Analysis of uncertainties in future climate projections for South America: comparison of WCRP-CMIP3 and WCRP-CMIP5 models

J. BLÁZQUEZ, M. N. NUÑEZ

Lima

Workshop; Workshop on Latin America and Caribbean (CORDEX LAC): Phase I; 2013

This paper identifies two sources of uncertainties in model projections of temperature and precipitation: internal and inter-model variability. Eight models of WCRP-CMIP3 and WCRP-CMIP5 were compared to identify improvements in the reliability of projections from new generation models. It was found for precipitation as well as for temperature that inter-model variability in both model packages increases with time. It was also found in all seasons and in both datasets that internal variability is more important in the early decades, whereas inter-model variability became important in the last decades of the 21st century. Regard to temperature, while no significant differences was observed between both datasets, some improvements were found in the new generation models. For example, in summer CMIP5 model variability of temperature was lower than CMIP3 inter-model variability over northeastern Argentina, Paraguay and northern Brazil, at the end of the 21st century. Reliability of temperature projections from both sets of models was analyzed by calculating the signal to noise ratio. It was found that reliability is high, with values greater than 1 over most of the study region. This is due to all the models project positive changes over most of the study region. Regarding precipitation, although no major differences were observed in both precipitation datasets, CMIP5 model variability was lower over northern and eastern Brazil in summer (especially in the last decades of the 21st century). Reliability of precipitation projections was low in both datasets, with values minor than 1 in most of the study domain. However, some improvements were detected in CMIP5 models. For example, in summer and autumn values of signal to noise ratio were close to 1 and even greater than 1 over eastern Argentina, Uruguay and southern Brazil. Besides, some values greater than 1 were observed in autumn. The no significant improvements mainly in the inter-model variability of CMIP5 models regards to CMIP3 could be attribute to the fact that new generation models include new feedbacks (like carbon cycle) or new parameterizations (especially those refers to cumulus and radiation schemes were improved or changed in some CMIP5 models). Maybe these changes would lead to new sources of uncertainties. It is interesting to note that internal variability is the lower limit of uncertainty, and difficult to reduce. However, the source of uncertainty that might be reduced in the future is the inter-model variability, as long as better representation is achieved of all the physical and chemical processes in the climate system.