Abstract:
This research addresses the need for reliable, repeatable, quantitative measures to differentiate informal (slum) from formal (planned) settlements using commercial very high resolution imagery and elevation data. Measuring the physical, spatial and spectral qualities of informal settlements is an important precursor for evaluating success toward improving the lives of 100 million slum dwellers worldwide, as pledged by the United Nations Millennium Development Goal Target 7D.
A variety of measures were tested based on surface material spectral properties, texture, built-up structure, road network accessibility, and geomorphology from twelve communities in Guatemala City to reveal statistically significant differences between informal and formal settlements that could be applied to other parts of the world without the need for costly or dangerous field surveys.
When information from satellite imagery is constrained to roads and residential boundaries, a more precise understanding of human habitation is produced. A classification and regression tree (CART) approach and linear discriminant function analysis enabled a variable dimensionality reduction from the original 23 to 6 variables that are sufficient to differentiate a settlement as informal or formal.
The results demonstrate that the entropy texture of roads, the degree of asphalt road surface, the vegetation patch compactness and patch size, the percent of bare soil land cover, the geomorphic profile convexity of the terrain, and the road density distinguish informal from formal settlements with 87-92% accuracy when results are cross-validated.
The variables with highest contribution to model outcome that are common to both approaches are entropy texture of roads, vegetation patch size, and vegetation compactness suggesting that road surfaces and vegetation provide the necessary characteristics to distinguish the level of informality of a settlement. The results will assist urban planners and settlement analysts who must process vast amounts of imagery worldwide, enabling them to report annually on slum conditions. An added benefit is the ability to use the measures in data-poor regions of the world without field surveys.
