Estimativas simplificadas da radiação global na Amazônia

Abstract

The Brazilian Amazon occupies a large territorial strip and presents a high diversity of fauna and flora associated with the abundance of rain and solar radiation. However, due to access and dimensions, there is a lack of information on meteorological elements for applications in research in the environmental, urban areas and their interactions. Among the different alternative ways of obtaining meteorological data, the use of simplified models based on measurements of other elements, can provide the generation of continuous and long data series, thus allowing a better

understanding of spatial and temporal variability, especially of radiation solar. The Angström- Prescott empirical models (which linearly correlates atmospheric transmissivity (Qg/Qo) with the

insolation ratio (n / N)) and the Bristow & Campbell model (which exponentially relates atmospheric transmissivity (Qg/Qo) by daily thermal amplitude (Tmax - Tmin) are widely known and applied in different climatic contexts. The objective of this work was to calibrate and validate the Angström-Prescott and Bristow & Campbell models, in different data groups (total, annual and seasonal) for conventional meteorological stations (EMC’s) distributed in the Brazilian Amazon biome. Twenty and 71 EMC’s were selected with sunshine data (n) and air temperature data, respectively. The data groupings considered were: total (all data) and location (each individualized EMC); these being subdivided into four seasonal groupings: drought (June to August), dry rain (September to November), rain (December to February) and dry rain (March to May). In all these groups, 70 and 30% of the data were selected for calibration of the parameterized coefficients of the models and validations of statistical performance, respectively. The statistical indicative of the mean relative error (MBE), the square root of the mean relative error (RMSE) and Willmott's agreement index (d) were used. The Angström-Prescott model in the radiation estimate showed that the interaction of the annual model was the one that showed

the best statistical performance, with relative deviations (MBE) ranging from -0.13 to 0.11 MJ m- 2 d

-1 , scatter (RMSE) from 1.13 to 2.14 MJ m-2 d -1 and adjustments from 0.88 to 0.98, in these cases, the calibrated linear coefficients (a) varied from 0.2251 to 0.3561, while the angular coefficients (b) ranged from 0.3452 to 0.5277 in all cases the correlation coefficients were greater than 0.80. The Bristow & Campbell model in estimating radiation showed that the interaction of the annual model was the one that showed the best statistical performance, with relative deviations (MBE) ranging from -0.35 to 0.27 MJ m-2 d -1 , spreading (RMSE) from 2.22 to 7.43 MJ m-2 d -1 and adjustment from 0.43 to 0.93; in this case, the calibrated coefficients (a) varied from 0.41 to 0.90, the coefficients (c) from 0.009 to 0.057, coefficients (c) from 1.37 to 2.81 and the correlation above 0.44. The Angström-Prescott and Bristow & Campbell models can be used to estimate global radiation for the Amazon region of Brazil, with the coefficients calibrated annually by EMC, and the use of unified coefficients for the entire biome is not recommended, regardless of the model.