Variabilidade Noturna do Vento Acima de Florestas Tropicais: Estudo Comparativo e Classificação Automática

Abstract

The study of meteorological variables in the atmospheric boundary layer (ALC) helps in understanding the forest-atmosphere interaction, more specifically during the night, since the nocturnal atmospheric boundary layer (NLL) phenomena are still poorly understood, especially in the Amazon. Given the specific differences presented by each forest region, the study aims to compare the characteristics of the CLN in two Amazonian regions. In addition, we sought to establish criteria for automatic classification of various nocturnal modalities of forest-atmosphere interaction, through the use of computer programs, whose elaboration is based on the experience obtained with the visual inspection of data developed in the first stage of the project. The methodology developed for automatic classification consisted in the development of the basic block diagram (DBB), which systematizes the main steps of the development process of the algorithm used in the automatic classification of nocturnal meteorological variables. In parallel, data from the Jarú Biological Reserve (Jarú REBIO), in Rondônia, were classified, based on the method of Cava et al. (2004), to be compared with those already existing for the Caxiuanã National Forest. Thus, the results reflected the existing differences, with the predominance of class I (no clouds), in REBIO Jarú, to the detriment of classes IV and V (with clouds), widely observed in Caxiuanã. Another distinct aspect between the two sites was the nighttime variability of the radiation balance, because in Caxiuanã, this variability presented, most of the time, amplitudes that exceeded 10W/m2, while in REBIO, no values exceeding 3W/m2 were found. Therefore, there were marked differences between the nighttime values of the radiation balance and relative humidity at the experimental sites, which were reflected in the proportions found for the classes of events characteristic of the turbulence structure. Furthermore, the programming methods point to the feasibility of using computational procedures for automatic classification.