition is: “Connectivity is defined as the degree to which a landscape facilitates or impedes movement of organisms among resource patches.” (Tischendorf & Fahrig 2000). The authors provide an additional wider explanation of connectivity in which they state that landscape structure must be provided from a species point of view. For that behavioral facets that affect “the degree to which a landscape facilitates or impedes the movement of organisms” have to be introduced. Therefore the final universal definition that is given by Tischendorf & Fahrig (2000) is: “Landscape connectivity encapsulates the combined effects of (1) landscape structure and (2) the species use, ability to move and risk of mortality in the various landscape elements, on the movement rate among habitat patches in the landscape.” This definition is subsequently split in to two coherent definitions; structural connectivity and functional connectivity. Correa Ayram et al., (2016) provides a clear explanation for both definitions. i) “Structural connectivity, corresponding to spatial relationships (continuity and adjacency) between the structural elements of the landscape (e.g. forest patches and continuous land masses).” ii) “Functional connectivity, which refers to landscape features that facilitate or impede the movement of species between habitat patches.” A demand for a proper assessment of connectivity is not to separate structural and functional connectivity (Tischendorf & Fahrig 2000; Goodwin 2003).
The changes of landscape features that affect the structural and functional connectivity for land vertebrates in lowlands and lower upland zones (below 600m), are studied with the application of a circuit theory migration model, available as Circuitscape linkage mapper (McRae & Kavanagh 2011). The model simulates the current of migrating organisms, based on the resistance that landscape features form to land vertebrates.
Research Question
In the discussion of conservation ecology, the effects of anthropogenic activity are a common topic. A comparison between the impact of anthropogenic activity and other influencing parameters will be valuable for conservationists and policy makers in their assessments and decisions. To contribute in the assessment of the impact of anthropogenic activity this research compares the presence of humans in the eastern Philippines with the effect of sea level rise. The resulting research question within this study is: What is the difference between the effect of sea level rise and the effect of Anthropogenic activity, on landscape connectivity in the eastern Philippines?
Methods
To analyze the difference in effects on landscape connectivity between sea level rise and anthropogenic activities both processes were modeled with the same method. The results were thereafter analyzed with a statistical analysis in matlab (appendix B). An overview of the acquisition of the results is displayed in a concept chart (figure 3).
The research areas of Paleo Luzon, Luzon and Mindanao were determined with a digital elevation model (DEM) obtained at marine-geo.org. The polygons were generated with the ArcGIS raster calculator and the raster to polygon tool. Polygons of areas above 600 meters were generated in a similar way and subsequently erased from the polygons of the research area.
With the DEM, a slope raster (appendix B, figure 1) that was required for the resistance raster was generated. The remaining requirements for the resistance raster were extracted from the open street map geofabrik.de. This data contains shape files of major water bodies (appendix A, figure 1), land use and roads (appendix A, figure 2). Raster files were generated from these shape files with the ArcGIS polygon to raster tool. The raster files, together with the slope raster where entered in to the Gnarly resistance mapper (McRae, Shirk & Platt, 2013), that generated a resistance map for each research area (appendix A, figure 3). An additional requirement for the Gnarly resistance mapper is an excel sheet, in which resistance values are indicated. An overview of the unique features and the corresponding resistance from these excel sheets is available in table 1.
The core area’s for Luzon and Mindanao between which the links and Euclidean distances (ED) had to be calculated where obtained from philgis.org, these where published by the
national mapping resource and information authority of the Philippines (NAMRIA). From these shape files the patches with at least 50% canopy cover where selected and intersected with the research areas.
For Paleo Luzon it was assumed that the island was entirely covered with suitable habitat. However the model Circuitscape linkage mapper (McRae, & Kavanagh 2011) requires specific locations to calculate the links. Therefore ArcGIS was used to generate 200 random points dispersed over the island of Paleo Luzon. With 200 random points a widespread distribution of core areas was simulated. A lower number of points would have disregarded parts of the island. A higher number resulted in an extremely long run time of the model, (estimated on more than 90h).
For every research area, a link map was calculated with the use of the Circuitscape linkage mapper. The input files where the core area’s, the Gnarly resistance map and a text file that contains the Euclidean distances between the core areas. The latter was generated with the Euclidean distance calculator from Conefor (Saura & Torné 2009) .
In the attribute table of the links are two important ratios displayed that identify the quality of every link. Firstly, the cost weight distance (CWD) / ED ratio, which indicates the effort to move from source to destination relative to the distance between source and destination. Secondly, the CWD/least cost path (LCP) ratio, which indicates the average resistance that is encountered along the path with the minimum resistance between source and destination. Optimal connectivity is represented by a ratio of 1, when the CWD is equal to the ED or LCP. The higher the ratio, the lower the connectivity (Dutta et al. 2016).
The distributions of the ratios for each island were analyzed with matlab. To conduct a reliable analysis two measures were taken on the raw data. Firstly, the ratio’s of which the denominator was zero or a negative number were considered useless, and therefore left out of the analysis. Secondly, the outliers were removed using the equations 1a and 1b.
highoutlier=3 thQuartile +1.5 × IQR
low outlier=1 st Quartile−1.5× IQR
(equation 1a and 1b, Where IQR is the inter quartile range of the data)To obtain a clear overview of the distributions the values were classified in value ranges of 5 and plotted in histograms (figure 5). The Wilcoxon rank-sum test was used to check whether the distributions were significantly different. The null hypothesis was that the
distributions originated from the same population, the alternative hypothesis was that the distributions came from different islands with different values in the parameters (different populations). A t-test was not appropriate because the sample size of the distributions is different.
GIS Layer Type Resistanc
e Description
Land use/cover Forest 0 Forest or woodland, considered not to contribute in obstructions. Land use/cover Nature reserve 0 Nature reserve, no influence of human activities.
Land use/cover Farm 40 Agricultural areas where crops are grown.
Land use/cover Grass land 40 An area where grass grows, no protection from vegetation. Land use/cover Orchard 40 Agricultural area where fruit crops are grown.
Land use/cover Heath 40 Heath area’s decreased protection from vegetation. Land use/cover Meadow 40 An area for grazing, no protection from vegetation. Land use/cover Scrub 40 Scrubland decreased protection from vegetation.
Land use/cover Park 80 Densely populated area, but sanctuary for small vertebrates. Land use/cover Cemetery 100 Progression of anthropogenic activities.
Land use/cover Industrial 100 Progression of anthropogenic activities. Land use/cover Military 100 Progression of anthropogenic activities. Land use/cover Quarry 100 Progression of anthropogenic activities. Land use/cover Residential 100 Densely populated areas.
Roads Footway 20 Unsuitable for cars.
Roads Path 20 Unspecified path, but unsuitable for cars.
Roads Track 20 Very small road, for agricultural use in forests etc. Often gravel roads. Roads Tertiary road 60 Third degree of national road.
Roads Secondary road 80 Second degree of national road.
Roads Living street 100 Minor street, pedestrians are in priority. Considered to be present only in densely populated areas.
Roads Motorway 100 Motorway/freeway
Roads Pedestrian area 100 Pedestrian only streets, do occur in densely populated areas. Roads Primary road 100 Considered as a crowded road.
Roads Residential road 100 Roads in residential areas.
Roads Service road 100 Very small roads that form access to buildings and parking lots. Roads Trunk 100 Important major road, typically divided.
Roads Unknown/unclassified road 100 Unclassified, considered as highest possible resistance
Waterbodies Wetland 40 Swamp bog or marshland. Permeable for small land vertebrates but definitely no habitat. Waterbodies Reservoir 500 Artificial lakes, physically not permeable for small land vertebrates
Waterbodies Major river 500 Large rivers, physically not permeable for small land vertebrates
Waterbodies Water 500 Unspecified bodies of water. Typically lakes, but can also be larger rivers or harbours
Table 1: Resistance values for every landscape feature. Values were determined based on description provided by Ramm (2017) and the Gnarly resistance mapper user guide (McRae, Shirk & Platt 2013) The resistance value of slope 0 was 0 and then incremented with 1.11 per degree, a slope of 90 degrees had a resistance value of 99.99.
Subsequently the effects of sea level rise and anthropogenic activities were compared. The equation that was used for the comparison calculates the difference in impact (DI; equation 2). The required input arguments are the three mean values of either the CWD/ED ratio or the CWD/LCP ratio. These are called (M1-3), where M1 is the average value of Paleo Luzon, M2 is the
average value of Luzon without anthropogenic activities and M3 is the average
values of Luzon with anthropogenic activities. After calculating the DI for Luzon, M2 and M3 where replaced with the average values of respectively
Mindanao without anthropogenic activities and with anthropogenic activities.
DI=
(
M 1 – M 2
M 1
)
× 100 %
(
M 2−M 3
M 2
)
×100 %
(equation 2: the calculation of difference in impact)
To obtain a clear overview of the distributions the values were classified in value ranges of 5 and plotted in histograms (figure 5). The Wilcoxon rank-sum test was used to check whether the distributions were significantly different. The null hypothesis was that the distributions originated from the same population, the alternative hypothesis was that the distributions came from different islands with different values in the parameters (different populations). A t-test was not appropriate because the sample size of the distributions is different.
Finally the effects of sea level rise and anthropogenic activities were compared. The equation that was used for the comparison calculates the difference in impact (DI). The required input arguments are the three mean values of either the CWD/ED ratio or the CWD/LCP ratio. These are called (M1-3), where M1 is the average value of Paleo Luzon, M2 the average value of
Luzon without anthropogenic activities and M3 the average values of Luzon
with anthropogenic activities.
Results
Paleo Luzon
The LCP’s based on the CWD map in Paleo Luzon (figure 4), do occur in lowland and coastal area’s of the island. Regions that are situated close to the mountain ranges (higher than 600m) are considered to have a relatively high CWD. Between the 200 random generated core points in Paleo Luzon 67 generated links for the CWD/ED ratio, and 272 generated links for the CWD/LCP ratio were suitable for the statistical analysis. The both right skewed distributions (figure 6) of the ratio’s ranged, for CWD/ED, from 1 – 40 and, for CWD/LCP, from 1 – 20. The means of the ratio’s resulted in 13 (CWD/ED) and 6
(CWD/LCP).
Luzon without human
activities
Sea level rise changed the shape of the coastline, which increased in distance. For example, two major bays now dominate the western coast, and created a peninsula (figure 5a). An outstanding result is that the LCP’s on this western peninsula are directed perpendicular on the west coast. Additionally there are relatively high CWD values found in the lower upland zones adjacent to the northwestern cordillera. (figure 5a).
The model generated 202 links with a suitable CWD/ED ratio and 237
Figure 4: Link map of Paleo Luzon with Least Cost Paths and Cost Weight Distance. The cost weight distance (CWD) is the minimum effort to the nearest core area, a cumulative value of distance and resistance.
links with a suitable CWD/LCP ratio. The right skewed distributions (figure 6) ranged, for CWD/ED, from 1 – 130 and for CWD/LCP, from 1 – 70. The means were respectively 35 and 22.
Mindanao without
human activities
The CWD values on the island are relativelyhigh in the lower upland zones. High CWD values are found on the northern coastline and southern peninsula as well. The LCP’s in the western peninsula of Mindanao are predominantly present in the coastal regions (figure 5b).
131 generated links had a suitable CWD/ED
ratio and 123 links had a suitable CWD/LCP ratio. The ratios of the distributions (figure 6) ranged respectively from 1 – 65 and 1 – 35, with an average value of 24 and 12.
Luzon with human
activities
In the current situation on Luzon, where anthropogenic activities are integrated the LCP’s follow straight lines. East and south of the northwestern situated cordillera, in vast
continuous lowland areas, the LCP’s intersect regions with a relatively high CWD value (figure 5c).
203 links with a suitable CWD/ED ratio, and 247 links with CWD/LCP ratio were considered to have valid values for the statistical analysis. The
ratios of the distributions (figure 6) ranged, for the CWD/ED ratio, from 50 -275 and, for the CWD/LCP ratio, from 35 – 130, with average values of 129 and 85.
Mindanao
with
human activities
The southwestern coastline of Mindanao is mainly determined by a large bay. When anthropogenic activities are integrated, a major part of the coastline has a relative high CWD value. The north eastern part of the island and the western peninsula are dominated by relative low CWD values. In the latter the LCP’s are centralized in the inlands (figure 5d).Figure5: Link map of Luzon(a) and Mindanao (b) without
anthropogenic
activities, Luzon(c) and Mindanao(d) with anthropogenic activities. The cost weight distance (CWD) is the minimum effort to the nearest core area, a cumulative value of distance and resistance.
c.
b. a.
50 suitable CWD/ED- and 53 suitable CWD/LCP ratios were generated with the model. The distributions of the ratios (figure 6) ranged respectively from 20 -195 and 15 – 100, with average values of 95 and 56.
Discussion
Decrease in connectivity by sea level rise
The distribution of the two types of values (CWD/ED & CWD/LCP) that determined the quality of the links between core points on Paleo Luzon is for both types right skewed (figure 6). The lower the value, the higher the quality of the link (Dutta et al. 2016). The right skewed distributions, with average values that approach to 1, indicate high quality links on Paleo Luzon.
The distributions of the ratios on Luzon and Mindanao without anthropogenic activity are right skewed as well, however the found average values are higher (figure 6; table 2), indicating that the quality of the links on both islands had decreased compared to the links on Paleo Luzon. A conducted Wilcoxon rank-sum test confirmed that the data of Luzon and Mindanao is in both cases significantly higher compared to the data from Paleo Luzon (all the p-values were < 0.01). This suggests that sea level rise during the past 20k years actually decreased landscape connectivity on Luzon and Mindanao.
This decrease in connectivity can be assigned to a substantial decrease in landmass for both Mindanao and Luzon. The effect of this decrease in landmass is that a LCP will encounter larger distances and more obstacles that increase the CWD. On Luzon for example two large bays, visible in figure 5a, dominate the western coastline. The length of a LCP parallel to the
western coastline is now increased relative to the situation on Paleo Luzon. When distances increase, the chance that areas with high resistance values are encountered increases as well (Dutta et al. 2016). Based on this hypothesis the expectation was that the quality of the links decreases together with the loss of lowland landmass as a result of sea level rise. The statistics of the CWD/ED and CWD/LCP ratios have confirmed this expectation and are therefore considered as helpful in the assessment of difference in impact between sea level rise and anthropogenic activities.
The effect of anthropogenic activities
The range of the CWD/ED and CWD/LCP ratio’s increased substantially (table 2), on average the range increased with 167% relative to the situation where anthropogenic activities are ignored. The wider variation of LCP quality suggests that the LCP’s are forced to cross regions with a high resistance on the one hand, and are incidentally able to avoid regions with high resistance on the other.
The average values of the CWD/ED and CWD/LCP ratio increased relative to the average values of both islands when land use and roads were ignored. This indicates that although not every LCP seems to be affected by the addition of land use and roads, the quality did decrease. To validate this hypothesis a Wilcoxon rank-sum test was conducted, resulting in the rejection
of the null hypothesis that the distributions of the quality ratios came from the same population (all the p-values were < 0.01).
The expected effect of the anthropogenic activities was that the CWD between two core areas increased, since more resistance values contributed to the cumulative value of distance and resistance that together form the CWD. When the numerator of the CWD/ED and CWD/LCP ratios increases, and the denominator value remains the same, the result is a decrease in quality of connectivity.
Since Euclidean distances remained the same, it is explained why the CWD/ED ratio’s did increase. The LCP’s did not always remained the same but found alternative routes. However when alternative routes are found it is likely that these will be longer, and that in turn increases the chance that a higher CWD is encountered (Dutta et al. 2016).
figure 6: The values of the ratios CWD/ED (left) & CWD/LCP (right) for each island and situation. The values are distributed over bins with bin width of 5. This visualization allows for determination, whether distributions are normal or skewed.
Table 2: Means of the Cost Weight Distance (CWD)/Euclidean Distance (ED) ratio distributions and Cost Weight Distance/Least Cost Path (LCP) ratio distributions for each island and period.
Mean of CWD/ED values Range of CWD/ED values Mean of CWD/LCP values Range of CWD/LCP values Paleo Luzon 13 1 - 40 6 1 - 20 Luzon, no humans 35 1 – 130 22 1 – 70 Mindanao, no humans 24 1 – 65 12 1 – 35 Luzon 129 50 – 275 85 35 – 130 Mindanao 95 20 – 195 56 15 - 100
The results of the Wilcoxon rank-sum test can thus be explained by the expectations. Consequently, the found qualities of connectivity are considered to be helpful in assessing the difference in impact between sea level rise and anthropogenic activities.
Difference in impact
The outputs of the model and the subsequent significance tests concretize that both sea level rise and anthropogenic activities had a significant impact on the previous state of connectivity in the Philippine archipelago. In order to assess the differences in impact of both events the results were stored in table 3a-d. The values in the most right column of table 3 (calculated with equation 2) reveal that for both islands the effect of anthropogenic activities was worse than the effect of sea level rise. If the latter was the case the values in the right column would be < 1.
The effect of anthropogenic activities that was modelled for Luzon appeared to be 1.6 times stronger, on average, concerning the CWD/ED ratios, and 1.1 times stronger concerning the CWD/LCP ratios. Contrary to Luzon, anthropogenic activities on Mindanao had a substantially higher impact on habitat connectivity than sea level rise. On average the values of respectively the CWD/ED & CWD/LCP ratio appeared to be 3.5 and 7.3 times larger when anthropogenic activities were involved. This result was considered outstanding since the overall quality of habitat connectivity is worse on Luzon than on Mindanao. However, the results have similarities with a case study on the north American east coast, where complex coastal regions encounter the same effects from sea level rise as from urbanization (Leonard et al. 2017).
Coastal regions are of special interest
The study of (Leonard et al. 2017) helps in explaining why the connectivity on Mindanao was found to be better than on Luzon. The results can be assigned to a difference in loss of landmass. The coastline of Mindanao is practically identical with the coastline of Paleo Luzon, whereas the coastline of Luzon became substantially more complex (figure 2). This aligns with the results of (Leonard et al. 2017), where a vast decrease in connectivity is predicted along the most complex regions of the coastline.
The similarity in these results may be a motivation for comprehensive research to landscape connectivity in geographically complex coastal regions. Whenever a trend can be identified that complex coastal regions are more vulnerable to loss in landscape connectivity due to sea level rise, the results
may be of substantial contribution to conservation programs. Especially because coastal regions are extremely vulnerable to anthropogenic settlement as well. This last argument is supported by the significant increase of CWD in southern Mindanao after addition of roads and land use to the resistance raster.
Improvements
Similar researches to the difference in impact between sea level rise and anthropogenic activities are scarce (Leonard et al. 2017). Therefore there is a lack in comprehensive method descriptions which causes that many improvements can be made concerning the methods of this research. Although the conducted approach of the circuit theory model (McRae & Kavanagh 2011) incorporates both structural and behavioral responses of organisms to landscape features (Imong et al. 2014), most of the improvements can be found on the topic of functional connectivity.
The resistance values that are used for this model are all theoretically determined. The benefits of this approach are that it is an efficient way of assessing changes in landscape connectivity (Imong et al. 2014), and that the results provide an overall view of changes in landscape connectivity. However an empirical determination of the resistance value for each feature that is involved in the resistance raster
3a
Luzon
Mean of CWD/ED
distribution
Difference in
connectivity
Decrease in
connectivity in %
Difference in impact
Paleo Luzon (SLR)
13
22
-169%
1.6
Luzon, no humans (SLR)
35
Luzon, no humans (AA)
35
94
-269 %
Luzon (AA)
129
3b
Mindanao
Paleo Luzon (SLR)
13
11
- 84 %
3.5
Mindanao, no humans (SLR)
24
Mindanao, no humans (AA)
24
71
-295%
Mindanao (AA)
95
3c
Luzon
Mean of CWD/LCP
distribution
Difference in
connectivity
Decrease in
connectivity in %
Difference in impact
Paleo Luzon (SLR)
6
16
-267%
1.1
Luzon, no humans (SLR)
22
Luzon, no humans (AA)
22
63
-286%
Luzon (AA)
85
3d
Mindanao
Paleo Luzon (SLR)
6
6
-50%
7.3
Mindanao, no humans (SLR)
12
Mindanao, no humans (AA)
12
44
-367%
Mindanao (AA)
56
will result in a more complete CWD map (Dutta et al, 2016).
Table 3: From left to right the average values of the ratio distributions (M1-3); the difference in connectivity (M1 -M2) & (M2 – M3); the decrease in connectivity in
Furthermore, is the output of the applied model is always isotropic (a 2 directional link) (McRae et al. 2008). A model that searches for two de directional links will improve the method concerning functional connectivity. Imagine that a connection contains a slope that is downhill, it will then be easier for animals to increase their range as when the animal goes uphill. So the way back for an animal is likely to be another way where the slope is less steep.
Finally the most important feature that is not incorporated in this research is population density. Similar researches, although focused on specific animals, did incorporate human population density (Dutta et al. 2016; Imong et al. 2014; Leonard et al. 2017). The reason why population density is not incorporated in this research is to emphasize the impact of habitat fragmentation due to the activities of humans only, since that will result in an overall judgment about the habitat connectivity. Whenever more complicated features are involved, the rating of habitat connectivity becomes species specific.
Conclusion
The 120 meter sea level rise since the LGM, divided the island Paleo Luzon in to Luzon and Mindanao. This decrease in landmass caused a negative change in landscape connectivity between core areas on both islands. In present time anthropogenic activities (roads and changes in land use) altered the landscape connectivity as well, with a severe negative impact.
The quality of connectivity can be expressed in two ratios: cost weight distance/Euclidean distance and cost weight distance/least cost path. The distributions of the values of these ratios were tested on significant difference with a Wilcoxon rank-sum test; the result was that both, the effect of sea level
rise and anthropogenic activities altered the landscape connectivity significantly.
The difference in impact that both processes had was measured by a comparison of the means of the ratio distributions. Although the effects of anthropogenic activity on Luzon are in fact higher, the results were found to be similar with the effects of sea level rise. The decrease in average CWD/ED ratio was 1.6 times larger and the decrease in the average CWD/LCP ratio was only 1.1 times larger. On Mindanao the effects of anthropogenic activity were found to be much stronger than the impact of sea level rise. The decrease in average CWD/ED value was found to be 3.5 times the effect of sea level rise, and even more outstanding was the decrease in average CWD/LCP value, that was 7.3 times the effect of sea level rise.
Complex coastal regions, that are determined by among others a: wide variation in elevation, peninsulas and oceanic bays, are subject to high impact of sea level rise. Considering the situation on Luzon, the impact is similar to the effect of anthropogenic activities. Combined with the global trend that urbanization occurs especially in coastal regions it can be concluded that landscape connectivity in the complex coastal regions is highly threatened.
Acknowledgement
For the formation of this BSc thesis I want to thank Ms. S. G. A. Flantua MSc from the institute for biodiversity and ecosystem dynamics (IBED), who helped me with the execution and running of the applied models. I also want to thank Mr. dr. K. F. Rijsdijk from IBED, who supervised this research and assisted with suggestions during the writing of this report.
References
1. Brown, R.M., Siler, C.D., Oliveros, C.H., Esselstyn, J.A., Diesmos, A.C., Hosner, P.A., Linkem, C.W., Barley, A.J., Oaks, J.R., Sanguila, M.B. and Welton, L.J., 2013. Evolutionary processes of diversification in a model island archipelago. Annual Review of Ecology, Evolution, and Systematics, 44, pp.411-435
2. Chan, L.M., Brown, J.L. and Yoder, A.D., 2011. Integrating statistical genetic and geospatial methods brings new power to phylogeography. Molecular phylogenetics and evolution, 59(2), pp.523-537.
3. Correa Ayram, C.A., Mendoza, M.E., Etter, A. and Salicrup, D.R.P., 2016. Habitat connectivity in biodiversity conservation: a review of recent studies and applications. Progress in Physical Geography, 40(1), pp.7-37.
4. Diamond, J.M. and Gilpin, M.E., 1983. Biogeographic umbilici and the origin of the Philippine avifauna. Oikos, pp.307-321.
5. Dutta, T., Sharma, S., McRae, B.H., Roy, P.S. and DeFries, R., 2015. Connecting the dots: mapping habitat connectivity for tigers in central India. Reg Environ Change. doi, 10, p.1007.
6. Geofabrik, 2017. Philippines latest free shapefiles [online] retrieved from: http://download.geofabrik.de/asia/philippines.html [18 may 2017].
7. Goodwin, B.J., 2003. Is landscape connectivity a dependent or independent variable?. Landscape ecology, 18(7), pp.687-699.
8. Heaney, L.R., Balete, D.S., Duya, M.R.M., Duya, M.V., Jansa, S.A., Steppan, S.J. and Rickart, E.A., 2016. Doubling diversity: a cautionary tale of previously unsuspected mammalian diversity on a tropical oceanic island. Frontiers of Biogeography, 8(2).
9. Heaney, L.R., Balete, D.S., Rickart, E.A., Veluz, M.J. and Jansa, S.A., 2009. Chapter 7. A new genus and species of small ‘tree-mouse’(Rodentia, Muridae) related to the Philippine giant cloud rats. Bulletin of the American Museum of Natural History, 331(1), pp.205-229.
10.Imong, I., Robbins, M.M., Mundry, R., Bergl, R. and Kühl, H.S., 2014. Informing conservation management about structural versus functional connectivity: A case‐study of Cross River gorillas. American journal of primatology, 76(10), pp.978-988.
11.Leonard, P.B., Sutherland, R.W., Baldwin, R.F., Fedak, D.A., Carnes, R.G. and Montgomery, A.P., 2017. Landscape connectivity losses due to sea level rise and land use change. Animal Conservation, 20(1), pp.80-90.
12.Marine Geo Data System, 2017. Digital Elevation Model, Philippine archipelago [online] retrievred from: http://www.marine-geo.org/tools/GMRTMapTool/ [18 may 2017].
13.Matisziw, T.C. and Murray, A.T., 2009. Connectivity change in habitat networks. Landscape Ecology, 24(1), pp.89-100.
14.McRae, B.H. and Beier, P., 2007. Circuit theory predicts gene flow in plant and animal populations. Proceedings of the National Academy of Sciences, 104(50), pp.19885-19890.
15.McRae, B.H. and D.M. Kavanagh. 2011. Linkage Mapper Connectivity Analysis Software. The Nature Conservancy, Seattle WA. Available at: http://www.circuitscape.org/linkagemapper.
16.McRae, B.H., A.J. Shirk, and J.T. Platt. 2013. Gnarly Landscape Utilities: Resistance and Habitat Calculator User Guide. The Nature Conservancy, Fort
Collins, CO. Available at:
http://www.circuitscape.org/gnarly-landscape-utilities
17.McRae, B.H., Dickson, B.G., Keitt, T.H. and Shah, V.B., 2008. Using circuit theory to model connectivity in ecology, evolution, and conservation. Ecology, 89(10), pp.2712-2724.
18.NAMRIA, 2010. Land cover 2010 [online] retrieved from: http://philgis.org/general-country-datasets/geology-philippines [18 may 2017].
19.Ramm, F., 2017. Open street map data in layered GIS format [online] retrieved from: http://download.geofabrik.de/osm-data-in-gis-formats-free.pdf [18 may 2017]
20.Rijsdijk, K.F., Hengl, T., Norder, S.J., Otto, R., Emerson, B.C., Ávila, S.P., López, H., Loon, E.E., Tjørve, E. and Fernández‐Palacios, J.M., 2014. Quantifying surface‐ area changes of volcanic islands driven by Pleistocene sea‐level cycles: biogeographical implications for the Macaronesian archipelagos. Journal of Biogeography, 41(7), pp.1242-1254.
21.Saura, S. and Torne, J., 2009. Conefor Sensinode 2.2: a software package for quantifying the importance of habitat patches for landscape connectivity. Environmental modelling & software, 24(1), pp.135-139.
22.Tischendorf, L. and Fahrig, L., 2000. On the usage and measurement of landscape connectivity. Oikos, 90(1), pp.7-19.
Appendix A
Residual maps, as background information for the conducted methods.
d
b
a
b
c
d
Figure 2: The land use/cover and roads GIS layers. a: Land use on Luzon, b: land use on
Mindanao, c: roads on Luzon, d: roads on Mindanao.These charts reveal that information about
land use and cover is less comprehensive than the information about roads. The resistance
rasters are therefore more affected by the appearance of roads.
Appendix B
Matlab script that was used for the statistical analysis of the CWD/ED and CWD/LCP ratios.
% Matlab script to conduct statistical tests on landscape connectivity in % the Phillipine archipelago
% BSc thesis Ranes Rioza 10675655
clear close all clc
% 10 datasets where generated for analysis
% The research area's are Paleo_Luzon, Luzon without anthropogenic % activity, Mindanao without anthropogenic activity, Luzon and Mindanao % Hence there are 5 scenarios in total:
% Paleo_Luzon, LuzonNH, MindanaoNH, Luzon, Mindanao (where NH == No Humans) % 2 datasets are available for each scenario:
% The CWD/ED dataset and the CWD/LCP dataset % load the 5 CWD/ED datasets
% load the dataset LuzonNH; MindanaoNH; Paleo_Luzon
PalLuz = xlsread('Paleo_Luzon_CWD_ED.xlsx'); LuzonNH = xlsread('LuzonNH_CWD_ED.xlsx'); MindaNH = xlsread('MindanaoNH_CWD_ED.xlsx'); Luzon = xlsread('Luzon_CWD_ED.xlsx'); Minda = xlsread('Mindanao_CWD_ED.xlsx');
% The excell sheets that contain the data have information stored in three % collumns, the first two contain information about the two core areas that % that are connected by the link that is described by the CWD/LCP ratio, % that data is availble for further analysis but not required for the % analysis about the overall landscape connectivity. Therefore we select % only the 3th collumn in which the values of the ratio are stored.
PAL = PalLuz(:,3)'; LNH = LuzonNH(:,3)'; MNH = MindaNH(:,3)'; MIN = Minda(:,3)'; LUZ = Luzon(:,3)';
% After analysis of the data it becomes clear that a lot of outliers are % present. We use the function 'findoutliers' to obtain an array with % with only outliers and one without outliers.
a
c
e
Figure 3: Resistance map created with Gnarly Resistance Mappera: Paleo Luzon, b: Luzon
without anthropogenic activities, c: Mindanao without anthropogenic activities, d: Luzon, e:
Mindanao
% see function at end of script
[outliers_MNH, not_outliers_MNH] = findoutliers(MNH); [outliers_PAL, not_outliers_PAL] = findoutliers(PAL); [outliers_LNH, not_outliers_LNH] = findoutliers(LNH); [outliers_MIN, not_outliers_MIN] = findoutliers(MIN); [outliers_LUZ, not_outliers_LUZ] = findoutliers(LUZ);
% The array with outliers is in this analysis not of any further interest % We continue with the arra not_outliers_...
% Find the average values of the array's not_outliers_...
meanLNH = mean(not_outliers_LNH); meanMNH = mean(not_outliers_MNH); meanPAL = mean(not_outliers_PAL); meanMIN = mean(not_outliers_MIN); meanLUZ = mean(not_outliers_LUZ);
% A clear overview of the data can be given when generated a histogram of % each distribution. The binwidth of the histogram is chosen on 5 to % provide the best overview.
% the histograms are plotted in the left collumn of the figure. The right % collumn will be availble for he CWD/LCP ratio
% In the histograms a red line is plotted that represents the mean of the % distribution
figure;
% subplot for Paleo_Luzon
abc = subplot(5,2,1);
histogram(not_outliers_PAL,'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([0 40 0 30]) hold on;
line([meanPAL, meanPAL], ylim, 'LineWidth', 2, 'Color', 'r'); title('Paleo Luzon', 'FontSize', 15);
xticks([0:5:40])
% subplot for LuzonNH
a = subplot(5,2,3);
histogram(not_outliers_LNH, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([0 130 0 30]) hold on;
line([meanLNH, meanLNH], ylim, 'LineWidth', 2, 'Color', 'r'); title('Luzon No Humans','FontSize', 15 );
% subplot for MindanaoNH
ab = subplot(5,2,5);
histogram(not_outliers_MNH, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([0 65 0 30]) hold on;
line([meanMNH, meanMNH], ylim, 'LineWidth', 2, 'Color', 'r'); title('Mindanao No Humans','FontSize', 15 );
xticks(0:10:65)
% Subplot for Luzon
a = subplot(5,2,7);
histogram(not_outliers_LUZ, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([50 275 0 30]) hold on;
line([meanLUZ, meanLUZ], ylim, 'LineWidth', 2, 'Color', 'r'); title('Luzon','FontSize', 15 );
xticks(50:30:270)
% subplot for Mindanao
ab = subplot(5,2,9);
histogram(not_outliers_MIN, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([35 195 0 10]) hold on;
line([meanMIN, meanMIN], ylim, 'LineWidth', 2, 'Color', 'r'); title('Mindanao','FontSize', 15 );
xticks(35:25:195)
% conduct wilcoxon ranksum test to check whether the distributions are % significantly different
% the conducted comparisons are: % Paleo_Luzon - LuzonNH
% Paleo_Luzon - MindanaoNH % LuzonNH - Luzon % Mindanao - MindanaoNH
% The function provides a P-value with a significance level of 5% and a H % value. When H = 0 the null hypothesis that the distributions are from the % same population can not be rejected, when H = 1 the null hypothesis is % rejected.
[P_PAL_LNH, H_PAL_LNH] = ranksum(not_outliers_PAL, not_outliers_LNH) [P_PAL_MIN, H_PAL_MIN] = ranksum(not_outliers_PAL, not_outliers_MIN)
[P_LNH_LUZ, H_LNH_LUZ] = ranksum(not_outliers_LNH, not_outliers_LUZ) [P_MNH_MIN, H_MNH_MIN] = ranksum(not_outliers_MNH, not_outliers_MIN)
% To keep a clear overview of the data, the mean vallues are printed one % more time. meanPAL meanLNH meanMNH meanLUZ meanMIN
% The entire process is repeated one more time for the CWD/LCP ratio % the commands remain the same, except for the histograms that are now % plotted in the right collumn of the same figure
PalLuz = xlsread('Paleo_Luzon_CWD_LCP.xlsx'); LuzonNH = xlsread('LuzonNH_CWD_LCP.xlsx'); MindaNH = xlsread('MindanaoNH_CWD_LCP.xlsx'); Luzon = xlsread('Luzon_CWD_LCP.xlsx'); Minda = xlsread('Mindanao_CWD_LCP.xlsx'); PAL = PalLuz(:,3)'; LNH = LuzonNH(:,3)'; MNH = MindaNH(:,3)'; MIN = Minda(:,3)'; LUZ = Luzon(:,3)';
[outliers_MNH, not_outliers_MNH] = findoutliers(MNH); [outliers_PAL, not_outliers_PAL] = findoutliers(PAL); [outliers_LNH, not_outliers_LNH] = findoutliers(LNH); [outliers_MIN, not_outliers_MIN] = findoutliers(MIN); [outliers_LUZ, not_outliers_LUZ] = findoutliers(LUZ); meanLNH = mean(not_outliers_LNH);
meanMNH = mean(not_outliers_MNH); meanPAL = mean(not_outliers_PAL); meanMIN = mean(not_outliers_MIN); meanLUZ = mean(not_outliers_LUZ);
% subplot for Paleo_Luzon
abc = subplot(5,2,2);
histogram(not_outliers_PAL,'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
hold on;
line([meanPAL, meanPAL], ylim, 'LineWidth', 2, 'Color', 'r'); title('Paleo Luzon', 'FontSize', 15);
xticks(0:5:20) yticks(0:40:130)
% subplot for LuzonNH
a = subplot(5,2,4);
histogram(not_outliers_LNH, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([0 70 0 50]) hold on;
line([meanLNH, meanLNH], ylim, 'LineWidth', 2, 'Color', 'r'); title('Luzon No Humans','FontSize', 15 );
xticks(0:10:70) yticks(0:25:50)
% subplot for MindanaoNH
ab = subplot(5,2,6);
histogram(not_outliers_MNH, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([0 35 0 50]) hold on;
line([meanMNH, meanMNH], ylim, 'LineWidth', 2, 'Color', 'r'); title('Mindanao No Humans','FontSize', 15 );
xticks(0:5:35) yticks(0:25:50)
% Subplot for Luzon
a = subplot(5,2,8);
histogram(not_outliers_LUZ, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([40 105 0 75]) hold on;
line([meanLUZ, meanLUZ], ylim, 'LineWidth', 2, 'Color', 'r'); title('Luzon','FontSize', 15 );
xticks(40:10:105) yticks(0:25:75)
% subplot for Mindanao
ab = subplot(5,2,10);
histogram(not_outliers_MIN, 'BinWidth', 5, 'FaceColor', 'none'); set(gca,'FontSize',12)
axis([15 100 0 10]) hold on;
line([meanMIN, meanMIN], ylim, 'LineWidth', 2, 'Color', 'r'); title('Mindanao','FontSize', 15 );
xticks(15:15:100) yticks(0:5:10)
[P_PAL_LNH, H_PAL_LNH] = ranksum(not_outliers_PAL, not_outliers_LNH) [P_PAL_MIN, H_PAL_MIN] = ranksum(not_outliers_PAL, not_outliers_MIN) [P_LNH_LUZ, H_LNH_LUZ] = ranksum(not_outliers_LNH, not_outliers_LUZ) [P_MNH_MIN, H_MNH_MIN] = ranksum(not_outliers_MNH, not_outliers_MIN) meanPAL
meanLNH meanMNH meanLUZ meanMIN
% the function that is used to find outliers
function [X1, X2] = findoutliers(X) IQR = iqr(X); Q1 = prctile(X,25); Q3 = prctile(X,75); High = Q3 + (1.5 * IQR); Low = Q1 - (1.5 * IQR); LCE = length(X); X1 = []; X2 = []; for i=1:1:LCE if X(i) > High X1(end+1)= X(i); end if X(i) < Low X1(end+1)= X(i); end
if High > X(i) && X(i) > Low X2(end+1) = X(i);
end end end
