High-resolution remote sensing data can predict household poverty in pastoral areas, Inner Mongolia, China

Abstract

The accurate prediction of poverty is critical to efforts of poverty reduction, and high-resolution remote sensing (HRRS) data have shown great promise for facilitating such prediction. Accordingly, the present study used HRRS with 1 m resolution and 238 households data to evaluate the utility and optimal scale of HRRS data for predicting household poverty in a grassland region of Inner Mongolia, China. The prediction of household poverty was improved by using remote sensing indicators at multiple scales, instead of indicators at a single scale, and a model that combined indicators from four scales (building land, household, neighborhood, and regional) provided the most accurate prediction of household poverty, with testing and training accuracies of 48.57% and 70.83%, respectively. Furthermore, building area was the most efficient indicator of household poverty. When compared to conducting household surveys, the analysis of HRRS data is a cheaper and more time-efficient method for predicting household poverty and, in this case study, it reduced study time and cost by about 75% and 90%, respectively. This study provides the first evaluation of HRRS data for the prediction of household poverty in pastoral areas and thus provides technical support for the identification of poverty in pastoral areas around the world.

Keywords

• Weighted relative wealth index
• Classification tree
• Inner Mongolia grassland
• Multi-scale