With the wide scope of data and information provided by the remote sensing process, determining changes or transformations in LULC has become increasingly easy, cost efficient and more importantly, a source of reliable information (Alphan, 2003; Lowry et al., 2007). However, the act of mapping these changes requires that the land in question be divided into specific land cover classes (Alphan, 2003; Anderson et al., 1976;
Lowry, 2005; Lowry et al., 2007). To accomplish this, LULC studies require a generalised LULC classification scheme which will help provide the desired consistency needed for LULC mapping and analysis (Lillesand et al., 2014; Thompson, 1996).
At the moment, there is no single universally acknowledged LULC classification scheme due to the fact that classification schemes tend to differ according to the country, organisation and also to the researcher’s preferences. However, one of the most commonly used land cover classification system was devised by the United States Geological Survey (USGS) in the mid-1970’s (Anderson et al., 1976; Lillesand et al., 2014).
The USGS Land Cover Classification System is noted as the basic structure of numerous land cover classification schemes; even for those classification systems which have been able to offer a more detailed and specialised mapping of land classes in recent studies (Lillesand et al., 2008).
It has long been foretold that the integration of LULC data should not take place especially on a single map;
however, studies have proven that by combining the various remotely sensed data the map becomes more reliable and resourceful in its use (Lillesand et al., 2008; Lowry et al., 2007; Thompson, 1996). The intermixing
of LULC data, studies conducted by Thompson (1996) and Lillesand (2008) noted that when generating and analysing LULC maps, it is crucial to acquire additional or supplementary data.
With data being generated from different sources, collecting supplementary data to provide improved accuracy to a remotely sensed map can range from processes such as site visits and the use of other updated maps (topographical or cadastral maps) (Dewan and Yamaguchi, 2009a; Lowry et al., 2007). By combining and integrating data of different types and from different sources, the maps generated are of a more accurate, reliable and constant quality. Furthermore, it is important to note that by using data which can be geometrically registered to a common geographic base; it increases the potential for greater information extraction (Dewan and Yamaguchi, 2009a; Dewan and Yamaguchi, 2009b).
There is no singular accepted land cover classification system due to the diversity of the landscape in question, however, a land use and land cover classification scheme was adapted from Thompson’s (1996) article ‘A Standard Land-cover Classification Scheme for Remote Sensing Applications in South Africa’ for the purpose of this research. This classification scheme suggested by Thompson (1996) is designed specifically to meet the needs of the South African user, while still being able to adhere to the condition set by international standards.
5.2.1 Identification of each LULC Class
As noted in previous chapters, differentiating between ‘land use’ and ‘land cover’ is increasingly difficult and has often resulted in the two being mistaken for each other due to their ability to interchange (Jansen and Di Gregorio, 2003; Jansen and Gregorio, 2002). In addition, “most conventional methods of assessing land- cover change only identify transitions between classes, while neglecting change within classes due to land- cover modification ... usually results in significant error, potentially underestimating the total area experiencing land-cover change and the magnitude of that change,” (Powell and Roberts, 2010: 185).
Using Thompson’s suggested scheme, it is also stated that the classification is based upon three hierarchical levels (Figure 5A), which enables the researcher to take broad level classes (Level I) and further the detail and specifics in each land cover class (Level II). Therefore, for the purpose of this research, the adopted land cover classification scheme consists of both Level I and Level II cases.
Figure 5A: The Three Level Hierarchy (Thompson, 1996)
In order to ensure that the correct land cover classes were used for this research endeavour, data such as Google Earth imagery and the eThekwini Municipality’s online aerial photographs were utilised to outline potential land cover classes prior to conducting the required field work. Although this process may seem tedious, it allowed for an overall visual assessment of the study site and permitted a shortlist of land cover classes to be identified with the area in question. As a result, the following land cover classes were identified for this specific study: Forest and Woodlands, Barren lands, Cultivated lands, Urban/Built-up lands and Grasslands.
Table 5A: The Land Cover Classification Scheme utilised in this study [Adapted from (Thompson, 1996)]
Abbreviation Class Name Description of the Land Cover Class
F & W Forest and Woodlands
This land cover class includes areas that are composed of greater than 10% tree cover, as well as areas encompassed with woody plants that occur greater than 5 metres in height.
B Barren Lands
This land cover is composed of all non- vegetated lands, or areas that consist of minimal vegetation. This class is composed of areas that are made up of
coastal dunes, Aeolian dunes and beach sands.
C
Cultivated Lands
This area is composed of lands that have been ploughed and/or prepared for the raising of various crop, however it excludes timber production. Included are crops, fallow lands, or lands that are prepared for planting.
U Urban/Built-up Lands
All areas composed of buildings and urban development; this includes residential, commercial, industrial land cover as well as that of communication systems.
T Transport Networks Roads and other transport derived routes, subclass of urban.
G Grasslands
Areas composed of grasslands that have primarily less than 10% of tree or shrub canopy cover but which has greater than 0.1% of the total vegetation cover present.
It is composed of the dominant grass like, non-woody as well as rooted herbaceous plant species.