Methods
Data Collection
The study site for this project covers the whole of Canada with the exception of Prince Edward Island. The data used to analyze the forest fires came from Natural Resources Canada's National Burned Area Composite (Fig 3), which was provided by Natural Resources Canada, and Provincial, Territorial, and Parks Canada agencies. Climate data has been generated with the ClimateNA v6.40b software package, available at http://tinyurl.com/ClimateNA, based on the methodology described by Wang et al. (2016). In order to analyze the climate change of the fire site more concretely, I reprojected the raw fire data into a form with a coordinate system, extracted the year, latitude and longitude, and obtained the climate change of the relevant year through the ClimateNA v6 software (Fig 4). For the summary of climate data, I first used ClimateNA v6.40b software to extract three sets of data, namely the current year data, that is, the current year climate data from 1986 to 2021. Previous data, i.e. data from the previous year. As well as normal data, i.e. data from 1961 to 1990. The base map used to describe the ecoregion where the fire occurred was also taken from Dr. Andreas Hamann's laboratory cloud file.
Statistical Analysis
Use R 4.1 as the main tool for statistical analysis, we were able to further analyze and explore the dataset.
The fire data is from 1986 to 2021, I try to simplify the 36 years of data by merging adjacent polygons. Due to the different amount of forest in the provinces or territories, I tried converting it to the same unit. Climate change is not weather change, it will be considered to reflect very long-term trends. We need to find out whether the climate conditions in the month before the fire month have an impact on the climate conditions in the fire month. So, to establish the connection between the month of fire occurrence and previous climate conditions (the average temperature, Hargreaves climate moisture deficit, Hogg's climate moisture index, maximum average temperature, and precipitation), I first calculated the difference between the current year's climate data and the normal year's climate data, and the difference between the previous year's climate data and the normal year's climate data, respectively. If the difference is positive, it means that the climatic conditions of the year are higher than those of a normal year; if the difference is negative, it means that the climatic conditions of the year are lower than those of a normal year. Then I extract the months that have the fire month records, and count the climate conditions of the fire month and the five months before this month. Then, I will use the ggplot() function to obtain a histogram to express the relationship between the anomalous temperature and the number of fire occurrences. The horizontal axis represents how the climate index deviates from 1961-1990 normal conditions in the fire month and the previous 5 months, and the vertical axis represents the number of fire occurrences. And in order to get the number of large-scale forest fires in each ecozone, I defined the fire area as a large-scale fire with a polygon area of more than 50,000 hectares to check the fire-occurring area. Finally, I summarized the relationship between the difference of each climate parameter and the increase in each month with the year. And in order to discuss the change of climate parameters more reasonably, through ggpubr, I created scatterplots using ggscatter for our correlation analysis. In addition, to plots, I have also added a regression line to demonstrate the slope between the relationship between the x and y variable.