Transcript P081104084
IPCC Model Classification and Regional Uncertainty Quantification in South America J.-P. Boulanger(1), L. Leggieri(2), A. Hannart(3), A. Rolla(4) and E. Segura(2) (1) IRD, (2) FCEN/UBA, (3) CNRS, (4) CIMA/CONICET INTRODUCTION Uncertainty in regional climate change projections is a combination of two kinds of uncertainties: 1. The uncertainty of the first kind is directly related to the uncertainty in the global mean temperature increase at the time of analysis, which affects the regional mean temperature increase (See Fig. 1; SRES A2) 2. The uncertainty of the second kind is directly related to the how the models converge/diverge in simulating the regional impact of climate change for a given global mean temperature increase (see Fig. 2; SRES A2). Mean global temperature change as a function of latitude for IPCC models (scenario SRES A2) (a) in year 2064 when the ensemble mean increase reaches 2°C (Fig.1; left panel) and (b) when each model reaches a global temperature increase of 2°C (Fig. 2; right panel). The year when each model reaches 2°C is represented in Figure 3 (below). STATEMENT We suggest that, in order to study the regional impacts of climate change, the uncertainty of the second kind should be investigated. To do so, model projections are selected not at a same date, but at a same value of global mean temperature increase (2°C in the following). The different years at which models reach 2°C are an indicator of the model feedback strengths (Fig. 3). METHOD • Then, Self-Organizing Maps (SOM) are used to classify regions where the models present similar types of response to climate change forcing. • Basically, each grid point is associated to a vector of 17 values corresponding to the grid point temperature increase for each model of the scenario (17 models for SRES A2) • The classification displays coherent regions meaning that in these regions, the models are distributed in a similar manner allowing to focus on model physics in objectively selected regions. • Moreover, the regions are climatically coherent (Patagonia, Southern and Northern parts of La Plata Basin, high-Andean plateau, Amazone region,…) suggesting that the model regional response to climate change forcing is coherent with 20th century observed climate phenomena. FIRST RESULTS •A first examination of the model dispersion in each region shows that some models are more extreme than others in their regional response to climate change. •A closer examination in each region allows to identifying dependence between the variables and to digging into the physics (see Region 3 and Region 1) CONCLUSIONS • Regional uncertainty is a combination of two kinds of uncertainties. In order to understand the model physics responsible of the regional projection of climate change, one should mainly focus on the uncertainty of the second kind. • SOM classification method is a powerful tool to analyze the coherent regions in the model ensemble response. • SOM allows to identifying “objective” regions to dig into the model physics and the possible causes of regional model uncertainty • Such a process is crucial to provide better estimate of regional scenarios before analyzing socio-economic impacts of climate change.