Introduction 

Heat-wave days cause adverse impacts on forests and economic activities ([21], [22], [40]). In addition, they play a key role on human health and death ([14]). The last 13 years (2000-2012) were characterized by frequent heat waves, which often triggered the occurrence of large wildfires ([5], [22], [8], [40]) in the Euro-Mediterranean region (Italy and Greece, 2007; Portugal, 2003 and 2005; Spain, 2006 and 2009) and overall the world (Australia, 1983 and 2009; Canada, 2004; Russia, 2010; USA, 2000, 2006 and 2007). In hot days, usually associated with very low fine dead fuel moisture content, the ignition probability is higher and wildland fire behavior could be extreme (e.g., increasing flame length, rate of spread, crown fire activity, and spotting activity). Therefore, fires can burn rapidly and intensely, and originate large and severe wildland fires difficult to extinguish, exceeding the fire-fighting capabilities ([23], [35]).

Climate change projections for the western Mediterranean Basin show a greater variability in weather conditions and an increase in extreme weather events, mainly in the summer season, when longer and more frequent heat waves are expected to happen ([4], [5], [26], [30]). This will result in an increase in wildfire activity ([4], [16], [19], [31]). For this reason, it is crucial to investigate the effects of temperature extreme events on fire activity and large wildland fires.

We chose the island of Sardinia (Italy) as case study for our analysis. Sardinia is a fire-prone area experiencing thousands of wildfires every year ([3], [28], [34]); some of them were very large fires as Bonorva’s fire (23 July 2009 - 9500 ha burned) and Nuoro’s fire (23-25 July 2007 - 9150 ha burned). Wildland fires larger than 100 ha (LWF) represent a small percentage of the total number of occurring fires but account for most of the area burned and cause the most of damage ([17], [34], [36]) with more severe fire behavior ([23]).

In this work, we assessed the historical relationship between high temperature days (HTD) and LWF in Sardinia, Italy. High temperature days were defined as those days in which the 850 hPa air temperature was equal or higher than 25 °C in at least two-thirds of north and south Sardinia. The 850 hPa pressure level is representative of the Earth surface and has the potential to identify unusually severe fire weather events ([22]). It is generally used by Forest Services to analyze past fire weather and forecast daily potential fire occurrence and behavior ([15], [40]). We analyzed the historical trends of LWF and HTD in north and south Sardinia from 1991 to 2009 and their relationships in terms of fire number, burned area, and mean fire size. We also assessed the probability of having LWFs in HTD and the role of other factors (wind speed and number of ignitions) influencing LWF occurrence and burned area in HTD. Finally, we analyzed the differences between north and south Sardinia considering fire number, burned area and mean fire size.

  Methods 

Study area

Sardinia is a large island in the Mediterranean Sea (Fig. 1 - 24 235 km²) located between 38° 51’ N and 41° 15’ N latitude and 8° 8’ E and 9° 50’ E longitude. The territory includes eight administrative provinces and a population of about 1.7 million inhabitants. In this work, we identified two reference areas (Fig. 1): northern Sardinia (including Sassari, Olbia-Tempo, and Nuoro provinces) and southern Sardinia (including Oristano, Ogliastra, Cagliari, Medio-Campidano, and Carbonia-Iglesias provinces). In a preliminary analysis of fire size data (larger fires in the northern part) and the number of high temperature days (higher number in the southern part) showed a marked difference between north and south Sardinia; for these reasons, the island was divided in two parts.

Fig. 1 - Geographic location of Sardinia (Italy) and northern and southern parts of the island.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Overall, the island has a complex topography with hills and low mountains ([32]). The average elevation of the island is 338 m a.s.l. and the highest point is Punta la Marmora with 1834 m a.s.l. in the center of the island.

The flora includes 2407 taxonomic species, of which 10% are endemic ([12]). Large areas are covered by scrub and/ or herbaceous vegetation associations (approx. 35% in the north and 36% in the south), comprised primarily of Pistacia lentiscus L., Arbutus unedo L., Erica arborea L., Myrtus communis L., Olea europea L., Phyllirea spp., Juniperus spp. and Cistus spp. ([34]). Woodlands and forest area is approximately 23% in the north and 17% in the south of Sardinia, with the main tree species being Quercus ilex L., Q. suber L., Q. pubescens Willd., and Q. congesta Presl. Pine plantations with Pinus pinea L. and P. halepensis Mill. only cover 3% of the island and are mainly concentrated along the coast ([34]). Pastures and agricultural lands represent about 36% in the north and 39% in the south, while the urban areas cover 3% of the island.

The climate is classified as Mediterranean, with dry hot summers and an important water deficit from May to September ([11]). The mean annual temperature ranges from 17.8 °C in the southern coast to 12.8 °C in the mountainous areas. Maximum temperature peaks are higher than 30.8 °C during the summer season. Average annual rainfall is 1300 mm in the mountains, but slightly less than 500 mm in the coast, and most of the annual rainfall occurs in fall and winter.

Fire data

We used the historical fire data records from the Sardinian Forest Service (CFVA - Corpo Forestale e di Vigilanza Ambientale) from 1991 to 2009 in north and south Sar dinia. The CFVA database has an entry for each wildfire ignition and provides information on date, municipality and location of the ignition, and area burned. On average, in the last years (1995-2009) Sardinia experienced approximately 2500 fires per year and about 17 000 ha burned per year. Wildfires are typically concentrated from June to September, with the maximum in July for both ignitions and area burned ([34]). Therefore, this work was focused on the June-September period. Similar to elsewhere in Euro-Mediterranean ecosystems ([23]), a few large wildland fires account for most of the burned area ([34]). For instance, in Sardinia fires larger than 100 ha accounted approximately for 60% of the total area burned. In the studied period, fires larger than 100 hectares were 806 and the area burned about 290 000 ha. We analyzed large wildland fires considering three different size classes: wildland fires larger than 100 ha (LWF₁₀₀), 500 ha (LWF₅₀₀), and 1000 ha (LWF₁₀₀₀).

High temperature days (HTD)

Re-analysis data from the National Centers for Environmental Prediction ([20]) were used to characterize the high temperature days on a synoptic scale. Daily air temperature maps at 850 hPa pressure level were analyzed to assess the influence of high temperatures on LWF for both north and south Sardinia from 1991 to 2009. We defined high temperature days when the 850 hPa air temperature was equal or higher than 25 °C at 00:00 UTC in at least two-thirds of both parts of Sardinia. Reversely, we defined as non-HTD those days characterized by 850 hPa temperature lower than 25 °C at 00Z in the period June-September. The 850 hPa air temperature was chosen as reference for several reasons. First, it is generally used by Forest Services to analyze past fire weather and to forecast daily potential fire occurrence and behavior ([15], [40]). Second, it is sufficiently close to the surface to be representative of it, while some of the problems that affect near surface reanalysis variables do not occur ([27], [39], [40]). Third, it provides a regional coverage of air temperature. A temperature equal or higher than 25 °C at 850 hPa is commonly associated to heat waves ([24]) and this condition is responsible of high temperatures at ground level in the territory, as occurred in Portugal in the summer of 2003 ([40]) or in Russia (2010). Furthermore, we analyzed duration and frequency of HTD, and we used HTD as a proxy for the potential occurrence of fires larger than 100 ha.

Statistical analysis

We analyzed the number of HTD, LWF number and area burned for northern and southern parts of Sardinia using the ANOVA analysis and considering the normalized values of: (1) annual area burned by LWF for 1 · 10⁶ ha; and (2) annual number of LWF for 1 · 10⁶ ha.

We analyzed the trends of number of HTD, LWF number and area burned and the differences in LWF number, LWF area burned and LWF size under HTD and non-HTD.

To investigate the relationship between HTD and large fires, we defined the following indicators: (1) number of HTD with at least one LWF with respect to the total HTD (%); (2) normalized LWF average number per HTD with LWF; (3) normalized LWF average area burned per HTD with LWF; and (4) area burned by LWF under HTD with respect to the total area burned by LWF (%). To assess significant differences between HTD and other days, we calculated the same indicators considering non-HTD. The number of days classified as HTD or non-HTD was calculated in the June-September time frame because the HTD were concentrated in these months, and about all LWF were observed in those four months ([34], [40]).

We also evaluated the influences of some explanatory variables on LWF. First, we calculated the number of fire ignitions for each HTD and the differences in ignition number considering days with and without LWF to understand how much the number of ignitions was associated to the occurrence of LWF in these extreme days. Second, we collected daily average wind speed data from several weather stations of the island to evaluate how wind velocity influenced HTD with or without LWF. We used four weather stations (Alghero and Olbia for north Sardinia, Capo Frasca and Decimomannu for south Sardinia). The weather data of these stations are available at ⇒ http:/­/­www.­tutiem po.­net. Fire data were coupled with the closest weather station (among the above mentioned ones) to the fire ignition point.

  Results 

The normalized annual number of LWF₁₀₀ in northern Sardinia was 16.2 fires per 10⁶ ha of wildlands and the normalized average annual area burned was 7553 ha per 10⁶ ha of wildlands, while in the south there were respectively 18.7 LWF₁₀₀ and 5121 ha burned. However, no significant differences between the two areas were found for both LWF₁₀₀ number (p-value = 0.604) and area burned (p-value = 0.299). In north Sardinia, the total number of LWF₁₀₀ calculated as five-year average decreased significantly along time from 38 to 10 fires (p-value=0.035) and in south Sardinia from 44 to 14 fires (p-value=0.028 - Tab. 1, Fig. 2). The number of LWF₁₀₀ under non-HTD also decreased in both northern and southern parts of Sardinia, but the number of LWF₁₀₀ under HTD did not decrease in the study period (Tab. 2). In terms of burned area, we observed a decrease in the burned area under non-HTD in the two parts of the island for LWF₁₀₀ (Tab. 2). Fig. 3 displays the LWF burned area in both north and south Sardinia.

Fig. 2 - Large wildland fire number (100+ ha) in north and south Sardinia from 1991 to 2009 (June-September).

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Fig. 3 - Large wildland area burned (100+ ha) in north and south Sardinia from 1991 to 2009 (June-September).

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Focusing on fires larger than 500 ha and 1000 ha, there were not significant differences between north and south Sardinia in both the normalized annual LWF number and the total area burned of LWF₅₀₀ and LWF₁₀₀₀ (Tab. 1). Moreover, not significant trends were also observed for normalized annual LWF number and total area burned considering days with fires larger than 500 and 1000 ha. However, the number of LWF₅₀₀ under non-HTD decreased significantly along time in both areas, and the normalized annual LWF₅₀₀ burned area also decreased in south Sardinia in these conditions (Tab. 2). These variables were not analyzed in LWF₁₀₀₀ due to the insufficient amount of data to perform a statistical analysis.

The normalized annual average values of LWF number and burned area were significantly different between north and south Sardinia in LWF₁₀₀ and LWF₅₀₀ under non-HTD (Tab. 3). However, not significant differences were observed between the two parts of the island in LWF occurred under HTD.

Tab. 4 shows the relationships between HTD and non-HTD and LWF in both northern and southern Sardinia. There were differences between HTD and non-HTD in terms of LWF daily area burned and LWF daily number. In the north, the area burned by LWF₁₀₀ under HTD equaled to 28.8 % of the total area burned and the normalized average daily area burned per HTD with LWF in LWF₁₀₀ category was significantly higher (2 503 ha day^-1 in 1 · 10⁶ ha of wildlands) than in non-HTD (541 ha day^-1 in 1 · 10⁶ ha of wildlands - Tab. 4, Tab. 5). Similar results were obtained considering the average daily number of LWF in HTD and non-HTD with values of 2.7 ha day^-1 vs. 1.4 ha day^-1, respectively (Tab. 4 and Tab. 5). The average size of LWF in north Sardinia was significantly different between HTD and non-HTD in LWF₁₀₀ (944 vs. 396 ha, respectively). In the southern part, differences between HTD and non-HTD were also important, as shown in Tab. 4 and Tab. 5.

Considering the LWF₅₀₀ and LWF₁₀₀₀, significant differences were observed between HTD and non-HTD in daily number of large fires, but not in relation to LWF daily burned area (Tab. 5) in north Sardinia. By contrast, in the southern part of the island, significant differences were found in LWF daily area burned in LWF₅₀₀ and LWF₁₀₀₀, while in the case of the LWD daily number only LWF₅₀₀ showed significant differences.

Our analysis highlighted that the annual number of HTD cannot significantly explain the annual LWF area burned and the annual number of LWF. However, in HTD the likelihood of having a LWF is higher than in other days. The percentage of HTD associated with LWF₁₀₀ was 41.6% in the north and 30.6% in the south of the island (Tab. 4). By contrast, under non-HTD, the LWF₁₀₀ were 8.6% and 12.0%, respectively. The same was observed for other LWF classes, with high differences between HTD and non-HTD (Tab. 4). In addition, the ratio between LWF area burned under HTD and the total area burned by LWF was very high in north and south Sardinia, as shown in Tab. 4. This percentage under HTD was 26.2% for LWF₁₀₀, 34.7% for LWF₅₀₀ and 39.4% for LWF₁₀₀₀ in the north. In the south, these values were 19.9% in LWF₁₀₀, 35.8% in LWF₅₀₀ and 45.8% in LWF₁₀₀₀.

We finally investigated on the possible influence of the daily number of ignitions and the average wind speed in the occurrence of LWF in days classified as HTD. HTD with LWF were characterized by a significantly higher number of fire ignitions than days without LWF, as shown in Tab. 6, with approximately the double of ignitions in all LWF categories for both northern and southern Sardinia. Additionally, the wind speed was higher in days with LWF₅₀₀ and LWF₁₀₀₀ in the northern part of the island, in particular considering the weather station of Alghero. On the other hand, we did not find significant influences of wind speed on LWF in the southern part of the island (Tab. 7).

  Discussion 

Whereas it is accepted that the major components for fire weather forecasts are low atmospheric humidity, high temperatures, and strong winds near the ground surface ([29]), meteorological indexes developed to evaluate temporal and spatial variations in meteorological conditions are not frequently used or available for all fire weather forecast agencies ([7], [10]). For that reason, we highlight the significance of discerning between HTD and non-HTD defined by the 850 hPa synoptic conditions in developing pre-suppression efforts to stand up to large fires. It would be extremely useful to be able to identify the simply “bad” and the “very bad” fire days with some reasonable lead time (i.e., 24 or 48 hours). We advise that this classification concerning HTD and non-HTD (at 850 hPa) can be used for that discrimination.

Some studies have established that mean, maximum and minimum temperatures have increased and will very likely increase in the next years in south Europe ([4], [9], [19], [26], [30]). Giannakopoulos et al. ([18]) studied possible differences between two reference periods (from 1961-1990 to 2031-2060) in terms of number of hot days (days with T_max>30 °C on the surface) and heat-wave days (T_max>35 °C) in the Mediterranean basin: in some of these areas, like for instance central Spain or north Italy, an increase in the occurrence of hot days and heat waves is expected, with 1-3 additional weeks per year. Therefore, in Mediterranean areas, HTDs are expected to become more frequent and to determine a decrease in air humidity and fuel moisture ([25]), along with an increase in the fire behavior potential ([4]). Overall, in the Mediterranean basin, most HTD are related to the weather regime bringing hot dry air masses from north Africa ([28], [33]).

In recent years, an increase in the frequency of heat waves identified using the 850 hPa synoptic conditions as indicator was observed in northern Spain ([6]). However, our work highlighted that the number of HTD did not change significantly from 1991 to 2009 in neither north nor south Sardinia. Furthermore, both the total normalized annual LWF number and area burned under non-HTD decreased in north and south Sardinia from 1991 to 2009 as in northern Spain ([6]). In particular, the normalized annual area burned decreased significantly under non-HTD conditions. Nevertheless, the normalized annual burned area and number of LWF under HTD did not change in the same period. In short, fire numbers and area burned have been reduced only on days of mild weather conditions in recent years. This fact could be explained with more efficient fire control activity due to important investments in fire suppression technology and training in the last years under non-HTD (⇒ http:/­/­www.­sardegnaambiente.­it/­protezionecivile/­).

Additionally, as expected, the normalized LWF average daily number with LWF, the normalized LWF average daily area burned with LWF, the percentage of days with LWF and the average LWF size were significantly higher under HTD than non-HTD conditions. Therefore, HTD influenced the occurrence of LWF and the area burned in those days. Moreover, this work suggested that HTD are critical for both fuel managers and firefighters, although the number of HTD did not increase in the study period. Probably, fires spreading under HTD can propagate faster and more intensely due to the low dead fuel moisture content and the water stress for live fuels.

Besides, this study shows that HTD with LWF had a higher number of ignitions than HTD without LWF. This could determinate a collapse in the efficiency of the fire suppression system during extreme weather conditions. Moreover, some fires remained smaller than 100 ha due to the quick and effective efforts of the firefighting forces. This is to say, some fires might have not grown larger in recent years because of a larger suppression power, in particular in non-HTD conditions.

Our work also highlighted relevant differences between north and south Sardinia regarding the relationship between fires and HTD. In the northern part of the island, the average frequency of HTD was lower than in the south (1.84 vs. 3.73 HTD per year, respectively), but the incidence of large fires burning in those days was clearly larger in terms of LWF area burned, number and size, particularly as far as LWF₅₀₀ and LWF₁₀₀₀ were concerned. This result suggested that the susceptibility to suffer large fires in the north of the island is higher than in the south, and that the use of 850 hPa synoptic conditions needs a careful evaluation and analysis to take into account other local phenomena (influence of topography, continentality effects, local winds, fuel types, etc.). From this point of view, for instance, wind speed was identified as a key factor to affect the occurrence of LWF₅₀₀ and LWF₁₀₀₀ during HTD in northern Sardinia.

In recent years, most of the largest wildfires in Sardinia happened in extreme HTD, as in 2007, 2009 or 2012 ([35]). This supports the statement that HTD provides more extreme conditions for fire propagation and more difficulties to suppress those fires. This also occurred in other countries ([22], [40], [5], [28], [8]) for the severe wildfires that affected Portugal (2003, 2005), Greece (2007), Spain (1994, 2006, 2009), Russia (2010), USA (2000, 2006 and 2007), Canada (2004), Australia (2005, 2006, 2009, 2011, 2012).

Using fine-scale simulation modeling on weather scenarios historically associated with large fires (i.e., heat waves and strong winds) could help fire managers to be more effective in addressing fire management in the Mediterranean basin, and to identify the priority areas in terms of extreme fire intensity or exposure of values of interest and assets ([13], [1], [34], [37]). It also allows for developing efficient methods and guidelines in a perspective of fire risk mitigation and budgetary planning ([2], [38]).

We are aware that in the future an important effort could be done in order to refine the division of the island in more pyro-climatic areas. Nonetheless, this study represents an important base for further investigations, and is so far the first study covering the complex relationship among temperatures at 850 hPa, large fires, and weather conditions in an Italian fire-prone area.

  Conclusions 

Results showed that neither the number of fires nor the area burned decreased under high temperature days (HTD), although a decrease of both indicators was observed on other days. Furthermore, the number of HTD did not significantly increase from 1991 to 2009 in Sardinia. However, we found a clear relationship between HTD and both LWF occurrence and LWF area burned.

Predicting HTD with reasonable lead time is critical for both fuel managers and firefighters to implement more efficient fire suppression tactics and strategies. The classification of HTD and non-HTD (at 850 hPa) could be used for discriminating those days in order to optimize fire control activities and the available firefighting resources.

Despite the increased amount of money invested in suppression resources in the period analyzed, the total normalized annual area burned by LWF under HTD did not decrease in either northern or southern Sardinia. Additionally, the normalized LWF average daily area burned, the normalized LWF average daily number, the percentage of days with LWF and the average LWF size were significantly higher under HTD than non-HTD conditions. Therefore, HTD conditions may be a useful indicator of an increased probability of the occurrence of severe wildland fires.

  Acknowledgments 

We are thankful to the Spanish Ministry of Economy and Competitiveness (ForBurn Ref: AGL2012-40098-CO3-01) for supporting this study. We are appreciative to the University of Lleida and Pau Costa Foundation for supporting this study through a partial grant to fund Adrian Cardil PhD studies. We are also grateful to the GEMINA Project (MIUR/MATTM number 232/2011) and the EXTREME Project (L.R. 7-2007, CRP-25405) for supporting this work. We appreciate the Corpo Forestale e di Vigilanza Ambientale della Regione Sardegna for providing the wildland fire dataset.

  References