Visible Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) data can be used to measure observed radiance (brightness of imagery pixels) from lights on during the nighttime. While this is not a direct measure of electricity used, it can serve as a proxy given a model linking observed radiance to measured electricity consumption. To accomplish this, monthly nightlight images are fetched and stitched as monthly global nightlight image. The SEZ site boundaries are then used to calculate basic statistics (mean, maximum, minimum, standard deviation) of the radiance for each zone. VIIRS data is relatively low resolution (at 750m per pixel), but this makes scaling any algorithm to multiple zones trivial because only a few pixels must be processed for each zone. Therefore, we can also easily track relative differences of these nighttime lights with high temporal resolution. Due to the high quality of VIIRS data and the fact that it is open, this is one of the easier indicators to calculate.

Conclusions

For estimating electricity consumption using monthly VIIRS Nightlights data was relatively straightforward compared to other indicators. Data from this open dataset is regularly preprocessed by NOAA to account for many of the artifacts including cloud cover, stray light, and lunar illumination. One minor complication, however, is that the monthly data is rather large (consisting of 12 gigabytes per scene) and must be downloaded for the entire globe. As mentioned, this does mean that the methodology used here will scale reasonably well as we can use the same data and processing pipeline for any number of zones. Another obstacle is that the observed radiance is not directly correlated with electricity consumption. In the future, we hope to find more accurate ways of relating radiance from a region with electricity used by exploring existing models from research.