Southeast Australia fire weather reconstruction using 2k Law Dome summer sea salt

This dataset contains data used to create Figure 5 in Udy et al., 2024 - Australia’s 2019/20 Black Summer fire weather exceptionally rare over the last 2000 years - DOI : 10.1038/s43247-024-01470-zThe 2000-year Law Dome summer sea salt (LDsss) record was used to reconstruct area average seasonal (November-February) FFDI over eastern Australia. This region incorporates the coastal and alpine regions of southeast Australia that were severely impacted by the 2019-20 Black Summer fires. The fire season for this region is generally between October to March. A linear regression model was trained over the 1960-1990 period and validated over 1994 – 2016 using observational data provided by the Bureau of Meteorology. The predict function in R with the trained linear model and the 95% prediction interval was used to calculate the 2000-year reconstruction using the LDsss values. The prediction interval reflects the prediction uncertainty around a single value in the reconstruction. All observed FFDI values in the calibration period (1960-1990) are within the 95% prediction interval. In the validation period (1994-2016), the observed FFDI values are within the 95% prediction interval for 21 of the 23 seasons. Two elevated fire weather seasons in the validation period (2002/03, 2006/07) are underestimated by reconstruction uncertainty (i.e. the upper prediction interval is less than the observed FFDI). The model fit was assessed by comparing the mean reconstruction and observations for the calibration, validation, 1960-2016 and 1951-2016 periods using Pearson and Spearman correlation score (r), adjusted r2, root mean square error (RMSE) and Nash-Sutcliffe model efficiency coefficient (NSE). RMSE is an absolute measure of the fit representing the variance of the model errors, with smaller values indicating a better fit. An NSE value greater than 0 indicates the reconstruction performs better than the climatology of the reconstructed period.Refer to the published paper for more details.