Incident shortwave radiation at a site in the Lower Amazon: Derivatives from measurements and from MERRA-2

This study evaluated the precision of the MERRA-2 reanalysis product in estimating the incident shortwave solar radiation in the Tapajós National Forest, Amazonia, comparing it with data observed in 2009. The objective was to analyze the performance of MERRA-2 in representing the variability of solar radiation in a region of high ecological importance and with complex climatic characteristics, especially regarding cloudiness. The results here show a monthly average of incident shortwave radiation of 300 to 350 W m−2 in the rainy season and 420 to 530 W m−2 in the dry season. In the hourly analysis, the maximum observed records were 1030 W m−2 for local measurements and 1085 W m−2 for reanalysis data, both occurring at 2 pm. The correlation method showed an overestimate of MERRA-2, and the analysis of the average and relative average deviation was greater in cloudy-sky conditions, with the error for some months varying above 100 % while the average error showed an overestimate at times of highest solar elevation around 400 W m−2 for clear-skies. The Willmont index had good agreement for clear-sky (0.92) and cloudy-sky (0.68) conditions. The results therefore point to the need for regional adjustments to the reanalysis models to improve the accuracy of the estimates of incident shortwave solar radiation, especially in regions with high atmospheric variability, such as the Amazon.