Heat wave over South America: spatial patterns and their relationship with the large scale circulation

ZAMBONI L., A. CHERCHI AND A. F. CARRIL

Foz do Iguassu

Conferencia; The meeting of the Americas; 2010

AGU

In this study we characterize summertime heat waves (HW) over South America (SA) with a focus over La Plata Basin (LPB). The occurrence of a HW is defined as a period of consecutive days in which the daily maximum temperature at 2m is above the 90th percentile. The 90th percentile is computed at each location for each calendar day: this allows the comparison among different regions and different months since the thresholds so defined are normalized to the local climate and automatically filter the submonthly and intraseasonal variability. Typical values of the 90th percentile are 26-28°C over the monsoon region, while higher values (~32-35 °C) are observed over northeastern Brazil and subtropical regions around 60W. The lack of a unique persistence criterion for defining HWs in the literature motivated us to conduct a statistical description of their duration. During the period 1961-2000, as found by using the NCEP-NCAR reanalysis, the typical timescale of HW in SA is 2-3 days, except for the Amazon where extreme temperatures are mostly isolate 1 day events. Over LPB there is on average 1 event lasting 2 days every summer, and 1 event lasting 3 days every other summer. The pattern and frequency of HW lasting 2 and 3 days over LPB is realistically captured by the SINTEXG atmosphere-oceansea-ice coupled GCM. For a parallel identification of the spatial patterns of HW we have implied a generalized Pareto distribution for the excesses of temperature above the 90th percentile: the outcomes confirm the existence of a northern (monsoonal) region where the mechanisms are different from the southern and subtropical western parts of the continent. The identification of co-varying spatial and temporal patterns allows the investigation of the large scale circulation associates with the interannual variability of HW. Results of this analysis together with the capability of SINTEXG in reproducing the observed relationships will be presented at the conference.