In recent decades, changes in climate have caused impacts on natural and human systems on all continents and across the oceans (IPCC, 2012). However, increases in climate variability have a greater effect on society than do changes in mean climate because it is more difficult to adapt to changes in extremes (Seneviratne et al., 2006). Europe was struck by record breaking extreme events, namely the mega-heat waves of 2003 in Europe (Trigo et al., 2005) and 2010 in Russia (Barriopedro et al., 2011; Bastos et al., 2014), and the large droughts in southern Europe in 2005 (Garcia-Herrera et al., 2007; Gouveia et al., 2009) and 2012 (Trigo et al., 2013). The last IPCC assessment on extreme events (IPCC, 2012) confirms that a changing climate can lead to changes in the frequency, intensity, spatial extent, duration, and timing of weather and climate extremes that combined with larger exposure can result in unprecedented risk to humans and ecosystems.
Several studies have also stressed the role played by recent climate change in the increase likelihood of occurrence of some of these extremes (Donat et al., 2013; Sillmann et al., 2013; Seneviratne et al., 2014; Ummenhofer and Meehl, 2017).
Conversely, compound climate extremes (concurrent occurrence of different extremes) have not been equally addressed, but are now receiving increasing attention because of their uneven impacts on humans and ecosystems (Lindner et al., 2010).
Risk assessments, however, generally focus on univariate statistics even when multiple stressors are considered. For example, concurrent extreme droughts and heat waves can substantially affect vegetation health, prompting tree mortality, and thereby facilitating insect outbreaks and fires (Mazdiyasni and AghaKouchak, 2015; Frank et al., 2015). Moreover, high temperatures may amplify drought effects by increasing water vapour pressure deficits and soil water stress (De Boeck et al., 2011). In a future climate, elevated CO2 may buffer effects of drought on vegetation productivity by increasing water use efficiency (Zhu et al., 2016). In addition, droughts and flash-droughts have the potential to trigger and intensify fires (Turco et al., 2017; Russo et al., 2017) and can cause severe economic damage (Mo and Lettenmaier, 2015), with no time for early-warnings in the case of flash-droughts (Mo and Lettenmaier, 2015). Thus, evidence on observed impacts as well as to climate change projected impacts on terrestrial ecosystems are mounting (Bastos et al., 2014; Gouveia et al., 2009; Lindner et al., 2010), suggesting significant vulnerability of both forest and agricultural ecosystems (Páscoa et al., 2017; Turco et al., 2017; Kurz-Besson et al., 2016; Gouveia et al., 2016).
While changes in regional temperature and precipitation patterns may create better growing conditions for forest ecosystems in large areas (Lindner et al., 2010), the extinction of certain species from affected areas allows for large losses of sequestered carbon thus fueling CO2 emissions (Frank et al., 2015). In semiarid/arid biomes water scarcity strongly limits carbon sequestration, whereas the response in temperate biomes is more uncertain (Bloor et al., 2010). In fact, the intra-annual variability of precipitation may have a stronger effect on carbon balance than precipitation amount (Bastos et al., 2016) and the combined effect of former factors is crucial on dry biomes, as it determines critical soil moisture thresholds (Vargas et al., 2012). While climate extremes such as droughts, heat waves, or flash-droughts, hereafter HDE (Hot and Dry extremes), can cause substantial changes in terrestrial C fluxes, extreme changes in C fluxes are often, but not always, driven by extreme climate conditions (Frank et al., 2015; Zscheischler et al., 2014). Furthermore, HDE are key drivers for vegetation stress (Gouveia et al., 2009; Gouveia et al., 2012) and potentially responsible for crop yield and wood losses (Granier et al., 2007; Popova et al., 2014; Eilmann et al., 2011; Kurz-Besson et al., 2016). So, continuous monitoring of vegetation activity and a reliable estimation of HDE’ impacts is crucial to reduce potential risks.
Agricultural and forest risk management aim to mitigate crop and forest growth losses (Iglesias and Quiroga, 2007; Adams et al., 2009; Dalezios et al., 2014; Albert et al., 2015), highlighting the severe damages that can occur in the case of a concurrent effect of high temperature (Adams et al., 2009). The concept of risk is associated with the analytical components of risk analysis, i.e.: hazard, exposure and vulnerability (IPCC, 2012). Recently, Carrão et al. (2016) shown global maps of drought risk and found that drought hazard is generally high for the semiarid areas such as the Iberian Peninsula. Hence, HDE predictability and forecasting plays a major role in risk management to promote adaptation and mitigation measures that contribute to minimize the impacts resultant from these extreme events.
Several studies have focused on agricultural and forest risk assessment (Popova et al., 2014; Carrão et al., 2016; Petr et al., 2014), addressing the adverse effects of HDE in agriculture and forests worldwide, mostly using statistical approaches. Alternatively, Artificial Neural Networks (ANN) models are promising tools often regarded as a good compromise between simplicity and effectiveness (Morid et al., 2007; Jiang et al., 2004).
Due to the complexity of the non-linear character of the agricultural systems under drought conditions, ANN are good alternatives to classical statistical algorithms (Jiang et al., 2004). Usually, local risk management strategies focus only on short-term climatic events without considering long-term climate changes (IPCC, 2012), such as vegetation and soil moisture changes. According to the latest IPCC report (Trigo et al., 2013) the increased summer dryness in semi-arid regions (e.g. Mediterranean) reduces plant growth and survival rates leading to an evapotranspiration reduction and increased warming. The use of methodologies which account for different variables and time-scales (Carrão et al., 2016; Petr et al., 2014) and include compound events were used for applications involving future climate scenarios or for the development of predictive tools (Quesada et al., 2012). General Circulation Models (GCMs) provide a powerful tool to evaluate recent trends in a broader temporal context and to investigate the underlying mechanisms (Soares et al., 2012). Nevertheless, currently used GCMs have coarse horizontal resolutions being unable to represent many land-atmosphere interactions and systems that drive regional and local climate variability (Soares et al., 2012). Alternatively, the Regional Climate Models (RCMs), namely the Weather Research and Forecasting model (WRF, 2km to 9km resolutions), are physically consistent with regional and local circulations at finer horizontal and temporal scales (Soares et al., 2012; Cardoso et al., 2013).