Ichnusa Fire Index: Development and preliminary evaluation at local scale

Fire danger models are commonly used in countries where wildfire occurrences are an important natural threat to forests and wooded areas (e.g., Spain, Canada, Portugal, and Italy). Although fire danger models are calibrated for specific geographic areas, they are often used outside of the calibration area. In addition, no comprehensive statistical evaluation procedures have been applied to fire danger models. In this paper, an integrated fire rating index (Ichnusa Fire Index IFI) developed using wildfire data from Sardinia (Italy), is presented. Depending on the sampling rate, IFI calculation requires meteorological, micrometeorological, topological and ecophysiological data. The IFI index includes (1) the Drought Code (DC), describing water status of plants; (2) the Meteo Code (MC), related to turbulence or weather conditions; (3) the Fuel Code (FC) that accounts density and moisture of vegetation; and (4) the Topological Code (TC), describing orography, slope and site exposure. To determine FC values, the first step defines fuel categories. Based on a vegetation cover map of Sardinia, six fuel types were initially defined as (1) broad leaf forest, (2) cork oak forest, (3) short shrubland, (4) tall shrubland, (5) garigue, and (6) grassland. In this study, the IFI performance was compared with three fire danger indices [i.e., Canadian Fire Weather Index (FWI), Portuguese Index (Port), and Italian Meteorological Danger Index (IMPI)] using data from the period 2002-2004. Index values were calculated for 57 zones of Sardinia corresponding to meteorological station of Agrometeorological Service of Sardinia. All fire events occurring within 20 and 40 km diameter circles around the meteorological stations were considered. Fire potential forecasting performances of all the indices were evaluated determining model capacity to discriminate days with or without fire events using Mahalanobis distance (DM) and the Score 1, Score 2 and Score 3 described by Mandallaz and Ye (1996). The Mahalanobis distance test showed the highest values for IFI for both 20 and 40 km diameters around the stations. Also Scores 1, 2 and 3 calculated for IFI were higher than those computed for the other models. Similar results were obtained when IFI was calculated using weather forecast. These results confirm the good performance of IFI in forecasting fire events in Sardinia. The good performance can partially be attributed to a preliminary study that defined an accurate Fuel Code.