Uncertainties in simulating extreme precipitation events with a Regional Climate Model

N. L. PESSACG; S. A. SOLMAN

Punta del Este

Workshop; CLARIS LPB, WP6 Meeting; 2009

CLARIS-LPB PROJECT

This work focuses on the evaluation of the Regional Climate Model MM5 in representing rainfall over South America. With the aim of quantifying the uncertainties associated with the model and defining an optimal model configuration, several sensitivity experiments were designed and evaluated. In particular, we focused on the representation of frequency and intensity of precipitation extreme events. Sensitivity experiments were carried on to evaluate different combination of convective and planetary boundary layer schemes, model version and treatment of boundary conditions using the nudging technique. The study period was November 1986 associated with rainy (dry) anomalous conditions over La Plata Basin-LPB (South Atlantic Convergence Zone-SACZ), and December 1986 with opposite characteristics. Results showed that the experiments achieved reproduce the variability between both months in LPB region. More frequency of events with more intensity in November than in December linked with the positive phase of SASS. Kain Fritsch convective scheme presented troubles for simulating daily precipitation in LPB region, that problem is associated with the deficiencies that this scheme showed for representing the circulation at low layers. These deficiencies were solved with the nudging option, and this also improves the simulation of daily precipitation in LPB region. Frequency of weak precipitation events was higher than the observed frequency in the most of cases, while the frequency of moderate and intense events was underestimated. Besides that, the experiments showed a level of uncertainties in SACZ region higher than in LPB region, with a major dispersion and range of daily precipitation.The level of uncertainties is bigger in wet months than in dry months. Experiments without nudging failed in representing the value of 90th percentile, in most of cases; this value was underestimated in both regions. Experiments with nudging improve the simulation of 90th percentile.