Parametrização de turbulência na previsibilidade de temperaturas mínimas em um modelo de mesoescala

Abstract

The present study aims to evaluate the quality of nocturnal temperature forecast made by a mesoscale numerical model and to understand the reasons behind the difficulties found. To do that, the Weather Research and Forecast (WRF) model is used, with the same configuration employed for operational weather forecast. The model has been ran for the 31 nights of July 2012, and temperature outputs have been compared to hourly observations measured by 26 weather stations scattered over the entire state of Rio Grande do Sul. Four different schemes for turbulence have been considered. Three of them, Yonsei University (YSU), Mellor Yamada Janjic (MYJ) and Bougeault-Lacarrere (BOU) are formulations available from WRF code, while the fourth, Bougeault-Lacarrere Modified (BOU-Mod) is a change made to BOU, aiming at making it less turbulent. A general analysis shows that the different formulations present similar root mean squared errors (EQM), with YSU showing slightly smaller errors than the others. An important discrepancy found refers to the fact that there is an appreciable difference between station real altitude and its altitude in the model, which is given by the height of the closest grid point. When such an altitude difference is corrected by a potential temperature, the errors are enhanced. In this case, the most turbulent formulations, YSU and BOU, tend to overestimate nocturnal temperatures, while the least turbulent ones, MYJ and BOU-Mod, tend to underestimate it. All schemes presented a tendency to underestimate the observed temporal variability. It means that they tend to overestimate the coldest observations and to underestimate the warmest ones. In the most stable nights, all parameterizations showed large EQM and overestimate the temperature. In the least stable nights there were some cases with reduced EQM, but all formulations tended to underestimate temperature, showing that it is necessary to increase the turbulent mixing in this cases. When the different stations are compared, it becomes evident that the height difference between station and model altitudes has a large influence in the nighttime temperature weather forecast. It happens mainly because stations lower than the nearest grid point the modeled winds tend to be larger than observed, causing more intense turbulent mixing and leading to warmer temperatures. The opposite happens in stations higher than the grid point. Such a situation occurs mainly in the more stable conditions, when the lower regions tend to have its surface decoupling from the higher atmospheric levels. The implications of these results and suggestions for improving nocturnal temperature forecasts are presented.