4.3 Assumptions about future trends

The work reviewed in this chapter is dependent on assumptions of various types that are important in assessing the level of confidence that can be associated with its results (Moss and Schneider, 2000), but can be challenging to quantify and aggregate. Assumptions and uncertainties associated with climate scenarios (Randall et al., 2007) are not considered here, other than to identify the greenhouse gas emission trends or socio-economic development pathways (e.g., SRES, Naki?enovi? et al., 2000) assumed in the literature we review (see also Table 4.1, especially scaling methodology and associated uncertainties). Since the TAR, many global or regional scenarios have become available to quantify future impacts (Christensen et al., 2002, 2007; Meehl et al., 2007), and confidence in future climate projections has increased recently (Naki?enovi? et al., 2000; Randall et al., 2007). However, many assumptions must be made, due to imperfect knowledge, in order to project ecosystem responses to climate scenarios. We provide here a brief outline and guide to the literature of those that are most relevant.

To project impacts of climate change on ecosystems there are basically three approaches: (i) correlative, (ii) mechanistic, and (iii) analogue approaches. For the correlative and mechanistic approaches, studies and insights from the present give rise to the assumption that the same relationships will hold in the future. Three modelling approaches in particular have provided relevant results since the TAR. Firstly, correlative models use knowledge of the spatial distribution of species to derive functions (Guisan and Thuiller, 2005) or algorithms (Pearson et al., 2004) that relate the probability of their occurrence to climatic and other factors (Guisan and Zimmermann, 2000). Criticised for assumptions of equilibrium between species and current climate, an inability to account for species interactions, lack of a physiological mechanism, and inability to account for population processes and migration (see Pearson and Dawson, 2003; Pearson, 2006), these methods have nonetheless proved capable of simulating known species range shifts in the distant (Martinez-Meyer et al., 2004) and recent (Araújo et al., 2005) past, and provide a pragmatic first-cut assessment of risk to species decline and extinction (Thomas et al., 2004a). Secondly, mechanistic models include the modelling of terrestrial ecosystem structure and function. They are based on current understanding of energy, biomass, carbon, nutrient and water relations, and their interacting dynamics with and among species such as primary producers. Such approaches generate projections of future vegetation structure, e.g., as the likely balance of plant functional types (PFTs) after permitting competitive interaction and accounting for wildfire (Woodward and Lomas, 2004b; Lucht et al., 2006; Prentice et al., 2007; but see Betts and Shugart, 2005, for a more complete discussion). Extrapolated to global scale, these are termed Dynamic Global Vegetation Models (DGVMs, see Glossary). An equivalent approach for oceans is lacking (but see Field et al., 1998). Thirdly, Earth system models have begun to incorporate more realistic and dynamic vegetation components, which quantify positive and negative biotic feedbacks by coupling a dynamic biosphere to atmospheric circulations with a focus on the global carbon cycle (Friedlingstein et al., 2003, 2006; Cox et al., 2004, 2006).

Ecosystem- and species-based models are typically applied at scales much finer than are resolved or reliably represented in global climate models. The requisite downscaling techniques of various types (statistical, dynamic) have matured and are increasingly used to provide the necessary spatio-temporal detail (IPCC-TGCIA, 1999; Mearns et al., 2003; Wilby et al., 2004; Christensen et al., 2007). Physically consistent bioclimatic scenarios can now be derived for almost any region, including developing countries (e.g., Jones et al., 2005) and complex, mountainous terrain (e.g., Gyalistras and Fischlin, 1999; Hayhoe et al., 2004). However, major uncertainties relating to downscaling remain in the impact projections presented in this chapter, centring mainly on soil water balance and weather extremes which are key to many ecosystem impacts, yet suffer from low confidence in scenarios for precipitation and climate variability, despite recent improvements (Randall et al., 2007).

Despite the recognised importance of multiple drivers of ecosystem change, they are rarely all included in current climate and ecosystem models used for assessing climate change impacts on ecosystems (Hansen et al., 2001; Levy et al., 2004; Zebisch et al., 2004; Feddema et al., 2005; Holman et al., 2005b; Pielke, 2005). The explicit inclusion of non-climatic drivers and their associated interactions in analyses of future climate change impacts could lead to unexpected outcomes (Hansen et al., 2001; Sala, 2005). Consequently, many impact studies of climate change that ignore land-use and other global change trends (see also Section 4.2.2) may represent inadequate estimates of projected ecosystem responses.