• Energy Research

Fire danger indices are descriptors of fire potential in a large area, and combine a few variables that affect the initiation, spread and control of forest fires. The Canadian Fire Weather Index (FWI) is one of the most widely used fire danger indices in the world, and it is built upon instantaneous values of temperature, relative humidity and wind velocity at noon, together with 24 hourly accumulated precipitation. However, the scarcity of appropriate data has motivated the use of daily mean values as surrogates of the instantaneous ones in several studies that aimed to assess the impact of global warming on fire. In this paper we test the sensitivity of FWI values to both instantaneous and daily mean values, analyzing their effect on mean seasonal fire danger (seasonal severity rating, SSR) and extreme fire danger conditions (90th percentile, FWI90, and FWI>30, FOT30), with a special focus on its influence in climate change impact studies. To this aim, we analyzed reanalysis and regional climate model (RCM) simulations, and compared the resulting instantaneous and daily mean versions both in the present climate and in a future scenario. In particular, we were interested in determining the effect of these datasets on the projected changes obtained for the mean and extreme seasonal fire danger conditions in future climate scenarios, as given by a RCM. Overall, our results warn against the use of daily mean data for the computation of present and future fire danger conditions. Daily mean data lead to systematic negative biases of fire danger calculations. Although the mean seasonal fire danger indices might be corrected to compensate for this bias, fire danger extremes (FWI90 and specially FOT30) cannot be reliably transformed to accommodate the spatial pattern and magnitude of their respective instantaneous versions, leading to inconsistent results when projected into the future. As a result, we advocate caution when using daily mean data and strongly recommend the application of the standard definition for its calculation as closely as possible. Threshold-dependent indices derived from FWI are not reliably represented by the daily mean version and thus can neither be applied for the estimation of future fire danger season length and severity, nor for the estimation of future extreme events. The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement 243888 (FUME Project). J.F. acknowledges nancial support from the Spanish R&D&I programme through grant CGL2010-22158-C02 (CORWES project). The ESCENA project (200800050084265) of the Spanish \Strategic action on energy and climate change" provided the WRF RCM simulation used in this study. We acknowledge three anonymous referees for their useful comments that helped to improve the original manuscript.

Despite regional warming, fire activity is decreasing in the Mediterranean region, blurring the well-established relationship between climate and wildfires. Here, we analyzed this relationship by focusing on the fire severity component of the fire regime. We determined the temporal trends of several climate, fire activity, and fire severity variables and the relationship of the latter two to the first in West-Central Spain (30,000 km2) for a 33 year period (1985 to 2017). Annually, fire variables at summer season were number of fires, burned area, fire size and fire severity (calculated using the relativized burn ratio (RBR) from Landsat satellite images). Fire severity was estimated for the whole area and for each of the main land use/land cover (LULC) types. Finally, the climate variables were maximum temperature, precipitation, and water deficit for all seasons (winter, spring, summer, and fall). Trends in those variables were assessed using the Mann–Kendal test, and the relationship between climate and fire variables was ascertained using autoregressive moving average (ARMAX) models. Main results indicated that number of fires and burned areas decreased, whereas drought conditions increased. Wildfires tended to burn preferentially in treeless areas, with conifer forests burning less frequently, and shrublands burning more so. Median RBR increased, as well as low (P5) and high (P90) percentiles. The percentage of burned areas at low severity decreased. All LULC types tended to burn at higher fire severities over time. The decreasing fire activity, but with increasing fire severity, coincides with rising maximum temperatures and drought (lower precipitation and higher water deficit). The temporal dynamics of fire activity and severity were well explained and predicted by spring and summer climate variables. Thus, while fire activity decreased, fire severity increased, driven by a more severe climate that was consistent with regional warming.

We present future fire danger scenarios for the countries bordering the Mediterranean areas of Europe and north Africa building on a multi-model ensemble of state-of-the-art regional climate projections from the EU-funded project ENSEMBLES. Fire danger is estimated using the Canadian Forest Fire Weather Index (FWI) System and a related set of indices. To overcome some of the limitations of ENSEMBLES data for their application on the FWI System?recently highlighted in a previous study by Herrera et al. (Clim Chang 118:827?840, 2013)?we used an optimal proxy variable combination. A robust assessment of future fire danger projections is undertaken by disentangling the climate change signal from the uncertainty derived from the multi-model ensemble, unveiling a positive signal of fire danger potential over large areas of the Mediterranean. The increase in the fire danger signal is accentuated towards the latest part of the transient period, thus pointing to an elevated fire potential in the region with time. The fire-climate links under present and future conditions are further discussed building upon observed climate data and burned area records along a representative climatic gradient within the study region. The research leading to these results has received funding from the EXTREMBLES project (CGL2010-21869) funded by the Spanish R&D programme and from the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement 243888 (FUME Project). The authors acknowledge the RCM data sets from the EU-FP6 project ENSEMBLES (http://ensemblesrt3.dmi.dk) and would also like to thank Erik van Meijgaard from the Royal Netherlands Meteorological Institute (KNMI) for making available ENSEMBLES RACMO2 climate model output verifying at 12:00 UTC. We are also grateful to Jesus Fernandez and three anonymous reviewers for their insightful comments that greatly contributed to the improvement of the original manuscript.

Abstract. Among the different meteorological hazards, droughts are those with the highest socio-economical impact on the Iberian Peninsula. Drought events have been largely studied in the instrumental period, but very little is known about the characteristics of droughts in the preinstrumental period. In this work, several series of rogation ceremonies are used to identify severe droughts within the period 1750–1850. The overlapping of the rogation series with some instrumental series served to identify some climatic characteristics of rogation ceremonies: (a) during spring, rainfall deficits needed to celebrate rogation ceremonies are smaller than in any other season; (b) the hydrological deficit in a particular region increases with the number of locations celebrating rogations simultaneously. On the other hand, it was found that between 1750–1754 and 1779–1783 are probably the driest periods of the 101 analyzed years. Both show an important number of rogations all over Iberia and during all the seasons. The most extended drought of this period occurred during the spring of 1817, affecting 15 of the 16 locations studied. This drought was influenced by the Tambora eruption (1815). The study of the climate footprint of this eruption and its comparison with similar situations in the series suggest that the spring drought of 1824 may be associated with the eruptions of the Galunggung and Usu volcanoes (1822). Further studies are required to confirm this fact and understand the atmospheric mechanisms involved.