Early Warning Wildfire System

Springtime burning is a significant threat to forests in Alberta. The majority of fires in Alberta are started by humans, and the frequency of those fires peaks around May 1 each year. We suspected that this spike in human-caused fire risk is the result of what is known as the spring burning window; a time when snow cover melts exposing dry, deciduous vegetation that does not yet have leaves. This time of year is critical for fire ignitions because sustained ignitions require relatively dry and combustible fine fuels, which are readily available in the spring.

We designed a system that uses the MODIS MOD09Q1 Version 6 product to track changes in foliar health from 2000 to 2016. We calculated the Normalized Difference Vegetation Index (NDVI) from the 8-day surface reflectance composites in order to generate intra-annual time series of vegetation greening. We then calculated the day-of-year of green-up using the date when NDVI reached half of the maximum value for the season.

Green-up dates varied spatially and from year-to-year across the province of Alberta. In order to develop an early warning system, we simulated incomplete time series by starting at the beginning of the year and iteratively adding images to simulate the availability of satellite imagery. At each time step, we modelled the green-up date using a thin plate spline from the available data and the average NDVI values for all remaining dates in the series. Through this innovative process, we were able to predict the green-up date using imagery up to March 30th for the majority of fire-protected forests in Alberta, which was accurate to within 7 days.

Read the article in Nature Scientific Reports.