Avaliação do desempenho do produto MSWEP no Rio Grande do Sul

Abstract

Precipitation is a meteorological variable with high spatial and temporal variability, requiring a dense and homogeneous ground monitoring network. In Brazil, this is challenging, due to resources scarcity for equipment implementation in suitable spatial density. Thus, alternatives to observed data on earth's surface can be obtained from Estimated Precipitation Databases (BDPE), which use information from alternative sources to provide precipitation estimates in a grid format. The Multi-Source WeightedEnsemble Precipitation (MSWEP) is an example of alternative database, and like other products, the estimated precipitation may be subject to detection and amount uncertainties. In this work, the performance of MSWEP as an alternative source of precipitation data in Rio Grande do Sul (RS) was evaluated. For this, data observed in rainfall stations were used to identify the quality of MSWEP product through categorical and quantitative indicators used as inference measures. Three spatial validation approaches for MSWEP product were investigated, including the closest cell (CP) of rainfall stations, arithmetic mean (MA) and inverse distance weighting interpolation (IDW) of four closest cells. MSWEP product quality was investigated on a daily, monthly, and seasonal basis. Different correction strategies to MSWEP data were evaluated, and a set of correction coefficients was proposed. MSWEP product was also investigated as a possible source for filling gaps in daily rainfall series observed in rainfall stations. The results show MSWEP tends to underestimate precipitation amounts. MA and IDW spatial validation approaches were more suitable in estimating precipitation amounts, while CP approach provided better results in precipitation detection. MSWEP provided greater uncertainties in estimating small amounts of daily precipitation, notably up to 2 mm.day-1. In general, uncertainties were lower in July, August and September, indicating better performance for winter months. Data estimated by MSWEP also showed the potential to fill up to 50% of gaps in series of rainfall stations. Results of data correction showed the use of correction coefficients related to months of the year and specific periods of data series produced better performance. The correction coefficients were spatialized in maps, which allows them to be applied to any point in RS.