10:00 〜 10:15
[ACG34-05] Analyses of the wildfire weather conditions in world's 14 regions for 20 years
キーワード：林野火災、衛星データ、地球システムモデル、GCOM-C/SGLI
The IPCC AR6 predicts that forest fires will increase in almost all tundra and boreal areas and some mountainous regions by the end of the century due to increased droughts and heat waves associated with global warming (IPCC, 2022). The UNEP Fire Risk Report (2022) also reports that even under the lowest emissions scenario, the probability of large fires will increase by 30-50% during this century (Sullivan, 2022). However, it has also been suggested that current climate models do not adequately account for the effects of material emissions, carbon cycling, and other factors caused by forest fires (Arora et al., 2020).
In this study, as an initial step toward building a fire model, we conducted a statistical analysis of forest fire occurrence and environmental conditions in the area by using satellite data and objective analysis data.
The relationship between the area burned and inter-annual changes in meteorological variables (temperature, specific humidity, precipitation, solar radiation, wind speed) was investigated. First, we used 20 years (1997-2016) of GFED (Global Fire Emissions Database) data on the area burned to create averages for each month and global anomaly data for each month. Global anomaly data for 20 years were also created in the same manner for five meteorological variables from Today's Earth (JAXA's land surface & river simulation system) : temperature, specific humidity, precipitation, solar radiation, and wind speed (all spatial resolutions were adjusted to 0.5 deg.). Using these data, scatter plots comparing the anomalies of burned area and meteorological variables for each of the 14 regions defined by climate, cover, etc (GFED definition).
The results showed that increases in temperature and solar radiation, and decreases in specific humidity and precipitation contributed to increases in burned area in many areas, which is consistent with previous studies (Schroeder et al., 1964; Crimmins, 2006). Furthermore, the magnitudes of the dependence of burned area on meteorological variables are different by region. These results suggest the need to take into account differences in regional characteristics such as cover and climate when building fire occurrence models.
In the future, we plan to utilize this information for estimating fire variables such as burned area by machine learning using GCOM-C/SGLI, etc., which can observe the entire globe at a higher resolution and frequency.
In this study, as an initial step toward building a fire model, we conducted a statistical analysis of forest fire occurrence and environmental conditions in the area by using satellite data and objective analysis data.
The relationship between the area burned and inter-annual changes in meteorological variables (temperature, specific humidity, precipitation, solar radiation, wind speed) was investigated. First, we used 20 years (1997-2016) of GFED (Global Fire Emissions Database) data on the area burned to create averages for each month and global anomaly data for each month. Global anomaly data for 20 years were also created in the same manner for five meteorological variables from Today's Earth (JAXA's land surface & river simulation system) : temperature, specific humidity, precipitation, solar radiation, and wind speed (all spatial resolutions were adjusted to 0.5 deg.). Using these data, scatter plots comparing the anomalies of burned area and meteorological variables for each of the 14 regions defined by climate, cover, etc (GFED definition).
The results showed that increases in temperature and solar radiation, and decreases in specific humidity and precipitation contributed to increases in burned area in many areas, which is consistent with previous studies (Schroeder et al., 1964; Crimmins, 2006). Furthermore, the magnitudes of the dependence of burned area on meteorological variables are different by region. These results suggest the need to take into account differences in regional characteristics such as cover and climate when building fire occurrence models.
In the future, we plan to utilize this information for estimating fire variables such as burned area by machine learning using GCOM-C/SGLI, etc., which can observe the entire globe at a higher resolution and frequency.