Skip to content

Developing a geo-data frame using dasymetric mapping principles to facilitate data integration

Literature › Works

Gerbrand Mans (2011). Developing a geo-data frame using dasymetric mapping principles to facilitate data integration. CSIR (Planning Support Systems, Built Environment), South Africa.
Type working paper
Scope Comparative and methods
Method descriptive quantitative
Period studied 2001–2010
Regions South Africa, eThekwini (Durban), City of Cape Town
Themes methods data sources, demography urbanisation
Citation key mans2011

Argument

The paper argues that integrating socio-economic and environmental data is hampered by incompatible spatial units, and that basic spatial units (such as the CSIR mesozones) become obsolete when administrative boundaries change, as they often do in South Africa. It therefore proposes a flexible point-based 'geo-data frame' built from the CSIR/ESKOM spot building count (SBC), a 2006-08 satellite-derived inventory of all formal and informal buildings. A five-step hierarchical dasymetric procedure assigns each point a population weight derived from 2001 census small-area data. The steps cover unchanged residential areas, new urban growth areas, informal settlements, commercial/industrial zones and rural areas. Validated against municipal population updates for eThekwini and Cape Town, the frame shows intraclass correlations of about 0.9 at sub-place level, so the author deems the method reliable for allocating population to any demarcation.

Evidence and method

Spot 5-derived spot building count (CSIR/ESKOM, 2008), the 2001 Statistics South Africa census (small area layers, sub-places), the 2001 land-use dataset, and population updates from eThekwini (2010, aerial images plus household surveys) and the City of Cape Town (2009, planning data, digitised informal settlements and backyard-dwelling surveys).

Findings

  • The ancillary dataset is a national building inventory from satellite imagery. (p. 4)

    The SBC was produced by the CSIR and ESKOM in 2008 and is a geo-referenced building frame developed using Spot 5 satellite imagery.

  • Frequent boundary changes undermine fixed spatial units in South Africa. (p. 3)

    Especially in a South African context this is a major issue with local municipality boundaries changing

  • Regression calibration between census and SBC was impossible because the two datasets come from different years. (p. 6)

    The incompatibility of the datasets arising from the different time lines of the datasets, however, rendered this impossible.

  • A hierarchical elimination procedure was used to distribute current population. (p. 6)

    A five-step hierarchical approach – incorporating a process of elimination – was followed in order to get a more detailed distribution of the current population figures for South Africa.

  • Informal settlement populations are imputed from the densest overlapping 2001 small area. (p. 7)

    The underlying assumption of the above mentioned process is that the density and household size of an informal area are the same as those of the densest SAL (2001) it overlaps.

  • Rural buildings are a weaker proxy for households. (p. 8)

    rural areas have a lot of buildings which may either be uninhabited or used for a different purpose, for example storage facilities on farms.

  • City-wide validation shows high agreement with municipal control totals. (p. 10)

    For the City of Cape Town the ICC is 0.907 and for eThekwini it is 0.899.

  • The method slightly underestimates population in the most populous sub-places. (p. 10)

    It is only for the areas with an exceptional high population figure that SBC-data frame underestimates the total population, but only slightly, for both the study areas.

  • Agreement is strongest in eThekwini's dense rural former tribal areas. (p. 12)

    The dense rural areas of eThekwini host about 8% of the total population.

  • Agreement is strongest in eThekwini's dense rural former tribal areas (ICC value). (p. 12)

    The ICC is 0.965.

  • The author stresses the method's value for former homeland regions where data are poor. (p. 13)

    it is these areas (think Eastern Cape, Limpopo, Mpumalanga) for which is difficult to get accurate and up-to-date population data.