Regionalization of the GOES-10 retrieval algorithm for tropical South America

Abstract

Meteorological satellites provide a unique opportunity to obtain thermodynamic profiles in regions of the globe that do not have a dense meteorological upper-air station network, such as South America. Temperature and mixing ratio profiles were retrieved every hour with a spatial resolution of 10 km over South America from July 2007 to February 2009 using radiances from the Geostationary Operational Environmental Satellite 10 (GOES-10). The GOES-10 retrieval algorithm for thermodynamic profiles was developed by the Cooperative Institute for Mesoscale Meteorological Studies (CIMMS) in the USA, so some adjustments had to be made for its application on South America, such as the construction of a new covariance matrix. In this context, the scientific focus of this research was the construction of a new covariance matrix adapted to the meteorological conditions in South America. In addition, the algorithm results were validated by the use of the original and the new covariance matrices. The variables validated were air temperature, mixing ratio vertical profiles and total precipitable water (TPW). The data set used comprised a total of 1100 radiosonde observations recorded in the tropical region of South America at 00:00 and 12:00 Universal Time Clock (UTC), as well as thermodynamic profiles from 12 h forecasts from the Centre for Weather Forecast and Climate Studies's (CPTEC's) Global Model, used as a first guess, and upwelling radiances of 18 infrared channels from GOES-10 for the period from July to November 2007. In general, the results indicated that with the regionalization of the covariance matrix, the algorithm performed better retrievals than with that of the original matrix. The greatest improvements were found in the mixing ratio profiles and in the values of the TPW. These results could be associated with the presence of the Amazon rainforest, which incorporated a greater amount of moisture into the new covariance matrix than the previous had the previous one.