Building Integral Gridded Carbon Emission Disaggregating Model (BIGCarbonEDM): Near real-time community-level CO2 emission evaluations for twelve cities in Egypt, South Africa, and Turkey

Cities generate the majority of CO2 emissions, the largest climate change contributor. Near real-time CO2 emission monitoring and modeling at relatively high resolutions are beneficial to fill the knowledge gaps for the spatial and temporal emission patterns in different regions, and to provide the public and the policymakers with accurate and timely information on major emission sources and emission amounts for better public awareness and decision making. On the other hand, there are even larger knowledge and information shortage in Global South counties for the city-/community- level CO2 emissions due to the lack of well-monitored official data with high latency and low transparency.

Data-driven models with big data embedded can be one of the most robust and efficient approaches for addressing those challenges in regions without well-documented data. Therefore, we developed the Building Integral Gridded Carbon Emission Disaggregating Model (BIGCarbonEDM) that disaggregates Scope 1&2 CO2 emissions to community-level (with a resolution at 500 meters) using machine learning models trained with building- to regional-level features. Multiple open datasets were used as model inputs, 1) building-level datasets: Microsoft Building Footprints, OpenStreetMap, and OpenStreetMap Building, 2) regional-level datasets: Global Human Settlement, World Settlement Footprint, VIIRS Nightlights, Local Climate Zone, Copernicus Digital Elevation, and land surface temperature, 3) economical and census datasets were collected from the national statistical report and world bank surveys, and 4）city-level near real-time CO2 emission: Carbon Monitor City (https://cities.carbonmonitor.org/), one of our previous projects.

BIGCarbonEDM for the first time proposed the approach for capturing the emission patterns for the community level based on building-level and regional-level features and provides the near real-time CO2 emission for twelve major cities in Egypt, South Africa, and Turkey. The cities are Adana, Trabzon, Ordu, and Manisa for Turkey; Cairo, Alexandria, Luxor, and Sheikh Zayed for Egypt; and Johannesburg, Tshwane, Ekurhuleni, and eThekwini for South Africa. BIGCarbonEDM was designed as a modular platform including modules for data collection, data fusion, spatial and temporal emission feature learning, emission estimation model, and data visualization. BIGCarbonEDM modules can be updated and modified separately, which simplifies the improvements and extensions of the final delivered dataset, also all the individual modules can be used by the research community for other relevant research.