Estimativa da radiação de onda longa atmosférica horária no estado de Mato Grosso

Abstract

The atmospheric long-wave radiation is essential for studies of energy balances in the biosphere-atmosphere interaction. However, the availability of these measurements still is precarious due to the high cost of the equipment that carries out these measures. Thus, it is necessary to use models that estimate long-wave atmospheric radiation from the most common data found in meteorological stations. The main aim of this thesis was to evaluate some models to estimate the atmosphereric long-wave radiation in the north and south region of the Mato Grosso state. The data were collected at the Baía das Pedras (2014 - 2015) experimental site, located in the Mato Grosso Pantanal, and at Fazenda São Nicolau (2002 - 2003), located in northwestern of Mato Grosso. Data were processed in hourly, daily, monthly, seasonal (dry and rainy seasons) and on clear and partly cloudy days averaged. The models of Swinbank (1962), Idso & Jackson (1969), Idso (1981), Prata (1996) and Duarte et al. (1996) were tested and parameterized for the studied regions. The evaluation of the models was based on the analysis of the mean error, square root of mean square error, absolute mean error, Pearson's coefficient (r) and Willmott's coefficient (d). All atmosferic of long-wave radiation models in the Pantanal region, excepted of Idso (1981), presented high errors before parametrization without classification for cloudiness or season of the year. After parameterization, the Swinbank model (1963) kept presenting the worst performance, while all the others presented lower errors and higher Pearson's correlation and Wilmott's coefficient. Regarding the separation on clear and partly cloudy days, all models had showed improvements, except the Swinbank model (1963). By separating the data according to cloudiness and seasonality, all models have showed worst performance. All models applied in the region of São Nicolau Farm, with no exception, required parameterization, as they presented high errors and low correlation coefficients (r) and Willmott (d). After parameterizing, all models reduced their errors and increased the correlation coefficients (r) and Willmott (d). The models of Idso (1981) and Swinbank (1962) ahd presented better and worse performance, respectively, in that region of Fazenda São Nicolau. There was no improvement in the performance of the models when classifying themselves in terms of cloudiness and seasonality in that region. Then, we conclude that the models of Idso & Jackson (1981), Idso (1981), Prata (1996) and Duarte et al. (2006), can be applied without the need to be classified as cloudiness and seasonality in the Pantanal region; And the Idso model (1981) in the region of Fazenda São Nicolau. The Idso's model have presented low difference among the post-parameterization errors, implying that it could be applied without reparametrization in both studied regions.