Skip to main content

Forest fires under climate, social and economic changes in Europe, the Mediterranean and other fire-affected areas of the world

Final Report Summary - FUME (Forest fires under climate, social and economic changes in Europe, the Mediterranean and other fire-affected areas of the world)

Executive Summary:
Forest fires result from a number of interacting factors like ignitions, conditions amenable for fire initiation and spread, and landscapes with vegetation (i.e. fuels) that can support the combustion process. Factors driving fire have not been stable during the last decades, mainly due to modifications in the territory caused by socioeconomic and climate changes. Changes in climate and socioeconomics are projected to continue. In FUME we have investigated the relationships between the various drivers of forest fires (socioeconomic, landscape and climate) across various scales and countries during the last decades. Additionally, future projections of these drivers were used to anticipate future risks. We assessed impacts of future changes by means of field experiments and modeling techniques, based on mean and extreme climate episodes, like droughts. The effects of changes in fire regime on the vegetation were also investigated. Restoration needs under changing conditions and for reducing fire hazard were also explored. Policy needs and procedures used in a number of countries were evaluated in regards coping with fire. The main focus was the Southern countries of Europe, although Northern Europe, Northern and South Africa, Anatolia, California and Chile were also investigated.

Following are some of the main results:
1. Fire activity has been changing in the Euro Mediterranean (EUMed) countries at a time when landscapes were highly dynamic. Fires responded to these changes, independently of whether they were planned (e.g. afforestation) or unplanned (e.g. land abandonment). Fires often burned where hazardous changes occurred, and did not burn equally all areas but preferentially burned certain surfaces over others, in particular recently burned ones. The rural-urban interface (RUI) is a particular area of risk and methods have been developed to map the RUI and model fire risk. Socioeconomic factors and climate were important in explaining fire occurrence in EUMed countries, and both should be considered in fire risk analysis. Dry spells and other weather anomalies (heat waves, strong winds) were main driving factors. Despite many factors contributing to fires, we developed procedures to isolate the climate signal, revealing that, despite reduced fire activity in EUMed in recent years and increased fire weather danger, climate was a positive factor affecting fires.
2. Future land use and land cover (LULC) at EUMed was modeled, showing that fire hazard will continue changing, but changes depend on a priori decisions (e.g. assigning land for conservation purposes changes the final outcome of LULC and hazard). Wildfire simulators proved most useful to evaluate future changes in hazard. With global warming, large increases are projected in mean fire-weather indices, length of the fire season and extreme values over large extensions of Europe, including areas in which fires were not prevalent until now. A case study in the Iberian Peninsula showed that fire risk can increase manifold. However, vegetation-fire models show that under scenarios of high climate change low productivity in parts of Southern Europe could limit burned area. Eastern Europe can become the new fire area. Change in fuel moisture, including production of necromass, can affect fire behavior under drought.
3. Regeneration by germination of Mediterranean species will suffer from changes in climate. The sensitivity to changes in germination conditions was idiosyncratic, thus generalizations of impacts will be difficult to make. Field experiments simulating future drought showed that post-fire regeneration will be subject to changes due to different sensitivity among plant functional groups. Changes in fire regime due to increased fire frequency compromise vegetation stability to fire. This occurred in low fire-frequency areas but also in high-frequency areas with resilient vegetation. Management actions and history in pine woodlands have left imprints in the post-fire vegetation that without local information may not be anticipated.
4. Mediterranean pine forests require active management to increase resilience. Pine thinning and introduction of hardwood resprouting species are recommended. Management actions have to be adapted to each stage of the pine-stand dynamics and site conditions. Post-fire restoration should consider fire-resilient, drought-tolerant species in the perspective of climate change. A structured approach has been produced for post-fire impact assessment and restoration under climate change, including options for improving restoration success: seedling acclimation to drought, soil preparation to increase water supply, and microhabitat conditioners to reduce water losses. Increased efficiency in investments in wildfire management operations, and resolving the disconnect problem between science, policy and management is needed in a context of increased fire risk.

Project Context and Objectives:
FUME, “Forest fires under climate, social and economic changes in Europe, the Mediterranean and other fire-affected areas of the world” is a project funded by the European Union’s Seventh Framework Programme (Grant Agreement nº: 243888). The Consortium is formed by 33 partners from 17 different countries from the Mediterranean, Southern, Central and Northern Europe, and other fire-prone countries, including some with Mediterranean-type climates, like Australia, Chile, South Africa and USA. The consortium includes experts in climate and climate change, remote sensing, fire ecology, ecosystem and land use modeling, restoration ecology, economy, social sciences and fire management.
Forest fires are a result of complex interactions between climatic, biological and socioeconomic factors operating at various scales. FUME aims to improve our understanding of forest fires in a context of global change that encompasses these interacting elements (Figure 1). The ultimate goal of FUME is to increase our understanding of how these three components interact to affect forest fires in order to better quantify the impacts of such human driven changes on future fire risk, fire regime and vegetation, among other. The strategy adopted is:
1) Document recent past fire activity and establish the relationships between fires and climate and other landscape and socioeconomic variables.
2) For various scenarios of future changes in fire drivers, including, in particular, changes in extreme climate, model future impacts on LULC, vegetation, fire risk and fire regime. Vegetation vulnerability to future changes in climate extremes and fire regimes will be also experimentally assessed.
3) Evaluate the capacity to adapt to the new conditions of fire risk by developing, among other, restoration strategies, appraising fire protocols to deal with future extremes, and assess the costs of fire impacts on forest services, and policies needed to deal with the new fire risks.
The rural-urban interface was considered a key spatial focus of analysis.
FUME is organized in three research modules plus one module for integration, knowledge transfer and management. Each of these modules comprises several work-packages that interlink to one another (Figure 2).
• Module 1 (Recent trends in landscape change and fire occurrence: disentangling the role of climate and other factors on fire) is dedicated to document and understand the recent past, to assess how landscapes changed in the past and what influence these changes had on past fires in interaction with climate, both in normal and extreme climate.
• Module 2 (Projections of future fire risk, fire regime and associated impacts due to climate and other social and economic changes) projects changes in future fire risk due to climate and other socioeconomic changes, and evaluates future impacts on the vegetation, landscapes and, ultimately, on fire regime, through field studies and modeling for a range of scenarios and mean and, in particular, extreme climate.
• Module 3 (Adapting to change: new approaches and procedures to manage risks and landscapes under climate and social change to reduce vulnerability to fire) analyzes the capacity to adapt to future conditions by developing restoration strategies, and reviewing current protocols and procedures for fire prevention, fire-fighting and the management of fire-prone areas under more extreme conditions. This analysis is complemented by an assessment of economic costs and policy implications of the expected changes.
• Module 4 (Integration, knowledge transfer and management) helps to achieve the former objectives by establishing a common data base, a network of study sites for model testing and validation, and promotes knowledge transfer through training actions with users, among other.


Project Results:
WP 1.1: Recent landscape change and fire regime
A first step in FUME project was to analyze main LULC (land use and land cover) and socioeconomic changes occurred in several study areas from 1950´s until present. Several study sites at different spatial scales (local, regional, EUMed) were established. First, a common legend to adapt all sources of LULC data was elaborated. LULC changes maps during the last decades were produced using historical maps, aerial photography, satellite imagery and other sources. Furthermore, partners defined the socioeconomic variables needed and they established the temporal frames for the analysis. A variety of methods to assess the factors governing LULC through modelling were applied, such as multilevel models, the LUC@CMCC model code suitable for regional applications (Santini and Valentini, 2010), and indexes like Shannon entropy and aggregation index, among other (Task 1.1.1). Additionally, collecting all the information needed to reconstruct the past and recent fire history, such as fire statistics, fire scars, remote sensing images was a next step (Task 1.1.2). Contemporary (1985-2005) fire regimes were characterized in the Euro-Mediterranean countries (EUMed) at different spatial scales using fire statistics (monthly fire number and burned area) derived from different sources. The existence of significant trends and shifts in fire occurrence was also investigated. Additionally, the availability of satellite image time series (mainly Landsat MSS, TM, ETM+ and NOAA-AVHRR) allowed the reconstruction of fire history through fire scars mapping in different study sites at local, regional and national scales.
From the fire datasets built in the various tasks, participants analysed the landscape components affected by fires, the interactions of landscape components and fire, and the impact of fires on landscape characteristics and dynamics (Task 1.1.3). Moreover, W.P 1.1 partners defined a standard rural-urban interface (RUI) mapping typology, and developed an easy-to-use-tool for RUI mapping for the evaluation of the changes in time of main RUI descriptors (Task 1.1.4). The main difficulties found in this WP were related to problems in acquiring and harmonizing past data such as LULC and socioeconomic information of past years, uncertainties in the existing maps of LULC changes, lack of remote sensing data for the reconstruction of fire history, etc.
Following are some of the details of each specific task within this work package:

Task 1.1.1 Landscape composition and dynamics, and socioeconomic factors of fire-prone areas
LULC changes, socioeconomy dynamics, and geophysical data were derived at several spatial scales, from 1950’s until present (Table 1). In many rural areas across Southern Europe, socioeconomic and geophysical factors interacted to increase landscape fire hazard since the middle of the 20th century. Until the middle of the 1980’s, LULC changes considerably enhanced landscape fire hazard in all study areas. The main LULC processes were: land abandonment (farms turned into pastures or shrublands), densification of shrublands and transitional forests (i.e. open woodlands), and afforestation (Figure 3). Depopulation of some villages was matched by concentration in others. In addition, in most areas farm and livestock density decreased from 1960’s until now (Figure 4). After the 1990’s, the dominant LULC change was the replacement of forests by shrublands caused mainly by fires. This, along with the encroachment of abandoned lands by pastures and shrubs, greatly contributed to increase fire hazard.
Results from the explanatory models about the factors governing hazardous LULC changes highlighted that environmental variables determined the suitability of farming and were important determinants of abandonment. Results indicated also that high distance to villages was a significant factor explaining land abandonment in several study areas. Finally, low farm size, opportunity costs of agricultural labor, aging of land-holders, high proportion of employees in other sectors different to the primary, and low degree of mechanization were positively related to land abandonment.

Task 1.1.2 Mapping fires and evaluating fire regimes and changes
In the framework of the reconstruction of contemporary fire regimes and the assessment of significant trends and shifts in fire occurrence, fires smaller than 0.01 ha were discharged from the analysis since their number was considered significantly affected by the variation in data recording and reporting systems during the last decades (Figure 5). The study area exhibited a general increase in fire number. In particular, upward change points in Portugal (1989 and 1994) and Greece (2000) and a downward shift in Italy (1994) were observed. The burned area had an opposite trend, with a generalized decrease throughout the period considered, significant in Italy, Greece, and Turkey. At NUTS2 level, a significant increase in fire number was observed only in Attica and Peloponnese (Greece) among the studied areas, while burned area followed a general decrease in all the study areas.
Operational mapping of burned surfaces and reconstruction of recent fire history at EUMed, regional and local scales was possible using low cost remotely sensed data. For the first time a complete historical series of burned area maps (> 500 ha) was produced for EUMed countries from 1981 to 1999 by processing low resolution NOAA-AVHRR satellite data (Figure 6). These new data allowed deriving spatially explicit information on fire regime, providing consistent information to assist in further investigations necessary for environmental planning and management. Furthermore, this satellite-derived information allowed performing innovative fire selectivity analysis, which incorporates the fire shape in the modeling process.

Task 1.1.3 Assessing burned landscape features and the interaction of fire on them
Built from previously developed fire datasets, the analysis of fire selectivity at different spatial scales over a wide geographical range (Table 1) provided an opportunity to summarize common patterns and distinguish important differences among the study cases. Through the study cases it has been shown that wildland fires in the Mediterranean Europe show clear evidence of selectivity towards certain LULC types. Fire events, especially large ones, generally, avoid burning croplands, whereas they usually prefer forests and shrublands. Additionally, it was shown that, at least at the EUMed scale, fires showed a preference for recently burned areas (Figure 7). Moreover, a pattern relating the small fires with human-influenced ecosystems and large, infrequent fires with natural ecosystems was revealed constant, with small variation, throughout the study cases. Knowledge of fire selectivity patterns will allow the development of landscape management policies in which fire prevention, pre-suppression and suppression strategies are fully integrated.

Task 1.1.4 The dynamics of the rural-urban interface (RUI)
The growing of RUI in all the Mediterranean study sites is a challenge for community leaders and managers. The high risk is due to numerous fire ignitions in a zone of human density and high valuable goods, rising issues for prevention, organization and fire-fighting and RUI distribution on the territory. In this context, an easy-to-use software designed for RUI mapping at different scales was developed to describe and map the territory at different resolutions. The tool gives the user a choice between three main methods for RUI mapping: two at local or regional scales and one at global scale (Figure 8). The choice between regional scale methods depends on the geographical context and the available data. When using the RUI tool to map the different types of RUI, it is possible to assess risk in RUI by using one of the risk models specified in the FUME project. Again, the choice between methods depends on the user’s objective, the geographical context and scale, and the data availability.
. 
WP 1.2: Interactions between landscape properties and fire under varied climate and weather conditions including extremes
In WP 1.2 contemporary and past relationships between climate/weather drivers and fire statistics, also including extreme conditions triggering to extreme or large fire events, were addressed and investigated at different scales (Table 1). In addition to climate and weather, the chief elements of risk were evaluated through fire modeling by identifying the main elements that drive fire propagation at the landscape level and their dynamics.
During the first phase of the project, the main emphasis was put on the collection, harmonization and pre-processing of available data (in cooperation with the activities carried out in WP 1.1) and on the development of common methodologies. During the second phase, climate/weather variables and fire-weather indices were used to determine their influence on fire occurrence applying different algorithms and methodologies (Task 1.2.1). Climate simulations, reanalysis and historical observation were further used to identify extreme weather/climatic conditions triggering large fire events (Task 1.2.2). Additionally, fire ignition and propagation models were calibrated and validated using a set of case studies based on real fires covering an East-West transect along the Mediterranean basin. The models were used to predict fire spread as a function of landscape characteristics and meteorological conditions (Task 1.2.3).
During the last phase, WP 1.2 participants prepared and finalized a number of papers related to the relationships between weather/climate variables -or their derived fire danger indices- and fire occurrence, including also fire-extreme situations. The potential of satellite data to provide relevant information regarding the dynamics of individual large wildfire events was also investigated. Additionally, wildfire simulators were applied to a number of case studies in the Mediterranean basin in a probabilistic configuration, in order to predict the effects through time (between 1950 and 2000) of the main key factors (fuels, weather conditions, and topography) affecting wildfire likelihood and intensity (Task 1.2.3).
Following are some of the specific achievements:

Task 1.2.1 Current relationships between fire and climate and weather
Despite the use of a wide range of methodologies and spatial scales, it was observed the importance of similar concurrent (and precedent) fire season climatic factors. Overall, seasonal droughts in the months prior to the fire season peak, occurrence of heat waves, and strong winds during the fire season have a strong influence on fire occurrence and seasonal severity across scales (e.g. Figure 9). It is clear that different areas have their own peculiarity. For example, Northern Italian areas showed a significant negative correlation also with minimum temperature. In the Iberian Peninsula, warm and dry fluxes associated with anticyclonic regimes were found to be the typical synoptic configurations favoring large burned areas (BA). In South Central Chile, strong relationships between fires and large-scale climatic oscillation patterns (e.g. ENSO) were found. In South Africa, the focus on event-driven fires (e.g. foehn wind conditions) provided more insight than exploring mean annual fire risk conditions. At local scale (Mt. Parnitha and Mt. Penteli in Attica, Greece) significant relationships between BA and summer precipitation were found, however accounting only for a small part of the burned area variability.
On the whole, a number of models, methodologies, and approaches to study fire/weather relationships were investigated. A number of partners based the analysis also on homogeneous units in terms of climate, fuel type characteristics, and fire occurrence, to test model performances under different conditions (Figure 10). Indeed, underlying homogeneous framework would provide more comprehensive outputs in exploring fire activity patterns at national and regional scales.
All these models, in addition to be incorporated in short-term and long-term predictive models of fire occurrence and risk, could be integrated into various aspects of fire planning activities, at both short and long term, such as budget considerations, resource allocations, and fuel management. This would contribute to the improvement of the effectiveness of fire suppression activities in fire-prone areas.


Task 1.2.2 Analysis of extremes, including meso-meteorological situations conducive to extreme fire risk
Results from this task contribute to the exploration of fire–climate relationships in extreme conditions, including meso-meteorological conditions (Figure 11), and could allow fire managers to more easily incorporate the effect of extreme weather conditions into long-term planning strategies. These results may become more important under climate change scenarios, since climate change is discussed on the basis of changes of extremes rather than changes in means. Furthermore, most discussion on climate change is focused on the effect of increasing temperature trends, but the findings of this task, as well as task 1.2.1 highlighted the importance of precipitation and especially the need to predict changes in seasonal precipitation more accurately.
Furthermore, partners provided a number of practical suggestions, related to the detection of extreme event indicators, on the correct and precise use of weather inputs for the definition of forest weather indices and within statistical models, and on the effectiveness of different reanalysis products often used as surrogate of observation data for the calculation of fire danger indices. All these suggestions and considerations are of practical use for stakeholders for fire modeling applications especially in extreme conditions.

Task 1.2.3 Evaluating the risk by identifying the chief drivers for fire propagation at the landscape level and their dynamics
Satellite-derived ignition areas appeared to be a very promising tool for the reconstruction of individual fire events, understand their complex spread patterns and their main drivers of fire propagation. The results showed a very good temporal agreement although the spatial errors were reflecting the trade-off between the spatial and temporal resolution of the MODIS sensors. An innovative algorithm was developed to derive the major fire spread paths that yield information about the most important spatial corridors through which fires spread, and their relative importance in the entire fire event. These major fire paths (Figure 12) were then used to extract relevant descriptors, such as the distribution of fire spread direction, rate of spread and fire radiative power.
The results of the modeling activity using wildfire simulators in a probabilistic configuration highlighted reductions of burn probability in recent time steps, in comparison with simulations performed in the 60’s and 70’s. The changes in land use affected significantly this trend (Figure 13), as well as the causes affecting the ignition patterns. The study confirmed the capabilities of fire simulators as a tool to integrate the multiple sources of variability. Furthermore, the experience of FUME modeling activities confirmed that wildfire likelihood and behavior could be studied and monitored over time, so that policy makers and management agencies can plan investments and activities and evaluate the responses at a fine scale, even in a perspective of future climate changes and/or other disturbances.

WP 1.3: Attributing fire occurrence to changes in climate and socioeconomics
The work in WP 1.3 was organized in three tasks; the first referred to the recent contribution of observed climate change to fire activity, the second to recent role of socioeconomic factors on fire, and the third to the integrated analysis for understanding the combined role of observed climate change and socioeconomic change. The three tasks needed input data on wildland fires, weather/climate observations, and socioeconomic data which were gathered in the previous reporting periods, in connection with WPs 1.1 and 1.2. Different study areas at different spatial scales were identified ranging from local, to national and EUMed scales.
Due to the diverse nature of the datasets at local, regional, national and EUMed scales it was not possible to establish a common statistical modeling framework. The individual relationships between fire occurrence and climatic data were investigated (in task 1.3.1). The changes of FWI in Europe since 1960 were studied using ERA- 40 (1960-1999) and ERA-Interim (1980-2012) datasets. The analyses were made for four regions in Europe, North, West, East and South Europe. Then, the correlation between the seasonal mean FWI and burned area was calculated for Greece and Spain. In task 1.3.2 relationships between fire regime and socioeconomic variables were analysed at different scales applying different approaches (Table 2). Finally, in task 1.3.3 more effort on the attribution problem for understanding the combined role of observed climate and socioeconomic changes on fire regime was put. Two approaches were followed; a physically-based modeling approach using the LPJmL-SPITFIRE model, and a more statistical one. A comparison with California, a Mediterranean-type area, was also included.

Task 1.3.1 Recent contribution of observed climate change to fire activity
The results related to the climate change detection signal in fire danger using ERA-Interim reanalysis data indicated that the fire danger has increased during the last 30 years and especially in Southern Europe. An interesting finding is that the burned area in the Southern Europe seems to have a tendency of getting smaller at the same time as the fire danger tends to increase (Figure 14). Additionally, changes in fire-weather danger indices and burned area were cross-correlated in Greece and Spain with a significant correlation of the order of 0.6 (Figure 15). Though the meteorological conditions for inducing fire danger have changed in some areas, forest fires have not necessarily followed suit, presumably due to other factors affecting them. This leads to a conclusion that human activities have had a positive impact on the occurrence of fires and especially in Southern Europe; without those activities the burned area would have been larger than what the current situation is.

Task 1.3.2 Recent role of socioeconomic factors on fire
Explanatory models of fire occurrence across different European areas indicated that, despite the well-recognized common traits in the Mediterranean Europe, important regional and local differences exist with respect to the fire environment. In this context, the role of the wildland-urban interface (WUI) in affecting fire regimes seems to be a common characteristic. This variable has shown to be positively correlated with historical fire occurrence at EUMed and regional scales (e.g. Figure 16) which confirms the importance of adopting prevention measures in those areas where human activity can be an important ignition source.
Additionally, the studies performed in Spain and Italy highlighted that better accessibility implies more human pressure on forest areas and consequently, an increase of potential ignition sources by accident and/or negligence. Agricultural and livestock activities have also shown to explain fire occurrence in EUMed countries, however, variables that represent those activities differ among the different study areas and scales and so the sign of the relationship with the explanatory variable. In some areas (e.g. Spain) models indicate that fire occurrence increase with a high proportion of the wildland-grassland interface and agrarian holders older than 55 years (the latter also at EUMed scale) (Figure 17). This would reveal a positive relationship between agricultural/livestock activities and fire occurrence probably due to the fact that farmers and shepherds who remain in the rural areas often practice traditional methods, including the use of fire as a tool to maintain pasture, eliminate harvest remains, thinning and land clearing, etc.
In general, models showed that fire occurrence is better explained by the combination of both landscape features and socioeconomic data than by each set of variables separated. This general picture illustrated how the presence of human habitat is a key driver for fire starts, but the spreading into large fire is much more influenced by the landscape composition and structure. This might be the result of either a facilitated fire rate of spread by specific land covers or landscape structure, or a lack of intervention of fire fighters in low value areas.
In this task, also a comparative analysis between Northern Africa and Southern Europe was performed. The findings showed that integrating potentialities for local conflicts (borders, change of political regimes, socioeconomic collapses) should be considered in future scenarios for fire risk assessment in the Mediterranean basin, a particularly instable region.

Task 1.3.3 Integrated analysis for understanding the combined role of observed climate change and socioeconomic change
At EUMed scale the application of LPJmL-SPITFIRE model showed that land use change played a major role in simulated burned area more than climate or changes in human population density. Specific pattern differed among countries, depending on the country and decade; landscape fragmentation due to urbanization was the dominant factor (Figure 18).
At EUMed and California, the statistical approach showed that climate and fires were correlated during the last decades and this relationship could be ascertained independently of changes in other drivers that occurred in the areas analysed. Furthermore, this signal has emerged despite the fact that the number of fires and burned area has decreased in EUMed in the last years. The regions with lower mean FWI (i.e. milder climatic conditions) were more responsive to a change in climate, that is, more climatically controlled. The relationships found for EUMed (Figure 19) were stronger than in California, where mean FWI values for the fire season were larger than their counterparts in EUMed. Therefore, the results obtained for California were consistent with the hypothesis that the climate-fire relationship tends to be reduced or disappear altogether the more extreme the climate of a region.
Finally, although strong relationships were found at both national and regional scale between weather-based indices or similar variables and fire activity, socioeconomic variables were also important, particularly at regional and local scales. Using these variables jointly with climate, the prediction of forest fires could be improved. The approaches used in FUME allowed linking both drivers. Furthermore, since both drivers (climate, socioeconomics) vary differently in time and space, one important achievement of the project is to have developed methods that allow investigating these relationships independently of the trends they experience with time.


Module 2: Projections of future fire risk, fire regime and associated impacts due to climate and other social and economic changes (10 pages)

WP 2.1: Scenario development
The main objective of WP 2.1 has been the construction and application of an exhaustive and integrated modeling approach, to provide scenarios about three factors strongly interacting with one another and influencing fire hazard. The modeling exercise comprised: i) future climate projections, at the European, EUMed to local scale, under alternative emission scenarios up to year 2100, and including extreme event analysis (Task 2.1.1); ii) coherent land use projections, assumed depending on a combination of biophysical (e.g. climatic) and socioeconomic future conditions impacting the spatial distribution of fire-prone areas (Task 2.1.2); and iii) potential vegetation scenarios, forced by climate, land use and socioeconomic projections (Task 2.1.3).

Task 2.1.1 Climate change scenarios
The most relevant outcomes of climate related activities have been the supply of projections of future climate useful for successive modeling of impacts related to land use distribution, vegetation status, fire hazard and risk, and thus fire restoration evaluations.
The state-of-the-art scenarios from ENSEMBLES (at 25km resolution) were first used to analyze the present and future climatic patterns for the whole European region. Then, new global and regional climate simulations (at 80 and 14km resolution, respectively) were provided to the project partners. All simulations covered the period 1970-2100 under the A1B (AR4-IPCC) emission scenarios. Further, a Statistical Analogue Re-sampling Scheme (STARS), based on the assumption that weather patterns observed in the past will very likely occur in the future and able to provide a large ensemble of simulations in a very short time, was tested and used to produce climate projections under new RCPs (AR5-IPCC) emission scenarios (Figure 20). For local scale analysis, projections of four fire climatic variables (temperature, precipitation, relative humidity and wind speed) strictly related to fire-risk were used to characterize the uncertainty of the multi-model scenario at nine locations of the FUME site network. The generic result is a projected increase in temperature between +1° and +4°C according to RCP scenario with a maximum value in Eastern Europe. Precipitation should mostly decrease in the summer period by 25% (except around the Black Sea where a precipitation increase would occur). This would result in an increase in FWI and length of the fire season with low standard deviation ensuring the robustness of predictions. Changes for other variables are small.
The focus of climate analysis also considered extreme events for temperature, precipitation and wind speed. The differences found in the estimation of extreme scenarios, using different downscaling and extreme definition methodologies, confirmed the importance of considering an ensemble approach in the projection of future regional climate change to quantify projection uncertainties.

Task 2.1.2 Socioeconomic scenarios (social and economic, land use and land cover, including the Land use and land cover
Results from the coupling of land use allocation and economic models through market-based land demands, still under A1B emission scenario, were used to generate land use simulations at the EUMed scale from 2000 to 2100, showing that around 10% of pixels (≈20 Mha) could be subjected to change (Figure 21). When preventing protected area from changes, land modifications reduce to 9% of the region. In terms of fire hazard related to land use types, the ensemble results show a spread of situations with medium to high hazard.
In the Autonomous region of Madrid, future land use and land cover change scenarios to assess the relative importance of anthropogenic factors linked with fire occurrence in Mediterranean areas were produced for 2025 and 2070, both by extrapolating past trends and by assuming an economic crisis. In this context, contact zones between land use categories (interfaces) appear of particular interest, which intensifies the pressure on forest cover.
Furthermore, models specialised in complex space dynamics, i.e. for rural-urban interface, were applied at regional and local scales (Figure 22). At the regional scale, scenarios of urban development mitigation and free urban development were compared, showing a convergence on the longer term. At local scale, three scenarios were tested: urban development, nature protection, and equilibrated scenario. Fire occurrence is closely related to the type of RUI concerned, where ignitions are concentrated in wildland surroundings densely built RUI which have low fuel load themselves.

Task 2.1.3 Potential vegetation scenarios
Finally, simulations of natural vegetation scenarios forced by new projected climate regime supported the modeling of fire propagation under extreme conditions. Vegetation was first related to expected climate conditions, analysed projected new climate zoning, with warmer winters in the future for Northern Europe, few changes in Central Europe, and drier and hotter climates in the Mediterranean area (Figure 23). Such changes will also drive modifications in forest composition as predicted by ecosystem models, with increased fire risk in the boreal forests.
Finally, new parameterized dynamical vegetation model simulations of changing vegetation patterns across Europe in response to a mix of factors (climate change, change in population density and land use) correspond with expectations of a north-ward (and upward) migrating timberline in the boreal and alpine areas, and an increasingly drought-adapted vegetation in parts of the Mediterranean. The presence or absence of fire alone, however, does not alter vegetation patterns visibly, likely as the simulate fire return-interval is too long to yield a shift in dominant species composition. It is needed to be investigated further to which degree enhanced atmospheric CO2 levels can act to compensate effects of fire.

WP 2.2: Impacts on fire risk, fire regime and on landscapes under climate and socioeconomic changes
WP 2.2 focused, among other, on understanding how changes in climate, and associated alterations in socioeconomics, would affect fire potential and fire risks. Models of fire danger based on climate inputs as well as models using global vegetation models with fire module were used to assess future fire regime in Europe and in other potentially new fire areas (Task 2.2.1). The sensitivity of plants and vegetation across several areas was also investigated in light of the anticipated climate and potential changes in fire regime (Task 2.2.2).

Task 2.2.1 Projecting impacts on fire danger, fire risk and fire regime
Regional to continental scale analysis of climate scenarios revealed an increase in temperature coupled with a decrease in rainfall and air humidity, overall leading to increasing Fire Weather Index across Europe (Figure 24). Modeling based on past fire relationships in a case study area (Iberian Peninsula), under the assumption that current patterns and relationships are maintained resulted in a several-fold increase in forest fires (Figure 25). Process-based simulations of coupled vegetation-fire models (LPJmL-SPITFIRE and LPJ-GUESS-SIMFIRE models) at the continental scale concluded on a longer fire season by 5 to 25 days in Northern Europe and up to 15-45 days in the Mediterranean basin, leading to an increase in burned areas. Eastern Europe appears to be highly sensitive to changes in climate, becoming a new fire area in Europe (Figure 26).
Among the different candidate variables contributing to fire risk, climate change appeared as the most significant, yet the role that CO2 could play was important and variable among regions. Atmospheric CO2 fertilization increased NPP (net primary productivity) and fuel load, overall leading to an increase in biomass combusted and the resulting CO2 emissions. Fuel limitation by increasing drought could however mitigate to generic trend on the most Southern part of the Mediterranean basin (Figure 27). Regional scale modeling illustrated the concomitant shift expansion of drought and fire tolerant species when drought increases. The continental scale socioeconomic impact, accounting for country specific management policies and population density, was pointed out as a bottleneck for future projection. Regional scale analysis provide information on the importance of the rural-urban interface and land use practices for agriculture or recreation that should be significant, if possible to predict accurately according to economic scenarios.

Task 2.2.2 Evaluation of systems vulnerability due to changes in climate and fire regime
Targeted field observations and modeling could identify key processes for future fire risk and consequences on vegetation vulnerability (Figure 28). Under increasing drought, we identified differential vegetation composition as a result of fire intensity (mostly controlled by local topo-climate and initial fine fuel amount), species resistance, fire return interval, as well as fire shape and distance to unburned plots providing a seed source. Seed germination from different provenances with contrasted climates, and grown under different drought conditions illustrated how increasing drought would affect germination success as a result of both seedling survival after emergence and seed viability resulting from the pre-fire seed maturation. Laboratory experiments with seed germination from various provenances across Europe showed that species differ in their sensitivity to water availability. Yet the responses were species-specific, which complicates making extrapolations. Additional experiments with seeds from various populations across various elevations showed as well that responses to changes in temperature varied among provenances. That included direct responses or indirect, i.e via seed dormancy release linked to cold temperatures. A case study using vegetation exposed to drought showed that maternal effects can also control germination (Figure 29). Overall, experiments demonstrated that anticipating plants responses at the germination phase to changes in climate will be difficult, and that the differential response by plants based on the locals they inhabit shows that changes in species are likely, even if the direction in which they will occur cannot be anticipated.
Additional experiments addressed the sensitivity of the vegetation to changes in fire regime. Two studies were made, one in a high-frequency fire area, one in a non-frequent fire area. Results showed that fire frequency within a given period and time elapsed since fire and the time when fire occurred are variables that significantly affected the post-fire community in the high-frequency area investigated (Table 3). Fires affecting non-fire areas can lead to tipping points in the post-fire vegetation, with major species replacements.

WP 2.3: Impacts of climate and weather extremes
Objectives for WP 2.3 were to figure how ecosystems respond to climate extremes (mainly prolonged drought or heat waves) through both field measurements and modeling exercises at different levels (stand, landscape, region and continent). Task 2.3.1 provided field information on 1) short-term and long-term vegetation responses to prolonged drought through rainfall interception experiments and gradient analysis, and 2) site-specific post-fire regeneration under similar experiments with experimental burning. Information from this task has been further used in task 2.3.2 for testing empirical drought and fuel moisture indices used in fire risk indices, as well as process-based models simulating the soil/plant water budget and biomass amount, with an emphasis on prolonged drought outside the range of usually observed drought periods. Task 2.3.3 was dedicated to the assessment of landscape and continental scale applications of coupled vegetation/fire models under extreme climate events.

Task 2.3.1 Measuring changes in vegetation and fuels in relation to drought
Plant water status and fuel moisture content
An extensive compilation of the seasonal course of fuel moisture content for different species across the Mediterranean basin and including rainfall interception experiment, illustrated the differential responses of species according to their functional traits (Figure 30). Additional information on plant water potentials and stomatal conductance indicated how, for a same soil water deficit, specific water use strategies allowed for species to desiccate more intensively and recover more or less rapidly during summer rainfall pulses. Plant growth and fuel amount.

Task 2.3.2 Modeling vegetation and fuels under droughts and heat waves
Extreme drought impact on plant growth was assessed through rainfall experiments on Mediterranean forests and shrublands, and with a regional analysis during the 2003 heat wave and the 2006 prolonged drought in Southern France. Twig elongation in shrubs was hardly affected by drought in most experiments but significant effects were observed on litterfall, leaf die back, and in turn ecosystem’s Leaf Area Index (LAI). Implication on fine fuel amount (leaves) and feedbacks on ecosystems water budget would then be significantly modified. Vegetation models were fairly able to capture this adjustment. However, very extreme drought induces ecophysiological process as changes in leaf life span for Quercus ilex which are hardly accounted for in vegetation models. Cavitation processes and branch/leaf die back are also an under-represented process in vegetation models that should be further investigated and implemented to increase model accuracy in simulating fuel amount and necromass production (Figure 31).
Comparisons with widely-used drought indices and fine fuel moisture codes concluded on the suitability of these indices to capture the seasonal pattern of the fire season, but differential correlation with actual fuel moisture according to the species (Figure 32). Process-based models, more difficult to calibrate and using more functional parameters, could however capture these differences. We concluded on thresholds of fuel moisture that can be reached at the end of the dry season depending on the species involved. Dangerous levelsmay not be exceeded when drought is prolonged under extreme events, unless cavitation and leaf/branch die-back appears.

Post-fire regeneration under extreme climate
Post-fire vegetation dynamics were assessed at two experimental sites in which rainfall total and patterns were modified (Figure 33). Simulations included extreme drought periods (i.e. 7 months drought, with rainfall at percentile 1 of the long-term climate series). Main results identified low impact of drought on resprouting species but significant effects on emergence and density of seeders (Figure 34). Seasonal plant water budget assessed with predawn water potentials showed significantly higher values for resprouters compared to mature vegetation indicating a lower drought constraint under dry spells due to lower LAI and in turn a prolonged soil water saving along the season. We concluded for a significant change in community composition after fire events under increasing drought according to regeneration strategies. Additional effects on nutrient availability and microbial activity affected by drought would also contribute to post-fire depletion on vegetation dynamics (Figure 35).

Task 2.3.3 Modeling fire propagation under various scenarios of extreme conditions
Fire risk under extreme events was evaluated as a response to extreme wind, dry spells and heat waves at the landscape scale. After evaluating extreme daily Fire Weather Index from historical and projected climate scenarios, fire spread was simulated for these events with the additional accounting of climate change impact on changes in vegetation types, and/or increases in the wildland-urban interface (WUI) where most fires start. Fuel moisture and air temperature appeared as the most contributing variables for increasing fire risk under climate change scenarios as change in wind is not clearly stated, and changes in vegetation fuel load is uncertain. Wind downscaling appeared however as a very sensible driver for the final fire shape and extend.
Continental scale analysis using a coupled dynamic global vegetation model/ fire model under climate change scenarios identified an increase in extreme fire events (30 and 120 year fire intervals). Temperate forests appear as the most sensible biome to increasing extreme events, while the Mediterranean basin should maintain or even reduce extreme fire events as a response to fuel load reduction. Fuel load then appears as the critical variable to refine in vegetation models to ascertain the conclusions (Figure 36).

WP.2.4 Impacts at the rural-urban interface
The objectives of WP 2.4 were to project future changes in RUI and associated change in fire risk, and to identify how fire settings were actually correlated to RUI development across the Mediterranean basin, by using a RUI analytical tool developed in WP 1.1 (Task 1.1.4). Based on this analysis, LULC change scenarios, including changes in urban and agricultural areas, were used to predict future changes in RUI and associated fire risk.
Fire risk at the RUI was analysed according to different analytical methods. In all study areas and independent of the method applied, fire risk appeared as highly controlled by RUI sprawl across the Mediterranean basin. This finding is independent of urbanization occurring or not. Projected changes in land cover and the resulting RUI showed an increasing fire risk. Fire risk will drastically increase in the next two or three decades due population growth, and potentially decrease afterwards. For risk mitigation, land planners have to take into account RUI sprawl in the next decades and to take care of scale consistency between the different planning levels (Figure 37).

Module 3: Adapting to change: new approaches and procedures to manage risks and landscapes under climate, social and economic changes to reduce vulnerability to fire (6.25 pages)
In Module 3 we developed approaches to adapt to changes according to assessments of change conducted in previous FUME modules. Research included all aspects of fire management, from prevention to suppression and post-fire restoration, and considered ecological, technical, economic and political perspectives (Figure 38). The main expected impacts of climate change framing the activities to be considered for adaptation hereafter were increased drought and more severe fire regimes.

WP 3.1: Managing risk under future climate and multiple extremes
Formal stakeholder groups established in Spain and Italy (Figure 39) produced the following adaptation recommendations for:
1. Institutional strengthening and political commitment: a) Definition of clear responsibilities between the actors involved in fire management, especially concerning the rural-urban interface (RUI); b) Introduction of legal obligations for land owners to conduct fuel reduction treatments.
2. Raising awareness: Future campaigns should target the following groups: a) Policy makers; while successful fire suppression results are immediately perceived by policy makers, as well as by the large public, the positive effects of long-term fire prevention activities are not easily and clearly recognizable, in the short term. This situation consolidates the general public and policy makers opinion that fire suppression is more important than fire prevention; b) People living in the RUI, who must be informed about wildfire risk at RUI and must understand the preventive benefits of fuel management to reduce fuel loads around buildings and critical infrastructures; and c) Farmers using fire for land clearing and agricultural practices; regulations of dangerous activities and training may help reducing this common cause of runaway wildfires due to negligence.
3. Fire prevention planning: Integration of wildfire prevention measures within other planning instruments like forest management and urban plans. More specifically: a) Plan duration should not exceed five years, given how quick fire scenarios change; b) Develop indicators to assess correct implementation of plan provisions; c) Prevention policies should be planned at national level for public and private areas; including a specific section on fire prevention for the RUI; d) Operational implementation of daily fire danger rating systems; and e) Development of fire risk zoning using new technologies, e.g. fire simulators (FARSITE, etc.) and integrated data collection systems to generate spatial information of fuel characteristics (optical remote sensing and Airborne Laser Scanning - ALS).
4. Fire-related research: Incorporation of fire risk analysis into fire prevention planning requires reliable data on fire environment and past wildfire events. A suitable time scale for fire prevention strategies requires future fire scenarios projections of no more than 20 years.
5. Capacity building: In the near future, conflagration like wildfires will stretch fire management organizations capacity to it limits. Dealing with these type of events requires changing strategies from direct attack to reducing and slowing fire behavior and defending RUI assets. Additional financial and technical fire fighting resources would be needed, particularly in the RUI. In this perspective, skilled fire specialists, notably RUI fire analyst experts in fire behavior forecast should be standard job positions in the agencies in charge of fire prevention programs.

The following measures are proposed for the adaptation of fire suppression:
a) Plans revisions yearly
b) Plans must include a risk and vulnerability map and a specific section for fire fighting in RUI
c) Plans must include specific protocols to handle participation of third-party resources
d) Standardization of communication frequencies
e) Increase use of fire for landscape management where needed
f) Prioritize investments in ground resources


WP 3.2: Managing and restoring landscapes under change and uncertainty
We conducted experiments to reduce fuel load and increase resilience and biodiversity in fire-prone Mediterranean ecosystems. Given the uncertain responses of species to combined changes in fire regime and drought, we explored species selection criteria for post-fire restoration. One operational output of this research is a Decision Support System (POSTFIRE-DSS) for post-fire restoration assessment.

Task 3.2.1 Integrating pre- and post-fire management, merging fire prevention and restoration approaches.
Target ecosystems were fire-prone shrublands and pine forests showing excessive vs low regeneration, and/or low biodiversity and resilience.
• Clearing Selective clearing accompanied by the introduction of woody resprouters significantly reduced fire hazard and increased ecosystem fire-resilience at the mid- term.
• Thinning combined with the introduction of several woody resprouters were successfully applied in a N-S transect from Southern France to Eastern Spain, down to Northwest Tunisia. In high density pine stands, experimental results showed the efficiency of thinning and planting in reducing fire hazard through reducing fuel load build-up rate and dead fuel accumulation, and, as a consequence, the requirements of further interventions for fuel control. In addition, the treatments improved ecosystem resilience, structure and quality by the introduction of woody-resprouter, late successional species.
• Mechanical chopping, soil scarification, controlled fire of low or high intensity were tested to enhance natural regeneration of mature Pinus halepensis stands in Southern France. High intensity fire was the most effective, whereas grass cover proved to reduce pine germination and seedling survival. Competition effects for introduced oak seedlings was analysed considering neighbourhood of pines, of a legume shrub, and the effect of manual weeding. Oaks survival and growth decreased with pine density, and the N-fixing shrub proved to be very competitive towards oak seedlings. Oak seedlings grown in open conditions tend to develop a large crown, a large number of shoots and to lose apical dominance, resulting in a bushy morphology. On the contrary, the presence of neighbours had favoured plant elongation and limited crown lateral expansion.
In mature Aleppo pine forests, light availability was more limiting in the moister site (Southern France) for the introduced seedlings than in Eastern Spain, which is more limited by water scarcity. In mature Aleppo pine, thinning had a positive effect in Southern France for introduced seedling survival, relative growth rates in diameter and in height for most species, but the contrary effect in Eastern Spain for survival (Figure 40) and relative growth rate in height. In the drier site (Eastern Spain), high density pine forests presented environmental conditions more suitable for growing and survival of seedlings, in contrast to medium and low density. Under the canopy, high density presents lower photosynthetic photon flux density, net precipitation and soil moisture than in the other densities. However, the values observed are not limiting factors for growing and survival in this early stage of development and results were globally better under high density stands. Of course, the situation may change at later development stages of introduced seedlings. Species identity had a strong effect on relative growth rate in both sites.

Task 3.2.2 Restoration under uncertain conditions
Remote sensing analysis of the combined effect of drought and large wildfires demonstrated that fire damage was significantly determined both by vegetation density before fire occurrence and by water-stress levels before the fire. A list of native Mediterranean species was elaborated for potential use in post-fire restoration, considering their regeneration ability after fire and their resistance to drought. In addition, we experimentally evaluated fire-resilient plant species and ecotypes for their capacity to acclimate to changing conditions, especially increased water stress. We selected woody resprouter species with an efficient use of water to avoid threshold negative water potentials (around -8 MPa), and good capacity to recover functionality after intense drought events (Figure 41). Our results showed that Arbutus unedo and Quercus ilex (and Fraxinus ornus for moister sites) were resprouter species with a good capacity to maintain low levels of mortality under water stress due to an efficient use of water. Pinus halepensis (Aleppo pine, obligate seeder) also showed high tolerance to water stress, although this species is eradicated by short-interval fires.

Provenance proved to play a significant role in the early development of Arbutus unedo under different water stress regimes, reinforcing the importance of seed source when planning ecological restoration actions. The wettest provenance showed the poorest drought adjustment and the provenance from the driest summer site developed seedlings with higher root dry weight.
We evaluated post-fire rehabilitation actions taken after mega-fires in Greece. The post-fire management of burned areas has been given lesser attention than fire itself. In none (0%) of the study cases on Pinus halepensis forests and mountain coniferous (mainly Abies cephalonica) forests have been taken actions for the prediction of vulnerable areas (step 1), with the exception of Olympia case, while only 3 of 8 cases (37.5%) have taken steps 2, 4 and 5 but without completing every path of the process. Emergency interventions (step 3) have been made in 62.5% of the cases but with low efficiency percentages. Post-fire management mainly concerns operations for controlling erosion, salvage logging of the burned trees and quite often reforestation. We concluded that most of the applied measures were empirical rather than science based. To advance in this area, we developed and tested a decision support system (POSTFIRE-DSS) to assist decision makers in managing burned areas (Figure 42). This application provides assessment procedures and recommendations for the restoration of burned areas in a context of climate change. POSTFIRE-DSS application and full description (D322) is freely available in the link: https://www.dropbox.com/s/xub2lck8fe9y3bg/Setup_postfire.exe


WP 3.4: Managing societal response to future fire risk conditions

Task 3.3.1 Economic impacts of future change in fire risk
Climate change scenarios show that by 2070 burned area will be on average higher. Monetized estimates of non use value losses for the Mediterranean area show an average annual loss of $ 200 million and of $ 97 million using the Representative Concentration Pathways (RCP) datasets 8.5 and 2.6 respectively. Subtracting an estimate of the value added generated by an hectare of forest in each Mediterranean country from the contribution that forestry sectors provide to national value added, the burned area can originate a loss from $ 848 million in 2035 to $ 2418 million in 2070 using RCP8.5. Finally, the average yearly social losses related to carbon release into the atmosphere by fire events for the 2020-2070 period amounts to $ 7587 million in the RCP8.5 and to $ 5286 million in RCP2.6. From these results we can conclude that forest fire events can potentially produce significant economic losses. Accounting for market and nonmarket losses across all Mediterranean countries, estimates show a potential annual loss equivalent to 3% of GDP of a rich EU country like Italy. These values should be considered lower bound estimates of the relevant cost and losses categories; e.g. they do not account for impact on health and loss of human life.

Task 3.3.2 Policy implications of changes in fire risk
Weaknesses have been identified in the EU support programmes for the forestry sector. EU legislation and programmes have failed to address landowners’ needs, concerns and objectives, leading to a lack of interest in forestry measures. The major requirements identified are:
• Improve the existing forest fire prevention programs to enhance climate change mitigation and adaptation.
• Increase the science-based approaches to forest fire prevention and forest fire fighting capacity. This will require a stronger nexus between research, policy setting, forestry management organizations and fire fighting institutions. Though now clearly stated in the EU forest strategy it requires a coordinated approach to implementation.
• Limit the administrative burden and simplify the reporting obligations under the Rural Development programs to increase the number of applications.
• The need to integrate the forestry sector into the Emissions Trading System for a more coherent climate policy.

Module 4: Integration, knowledge transfer and management (1.25 pages)
The objectives of this module were to organize the use of data among various partners, by organizaing a data management center, to establish a network of sites where various partner would cooperate in field and modeling experiments, to organize dedicated activities to training and knowledge transfer and to organize the dissemination of the results, including its exploitation. Following are the activities and results of each of the work packages included in this module:

WP 4.1: Data management
In the context of the FUME project, different partners have made available a number of datasets as an outcome of their research activity during the project. These datasets, among other datasets, climate simulations, post-processed observational data, satellite-derived fire scar maps, socioeconomic and LULC data, at different spatial scales and varying temporal domains, leading to a heterogeneous collection of data of different sizes, formats and characteristics, all relevant in the context of fire research. These new datasets are now available through the FUME-THREDDS Data Server, a web server intended for earth system data, that provides metadata and data access for scientific datasets, using OPeNDAP, HTTP and other remote data access protocols (Figure 43 ) . These databases are accessible without further login steps or credentials through this link: http://www.meteo.unican.es/thredds/catalog/FUME/
A user manual for accessing and downloading the data was also developed, with worked examples of the web-based user interface and also using R, a popular open-source environment and language for statistical computing and data analysis.

WP 4.2: Sites network
A sites network with observational, experimental and modeling sites was established in FUME, including 20 sites located in 6 different Mediterranean countries (Figure 44). In experimental and observational sites data were collected at a local scale through manipulative and observational experiments, afterwards they were used for scaling up processes (to regional and continental scales). The compilation and presentation of the information (meteorology, vegetation, soil variables, socioeconomic) available at each experimental site was carried out following a standard format that was defined and agreed by all the partners. Besides, the protocols implemented in the different sites were compiled following a detailed template. A “Sites network” section was implemented in the FUME project website. The information compiled for every site of the project, including the final protocols, was made available to all partners of the FUME consortium through the website. This tool was crucial for supporting and promoting an effective and permanent communication and data exchange between data providers (e.g. field data collectors, modelers) and data users (e.g. modelers), but also between modeling teams and, particularly, for promoting the establishment of synergies among partners. A high number of synergies was established indeed among FUME partners and some synergies involved a large number of teams.

WP 4.3: Training and knowledge transfer
The objective of WP 4.3 was to transfer the knowledge produced in FUME to the scientific community and to the stakeholders, especially fire managers. Two advanced courses were organized by Partner 18-IAMZ/CIHEAM for this purpose.
The first course, titled "Forest Fires in the Perspective of Global Change" (Zaragoza, Spain, 13-17 February 2012), was addressed to 29 participants of academic background; the second course, "Managing Forest Fires to Face Climate and Socioeconomic Change" (Zaragoza, Spain, 20-25 May 2013) was attended by 25 managers and decision makers related to forest fires. The participants, selected among more than 150 candidates, represented 18 different countries (mainly from the Mediterranean basin). The lecturers were members of FUME’s consortium.
The courses offered updated knowledge on the interactions between climate change and other relevant factors of forest fires, on fire risk and fire regimes in general, and their consequences on ecosystems, as well as the goods and services they provide, putting forward some of the approaches and findings developed in FUME. The programme was designed to combine lectures, practical sessions and debate in groups; the second course also included a field visit.
Both courses were positively evaluated by lecturers and by attendants with a global appreciation score of 4.3 points (over 5). Besides the formal knowledge transfer, the interaction and exchanges within the group were considered very fruitful.

Potential Impact:
Societal implications
Gender balance
The project coordinator established as a target an equal gender balance at the project coordination level, which was successfully achieved, with 40% of the work packages being under the responsibility of women. At the level of researcher there was some imbalance, however, the lowest proportion of female researchers being the level of senior researchers, with 30% of all experienced researchers being women scientists. An equal balance of 52% percent was found among the PhD students, and even 54% of all other staff members (see Table 1). These numbers were achieved despite the fact that most partners did not establish gender equality actions, but a few partners did and reported that they were only in part effective. These results on the gender balance of the work force as well as the implementation of gender equality actions could be interpreted that scientist are already aware of the necessity of an equal gender balance in the environmental and socio-economic sciences community.

EC project as opportunities to facilitate scientific career of younger scientists
EU projects offer good opportunities to younger scientist to assume responsibilities in the complicated framework of the multi country cooperation of these projects. In this regard, the FUME project was a particular challenge since it involved participants from 17 countries and all continents. At the commencement of negotiations a goal was set to assign younger experienced scientists, in the middle of their career, the opportunity to lead work packages and gain experience in scientific coordination in an international context. Additionally, an ombudsperson was elected to mediate possible conflicts between partners; the person being elected fell also within this category. While the burden of coordination is heavy, the FUME experience has been most successful, with full achievements of all objectives being set irrespective of who was responsible for them.

Employment
The net effect on employment was a bit difficult to estimate. Relative to the total number of staff, few staff members were additionally recruited, of which more than half of them were female. Most partners re-employed experienced staff, which actually was necessary given the nature of the project. An exception might be the employment of new PhD students, but this could not be ascertained give that information could not be separated based on the origin of this personnel.

Table 1: Workforce statistics in the FUME project
Type of Position Number of
Women Number of Men
Scientific Coordinator 1
Work package leaders 9 15
Experienced researchers (i.e. PhD holders) 33 63
PhD Students 16 18
Other 19 17
How many additional researchers (in companies and universities) were recruited specifically for this project? 23
Of which, indicate the number of men: 14


Other societal implications
The project faced important coordination challenges due to the variety of countries and cultures of the participants. FUME offered an opportunity to integrate partners from the entire Mediterranean basin and partners working in overseas Mediterranean-type ecosystems. This involved partners from Morocco, Algeria, Tunisia, Turkey, Chile and South Africa. Additionally, a number of partners were from the USA, one of them being a member of the advisory board. The challenges for integrating scientists from different scientific and administrative backgrounds into the project were manifold, starting from organisational and financial, to reporting issues. On the scientific side, scientists working in these countries were involved in a lot of field work, working on experiments as well as regeneration and restoration studies, but also modelling exercises. Through that interaction on new scientific knowledge and state-of-the-art science could be established. This also involves the chance for scientists and end-users from non-EU partners to widen their network. Two of the project meetings were held in non EU countries and field experiments were conducted in close collaboration with local stakeholders. In situ discussions were held with the local experts to analyze local problems, including the consequences of very large fires and strategies to guide postfire management. Experts and PhD students from outside the project partner institutions were trained in courses specifically targeted to scientists and decision makers with an impressive number of participants from virtually all countries around the Mediterranean Basin.
The project was not immune to the changes undergoing some partners whose countries were involved in the Arab Spring. This brought in a new dimension on organizational issues within some institutions. In parallel, the northern EU Mediterranean partner institutions were increasingly affected by the economic crisis and were additionally faced with institutional and re-organizational problems. This posed additional management problems and burden on the side of the scientists involved. The lack of flow of money, even if provided from the EC, meant delays in hiring the personnel and other drawbacks. The scientists affected, nevertheless, held on to the scientific plans in FUME to achieve the project goals. This was so despite the fact that, sometimes, their future in their respective institutions remained unclear, irrespective of any scientific achievements.

Dissemination activities and exploitation of results
FUME strategy for dissemination and exploitation of results was supported on four main pillars:
1) Quick communication of FUME results to the scientific community by attending and presenting them at the most relevant scientific and land management meetings, and through publications in high impact international specialized journals. The FUME consortium had published over 80 scientific publications in peer reviewed international journals at the time of completing the project, with many additional manuscripts being in the process of being submitted. The large number of journals and thematic areas associated demonstrated the high multidisciplary character of FUME.
2) Permanent and bi-directional line of communication with stakeholders through the users group. This activity has been realized especially in relation to Module 3 – Adaptation to change. Stakeholders feedback is an essential component of adaptation to the projected new fire regime. Therefore, FUME considered several activities involving stakeholders, both for transferring scientific and technical results and for capturing and make use of local experiences and perception. The main objective was to promote feedbacks between FUME scientists and relevant stakeholders in the field of fire management and post-fire restoration. This is considered critical to optimize the exploitation of FUME results, especially for adaptation to changing climate and fire regime. The main activities related to stakeholders were:
• Interactions with stakeholders for the analysis of fire prevention and fighting plans and for the elaboration of adaptation proposals. Translated questionnaires into local languages (Italian and Spanish) where distributed to selected stakeholders in order to collect the information of existing fire management plans. Main FUME results on the projections of climate and fire regime were presented to the selected stakeholders for analysing adaptation needs. Formal and informal meetings with stakeholders were celebrated in several regions of Spain and Italy.
• Focused meetings and informal discussions with relevant stakeholders were organized in Spain, France and Greece to specifically discuss post-fire restoration approaches and POSTFIRE-DSS. In addition, the post-fire assessment protocol and decision support system was presented to stakeholders in several countries.
• To convey FUME results to a wide audience, and especially to end users related to fire management in a wide sense, we produced factsheets summarizing our most outstanding results in layman language, for each relevant FUME deliverable. This initiative was based on an exercise specifically carried out with stakeholers in which they were asked to assess FUME deliverables with regard to their own needs.
• The IAMZ-CIHEAM organized a FUME training course oriented to fire managers: Managing forest fires to face climate and socioeconomic change, Zaragoza (Spain), 20-24 May 2013. The course produced training material issued from FUME research. Taking advantage of the course, we distributed a questionnaire among the participants to analyse deficiencies or gaps detected by managers when planning for forest fire control (at prevention and fighting level) in the different countries taking into consideration the future as envisaged by researchers from the FUME Project. The questionnaire was translated to English, French and Spanish. We collected information from fire managers from Albania, Algeria, Bulgaria, Colombia, Greece, Morocco, Mexico, Portugal, Tunisia and Turkey.
• Several Spanish stakeholders were invited to the last FUME coordination meeting to the focused session: The way forward (November 28th, 2013) Toledo (Spain). In this session, several proposals were suggested to further exploit FUME results in the near future. Proposals for discussion were suggested in the previous workshops with stakeholders. Main lines of results exploitation included: a) Mapping historical fire scars in the EU. Remote sensing reconstruction of fire scars using LANDSAT images available since the 1970s, complemented with ancillary information of fire causes (National databases), land use and land cover, digital elevation models and other GIS layers such as forest inventories and maps; b) Fire management in natural protected areas; c) Implementation of FUME POSTFIRE-DSS for helping decision making in postfire management in the perspective of climate change.
3) Addressing policy and decision makers, particularly those involved in international negotiations around climate change and national as well as EC agencies involved in planning for adaptation to climate change. José Manuel Moreno, FUME coordinator, has been actively participating in the IPCC WG2 for the AR5 report. In addition, he has collaborated with the Spanish Office for Climate Change. J.M. Moreno and V.R. Vallejo have contributed to the White Paper directed to the United Nations and International Organizations: Vegetation Fires and Global Change – Challenges for Concerted International Actions, J.G. Goldammer ed., in press ( Chapter 9: Current Fire Regimes, Impacts and Likely Changes – VI: Euro Mediterranean).
4) Communication and dissemination of the main results to the public at large through accredited press in the topics of FUME, brochures and web portal open information. Major scientific findings of the FUME project have been published in a brochure, the FUME “Lessons Learned and Outlook”, which addresses the scientific experts outside the FUME field of expertise, decision-makers, including fire managers and respective stakeholders, policy makers as well as the wider public. Results are described and displayed in a general and understandable way. The booklet is freely accessible at the web page of the project (http://www.fumeproject.eu).

List of Websites:
José M. Moreno
Coordinator of FUME project
Departamento de Ciencias Ambientales
Universidad de Castilla-La Mancha
Avda. Carlos III s/n
45071 Toledo, Spain
E-mail: josem.moreno@uclm.es
Tel: +34 925 26 88 00 E5490

Web page: www.fumeproject.eu