Solar resource density (solar panel area per unit area of sq m/sq mile) at state, county and census tract levels, with examples of detected solar panels. Source: Stanford University

Where exactly are the solar photovoltaic installations in the contiguous U.S.? Researchers and citizen scientists can delve into DeepSolar for such information. The publicly accessible database was developed at Stanford University by training a machine learning algorithm to analyze 370,000 satellite images, each labelled as either having or not having a solar panel present.

The system identified color, texture and size features associated with solar panels. After DeepSolar correctly identified images with solar panels 93% of the time, the system was applied to the analysis of a billion satellite images, a task completed in one month.

The group then had DeepSolar analyze the billion satellite images to find solar installations — work that would have taken existing technology years to complete. The inventory contains the locations of residential, commercial and utility-scale photovoltaic systems. A total of 1.47 million installations was identified.

The geographic data were combined with U.S. Census data to explore socioeconomic impediments to and motivating factors behind solar power adoption. Low- and medium-income households do not often install solar systems even when utility bill savings would exceed the monthly cost of the equipment. The influence of household income is a factor up to a point; above $150,000 a year, income ceases to play much of a role in user’s decisions.

The research is published in Joule.