Hydrological Research Institute (now Institute for Water Quality Studies), Department of Water and Sanitation, Private Bag X313, PRETORIA 0001
Copyright Southern African Society of Aquatic Scientists (1996)
Published as: Silberbauer, M.J. & J. Moolman (1993) Changes in urban residential
land in the Rietspruit catchment, southern Transvaal. Southern African
Journal of Aquatic Sciences vol. 19(1/2) 89-94.
KEYWORDS
urban land cover, satellite image, aerial photograph, informal settlement, water quality
SUMMARY
Informal settlements have become an important and rapidly-expanding component of urban development in South Africa. In the Rietspruit catchment, south-west of Johannesburg, average growth rates in urban land cover of up to 2.7 km2 per year (1972-1991) were measured from satellite images. Aerial photographs show that in 1991 more than a third of the residential land consisted of informal housing. Elevated concentrations of dissolved nitrogen, phosphorus and faecal coliforms in the Rietspruit point to a possible eutrophication and health risk.
INTRODUCTION
Informal settlements can take the form of new shanty towns or older established residential areas where planned housing densities have been exceeded by the construction of shacks and lean-tos between the original buildings. The informal settlement has long been a response to the Third World need for affordable mass accommodation near large cities, and in South Africa informal settlements have sprung up around many urban areas. The rate of increase in these settlements and their effect on the surrounding aquatic environment are the subjects of this research note.
To determine the rate at which informal and formal residential areas are expanding, with a view to evaluating their potential influence on aquatic resources, the Rietspruit catchment south-west of Johannesburg was selected for analysis (Figure 1). The Rietspruit drains into the Vaal River upstream of Vaal Barrage. The residential areas in the catchment include various gold mining towns, smallholdings and the towns of Sebokeng, Evaton and Orange Farm. An iron and steel works is situated in the lower part of the catchment. The changes in urban land use which occurred during the 70's and 80's were mapped, and areas of formal and informal development were identified, where possible.
At the same time as this mapping exercise was under way, other water research and management organisations (see Acknowledgments) were studying the chemical and microbial water quality in the catchment. They made their preliminary results available for comparison with the land use data.
METHODS
A combination of satellite image and aerial photograph interpretation was used for determining land cover. Initially, various automatic or partially automated computer classifications of satellite images were tried, but with little success. The main reason for this is that the predominant features of urban land-cover are the rectangular and linear structures showing human activity. Digital reflectance values, used with success for classifying rural land uses, are of less use in analysing the finer texture of urban land cover. For this reason, the images were first edge-enhanced and contrast-stretched to make linear features stand out, then visually classified using an interactive drawing technique (ROI) on the image processor (IIS, 1986).
Three images were selected to span the period between 1972 and 1991 (Table 1). Before visual analysis, the two earlier images were geographically registered to the 1991 Landsat Thematic Mapper (TM) image, which had been geometrically corrected to the South African Gauss Conform map projection system. To ensure an exact fit, the Landsat Multispectral Scanner (MSS) images had to be resampled to a pixel size of 30 metres. The land cover types classified from the images were residential, smallholding, mining, commerce / industry and agriculture / "other". The agriculture / "other" class is a catch-all or reject class which includes, besides agriculture, everything that does not fit into one of the other four categories.
Table 1 The Landsat Multispectral Scanner (MSS) and Thematic Mapper (TM) satellite images used for determining land cover.
| Sensor | Date | Pixel size |
|---|---|---|
| MSS | 21-Nov-72 | 80m |
| MSS | 16-Nov-82 | 80m |
| TM | 07-Apr-91 | 30m |
Aerial photography of most of the residential areas in the eastern portion of the catchment was available for 1991, with a few settlements photographed in 1992. The aerial photography was interpreted by specialists in urban mapping at Boutek (CSIR), who classified urban land use into the residential categories of formal housing, informal housing, small scheme housing (interspersed with shacks) and the non-residential categories of schools, hostels and commercial sites.
The satellite and aerial photograph classifications were imported into a Geographic Information System (GIS), where the classifications derived from aerial photographs were registered to the geometrically correct 1991 TM image. The aerial photographs thus served to refine those portions of the satellite image of most interest.
RESULTS
The sequence of satellite images shows a steady increase in residential land use in the Rietspruit catchment between 1972 and 1991 (Figure 2). The areas where the greatest change occurred were those which were identified as informal housing and shacks from the 1991 aerial photography. Residential land cover made up 86% of the urban land cover and nearly half of the residential land was used for informal settlements.
The total urban area in the catchment increased from 22.7 to 42.7 km2 between 1972 and 1982, and had reached 67.1 km2 by 1991. The rate of increase therefore averaged 2.0 km2 per year for the first decade and 2.7 km2 per year for the second.
DISCUSSION
The satellite images used in this study provided adequate information for the construction of maps of land use change at a coarse scale and for broad categories of land use. Only the 1991 TM image data contained sufficient information for discriminating between high density and low density housing, but this resolution was not available earlier than 1989. In order to make clear distinctions between different types of urban land cover and to accurately measure their surface area, it was necessary to make use of aerial photographs. The interpreted aerial photographs presented problems in geographic registration, but when these had been solved they provided essential details of the quality of residential land use. In both aerial and satellite data interpretation there was much room for subjectivity, and it is likely that different interpreters could make different decisions based on experience, local knowledge and visual acuity (Trolier & Philipson, 1986). Classification should therefore be done by a specialist with close involvement in the management of the project.
Rapid urban development is clearly occurring in the Rietspruit catchment. The rate of expansion is comparable with the 2.8 km2 per year measured in Lagos, Nigeria, from 1962 and 1974 aerial photography (Adeniyi, 1980). More than 57% of the urban land increase in Lagos was attributed to "unplanned", high-density residential areas. In 1991, about 40% of the urban land cover in the Rietspruit catchment consisted of informal housing.
Informal development at the rate indicated by this study could be expected to have an influence on many environmental factors, including the aquatic environment. Results for samples of stream water taken at RV1 (Figure 1) downstream of one of the high-density residential areas (Rodda et al. 1992, Rand Water Board, pers. comm.) show increases in coliforms, conductivity, phosphate, nitrate and nitrite (Table 2). Faecal coliform counts exceed the guidelines for swimming by a large margin. These water quality changes could be the result of untreated domestic effluent and solid waste reaching the stream.
Table 2 The mean annual chemical water quality in the Rietspruit catchment at RV1. Conductivity (EC) is expressed as mS m-1; nitrite (NO2), nitrate (NO3) and ammonia (NH3) as mg N l-1; phosphate (PO4) and total phosphorus (TP) as mg P l-1; faecal coliforms as counts 100 ml-1. The number of samples used to calculate the mean is given in brackets. The guidelines are from DWAF (1993a & b). See Acknowledgments for sources of raw data.
| 1974* | 1982 | 1991 | Domestic guideline | Swimming guideline | |
|---|---|---|---|---|---|
| EC | 34.6 (9) | 52.8 (12) | 76.6 (9) | 70 | - |
| NO2 | - | 0.07 (3) | 0.60 (6) | - | - |
| NO3 | - | 3.0 (8) | 4.6 (8) | 6 (NO2+NO3) | - |
| NH3 | - | 0.8 (1) | 8.6 (9) | - | - |
| PO4 | - | 0.49 (2) | 1.37 (9) | - | - |
| TP | - | - | 4.72 (13) | - | - |
| FC | - | - | **3.0x104 (11) | 0 | **150 |
On another level, this project gave the participants an introduction to the use of GIS for combining diverse environmental data with the capacity for viewing a wide range of information covering a whole catchment. The problems of inter-agency co-ordination and data transfer between different computer systems emerged, and it seems inevitable that the use of GIS will lead to changes in the way in which multi-disciplinary projects are managed.
ACKNOWLEDGMENTS
The information contained in this research note is published with the permission of the Director-General of the Department of Water and Sanitation. Aerial photograph interpretation was performed by Mr and Mrs F du Toit of Boutek, CSIR. Access to unpublished water quality data was provided by the Rand Water Board, Watertek (CSIR) and the Water Research Commission.
REFERENCES
Adeniyi, P. O. (1980) Land-use change analysis using sequential aerial photography and computer techniques. Photogrammetric Engineering and Remote Sensing, vol. 46(11), pp 1447 - 1464.
DWAF (1993a) South African Water Quality Guidelines. Volume 1: Domestic Use. Department of Water and Sanitation / CSIR Environmental Services, Pretoria.
DWAF (1993b) South African Water Quality Guidelines. Volume 2: Recreational Use. Department of Water and Sanitation / CSIR Environmental Services, Pretoria.
IIS (1986) System 600 Version 1.4. International Imaging Systems, Milpitas, California.
Rodda, N., D. Hohls, J. Harris, R. Kfir, M. Steynberg & C. de Wet (1992) Techniques for microbial water quality investigation of South African rivers. Progress Report 1 submitted to the Water Research Commission by the Division of Water Technology, CSIR, and the Rand Water Board. (January, 1992.)
Trolier, L. J. & W. R. Philipson (1986) Visual analysis of Landsat Thematic Mapper images for hydrologic land use and cover. Photogrammetric Engineering and Remote Sensing, vol. 52, pp 1531 - 1538.