A classification scheme to determine wildfires from the satellite record in the cool grasslands of southern Canada: Considerations for fire occurrence modelling and warning criteria

Abstract

Daily polar-orbiting satellite MODIS thermal detections since 2002 were used as the baseline for quantifying wildfire activity in the mixed grass and agricultural lands of southernmost central Canada. This satellite thermal detection record includes both the responsible use of fire (e.g. for clearing crop residues, grassland ecosystem management, and traditional burning) and wildfires in grasslands and agricultural lands that pose a risk to communities and other values. A database of known wildfire evacuations and fires otherwise requiring suppression assistance from provincial forest fire agencies was used to train a model that classified satellite fire detections based on weather, seasonality, and other environmental conditions. A separate dataset of high resolution (Landsat 8 thermal anomalies) of responsible agricultural fire use (e.g. crop residue burning) was collected and used to train the classification model to the converse. Key common attributes of wildfires in the region included occurrence on or before the first week of May with high rates of grass curing, wind speeds over 30 km h-1, relative humidity values typically below 40 %, and fires that are detected in the mid-afternoon or evening. Overall, grassland wildfire is found to be restricted to a small number of days per year, allowing for the future development of public awareness and warning systems targeted to the identified subset of weather and phenological conditions.