Home Articles A landslide potentiality mapping on Mauritius Island

A landslide potentiality mapping on Mauritius Island

Department of Physics
Faculty of Science
University of Mauritius
Email: [email protected]
Tel: + 230 9135916
Fax: + 230 4662012

Soonil D.D.V. Rughooputh
Professor of Physics
University of Mauritius

* Corresponding Author

A landslide potentiality assessment method that is based on a factorial indexing system has been created and tested for Mauritius Island. The assessment aims at finding lands that can potentially slide by making use of an equation that has as input five parameters decisive for potential landslide namely; rainfall amount, terrain slope angle, permeability of geology strata, soil and land cover type. The equation output are indexes ranging from 1 to 10 such that 1 is associated with the lowest risk and 10 with the highest risk of potential landslide. A region that has not been assigned an index have no risk of land sliding there even though that region might be vulnerable to sliding lands from uphill. Results show that the method succeeds in mapping land that can potentiality slide and in evaluating the corresponding regions that are at risk from the latter. Keywords: Potentiality mapping, factorial indexing, Landslide, GIS, Mauritius

1. Introduction
Mountain and foothill slopes occupy a significant percentage of the surface relief of Mauritius Island and according to Saddul (1995), the landslide prone areas are those that have terrain slope angle exceeding 20°, where specifically areas with slope = 30° are definitely prone to landslide while areas with slope in the range 20-30° are prone depending on the soil and land cover type. Mauritius has a total surface area of 1868 km2 and regions with slope = 20° make up 108.83 km2, which is 5.83 % of the Island’s total surface area (Figure 1).

Figure 1 – Terrain slope angle in degrees.

Although the landslide prone areas are known, what is unknown so far is the potentiality of land sliding for these prone areas. The objective of this work is to map the landslide potentiality for such landslide prone areas. An initial methodology based on slope and rainfall amount was devised to create a preliminary landslide risk map for Mauritius (Rughooputh et al., 2008). The methodology used in work is based on the work accomplished so far by Saddul (1995). Specifically the latter has concluded that in addition to slope = 20°, the factors crucial for landslide are: when the soil layer is weathered, unconsolidated, clayley, tuffaceous and consists of colluvial deposits; when the geology strata is hard, compact and consolidated; when rainfall amount received is high and/or frequent; when the land cover is barren and is not protecting the soil and finally; when land use practices, such as, construction, roads traffic vibration and agricultural practices have disturbed the soil.

Thus, by taking into consideration the latter information, a Landslide Potentiality Index (LPI) map has been created based on a factorial indexing equation with five geospatial input parameters: rainfall amount, terrain slope angle, permeability of geology strata, soil & land cover type. The equation was run in a GIS, which enable the handling, manipulation, analysis and combination of the input geospatial data. The LPI equation output is an index ranging from 1 to 10 such that 1 is associated with the lowest risk and 10 with the highest risk of landslide. A region that has not been assigned an index has no risk of land sliding. As such, the indices on the LPI map show lands that can potentially slide and do not specifically show all lands that are at risk to sliding lands from uphill.

2. Study area description

2.1 Location, geology and geomorphology
Mainland Mauritius is situated southwest in the Indian Ocean between latitudes 19° 58.8′ and 20° 31.7′ S and longitudes 57° 18.0′ and 57° 46.5′ E, approximately 850 Km east of Madagascar (see location plan in figure 2). The Island has an elliptical shape with its major axis about 63 Km long, its minor axis about 43 Km and its surface area 1868 Km2.

Figure 2 – Upper right inset shows location map of Mauritius in the Indian Ocean. Main figure shows a DEM hillshade

Various authors (e.g., Willaime, 1984; Saddul, 1995) have reported that the island is entirely of volcanic origin with a four stage geochronology consisting of: The Breccia Series (10 – 7.8 Million Years (MY) ago); The Old Lava Series (7.6 – 5 MY ago); The Early Lava Series (3.5 – 1.7 MY ago); and The Recent Lava Series with Intermediate Lavas (0.7 – 0.5 MY ago) and Late Lavas (0.1 – 0.025 MY ago). The Breccia Series caused the emergence of the island but were later entirely covered with the Old Lava Series. The latter gave rise to the mountain ranges and at present Old Lavas cover 21% of the island. Thereafter, Early Lavas Series consisting of compact olivine basalts were emitted and at present they cover 4% of the island. After an erosional interval of 1 MY, Recent Lava Series were emitted and shaped the island. The latter series started with compact Intermediate Lavas, which make up 35% of the island, followed by Late Lavas which occupy 40% of the island. Late Lavas are different from the Intermediate Lavas in that they are highly vesiculated and are characterised by many rocky areas with an almost complete absence of surface drainage. The only materials that are not of volcanic origin are coral reefs, sandy beaches, and some consolidated coral and shell debris in isolated remnant raised beaches (Saddul, 1995).

2.2 Climate, soils, surface drainage and land use
The climate of mainland Mauritius is of the humid tropical type and is under the influence of atmospheric circulations and static factors such as altitude, exposure to the South East Trade Winds and distance from the sea (Willaime, 1984). In the summer period from November to May, tropical cyclones are the most important climatologic features and during the winter period, from April to October, the South East Trade Winds predominates accompanied with few rainfall amounts that are associated with frontal systems. Long-term (30 years) mean annual rainfall depth varies from 1500 mm on the eastern coast to 4000 mm in the central uplands and 800 mm on the western coast. This spatial variation is attributed to orographic effects which are caused by an eastern mountain range and the ‘ridge’ of the central uplands. Inter-annual variation in rainfall amount depends on the passage of cyclones which can multiply the “normal” monthly rainfall depth by 2 to 3 (Willaime, 1984; Padya, 1989). Concerning intra-annual variability, on average 80% of mean annual rainfall is received during the summer period. February is the wettest month and has the highest probability of cyclone formation and October is the driest month of the year. There is also a strong spatial intra-annual variation in rainfall especially in the western and northern regions where very little rainfall is received in winter. The rainy regions in the central uplands shows, however, a less pronounced intra-annual variability and where, even in winter, there is relatively high rainfall recorded (compared to the northern and western regions).

Halais and Davy (1969) used the formula of Thornthwaite of 1948 to map the local climate and found that the super-humid class A occupies 46% of the total area; the humid classes B occupies 35%, the sub-humid class C occupies 19%; and the semi-arid class D occurs in only one small area in the west and is of negligible area.

Parish and Feillafé (1965) mapped three soils order namely; Zonal, Intrazonal and Azonal soils. Zonal soils are matured latosols that have permitted the fullest expression of climate and vegetation as soil forming factors such that there are almost no undecomposed minerals left in the soil complex. These soils make up approximately 33% of island’s surface area. Intrazonal soils are developing under conditions where the effects of climate and vegetation are masked by local factors of environment such as relief, drainage and composition of parent material. The Intrazonal soils have materials still in the process of weathering and they occupy about 47% of the island’s surface. The Azonal soil groups are located mostly on Old Lavas and have little or no profile development and they are: eroded rough broken land of mountains and gorges; soils from recent alluvium; and coral deposits or soils of unconsolidated deposits other than alluvium (Parrish and Feillafé, 1965).

Surface drainage is essentially radial from the central tableland. Most rivers are deeply incised with steep gradients and consequently with many cascades and waterfalls. Torrential flows with severe bank erosions are common during the storms and cyclones of the rainy seasons. The two biggest river basins have sizes of 166 and 116 Km2 (Arlidge and Wong You Cheong, 1975). Most of the island’s indigenous vegetation has been removed to make room for an extensive sugarcane cultivation which to-date occupies 55% of the island and 98% of its cultivated lands. The other cultivations are tea, vegetables and fruits plantations. Only about 4% of indigenous vegetation remains. Scrubs cover about 11% of the island, forest occupies 27% (including indigenous vegetation), built-up areas occupy 6% and the rest are water bodies, wetlands and sandy beaches (Ministry of Agriculture & Natural Resources, 1999; Venkatasamy, 1991).

3. Materials and methods

3.1 The factorial indexing system
The factorial indexing system takes the form of equation 1 below and has as input slope, rain, geology, soil and land cover and produces Landslide Potentiality Index as output.

The factorial indexing system works in a way such that, a higher index is given to a class that has a higher potentiality and a lower index is given to a class that has a lower potentiality of landsliding. Each indexed input map is then multiplied with a factor weight. Applying a factor weight has the effect of giving more importance to certain maps than others. For example, if slope is the most determinant factor for LPI followed by soil, then slope is given say 40% weight, compared to a lower value given to soil (say 15% weight). Thus, the slope map will be multiplied by 0.4 (40%) prior to addition. The values chosen for factor weights are such that their sum is one (1). The minimum and maximum indexes are set to 1 and 10 respectively. These two conditions allow the creation of an LPI map having indexes ranging from 1 to 10, which enable an easy interpretation.

3.2 Factor weights and indexes
Slope being the most dominant factor is given a high factor weight, namely 40%. The indexes assigned to slope classes are shown in table 1.

Table 1: Indexes for slope classes


SlopeClass (degress) Landslide vulnerability Index
> 30 ° Very high 10
20 – 30 ° High 8
0 – 20 ° No Risk No Data

A value of NoData is thus given to all slopes less than 20° such that these regions are given no risk when running equation 1. According to Saddul (1995) the usual mechanism that triggers landslide is water that run between an unconsolidated soil layer and an impermeable geology strata. As such, the indexes for rainfall (table 2) are increased with an increase in amount of rainfall received. The factor weight assigned is 0.15.

Table 2: Indexes for rainfall

Rainfall (mm) Index
400 – 500 10
200 – 400 9
125 – 200 8

According to Giorgi et al. (1998), the impermeability of the geology strata increases with an increase in the age of the strata. The indexes chosen for geology (table 3) are thus increased with an increase in the age of the geology strata and the factor weight assigned is 0.15.

Table 3: Geology strata and indexes assigned

Geological Classes (Chronological order) Index
Old Lavas_10 10
Early Lavas_9 10
Intermediate Lavas_8 6
Intermediate Lavas_7 5
Intermediate Lavas_6 4
Late Lavas_5 3
Late Lavas_4 3
Late Lavas_3 2
Late Lavas_2 2
Sandy Coasts_1 1

Clayey and tuffaceous soils swell during rainfall (Saddul, 1995) and then slide downhill. By considering the percentage of clay content and infiltration rate from Balaghee (2001) and Kremer (2000) respectively, the indexes applied to each of the soil group are as shown in table 4 and the factor weight assigned is 0.15.

Table 4: Indexes assigned for soil groups

Soil Group Index
Low Humic Latosols, L 7
Humic Latosols, H 7
Humic Ferruginous Latosols, F 6
Lathosolic Red Prairie Soils, P 3
Lathosolic Brown Forest Soils, B 7
Dark Magnesium Clay, M 10
Ground Water Laterite, W 5
Grey Hydromorphic Soils, D 9
Low Humic Gleys, G 5
Regosols, C 1
Lithosols, T 3
Mountain Slope Complexes, S 9
Alluvial Soils, A 5

Landslide potentiality is increased when the land cover is barren and is not protecting the soil or when land use practices, such as, urban/built-up, roads traffic-vibration and agricultural practices have disturbed the soil. As such, the indexes applied to each of the land cover class are as shown in table 5 and the factor weight assigned is 0.15.

Table 5: Indexes assigned for land cover classes

Land Cover Class Index
Urban/Built-Up/Settlements 10
Marshy /Wetlands 1
Lakes, Reservoirs 1
Agricultural fields 6
Barren Lands, Shrubs 5
Forested areas 2
Sandy areas 1

3.3 Sources of input factor maps
The slope map used for the study (Figure 1) was derived from a 75X75 m Digital Elevation Model (DEM), the latter that was produced by interpolating Ordnance Survey’s (1991) 1:25,000 10 m contours using Kriging interpolation algorithm of ArcGIS software. The geology map (Figure 3) has been produced by digitizing the 1:200,000 geology map of ORSTOM and MSIRI (1983b).

Figure 3 Geology strata in chronological order

The rainfall map used is for the month of February (Figure 4) and has been created using 194 station data points averaged over the period 1961-1990 and interpolated using the kriging method.

Figure 4 Rainfall map for February

The soil map (Figure 5) has been produced by digitizing the 1:100,000 soil map of DOS and MSIRI (1962).

Figure 5 Soil map showing main soil group

The land cover map (Figure 6) has been produced by digitizing the 1991’s Ordnance Survey 1:25,000 map series of the island.

Figure 6 Land cover map

3.4 Processing of the inputs and running the equation
After the paper maps were digitized into vector maps, the classes of each digitized map were given their indexes and the new indexed maps were converted into raster maps to be used for map algebra (of equation 1) in ArcGIS software. During conversion of the vector maps into raster maps, a cell size is chosen based on the scale of the original paper map. For example a 1:100,000 map is given a cell size of 100X100 during rasterization. Maps already in raster format (such as slope and rainfall) are simply reclassified to their corresponding indices, for example the slope class 20-30° is directly converted to its index of nine (9). After all the rasters have been produced, equation 1 is run using the map algebra function of the ArcGIS software (version 9.2).

4. Results and discussion
Figure 7 shows the LPI map produced. Most values of landslide potentiality are above 6.2. Only 6% (109 km2) of the total land has a potentiality of undergoing landsliding and/or soil creep. 0.54 km2 of settlements are found directly on these landslide prone areas and significant portion of settlements are also found on the path of sliding lands. In addition to that, some landslide prone areas are found close to dams which in turn are found close to settlements. In case of landsliding, overtopping of these dams can be disastrous for these nearby settlements. Finally another significant portion of land transportation networks are found on the path of sliding land or even found directly on the landslide prone areas.

Figure 7 LPI map

Originating from the work of Saddul (1995), the landslide prone areas that was defined has been successfully converted into a map of the actual potentiality of these areas to slides and this by taking into consideration the determinant factors for landslide. Now it is possible to make a concrete evaluation of the regions that can potentially slide and also what are the regions downhill that are at risk to sliding lands. What are needed now are quantitative statements like: how much, for how long and when will landslide occur. Measurement of these quantitative values require field surveys together with physical models and the LPI result directly provides the structure for planning these field surveys as well as guiding in the choice of the most appropriate physical models that can be used.

For the interpretation of the results, the indices are relative categories and would be interpreted as such. The indices can also be grouped into descriptive classes such as 5 – 6 (moderate), 7 – 8 (high) etc… The resulting LPI map is in raster format with a cell size of 25X25 m, the minimum of the input raster cell sizes (land cover map).

The more accurate the data inputs are, the more accurate will be the final output. A geology map at a scale of 1:50,000 (Burgeap-Geolab et al., 1999) can be used to substitute for the geology map of 1:200,000 (ORSTOM and MSIRI, 1983b). A soil map at a scale of 1:50,000 (ORSTOM and MSIRI, 1983a) currently exist and can be used to substitute for the soil map of 1:100,000 (DOS and MSIRI, 1962). But before, these maps need be put in the desired GIS format, which was not done in this study due to time and man-power constraints.

Moreover, having larger scales of contours map than the one currently used to produce the slope map would have been a tremendous advantage. According to Nathire (2003), large scale 2 m digital contour maps at 1:2,500 are being produced at the Cartographic Section of The Ministry of Housing and Lands, Mauritius. Actually only a small part of the island has been covered, and when the work will be completed, then finer resolution slope map (such as 5X5 m) can be produced using these 2 m contours. Good resolution DEM (thus slope map) can be produced using radar interferometry and stereo-photogrammetry techniques while land cover map can be more accurately produced using methods of classification for the most recent optical and/or radar polarimateric remote sensing images (Jensen, 1996; Lillesand and Kieffer, 2000; Bamler and Hartl, 1998; Cloude and Papathanassiou, 1998).

5. Conclusion
LPI is a method for computing the potentiality of a piece of land to slide. It has the form of an equation that is based on a factorial indexing system and which relates landslide potentiality index to slope, rain, geology, soil and land cover. The tool required to run the equation is GIS that have map algebra (raster) capabilities. The output Landslide Potentiality Index map indicate areas that can potentially slide through interpretation of the indices represented on the map. The lowest and highest indexes (1 & 10) correspond to the lowest and highest potentiality of land to slide and lands with no indexes have no risk of sliding even though such lands can still be vulnerable to sliding lands from uphill. An LPI map has been successfully produced for Mauritius and it is enrichment for the already defined landslide prone areas which are those lands that have slope = 20°. The produced LPI map as is, does not explicitly define regions that are vulnerable to landsliding. Indeed, lands that have no potentiality of sliding are still at risk from uphill sliding lands. But with the LPI method coupled with its supporting geospatial platform, it is possible to at least visualise those downhill areas and land features that are at risk from uphill sliding lands.

We want to thank the University of Mauritius for providing the necessary facilities for carrying this research. The Cartographic Section of the Ministry of Housing and Lands and Meteorological Services are gratefully acknowledged for providing the digital 10 m contours data and the rainfall data respectively. This work was given support under a scholarship at the University of Mauritius from the Tertiary Education Commission (TEC).

Arlidge, E.Z. and Wong You Cheong, Y., 1975: Notes on the land resources and suitability map of Mauritius 1:50,000. MSIRI Occasional Paper 29, 1-137.

Balaghee, P.D., 2001: Soil Aggregation and porosivity in relation to contents of clay, organic matter and iron oxide. BSc thesis, University of Mauritius, Mauritius.

Bamler, R. and Hartl, P., 1998: Synthetic aperture radar interferometry. Inverse Problems 14, R1-R54.

Burrough, P.A., 1986: Principles of Geographical Information Systems for land resources assessment. Oxford University Press, New York.

Cloude, S.R. and Papathanassiou, K.P., 1998: Polarimetric SAR interferometry. IEEE Transaction in Geosciences and Remote Sensing 36, 1551-1565.

DOS and MSIRI, 1962: Soil map of Mauritius (provisional classification) at 1:100,000. Directorate of Overseas Survey (DOS), London.

Geolab-Burgeap, WRU and Sigma-Arup, 1999: Geologic map of Mauritius at 1:50,000: Hydrogeologic schema . Geolab-Burgeap, Mauritius [in French].

Giorgi, L., Borchiellini, S. and Delucchi, L., 1999: Notes on the geologic map of Mauritius at 1:50,000: Hydrogeologic schema. Geolab-Burgeap, Mauritius [in French].

Halais, P. and Davy, E.G., 1969. Notes on the 1:100,000 agro-climatic map of Mauritius. MSIRI Occasional Paper 23, 1-29.

Jensen, J.R., 1996: Introductory digital image processing. Prentice Hall, New Jersey.

Kremer, M., 2000: Aspect of soil erosion in Mauritius-Analysis using GIS. MSc thesis, Philips-University Marburg, Germany [in German].

Lillesand, T.M. and Kieffer, R.W., 2000: Remote sensing and image interpretation. 3rd ed. John Wiley & Sons, New York
Ministry of Agriculture & Natural Resources, 1999: Annual report of the forestry service for the year 1995. The Government Printer, Mauritius.

Morgan, R.P.C., 1995: Soil erosion and conservation. Longman Group Ltd, London.

Nathire, F., 2003: Topographic data at the Ministry of Housing and Lands. MNODC Bulletin Dec 2003, Mauritius.

Ordnance Survey, 1991: 1:25,000 Maps for Mauritius. UK Government (Ordnance Survey) for the Government of Mauritius, London.

ORSTOM and MSIRI, 1983a: Pedology map of Mauritius at 1:50,000. OSTOM and MSIRI, Mauritius [in French].

ORSTOM and MSIRI, 1983b: Geologic strata and rainfall map at 1:200,000 for Mauritius. OSTOM and MSIRI, Mauritius [in French].

Padya, B.M., 1989: Weather and climate of Mauritius. Mauritius: The Mahatma Ghandi Institute Press.

Parish, D.H. and Feillafe, S.M., 1965: Notes on the 1:100 000 soil map of Mauritius. MSIRI Occasional Paper 22, 1-45.

Proag, V., 1995: The geology and water resources of Mauritius. The Mahatma Ghandi Institute Press, Mauritius.

Rughooputh, S.D.D.V., R. Nigel, H. Runghen, M. Bhuruth, K. Gooniah and F. Nathire, 2008. Integrating Disaster-management Information: Mauritius. In: Sharing Innovative experiences, Vol 12 – Examples of Natural Disaster Mitigation in Small Island Developing States. NY: UNDP. pp 99-118. (A Proc. of the International workshop on : Sharing Innovative experiences on Natural Disaster Mitigation in Small Island Developing States, UNDP, SSC, TWAS, October 2005, University of Mauritius, Réduit, Mauritius)

Saddul, P., 1995: Mauritius – a geomorphological analysis. The Mahatma Ghandi Institute Press, Mauritius.

Sentenac, R., 1964: Undergound water research for Mauritius. MSIRI Occasional Paper 17, 1-21.

Venkatasamy, D., 1991: Atlas for Mauritius. Edition de L’ocean Indien (EOI) & George Philips Ltd, London.

Water Resources Unit, 2003: Hydrology Data Book 1995 – 1999. The Government Printer, Mauritius.

Willaime, P., 1984: Simplified notes on the pedology map of Mauritius at 1:50,000. MSIRI Occasional Paper 33, 1-19 [in French].