The upward trend of burnt area is shown below.
figure 1: On July 28th (red) and log linear regression fitting (gray), the prediction standard error (light purple small x) was made according to the logarithmic first-order difference regression 2012-18 (purple x) and plus/minus 1. source: NIFC1, NIFC2 , And the author’s calculation.
If you want to use year-to-date numbers to extrapolate the severity of the upcoming season (at least A commentator – This The same person also has a question about the confidence interval), a person may make a serious mistake. All people have to do is to absorb what is often mentioned, that the fire season is expanding so that it lasts throughout the year. The trend of discrepancies between the annual and annual data—that is, the gap between the black line and the red line—is consistent with this point.
How to use this fact to deal with the burned 2021 acres of land? It is instructive to check the first difference in logging variables.
figure 2: The first difference between the number of log acres burned on July 28 (red) and year-to-date (YTD) (black). source: NIFC1, NIFC2 , And the author’s calculation.
The correlation is broken down in 2019 and 2020, but overall it seems to be strong. I used this correlation between 2012-18.
The regression equation is:
d (log (acre)) = 0.046 + 0.894×d(log(ACRES_YTD))
bold Indicates the significance of using HAC robust standard error at 1% msl. Adjustment-R2 = 0.86. DW = 2.45, n=7.
Therefore, even if you do not believe in any regression results, you should be cautious when using YTD numbers to extrapolate the final annual numbers early in the fire season, especially in the new climate normal I think we are in.
