Formulating a Geolocation Bias Correction for DMSP Nighttime Lights of Global Cities

Abstract

This study examined the potential geolocation biases/errors in the Defense Meteorological Satellite Program (DMSP) nightlight data that has been used in carbon emission models, such as the global high-resolution emission inventory Open-source Data Inventory for Anthropogenic CO2 (ODIAC). Quantifying and mitigating the bias has become an urgent, critical task in obtaining robust emission estimates for urban areas, the major sources of the global carbon emissions, from nightlight-based emission estimates. We first attempted to characterize the bias by changing the location of urban cores identified by the DMSP data and their orientation using the city boundary data from the Open Street Map (OMS) as a reference. We hypothesized that a geolocation bias-free emission map should maximize the total emission within the administrative boundaries, and developed an iterative optimization algorithm to obtain correction vectors (distance and angle) for shifting emission data (or DMSP data). We implemented the proposed algorithms for many global cities from different continents, and discovered the latitudinal dependence of the bias. The paper is focused on the methodological aspects of the bias correction, and the implementation and applications to global cities.KeywordsSatellite remote sensing dataNight-time lights dataCarbon dioxide emission dataSatellite data biasGeolocation bias correction