Reconstruction of fire spread within wildland fire events in Northern Eurasia from the MODIS active fire product

Russian boreal forests have been reshaped by wildland fire for millennia. While fire is a natural component of boreal ecosystems, it impacts various aspects of the environment and affects human well-being. Often fires occur over large remote areas with limited access, which makes their ground-based observation difficult. A significant progress has been made in mapping burned area from satellite imagery, which provides consistent and fairly unbiased estimates of fire impact on areas of interest at multiple scales. Although the information provided by burned area products is highly important, the spatio-temporal dynamics of individual fire events and their impact are less known. In high northern latitudes of Northern Eurasia, MODIS (Moderate Resolution Imaging Spectroradiometer) makes up to four daily observations from each of the Terra and Aqua satellites providing consistent data on fire development with high temporal frequency. Here we introduce an approach to reconstruct the development of fire events based on active fire detections from MODIS. Fire Spread Reconstruction (FSR) provides a means for characterization of fire occurrence over large territories from remotely sensed data. Individual fire detections are clustered within a GIS environment based on a set of rules determining proximity between fire observations in space and time. FSR determines the number of fire events, their approximate size, duration, and fire spread rate and allows for the analysis of fire occurrence and spread as a function of vegetation, fire season, fire weather and other parameters. FSR clusters were compared to burned scars mapped from Landsat7/ETM+ imagery over Yakutia (Russia). While some smaller burn scars were found to be formed through a continuous burning of a single fire event, large burned areas in Siberia were created by a constellation of fire events incorporating over 100 individual fire clusters. Geographic regions were found to have a stronger influence on the rates of fire activity in the area compared to vegetation zones. In addition, fire spread rates do not directly correlate with the intensity of a given fire season. FSR is also used to identify the points of ignition for individual fire events in spatio-temporal domain for fire danger and fire threat modeling. This approach presents another step towards the more complete characterization of fire events from remotely sensed data.