Rooftop solar enables carbon-neutral transition in Hong Kong's building sector

Data Description 1. High-resolution (0.5m) Global Solar Radiation Dataset Annual_GSR_final.tif is the pixel-scale annual global solar radiation (GSR) for the building rooftops in Hong Kong, with a spatial resolution of 0.5m, in GeoTIFF format. It is derived by our developed machine learning model, which has considered the solar attenuation due to roof geometry, building shadow, and cloud attenuation effect. 2. Building-scale Rooftop Solar Energy Potential (SEP) Dataset Building_Scale_Annual_SEP.geojson.json contains the georeferenced (WGS-84) building footprint polygons with ancillary data, including the estimated annual SEP, financial payback periods, and carbon payback periods for each building rooftops in Hong Kong. 3. District-scale Rooftop SEP Dataset District_Scale_Annual_SEP.geojson.json contains the geo-referenced (WGS-84) administrative district boundary in Hong Kong with ancillary data, including the absolute annual SEP and normalized annual SEP for each district. 4. Building-scale Decarbonization Assessment Dataset Decarbonization_Analysis_Commercial_Buildings.geojson contains the geo-referenced (WGS-84) point feature for the commercial buildings in Hong Kong with ancillary data, including the building-level annual energy consumption and percentage reduction in annual energy consumption. 5. Pre-trained Machine Learning Model Pretrained_RF_Model.joblib is our pre-trained machine learning model. The model requires daily clear-sky GSR and daily cloud optical thickness datasets in GeoTIFF format as a model input to estimate the daily cloud-attenuated GSR. The sample testing can be retrieved via our GitHub repository at https://github.com/Jimmy-Tong/Rooftop_SEP/. 6. Web Application for Visualizing Multi-scale Annual SEP and High-resolution (0.5m) GSR Datasets https://Jimmy_Tong_RooftopSEP.users.earthengine.app/view/sepgsr is the web application for visualizing our annual SEP and high-resolution GSR Datasets. Even without specific GIS software and license, the web users can still visualize our map-based research findings.