There are several problematic issues related to defining a metric based on the general formulation given above (Fuglestvedt et al., 2003). A major problem is to define appropriate impact functions, although there have been some initial attempts to do this for a range of possible climate impacts (Hammitt et al., 1996; Tol, 2002; den Elzen et al., 2005). Given that impact functions can be defined, AM calculations would require regionally resolved climate change data (temperature, precipitation, winds, etc.) that would have to be based on GCM results with their inherent uncertainties (Shine et al., 2005a). Other problematic issues include the definition of the temporal weighting function g(t) and the baseline emission scenarios.
Due to these difficulties, the simpler and purely physical GWP index,
based on the time-integrated global mean RF of a pulse emission of 1 kg of some compound (i) relative to that of 1 kg of the reference gas CO2, was developed (IPCC, 1990) and adopted for use in the Kyoto Protocol. The GWP of component i is defined by
where TH is the time horizon, RFi is the global mean RF of component i, ai is the RF per unit mass increase in atmospheric abundance of component i (radiative efficiency), [Ci(t)] is the time-dependent abundance of i, and the corresponding quantities for the reference gas (r) in the denominator. The numerator and denominator are called the absolute global warming potential (AGWP) of i and r respectively. All GWPs given in this report use CO2 as the reference gas. The simplifications made to derive the standard GWP index include (1) setting g(t) = 1 (i.e., no discounting) up until the time horizon (TH) and then g(t) = 0 thereafter, (2) choosing a 1-kg pulse emission, (3) defining the impact function, I(∆C), to be the global mean RF, (4) assuming that the climate response is equal for all RF mechanisms and (5) evaluating the impact relative to a baseline equal to current concentrations (i.e., setting I(∆Cr(t)) = 0). The criticisms of the GWP metric have focused on all of these simplifications (e.g., O’Neill, 2000; Smith and Wigley, 2000; Bradford, 2001; Godal, 2003). However, as long as there is no consensus on which impact function (I(∆C)) and temporal weighting functions to use (both involve value judgements), it is difficult to assess the implications of the simplifications objectively (O’Neill, 2000; Fuglestvedt et al., 2003).