Relative Spatial Poverty Analysis using Remote Sensing and Points of Interest
Elysabeth Nindy Nasing1*, Agung Budi Harto2,3, and Anjar Dimara Sakti2,3

1. Master Program in Geodesy and Geomatics Engineering, Institut Teknologi Bandung,
2. Sains and Technology Information Geographics, Faculty of Earth Science and Technology, Institut Teknologi Bandung, Indonesia,
3. Center for Remote Sensing, Institut Teknologi Bandung, Indonesia
*nasingnindy[at]gmail.com

Abstract

Poverty mapping is a crucial aspect in designing targeted policy interventions. However, conventional approaches based on household surveys are often constrained by high costs. This study proposes the development of a Relative Spatial Poverty Index (RSPI) derived from multisource geospatial data, including nighttime light intensity (NTL), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and accessibility to Points of Interest (POI). Variable weighting was performed objectively using Principal Component Analysis (PCA) on a 1.5 x 1.5 km spatial grid to estimate poverty levels on Timor Island, East Nusa Tenggara. Furthermore, hotspot analysis using the Getis-Ord Gi method was used to identify poverty clusters, while model validation was performed using simple linear regression. The results showed that the RSPI performed strongly, with a high correlation to official poverty data from the Central Statistics Agency (BPS) (Pearsons Correlation Coefficient = 0.84- R^2 = 0.70) with an RMSE of 19.65. Morans I analysis confirmed the presence of significant positive spatial autocorrelation (Morans I Index = 0.924). The spatial pattern revealed shows a concentration of poverty in rural and urban areas. Overall, this study offers an effective methodological framework for more granular spatial poverty mapping, which can be the basis for formulating more targeted intervention policies.

Keywords: relative poverty, remote sensing, PCA, hotspot mapping

Topic: Topic B: Applications of Remote Sensing