Comparisons of fire weather indices using Canadian raw and homogenized weather data

Modifications to the environment around a weather station or changes in instrument result in discontinuities or shift in weather data. This paper asks the often ignored questions such as, “what are the impacts of inhomogenized data?” and “does using homogenized weather data affect the conclusions of environmental research?” To answer these questions, we used the Canadian Forest Fire Weather Index (CFFWI) System for our studies. Weather station data are used to calculate wildfire danger indices. The homogenized data and raw (inhomogenized) observations for sixteen weather stations spread across Canada were used to calculate the CFFWI indices. The sixteen weather stations were further divided into three subset of stations based on the length of the accessible data during the fire season (April to end of September). The first set included stations that covered just 27 years of data and the second data sets had 49 years of data, while the third set included only five stations with the longest time period of 66 years. The majority of the stations, as measured by the Wilcoxon signed-rank test, rejected the null hypothesis (difference between the pairs follows a symmetric distribution around zero). The rejection rate increases to 100% as the length of data record increases from 27 to 66 years. Homogenization of 66 years data reduced the indices values approximately 0.7-8.4% and also reversed the long-term trend in some stations such as Kapuskasing.