Burned area and climate extremes in different land covers in southeastern Australia

Large burned areas (BA) in southeastern Australia were regularly registered during hot and dry years, such as the Black Saturday (2009) and the Black Summer (2019-2020) extreme bushfires. These types of extreme climate conditions are expected to become more frequent, leading to an increased risk of large BA in this region. In this work, the influence of drought conditions and hot events on the BA in southeastern Australia was assessed, using correlation and copula functions. Bivariate copula functions were fitted, and conditional probabilities of large BA given climate extremes were computed. Three classes of drought intensity were studied, namely moderate, severe, and extreme, as well as three thresholds for temperature extremes, namely the 80th, 90th, and 95th percentiles. Monthly BA were computed as the sum of the burned pixels in the fire season (from October to March), using data from the MODIS Burned Area product. The analysis was performed on forests, grasslands, and savannas separately. Drought conditions were assessed with SPEI at several time scales, computed with data from the CRU TS4.07 dataset. Maximum and minimum daily temperature were retrieved from the ERA5 dataset. Results showed that the correlation between BA and SPEI was high in the current and previous 1 month for all land covers, being highest in savannas and lowest in grasslands. Short time scales of SPEI had the highest correlation on grasslands, and the opposite was observed in forests and savannas. The correlation with maximum temperature increased until 10-15 days before the fire event and surpassed 0.6 over forests. Minimum temperature presented much lower correlations and there was not a pronounced increase in the previous days, as observed with the maximum temperature. The conditional probability of large BA increased with the intensity of the drought on all land covers, and it reached almost 100% probability of exceeding the 50th percentile of BA under extreme droughts on forests and savannas. For the case of the 80th percentile of BA, the probability was lower, but the difference given drought and non-drought conditions was larger than for the 50th percentile. On savannas and forests, the conditional probability was still high when considering SPEI in the previous 2 and 3 months. Maximum temperature yielded a higher probability of BA for the two highest percentiles. Savannas presented the lowest probability of BA given hot events, and forests the highest. The probability increased up to 10 days before the fire. Overall, the probabilities obtained given drought conditions are higher than given hot events, particularly for larger fires. Moreover, high probabilities obtained with large time scales and longer lead times are indicative of the importance of drought conditions before the fire season and may help predict the occurrence of large BA.   Acknowledgments: This study was partially supported by FCT (Fundação para a Ciência e Tecnologia, Portugal) through national funds (PIDDAC) – UIDB/50019/2020, by project Floresta Limpa (PCIF/MOG/0161/2019), and by project 2021 FirEUrisk, funded by European Union’s Horizon 2020 research and innovation programme under the Grant Agreement no. 101003890). A.R. was supported by FCT through https://doi.org/10.54499/2022.01167.CEECIND/CP1722/CT0006.