Previsão micrometeorológica no Pantanal mato-grossense pela teoria de sistemas dinâmicos

Abstract

Micrometeorological variables measured in ecosystems possess nonlinear behavior and represent the dynamics of the physical and biological processes that occur between the vegetated surface and atmosphere. Thus, the aim of this study was to understand the behavior of a micrometeorological seasonally flooded forest in Pantanal, from the time series analysis of the net radiation, temperature and humidity (with data from 15 to 15 minutes), measured above the canopy, and make predictions using the theory of dynamical systems. In order to justify the approach, we performed tests of nonlinearity. Then the phase space containing the whole attractor associated to each time series is reconstructed by time delay and dimension of immersion respectively by the methods of Mutual Information and False Neighbors Next. Aiming to extract information for the prediction, we characterize the variables according to the analysis of nonlinear time series, the quality of charts presenting recurrence texture seasonal nonlinear quantified by Quantitative Analysis of Recurrence. Also evaluated are the dynamics and geometry of the series, respectively the largest Lyapunov exponent for the correlation dimension. Finally the method is applied in simple non-linear prediction that uses a series measurement to estimate an expected range. It is verified the correlation, accuracy and errors within the time set by parameter reliable forecast. The results show the existence of strange attractors of low fractal dimension and sensitive to seasonality, and that the prediction method captures the simple non-linear dynamics of the overall series micrometeorological measurements.