Using Endorsement Theory to Validate Annual Burnt Areas of a Large Region in Central Siberia

摘要

Validation of remote sensing products can be very difficult, especially if the study area is large and inaccessible. In this paper, we move away from the more traditional methods of validation, to investigate whether a non-numerical approach can be utilized. Endorsement theory is useful in this context because it allows inference to be made from partial (and incomplete) knowledge, and can represent common knowledge in a natural form without being necessary to translate an experts' opinion into a numerical value. In this study we use endorsement theory to validate a fire disturbance map covering over 3 million km~2 and 12 years using four land cover maps from years 2000 to 2004. The results show that while the method shows potential, as with other validation techniques, the results are dependent on the quality of the reference data.