The Detection Of Burned Forest Area Using Annual Metrics Constructed From MODIS Data In Manitoba Of Canada

The monitoring of burned forest area is used to evaluate forest fire carbon release and forest recovery and can provide data in support of the evolution of global carbon sources and sinks. Remote sensing satellites, which can repeatedly observe the earth at large scale, are the main means for now to run the forest burned area monitoring.A new method for mapping annual burned area using four types of Metrics constructed from Moderate Resolution Imaging Spectroradiometer (MODIS) data for Manitoba was developed for the2003-2007period. The method combined the following:1) four types of Metric that are not sensitive to months in which fire happened;2) the TreeNet software with the SGB algorithm to build the forest burned area predict model with the annual universal characteristics; and3) the best grow threshold and cutoff threshold ascertained by balancing Omission and Commission. In the end, Canada National Fire Database(CNFDB)、Landsat/TM data and MODIS Burned Area product MCD45A1were used to evaluate and validate the algorithm and the product.The result showed that:(1) Metric constructed from MODIS data that are not sensitive to months in which fire happened is the key to build the annual universal forest burned area predict model, which can detect fire events happened in the year of training data and other years. Meanwhile, multi-temporal Metric can retain or enhance the spectral information while removing the effects of clouds and other noise in single-date imagery.(2) The predicted burned area results match well with MODIS Fire Product MCD45A1in spatial location. The predicted burned area and CNFDB were similar for the same area and correlated well for large burned scars (r2=0.913). A comparison with TM images indicated that this mapping product was a significant improvement relative to the CNFDB burned area map product.(3) Fire events from2003to2007are mainly distributed from54°to57°N,96°to99°W in Manitoba, and it has growing and fading distribution characteristics.