Get Complete Project Material File(s) Now! »
Chapter 2 LITERATURE REVIEW
Introduction
The African elephant (Loxodonta africana) has become an endangered species and the dwindling population numbers need to be actively conserved. This decline in elephant numbers over most of Africa is mainly the result of habitat loss and fragmentation, as well as poaching (Blanc, Barnes, Craig, Dublin, Thouless, DouglasHamilton & Hart, 2007). Elephant numbers are only on the increase in Southern Africa with the population currently standing at about 321 000 elephants, of which 18 507 are found in the enclosed protected areas of South Africa (Blanc et al., 2007).
Elephants represent a source of revenue for local communities from the tourist industry through game reserves. For this reason, many of the smaller private conservation areas have re-introduced elephants. This can have a major impact on habitat conservation as increasing the available water points, fire control and the building of tourist infrastructure all result in environmental modification. In confined areas, high elephant numbers can cause unsustainable and often irreversible change to vegetation. A balance between maximum elephant numbers and minimum ecological damage needs to be found. In order to find this balance managers of protected areas need ecological data on which to base management decisions.
At the beginning of 2008, the South African Minister of Environmental Affairs and Tourism (DEAT), Mr. Marthinus van Schalkwyk, issued Norms and Standards for the Management of Elephants in South Africa. The Norms and Standards now make it mandatory for elephant owners to prepare and submit an elephant management plan (DEAT, 2008). This decision by Minister Van Schalkwyk has further increased the need for accurate and up to date ecological information.
Reliable ecological data can only be produced using accurate environmental monitoring methods. Field based monitoring methods produce detailed ecological data on a small scale but can be uneconomical and impractical when large scale data sets are required at regular intervals. Remote sensing, especially images derived from satellite sensors, provides a cost effective way to assess habitat structure and land cover changes regularly over large areas. The large spatial and temporally repeated observations provided by satellite images can significantly improve the quantity and quality of environmental data collected for a protected area. This information can be used to map and monitor land cover changes taking place over time (Paolini, Grings, Sobrino, Muñoz & Karszenbaum, 2006).
Elephants as agents of environmental heterogeneity
Savannah ecosystems are complex and dynamic with many environmental factors interacting over time to produce changes in the ecosystem (Tews, Ester, Milton & Jeltsch, 2006). No one factor, such as elephants, is solely responsible for changes in savannah ecosystems (Dudley, 1999). Disturbances such as fires, droughts and browsing/grazing by herbivores all help to maintain the heterogeneity of savannahs (Dudley, 1999; Owen-Smith, 2006; Chongo, Nagasawa, Ahmed & Perveen, 2007). The spatial and temporal availability of water is another important factor in maintaining habitat heterogeneity. Adler, Raff and Lauenroth (2001) define spatial heterogeneity as the relationship between the values of one variable observed at different locations. The more spatial heterogeneity is present, the more random the pattern of ecosystem variables and the wider variety of habitats will be available. This is an important buffer for the effects of the temporal variability seen in resource availability, as a varied habitat produces an overlap in resources.
There is a delicate balance between environmental disturbances creating heterogeneity and causing habitat degradation. The impacts of disturbances are all interlinked and the effects of each disturbance may be amplified or masked by others (Guldemond, 2006). ‘The intermediate disturbance hypothesis’ as expounded by Owen-Smith (2006) states that; overall species diversity is increased by disturbances that are not too severe or frequent. Even though diversity may initially be reduced in the disturbed patches, a variety of habitats with their associated species will result at the landscape scale thus increasing heterogeneity. It is persistent or pervasive disturbances that can reduce diversity, as species that are unable to tolerate such pressure will disappear. Such disturbances are often found when areas are enclosed and manipulated by humans and in so doing forcing unnatural utilisation patterns on the ecosystem.
Elephant impacts is one of the most obvious and hotly debated disturbances seen in Southern African savannahs as increased feeding pressure by elephants result in very visible damage. Elephants, being a keystone species, are also regarded as important elements in natural savannah systems as their destructive feeding habits create and maintain heterogeneity (Du Toit, Rogers & Biggs, 2003; Owen-Smith, 2006; Chongo et al., 2007). Despite the damage elephants do to vegetation, this is a necessary part of the development of a savannah ecosystem. Elephants open up woody areas allowing easier access to densely bushed areas thus improving the habitat for some species and in this way help to regulated bush encroachment into grasslands (Augustine & McNaughton, 2004). New palatable growth is stimulated in damaged trees at a height that is within the feeding range of smaller browsers (MacGregor & O’Connor, 2004) and nutrients are made available that would otherwise be locked away in tree bark and wood. It is when feeding becomes locally concentrated for long periods of time that vegetation damage becomes a threat to biodiversity (Dublin, Sinclair & McGlade, 1990; Birkett, 2002).
There is, however, still much uncertainty about the long-term impact that elephants in confined areas have on their environment, as the effect of such sustained damage in an enclosed system is unknown. In a literature review conducted by Guldemond (2006), it was found that most short-term studies of elephant impact on vegetation showed a decrease in the abundance of woody vegetation. However, long-term studies found that elephants resulted in an increase in vegetation abundance or otherwise no change in vegetation abundance was seen. Guldemond (2006) also suggests that elephants may affect different habitats differently. He found that elephants in Tembe Elephant Park, South Africa, enhance the spatial heterogeneity of closed woodlands while open woodlands were homogenised. As elephants are visibly destructive, they have often been the only source of disturbance taken into account when investigating vegetation change. This has resulted in the effect of other animals being overlooked or minimised. Recent studies have shown that other browsers also negatively impact vegetation, though their exact contribution to the savannah dynamic has not yet been established (Styles & Skinner, 2000; Birkett, 2002; Augustine & McNaughton, 2004; Birkett & StevensWood, 2005; Owen-Smith, 2006). Even small browsers such as steenbok, have been found to effect vegetation density by reducing the recruitment of woody seedlings (Barnes, 2001; Augustine & McNaughton, 2004; Owen-Smith, 2006). This can reduce the number of seedlings that reach reproductive maturity and over the long term reduce vegetation diversity. Browsing pressure from giraffe has also been found to have a marked effect on vegetation density and distribution with giraffe having the greatest impact at 3–5 m heights (Birkett, 2002). While they usually do not kill trees outright, they can reduce growth and make the trees more susceptible to disease, drought and other stresses (Bond & Loffell, 2001).
Elephants have traditionally been held responsible for the hedging of Mopane trees (Colophospermum mopane) in Botswana’s Tuli Block. However, Styles and Skinner (2000) found that eland are also major contributors to the formation of hedges in the 1.5–3 m range. Styles and Skinner (2000) went as far as suggesting that the dependence of elephants on Mopane trees may have been overemphasised and that the role of other large herbivores, such as eland, is much bigger than currently recognised. In a three-year study in Kenya, Birkette and Stevens-Wood (2005) found that while elephants were responsible for 40% of the recorded tree deaths, 33% of the dead trees could be attributed to browsing by Black Rhino and 27% to drought.
Surface water availability is another important driver of ecological heterogeneity (Thrash, 1998; Chamaille-Jammes, Valeix & Fritz, 2007). Trampling and overgrazing around water points due to the local concentration of game has visible effects radiating out as far as 200 m from the water point (Thrash, 1998). The provision of permanent artificial water points has increasingly been seen as one of the main culprits for degradation of natural vegetation. By altering natural distribution patterns of watering points, impacts are moved to and/or concentrated in areas that might not be robust enough to handle the new pressure (Thrash, 1998; Leggett, 2006). This results in erosion around water points and the reduction of biodiversity, as species unable to tolerate the sustained pressure are eliminated.
The provision of artificial water points also affects the behaviour and distribution of animals. This allows water dependent animals to move into and remain in areas, which previously would only have been utilised seasonally, and in so doing alters the natural distribution and abundance patterns of game, as animals are no longer dependent on seasonally variable water sources (Thrash, 1998; Leggett, 2006). Chamaille-Jammes et al. (2007) found that the provision of artificial water points was a major cause of local overabundance of elephants, as they now did not need to move far in order to find food and water.
A very uniform utilisation of vegetation results if the seasonal variation in grazing pressure is no longer present. Irregular grazing allows areas of dense and tall stands of grass or shrubs to develop, and the loss of these habitats displaces organisms dependent on them. The effect of too many artificial water points has been seen in Kruger National Park where the provision of artificial water points has resulted in no area in Kruger being further that 10 km from permanent water. Little or no areas of tall stands of ungrazed grass remain. The reduction of Roan antelope numbers in recent years has been attributed to the reduction in the amount of thick stands of grass, which serves as hiding places for the altricial young of this species (Thrash, 1998).
Abundant permanent water also means less competition for this resource in especially the dry seasons, removing one of the natural checks on animal populations. As a result, water dependent animals such as elephants, wildebeest and zebra can increase to the point where the availability of food becomes the main limiting factor of population growth. This shift in environmental dependence results in food sources being over-utilised to the detriment of other organisms.
Studies in the Serengeti have shown that although high elephant density can prevent woodlands from expanding, a further external perturbation such as frequent or severe fire was necessary in order to change the vegetation over from woodland to grassland (Ben-Shahar, 1996; Dublin et al., 1990). Once grassland formed, elephant feeding pressure would prevent the vegetation cover from reverting back into woodland (Dublin et al., 1990). Skarpe (1991) found that in Botswana, Acacia erioloba are vulnerable to fire until their canopies are above 2-3 m. If fire occurs too frequently the woody seedlings never get an opportunity to grow out of such vulnerable height class. An increase in fire incidences can result in a reduction in the density of trees and an increase in the density of the lower height classes (BenShahar, 1996). Combined with elephant feeding pressure, fire can have a farreaching effect on woody vegetation cover.
While elephants are not solely responsible for changes in vegetation, they are acknowledged as one of the key drivers behind vegetation change. In confined areas elephant impacts must be carefully monitored to ensure that these stay within acceptable thresholds. Exceeding these thresholds can cause excessive damage to the vegetation with the resulting destruction of habitat and loss of biodiversity. This threshold of acceptable damage is difficult to determine, especially at landscape level in a heterogeneous environment. Factors such as vegetation type, animal densities, topographic features and water availability all affect the threshold of acceptable damage of a particular area. This threshold may even vary from year to year as temperature and rainfall fluctuates. The impact of elephants also depends as much on where they are, as on how many they are (Henley & Henley, 2007; Van Aarde & Jackson, 2007).
A variety of factors thus have to be considered when establishing elephant impact thresholds. It is not enough to just monitor actual elephant numbers, but the affect they have on the vegetation of the area needs to be monitored as well. Many of the impacts are cumulative with their full effect only becoming visible over a period of a few years. Long-term ecological monitoring programs are needed to assess trends in vegetation change and the stability of the ecosystem so that the full scale of the impacts can become visible. The more effective and accurate this ecological monitoring is, the better the management policies that can be implemented. Ecological information needs to be objective, economical and easily available in order for management to effectively plan environmental and economic objectives (Gorden, Hester & Festa-Bianchet, 2004).
Satellite images as a tool for environmental monitoring
Satellite sensors use reflected energy in the visible and infrared regions of the electromagnetic spectrum to produce images. For any given material, the amount of solar radiation that it reflects, absorbs, transmits, or emits varies in each wavelength. When these amounts are plotted over the wavelength range, the connected points produce a spectral signature or curve that is unique to each material (Lillesand, Keifer & Chipman, 2004; Jensen, 2005). It is this important property of matter that makes it possible to identify different objects and distinguish them from one another.
Because remote sensing devices traditionally operate in the green, red, and near infrared regions of the electromagnetic spectrum, they can be used to discriminate vegetation. Chlorophyll pigment in green-leaf chloroplasts absorbs radiation centred at about 0.65 µm (visible red) and also in the blue range (about 0.55 µm). Most vegetation has a green colour as chlorophyll reflects the green wavelengths of visible light. Green vegetation however, most strongly reflects light of wavelengths between 0.7 and 1.0 µm (near IR). As the intensity of this reflectance is usually bigger than from most inorganic materials, vegetation appears bright in the near-IR wavelengths. In multi-spectral images vegetation is characteristically dark in the blue and red bands, lighter in the green band and very noticeably light in the near-IR bands (Short, 2007). This distinctive response seen in vegetation allows vegetated areas to be relatively easily identified and analysed using remote sensing techniques and thus remote sensing, or more specifically satellite images, can be a useful natural resource management tool for gathering ecological information on a large scale at regular intervals.
Plant species richness is a key indicator of biodiversity at the community and regional scales and quantitative as well as qualitative information about plant species richness can be gathered using remote sensing. Rocchini (2007) used spectral heterogeneity of satellite images to predict species richness, which is known as the Spectral Variation Hypothesis. However, the satellite sensors currently available are limited to specific scales of investigation as spectral variability is scene and sensor dependent. Coarser resolution data tend to have mixed pixel problems and are thus less sensitive to spatial complexity. However, the spectral response from different landcover features in images with higher spectral resolution exhibit higher complexity.
The effects of scale when measuring spectral and spatial heterogeneity and relating it to field data must be kept in mind when using satellite images (Rocchini, 2007).
Assessing the effect that environmental changes have on vegetation and thus also on animal populations, is an important component of environmental management allowing better predictions of the effects of biodiversity reduction or habitat degradation. With limited vegetation information at large temporal and spatial scales, it is difficult to discern the direct and indirect effects of environmental change. The use of Vegetation Indices such as the Normalised Difference Vegetation Index (NDVI) as a means of gathering vegetation information has to a large extent provided sufficient information (Pettorelli, Vik, Mysterud, Gaillard, Tucker & Stenseth, 2005). As ecologists become more aware of the benefits of using satellite images, their use in ecological studies has increased (Nagendra, 2001; Kerr & Ostrovsy, 2003; Pettorelli et al., 2005; Revenga, 2005).
While remote sensing is not perfect it is a powerful tool for identifying and classifying habitats, allowing predictions about species distribution to be made and changes from landscape to global level to be detected (Kerr & Ostrovsy, 2003). Information derived from remote sensing can be used for mapping and monitoring land cover changes, especially in the forestry sector (Jha, Goparaju, Tripathi, Ghari, Raghubanshi & Singh, 2005; Katsch & Kunneke, 2006) with vegetation degradation due to disturbances such as livestock grazing, fire, drought and anthropological impacts also being tracked (De Stoppelaire, Gillespie, Brock & Tobin, 2004; Dukiya, 2006).
Traditionally the measuring of species richness is conducted at the species level and while this provides useful information, it is spatially constrained. The use of remote sensing allows for large area characterisations of biodiversity in a systematic, repeatable and spatially exhaustive manner (Duro, Coops, Wulder & Han, 2007). A combination of direct and indirect approaches can derive four key indicators of diversity namely: productivity, disturbance, topography, and land cover. By monitoring these indicators over time at an ecosystem level, can provide an early warning system of potential biodiversity changes. Large area biodiversity monitoring systems can thus provide an initial stratification of key areas where further analysis at a local scale can be focused (Duro et al., 2007).
Worldwide land degradation is a serious environmental problem that needs to be monitored. The land degradation of grassland in northeast China was monitored using LANDSAT TM/ETM 6 data, the Normalized Difference Vegetation Index (NDVI), and variables (brightness, greenness, wetness) generated by the Kauth– homas Transforms (KT) algorithms as the feature nodes of a DT classifier. An overall accuracy of more than 85% was obtained for the distribution maps of land degradation that were generated (Chen & Rao, 2008). By choosing sensor bands sensitive to vegetation, quick, accurate and economical data sets targeted specifically at vegetation and vegetation change may be generated (Mas, 1999).
Species distribution patterns techniques using remote sensing usually fall into one of three categories. The first category is the direct mapping of individual plants or associations of single species in relatively large, spatially contiguous units. The second technique is habitat mapping using remotely sensed data, and species habitat requirements. The third category is the use of direct relationships between spectral radiance values and field based species distribution patterns. Direct mapping is applicable over smaller extents for detailed information on the distribution of certain canopy tree species or associations. Estimations of relationships between spectral values and species distributions may be useful for the limited purpose of indicating areas with higher levels of species diversity, and can be applied over spatial extents of hundreds of square kilometres. Habitat maps appear most capable of providing information on the distributions of large numbers of species in a wider variety of habitat types. This is strongly limited by the variation in species composition, and best applied over limited spatial extents of tens of square kilometres (Nagendra, 2001).
Further problems incurred when mapping natural vegetation using mid-resolution satellite images and conventional supervised classification techniques are: (1) defining the adequate hierarchical level for mapping; (2) defining discrete land cover units discernible by the satellite; and (3) selecting representative training sites (Cingolania, Renisona, Zaka & Cabidoa, 2004). These problems can be limited through the use of spectral information to objectively select the best training sites. Chust, Ducrot and Pretus (2004) found that as automated-classification procedures of satellite imagery are based on surface reflectance it generally ignores other properties such shape and size of landforms.
In Africa remote sensing has also been successfully implemented in a number of ecological monitoring projects. Serneels, Said and Lambin (2001) have used MODIS images to characterise short-term land cover change in East Africa. They evaluated land use, fire and livestock grazing and found that there was a strong correlation between land use and vegetation response to rainfall variability. Yang and Prince (2000) estimated canopy cover in Zambia using satellite imagery derived from the LANDSAT MSS scanner allowing changes in the vegetation structure to be monitored. Munyati (2000), using LANDSAT MSS and TM sensors, successfully tracked changes in the wetlands of the Kafue Flats.
Botswana has yielded a number of ecological studies using remote sensing with McCarthy, Gumbricht and McCarthy (2005) mapping eco-regions in the Okavango Delta. Cassidy (2007) and Moleele, Ringrose, Matheson and Vanderpost (2002) mapped the burned areas in the Okavango Delta Panhandle by monitoring bush encroachment. Baldyga, Miller, Driese and Gichaba (2007) used Landsat TM images to quantify the timing and rate of these changes in and around the River Njoro watershed located near the towns of Njoro and Nakuru in Kenya’s Rift Valley. Vegetation diversity and temporal variability, common to tropical and sub-tropical areas, posed several challenges in disaggregating classified data into subclasses. Campbell, Lusch, Smucker and Wangui (2005) conducted a landscape-scale study combining the analysis of multi-temporal satellite imagery spanning 30 years and information from field studies extending over 25 years to assess the extent and causes of land use and land cover change in the Loitokitok area in the south-east Kajiado District, Kenya.
Relatively little use has been made of remote sensing in South Africa for ecological monitoring. Vegetation degradation detection and mapping using Advanced Very High Resolution Radiometer (AVHRR) data has been conducted by Wessels, Prince, Malherbe, Small, Frost and Van Zyl (2007). In the Eastern Karoo, Archer (2004) used NDVI and rainfall data to track the effects of commercial stock grazing practices while wetlands were mapped in the Western Cape by De Roeck, Verhoest, Miya, Lievens, Batelaan, Thomas and Brendonck (2008).
While some studies have utilised satellites and GPS technology to track elephant movements (Henley & Henley, 2007), few studies have used satellite images to track elephant related vegetation change. In northwestern Zimbabwe, Murwira and Skidmore (2005) made use of NDVI data to predict elephant movements due to changes in spatial heterogeneity. They found that the presence of elephants could be reliably predicted using changes in intensity and dominant scale of spatial heterogeneity. While the study by Murwira and Skidmore touches on the use of remote sensing for monitoring elephant induced vegetation change, the full potential of this tool has not yet been recognised.
Elephant conservation areas cover large areas and are usually highly heterogeneous making accurate vegetation surveys difficult. Ground surveys, while producing reliable small-scale data, is labour intensive, time consuming, expensive and may not cover the entire area. Without an accurate vegetation map, other ecological data derived for the area will be incomplete and thus unreliable. While the use of satellite images cannot do away with the need for fieldwork, it is a useful tool for supplementing field observations and in this way can reduce both the time and manpower needed to produce results (Jensen, 2005).
Table of Contents
Declaration
Acknowledgments
Abstract
Table of Contents
List of Figures
List of Tables
Chapter 1 INTRODUCTION
1.1 Background
1.2 Location for the study
1.3 Research problem
1.4 Aim & Objectives
1.5 Research design
Chapter 2 LITERATURE REVIEW
2.1 Introduction
2.2 Elephants as agents of environmental heterogeneity
2.3 Satellite images as a tool for environmental monitoring
2.4 Conservation management on Welgevonden
2.5 Conclusion
Chapter 3 METHODOLOGY
3.1 Introduction
3.2 Methods and processes
3.3 Conclusion
Chapter 4 ANALYSIS OF RESULTS
4.1 Introduction
4.2 Classification
4.3 Change detection
4.4 Further analysis
4.5 Conclusion
Chapter 5 DISCUSSION
5.1 Introduction
5.2 Vegetation change
5.3 Conclusion
Chapter 6 CONCLUSIONS
6.1 Introduction
6.2 Elephant induced vegetation change
6.3 Further research
6.4 Conclusion
REFERENCES
GET THE COMPLETE PROJECT
The utilisation of satellite images for the detection of elephant induced vegetation change patterns