OpenStreetMap has proven to be a great mapping platform, but it’s main goal has always been to create the most recent accurate map. While features are tagged with some date information, there is little effort to align the dates with the actual dates these features are first created, or when they are removed. We wanted to map changes over time to very high accuracy by employing the same approach used by OSM - with high resolution imagery we would map several points in time. But this posed a few challenges - what’s the best way to manage map data from multiple dates, and how do we store the data - we can’t just edit OSM with a bunch of historical information.

So we used our own OSM instance, called osm-seed. First we mapped out our areas of interest in the real OSM so that it contained current information for our regions. Then we identified historical imagery and worked backwards, at each time date deleting, or adding features as needed, until we got back to our earliest date. Finally with our OSM data in hand for multiple dates we were able to calculate statistics over time, such as change road lengths and building footprints.