The ensuing analysis passed a major threshold last month.  Surprisingly, the ground data was not strongly correlated to the satellite burn severity data.  This was a surprise given the method’s very good performance on eight fires investigated on National Park Service lands in Alaska.  The regional fire ecologist and regional refuge biometrician recently reviewed all potential sources of errors in the data collection and processing: checking the ground plot selection methods for bias, the ground data for errors, the satellite data for errors, checking all calculations, reviewing the methodology, and even tested one of the fires for observer bias.  None of these potential errors affected the analysis results, suggesting that the poor performance on Refuge lands is “real”. The Fish and Wildlife Service has now begun exploring other potential explanations with partners at Universities and the US Geological Survey (USGS).