Abstract:
Periodic electoral redistricting is vital in a representative system of government. It uses updated census data to ensure fair voting power, eliminating population biases among districts. Consistent census frequency is a costly endeavor, often unattainable for developing countries. Consequently, it opens the door to gerrymandering, where electoral districts are strategically arranged to benefit specific candidates. This study proposes a cost-effective alternative for estimating updated census data by leveraging previously available census data and establishing its correlation with built-up areas through medium resolution remote sensing datasets. Firstly, an integrated approach of object-based and pixel-based image analysis, using Sentinel data, was employed to extract Built-Up Areas (BUA). This BUA extraction served as the foundational footprint for extracting BUA in the census year using Landsat data. The findings demonstrated the effectiveness of integrating census and remote sensing data, capturing 97% of population growth while depicting its spatial growth pattern. The developed relationship was utilized to perform a pixel-level disaggregation of the population which proved instrumental in electoral redistricting up to an accuracy of -1.7 to 4.1%. Additionally, this paper examined the impact of different hierarchical levels of aggregated census data. The analysis revealed a variation of 3.06% at the district level, ranging from -0.53% to 14.94% at the sub-district level (Tehsils), and a wider range of -21% to 41% at the local district level (Union Councils). By utilizing pixel-level population data, multiple parcels can be generated in accordance with the current redistricting guidelines. This approach enables efficient and quick review of existing electoral boundaries.
