Targeting and measuring housing problems in developing countries through urban texture analysis

Latin-american countries passed from predominantely rural to predominantely urban within few decades. The level of urbanisation in Brazil progressed from 36% in 1950, 50% in 1970, and scalating to 85% in 2005. This rapid transformation resulted in many social problems, as cities were not able to provide appropriate housing and infrastructure for the growing population. As a response, the Brazilian Ministry for Cities, in 2005, created the National System for Social Housing, with the goal to establish guidelines in the Federal level, and build capacity and fund social housing projects in the State and Local levels. This paper presents a research developed in Gramado city, Brazil, as part of the Local Social Housing Plan process, with the goal to produce innovative tools to help social housing planning and management. It proposes and test a methodology to locate and characterise/rank housing defficiencies across the city combining GIS and fractal geometry analysis. Fractal measurements, such as fractal dimension and lacunarity, are able to differentiate urban morphology, and integrated to infrastructure and socio-economical spatial indicators, they can be used to estimate housing problems and help to target, classify and schedule actions to improve housing in cities and regions. Gramado city was divided in a grid with 1,000 cells. For each cell, the following indicators were measured: average income of households, % of roads length which are paved (as a proxy for availability of infrastructures as water and sewage), fractal dimension and lacunarity of the dwellings spatial distribution. A statistical model combining those measurements was produced using a sample of 10% of the cells divided in five housing standards (from high income/low density dwellings to slum's dwellings). The estimation of the location and level of social housing deficiencies in the whole region using the model, compared to the real situation, achived high correlations. Simple and based on easily accessible and inexpensive data, the method also helped to overcome limitations of lack of information and fragmented knowledge of the area related to housing conditions by local professionals.