Climate Change in Argentina: Third National Communication- The framework for model selection and development of high resolution climate projections

CAMILLONI, INÉS; SOLMAN, SILVINA; BARROS, VICENTE; CARRIL, ANDREA; NUÑEZ, MARIO

Sao Jose dos Santos

Workshop; IPCC Workshop of WG I on Regional Climate Projections and their Use in Impacts and Risk Analysis Studies; 2015

IPCC

One of the main objectives of the Third National Communication on Climate Change for Argentina was assessing the climate change projections for the twenty first century. Taking advantage of the CMIP5 initiative (Taylor et al., 2012) thelarge number of GCMs available allowed exploring not only the capability of GCMs in reproducing the main features of current climate over Argentina but alsoquantifying the uncertainty in both, present climate and future projections. However,due to the horizontal resolution of GCMs is still insufficient for the impactusers community, GCM-driven Regional Climate Models (RCMs) operating at higher spatial resolution (50 km) performed for the South American continent under the framework of the CLARIS-LPB Project were alsoused. All RCMs were forced by CMIP3-GCMs under the SRESA1B scenario. The number of simulations available, from both the CMIP5 GCM ensemble and CLARIS-LPB RCM ensemble, includes more than 50 models. However, it is well known that model evaluation is the first step in order to build objective criteria for model selection, before exploring the future climate. In this context, the main aim of this work was to assess the capability of both GCMs and RCMs in simulating present climate conditions over Argentina and developing objective criteria for model selection. Evaluating models' performance allowed identifying systematic model biases. Hence, another objective of this work was to remove systematic climate biases in the selected GCMs and RCMs in order to produce a set of high-resolution climate projection dataset (0.5°lat x 0.5°lon) appropriate for the national climate change impact studies.An initial model selection was made on the basis of data availability. The needfor daily data of temperature and precipitation under both current climate andtwo RCP future scenarios (RCP4.5 and RCP8.5) allowed identifying only 14 GCMs meeting the criteria. All participating RCMs were included in the analysis. The model evaluation was made for 4 selected regions within Argentina: Humid Region, encompassing northeastern Argentina, Central Region, Andean and Patagonia.The reference dataset used for evaluating models' behavior for both temperature andprecipitation was the Climate Research Unit (CRU) dataset and the period considered was 1961-1990. An objective integral metric for assessing models?performance was built based on 5 indices quantifying the seasonal mean bias for summer and winter seasons, the annual mean bias, the correlation coefficient between the modeled and observed annual cycle and the ratio between the modeled and observed interannual variability for each variable. The index was computed for each individual model and for each region. After computing the objective integral metric for all the models, a ranking of model was built for each region. The best four models for each region were then selected for evaluating future climate scenarios. Overall it was found there is not a best model identified for every region, but depending on the region the selection of models wit hbetter performance differs. Overall, GCMs tend to perform better than RCMs,except over regions characterized with complex topography. However, even for the outperforming models, both GCMs and RCMs are characterized by systematic errors in the representation of the atmospheric circulation and related variables.Accordingly, systematic biases in daily temperature and precipitation were removed by applying the quantile-based mapping bias correction method(Wood et al., 2002). This methodology consists in constraining thedistributions of these variables produced by climate models to the observed climatology for a target period. This method was already found to be adequate to produce high-resolution bias corrected meteorological information for climate change impact studies (Vidal and Wade, 2007; Saurral, 2010; Saurral et al.,2013). The analysis allowed building a dataset with bias-corrected model outputs for daily temperature and precipitation for both present and future climate conditions for the impact analysis community.