Desempenho de modelos para a estimativa da radiação de onda longa atmosférica em Petrolina-PE.

Abstract

Atmospheric longwave radiation is of paramount importance in studies of energy balances in the Biosphere-Atmosphere interaction, being decisive in research on climate change at regional and global level, since it is directly related to the temperature of the atmosphere. It is the input parameter of several models of atmospheric circulation and weather forecasting. However, it is rarely measured in Brazilian observational networks because it is one of the components of the most expensive and difficult radiation balance to be monitored in micrometeorological studies. Thus, historically many efforts have been made to create empirical and analytical models to estimate this radiative component through more easily measured variables. The main objective of this work was to evaluate the performance of twelve models for the estimation of atmospheric longwave radiation in Petrolina - PE using data from the SONDA project. The calculations were processed in hourly and daily averages for different cloud conditions (clear sky, partly cloudy sky and cloudy sky) and seasonality (dry and rainy period), using the meteorological variables of water vapor pressure and air temperature at as input parameters of the models. The study period comprised the year 2015. The statistical parameters for validation of the models consisted of: Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Percent Mean Relative Error (PMRE). In addition, Pearson correlation coefficient (r) and coefficient of determination (R²). Therefore, the models that presented the best results from the statistical analysis, being the same indicated for the measurements of the long-wave atmospheric radiation in Petrolina were: Brunt (1932), Brutsaert (1975) Prata (1996), Berdahl and Martin (1984), Berger et al. (1984), Angstrom (1918), Idso (1982) and Lommer et al. (2007).