Projections of future meteorological forest fire danger in the European Alps

In this study we simulated the impacts of a changing climate on the meteorological forest fire danger in the Alpine area. The analysis was based on simulations from a regional climate model and subsequent statistical downscaling. The model simulations were performed with the regional numerical model COSMO-CLM which was driven by the global model ECHAM 5 (scenario A1B). The grid size of the regional climate model was 0.2° (~ 18km). Due to the very complex topography in the investigation area (altitudes range from 0m to 4800m asl.) it was necessary to post-process the model output. Therefore we applied a relatively simple downscaling method which is only adequate for climatological purposes. In this technique the direct model output of the COSMO-CLM model was compared with observations at selected stations (25 across the Alps) in the control period 1991 – 2010. The stations were chosen to represent different climate regions of the Alps. Average deviations between the meteorological stations and the respective 4 surrounding gridpoints of the model were calculated separately for each month (Jan. – Dec.) and for 4 times a day (00, 06, 12, 18 UTC). The deviations were then applied to the future scenario runs of the model. In this method we supposed that the climatological deviations between the grid points and the stations would not change under future climate. This procedure was performed for temperature, precipitation, relative humidity and wind speed. Through the application of this downscaling method, the bias of the regional climate model was corrected. The output comprised future climate projections at the 25 selected stations. To assess future meteorological forest fire danger we calculated different forest fire danger indices (FWI and sub-indices of the Canadian Forest Fire Danger rating system (CFFDRS), Baumgartner, M-68, Nesterov, Angstrom, McArthur) with the meteorological parameters from the downscaled model output. The time period under consideration spanned the 60 years between 1991 and 2050. The statistical analysis was based on a simple percentile analysis. Therefore we divided the dataset into single years and computed index values at certain percentiles (e.g. 50th, 75th, 90th, 95th) for each year. To assess a potential temporal trend we examined for statistically significant changes of the percentile values within the scenario period. With this method it was possible to differentiate between the trends of low and high index values. Additionally, other statistical summaries such as seasonal averages of fire danger or number of days above a locally adapted threshold were also derived from the dataset. Initial results showed that climate change is likely to have different impacts on potential forest fire danger in the Alpine area. There was hardly any temporal trend of fire danger evident north of the Alps. This was mainly due to the fact that beside higher future temperatures, the climate model also predicted an increase of precipitation amount and frequencies. In contrast, model projections for the regions south of the Alps showed a significant trend towards more events with extreme meteorological fire danger because of higher temperatures, less precipitation and longer dry spells.