Multi-sourced data modelling of spatially heterogenous life-cycle carbon mitigation from installed rooftop photovoltaics: A case study in Singapore

Accurately quantifying carbon mitigation of operational photovoltaics (PVs) influenced by dynamic geo-environment is crucial for developing suitable initiatives on renewable energy transition. However, previous studies made strong assumptions to avoid modelling spatial heterogeneity of carbon footprints or ignore weather and shadowing effects on PV electricity generation, making the estimated results unreliable and even causing false policymaking. To tackle this challenge, we developed a novel model coupling multi-sourced data modelling and life-cycle assessment to estimate spatially heterogenous carbon mitigation of all the operational rooftop PVs in an entire city. It is built by three hierarchal modules: (i) segmenting PV areas from high-resolution satellite imagery, by using Deep Solar PV Refiner, an advanced semantic segmentation network; (ii) estimating electricity generation in the segmented PV areas, by using a well-developed 3D solar irradiation model that considers the effects of land surface solar irradiation influenced by weather and shadowing effects produced by 3D buildings; (iii) quantifying carbon mitigation potential of PVs, by developing a spatial-aware life-cycle model to track the life-cycle carbon footprints of PVs from production, transportation, operation, to decommission. Investigating Singapore by 2020, we reveal that industrial, airport, and residential areas have the largest rooftop PV installation. We also suggest a carbon emission rate of 13.20 g-CO2/kWh, a carbon payback time of 0.81 years, and an energy payback time of 0.94 years, showing an improved carbon mitigation capability compared to the past years. This study contributes to GIS data modelling and helps understand the geospatial characteristics of urban-scale PV carbon mitigation.