Overview of the Data

Good data is essential for good urban indicators. Several key issues arise immediately when dealing with data from developing nations. First, the availability of data can be severely limited. Developing nations lack the information infrastructure necessary to collect, analyze, and interpret large amounts of information. Thus, the analyst is often bound by real constraints on the amount of data with which investigation can be conducted. Second, the quality of data is somewhat suspect. Political considerations, lack of resources, and limited training all interfere with data quality. Additionally, very few developing countries have national data standards, not to mention national digital data standards. Finally, accessing available data can be difficult due to language differences and sheer difficulty in transfer. Internet resources are often scarce and poor in quality where available and courier service incurs great expense. Translating digital data is a further difficulty, especially when the data has been collected in a less than optimal method, with meta-data often non-existent or incomplete.

We have attempted in this project to ameliorate some of these issues by working closely with our partners, who both collected and ground referenced much of the data used here. Working with CIDDI at the Catholic University of Mozambique, we obtained GIS coverages on population, land cover and land use, socio-economicTwo types of data are used in this project, GIS coverages and remotely sensed imagery, both air photography and a satellite image (Landsat Thematic Mapper).

Two types of data are used in this project, GIS coverages and remotely sensed imagery, both air photography and a satellite image (Landsat Thematic Mapper). development, housing, and infrastructure.

GIS coverages were obtained from both external sources and were generated internally. Of the images, air photographs were obtained from the state mapping agency, CENECATA and the satellite image was obtained from the Global Environmental Change Program at the University of Virginia.

Initial data generation for the entire city of Beira, within the UN framework on urban indicators, presented us with two challenges. First, a fundamental concept regarding the operability of the UN model for urban indicators was problematic, namely the scale of analysis. Intra-urban, or sub-city indicators, are critical in understanding the dynamics of rapidly growing urban areas, especially in the developing world. The UN model lacks sensitivity to scale, thus reducing its overall applicability in a data poor city, such as Beira where internal, as well as external, dynamics drive urban change. Second, we also determined early on in the project that monitoring all of Beira was not feasible due to limited data availability and data quality concerns.

The project database includes five information types:

Population

With population data, we can begin to build predictive models for areas that lack sufficient data. Population data is the most basic form of data used here. Population data was obtained from the Mozambique census for the year 1997. This is a robust data set including population totals, density, and several descriptive variables.

Land Cover and Land Use

Land use data includes the city of Beira at selected sample points. The major categories include the urban-built environment, gardening/farming, and open-land. A satellite image was obtained from the Global Environmental Change Program at the University of Virginia from which land cover for the city was digitized. Land use is particularly relevant to this project as the impact of urban agriculture is great in the poorer areas of the city. By identifying these areas of urban agriculture we can make assumptions regarding service provision and livelihood security. Areas with higher concentrations of urban agriculture tend to be those in poorer areas. Urban agriculture also reduces vulnerability of the city's poor.

Socio-economic Data

Socio-economic data was collected for the three study barrios, Matacuane, Macurungo, and Maluti by CIDDI and incorporated into the GIS databases. The sample size per barrio was approximately 100. This database includes detailed information on water and sewerage utilization in the three barrios and important demographic data. Variables include gender, age, region of origin, level of education, type of housing, and access to electricity and phone service. Of specific interest to this project are the variables on water connections, particularly access and source of water, water usage and expense, and distance to the nearest water source.

Housing and Buildings

A coverage of buildings was captured and referenced by CIDDI. These buildings include housing and housing type. The density of housing by barrio was plotted and, at the time of writing, is being verified off of recently obtained air photography. This coverage is fundamental in the modeling exercise, as we are concerned with infrastructural connections to houses and differences in connections by housing type.

Air Photography

The aerial photographs obtained from CENECATA, the national mapping agency of Mozambique, are being digitized and will be used to enrich existing data and create new data sets. Coverages were obtained for the years 1982, 1993, and 1996. We anticipate complete digitization of these photographs in the early part of Year Two. Air photography data will be used as an underlay to the digital coverages obtained from CIDDI. Specific data layers that will be improved are roads, houses and building stock, and water. As a developing city, Beira exhibits some fundamentally different spatial patterns and methods of service provision from cities in more developed industrialized regions. A network system of underground tunnels and open canals facilitates city drainage (Map 6). In addition to verification of existing data, new layers of information will be obtained from the photography. For instance, the road network on the periphery of the city in the GIS coverage is incomplete. Using the air photos, we have added to the existing data. A second example, land use, provides further clarity in intended data creation. Currently, the land use information for Beira is a point file. Using this coverage and the air photographs we will be able to map land use parcels or polygons.

Methods and Analysis

Methods

For these data to be useful, they all had to be transformed, referenced to the same co-ordinate system, and checked for errors and cross-compatibility. All files were referenced to the UTM system (WGS 84 Zone 35 South). Errors in projection and referencing were found and corrected for cross-compatibility (do the different coverages match when overlaid?) and graphically enhanced. Once each coverage and image had been corrected, overlay analysis began. Several coverages were overlaid to build the urban indicators. As this project is primarily interested in connectivity, housing and water access were analyzed first. Next, population and housing density were overlaid to determine concentrations in the barrios. Proximity to electrical lines was assessed, although inadequately due to the poor quality of data on such lines.

Land use was overlaid with housing to identify areas of high urban-agricultural activity (garden plots). The satellite image was then overlaid with each GIS coverage for empirical validation. Finally, GIS modeling of the infrastructure data was undertaken.Due to the data quality issues uncovered in year one, we began experimenting with GIS and remote sensing methods to enhance the utility of existing information. If funded, this will be a major focus for year two and will include the creation of a land cover and use change map for the city and region.

Air photographs will be geo-referenced and rectified, enhanced, and digitized. Concurrently the air photography will be overlaid with existing GIS coverages and additional features, e.g. houses, streams, land cover types, will be incorporated into the existing data. Only initial processing of this type has begun.

Initial Findings: Descriptive GIS Analysis

Initial findings shown in maps 2-9 are the result of geographic overlay and querying, buffering, and image overlay and querying. Basic mapping will not be included in this section.

a) Landsat Thematic Mapper Image: Map 2 shows the land cover of the city of Beira in 1991. Although generally urban in structure, several parts of the city display large patches of vegetation. To the southwest, large parks and open spaces dominate (shown in red on the false-color image), however, the vegetation in both Macurungo and Macuti are urban gardens, rather than parks or open spaces. Currently, there is no information on the exact use of these gardens. The heaviest concentration of industry is on the west at the rail yards, while the center of the city is marked "CBD"(map 2). The darker green and blue areas to the north of the city are exposed areas (soil, open water,etc.). Streets are manifest as linear, light lines. The white areas along the outer boundaries are beaches. The airport is located in the extreme northeast part of the city, evidenced on the map as a star pattern with the runways in dark blue.

b) Population Density and Infrastructure : High population densities in the city of Beira correlate to the type of housing. Less formal housing areas were identified from the satellite image (and some initial analysis of the air photographs).The less formal housing areas appear to be more densely populated (map 3). However, an inverse relationship was found between service provision and housing density. In short, those areas with the most need for infrastructure are those with less access to it and this helps explain the widespread unsanitary conditions.

By far, the central and smaller districts have the highest population densities. The concentration of development around the CBD is typical of a developing city.

The periphery of the city is rather close; denoting the poor transportation network In fact the city is skewed toward the transportation lines that run to the next city to the northwest, Dondo. Map 3 clearly shows higher densities in the center-west. These districts are rather large in area, but population distribution is not even. There are clear population clusters near the transportation lines.

Map 4 shows the dramatic increase in population in Beira from 1980-1997. A major influx of refugees from Mozambique's civil war flooded the city in the early and mid-1980s. Mauve district shows the greatest increase as it had the lowest overall population in 1980. The influx of poor, dominantly rural people settled largely in the northeast districts. On initial inspection of aerial photography, Mauve shows a great deal of rapid urbanization, but constrained to the southern part of the district.

One possible indicator of urban economic growth is the gender ratio in a city (map 5). A large influx of males seeking employment can lead to widespread unemployment and a host of service provision issues. Beira does not display an unbalanced gender population. All districts show relatively equal amounts of males and females in the population.

c) Population and Land Use: Analysis of the land use coverage overlaid on the satellite imagery showed the greatest concentration of urban agriculture in those areas with informal housing, but no strong relationship exists.

Informal gardens can also be identified in the areas with better housing stock and larger housing plots. Specific relationships can only be determined after processing of the air photographs has been complete.

d) Socio-Economic Data: The most significant analysis of the socio-economic data was in mapping rather than GIS analysis (See Appendix A for data tables). Specifically, density was queried against provision of water service and distance to water source was utilized as a buffer with the GIS coverage of water lines and canals. As map 6 shows, the area of the city within 0.25 km of both water and transportation is rather small. In developing cities, not only is access to water important, but also often getting to the water source point can be a challenge. Most of the areas with easy access to both water and transportation are in the central city area, however, recent water provision to the new settlements in the northeast is also evidenced on this map.

Beira sits on a very swampy, low part of the coastal plain. A major health challenge in a city of this size is drainage of water to prevent water-borne disease. Map 7 shows that the drainage system services a large part of the downtown, with some notable exceptions.

e) Housing and Infrastructure

GIS Modeling

Our goal here was to investigate means of estimating urban indictors at the intra-urban scale using GIS where direct measurement by survey, official record, etc., is unavailable. For example, access to water is one of the 23 key urban indicators defined by UNCHS. The UN Statistics Division defines access to water as having water located within 200 meters of the dwelling; hence the indicator has a specific spatial component.

It refers to piped water either within the building or accessible within 200 meters _ without extreme effort - to provide for household needs. The indicator is reported as the percentage of households with access to water.

No areal unit is specified, so this statistic may be reported for any geographical zones for which data are available. One method to estimate percentage of households with access to water using a GIS, given the dwelling locations and water pipeline infrastructure is shown (maps 8 and 9).

Inputs are a map layer of house footprints (polygons) and a water pipeline layer (arcs). A buffer is created from the water pipelines at convenient distances, for example 50 meters, 100 meters or 200 meters, etc., and this is overlaid on the house layer. Houses within 200 meters of a pipeline are selected as having access to water. A map of administrative zones (census tracts, barrios, neighborhoods) is overlaid on the access to water layer and the proportion of houses in each zone with access to water is calculated. A map showing differential access to water over the entire urban area or of specific dwellings with or without access may be created, for instance, based on map 8.

Similar methods may be applied for access to sewerage, electricity, gas, and other utilities. The main disadvantage of this approach lies in its reliance on the physical manifestation of the infrastructure. For instance, housing may be mapped from aerial photographs or satellite images, but other types of urban infrastructure such as water mains and sewers are subterranean and therefore cannot be mapped using remotely sensed images. Access to engineering plans or similar data is necessary.