Back in the early 1990's we were asked to bid on a small satellite network to do this while I was working at AeroAstro. Our CTO and CEO ran the numbers and figured out it would be far more economical to build ground based fire sensors that would report via pager or cell data and distribute those instead. We did not win that proposal.
Fast forward to today's interest in IR sensors for tracking things of military interest and it's actually a big deal to try to figure out how to identify the things you care about instead of the small fires (campfires) scattered around the world. So, if the military were so inclined, the same global system that watches for militarily interesting things could be used to spot forest fires starting. In fact, with the modern data systems being proposed, it would just be a matter of designing a database query to periodically scrape the central database for small signals that grow past a certain limit to forward to the appropriate forestry agency.