Abstract
The use of data from global atmospheric models is an alternative to missing surface meteorological data, but few studies have determined the accuracy of these models and evaluated their use in water balance modeling. We determine the accuracy of the ECMWF ERA-Interim reanalysis dataset and its application in the calculations of water deficiency in Paraná, Brazil. We used 10-day period mean air temperature (ºC) and rainfall (mm) monthly data from 1989 to 2014 from ECMWF ERA-Interim and surface data (INMET). For a more careful evaluation, the meteorological data were stratified by macro regions and seasons. We calculated the climatological water balance according to Thorthwaite and Mather (1955) for both data sources with CAD = 100 mm. The maps were generated using geographic information system (SIG) and the comparisons between the ECMWF ERA-Interim data and the meteorological station data were performed by the indexes that evaluate the accuracy, precision and trend of the data. The TAIR of ECMWF ERA-Interim showed high accuracy in all seasons of the year, with MAPE > 5% and lower than < 2ºC for all Paraná State. The Tukey test proved the high accuracy, as it evidenced by the equality between the TERA-I and TINMET data. The PERA-I consistently underestimated the PINMETs that occurred in the Northwest, North and Southwest macro-regions, showing low accuracies (R2 ≅ 0) and high MAPE (> 15%). The Tukey test proved that the PERA-I and PINMET are different, and thus demonstrates that the PERA-I needs corrections to be used. The accuracy of the DEFERA-I was reasonable demonstrating a confidence index of C = 0.69, which corresponds to a “good” performance.
Keywords:
water balance; bigdata; climate modeling; rainfall; air temperature
