Table 1 Description of disparate data sets used in the study. The first two rows corresponding to census and survey data are used to create targets for our model. The next three rows, NTL, AOD and L8 are EO datasets used to extract covariates. The last row corresponding to population data is used in getting settlement information and is used in weighting the covariates of EO features. DMSP/OLS/VIIRS refers to the Defense Meteorological Satellite Program/Operational Linescan System (OLS) and the Visible Infrared Imaging Radiometer Suite (VIIRS)
From: Accurate intercensal estimates of energy access to track Sustainable Development Goal 7
Datasets & resolution | Frequency of availability | Source | Raw covariates per pixel per year | Details | Cost for data collection & preparation |
|---|---|---|---|---|---|
Targets | Â | Â | Â | Â | Â |
Senegal Census | Every 10 years | National Statistics Office of Senegal (ANSD) | NA | Energy access indicators | $$$ (USD 29 million) |
DHS Survey Data | 2-5 years | United States Agency for International Development (USAID) | NA | Energy access indicators | $$ (but publicly available) |
EO Data | Â | Â | Â | Â | Â |
Harmonized Night-time Lights (NTL) (1Â km/pixel resolution) | Monthly | Harmonized Nightlight Dataset [38] | Mean NTL averaged annually | NTL data DMSP/OLS/VIIRS | Low/no cost (data exhaust) |
Aerosol Optical Depth (AOD) (1Â km/pixel resolution) | Monthly | Google Earth Engine | Median AOD averaged annually | Blue band (0.47 ÎĽm) AOD over land | Low/no cost(data exhaust) |
Landsat-8 multi-spectral imagery (L8) (30 m/pixel resolution) | Monthly | Google Earth Engine | 512 length vector extracted using a deep neural network [31] | Bands – Red, Green, Blue, Near Infrared (NIR), Shortwave Infrared 1 & 2 (SWIR1/2), Thermal (TEMP1) | Low/no cost (data exhaust) |
Human settlement | Â | Â | Â | Â | Â |
Population estimates (100Â m/pixel resolution) | Yearly | WorldPop [39] | NA | High resolution gridded population | $$ (but publicly available) |