Urban poverty maps: From characterising deprivation using geo-spatial data to capturing deprivation from space

Most earth observation (EO) approaches only yield a binary delineation of deprived/non-deprived areas – an oversimplified characterisation with little information inferred regarding the diversity of intra-urban deprivation. In this study, we attempt to explore the potential of using VHR EO-based data to predict the degrees of intra-urban deprivation in Nairobi, Kenya. This involves a two-step workflow of characterising and predicting a continuous index of deprivation degrees. First, a principal component analysis (PCA) is conducted to characterise the multi-dimensionality of deprivation using open geospatial datasets as a set of continuous indices. Next, a convolution neural network (CNN) based regression model is trained to directly predict the deprivation indices using SPOT-7 images. The best prediction of the proposed CNN regression model is obtained in the morphology-based domain, with an R² of 0.6543. We demonstrate the potential of an EO-based method to directly capture the degrees of morphological deprivation with relatively high accuracy, while also acknowledging its limitations in predicting the complexity of deprivation.