Leonardo D. A. Sá

Non-extensive statistics and three-dimensional fully developed turbulence

In this paper, we present further evidence, based on new data from the Large Scale Biosphere Atmosphere Experiment in Amazonia (LBA), that the generalized thermostatistics provides a simple and accurate framework for modeling the statistical behavior of fully developed turbulence.

Interannual variations of rainfall and corn yields in Northeast Brazil

Abstract Interannual variations of rainfall in Northeast (NE) Brazil are studied giving emphasis to the unusually heavy rainfall during 1984 and 1985. It is found that a more southward location of the Inter-tropical Convergence Zone is probably responsible for the higher rainfall in 1984 and 1985. In 6 of the 9 states, rainfall correlates significantly with com yields. In 7 of the 9 states, the Southern Oscillation (SO) index shows a strong positive correlation with annual corn yields. This suggests that the SO index can be used to predict annual corn yields in NE Brazil. The feasibility of predicting annual corn yields in Northeast Brazil from prior observations of the Southern Oscillation index is examined. It is found that good success can be obtained in some states.

Variability of Carbon and Water Fluxes Following Climate Extremes over a Tropical Forest in Southwestern Amazonia

The carbon and water cycles for a southwestern Amazonian forest site were investigated using the longest time series of fluxes of CO2 and water vapor ever reported for this site. The period from 2004 to 2010 included two severe droughts (2005 and 2010) and a flooding year (2009). The effects of such climate extremes were detected in annual sums of fluxes as well as in other components of the carbon and water cycles, such as gross primary production and water use efficiency. Gap-filling and flux-partitioning were applied in order to fill gaps due to missing data, and errors analysis made it possible to infer the uncertainty on the carbon balance. Overall, the site was found to have a net carbon uptake of ≈5 t C ha−1 year−1, but the effects of the drought of 2005 were still noticed in 2006, when the climate disturbance caused the site to become a net source of carbon to the atmosphere. Different regions of the Amazon forest might respond differently to climate extremes due to differences in dry season length, annual precipitation, species compositions, albedo and soil type. Longer time series of fluxes measured over several locations are required to better characterize the effects of climate anomalies on the carbon and water balances for the whole Amazon region. Such valuable datasets can also be used to calibrate biogeochemical models and infer on future scenarios of the Amazon forest carbon balance under the influence of climate change.

Multiscale analysis from turbulent time series with wavelet transform

We present a multiscale signal analysis based on the multifractal spectrum obtained by the Wavelet Transform Modulus Maxima technique. We analyze time series from turbulent data: the first step is to obtain the PDF of the flutuations for velocities records and then to fit them by means of the Tsallis generalized thermodynamics (Tsallis, J. Stat. Phys. 52 (1998) 479) the second step is to obtain the multifractal spectra of the time series by the wavelet transform (Muzy et al., Phys. Rew. Lett. 67 (1991) 3515). The aim of this approach was to investigate a possible phenomenological connection between the entropic parameter (q) and the multifractal spectrum for turbulence.

Estimating Buoyancy Heat Flux Using the Surface Renewal Technique over Four Amazonian Forest Sites in Brazil

The surface renewal (SR) method was applied for the first time to measurements of air temperature over four Amazonian forest sites and different seasons in order to obtain estimates of buoyancy heat flux. The required calibration of this method against eddy covariance resulted in a value for a specific parameter that is close to the range reported in other studies, contributing to the generalization of the SR method to different kinds of canopies. The comparison with fluxes obtained using the eddy covariance technique revealed a good match between the two methods for different sites, heights and seasons. Sites with high levels of non-stationarity in the signals of temperature and wind speed presented higher scatter in the regression with fluxes from eddy covariance. For a particular site with previously reported influence of low-frequency motions, the regression was only satisfactory, i.e., slope parameter close to unity and small offset, when oscillations with periods longer than

Generalized thermostatistics description of probability densities of turbulent temperature fluctuations

\approx

REACTIVE AND NON-REACTIVE TRACE GAS EXCHANGE WITHIN AND ABOVE AN AMAZONIAN RAINFOREST

≈13 min were filtered out. The SR method has a great potential due to the simplicity of the instrumentation required. However, care should be taken when measuring under the influence of mesoscale motions, which can lead to high levels of non-stationarity, compromising the fundamental concepts of the SR theory.

VARIABILIDADE VERTICAL DE ESTRUTURAS COERENTES NA CAMADA LIMITE CONVECTIVA DA AMAZÔNIA

Neural networks were used to predict the anomalies of the time series of monthly rainfall of the Northeastern Region of Brazil. The forecasts made using a feedforward network with backpropagation algorithm from the original data were not satisfactory. We have therefore tried to combine two advanced methods, Wavelet Transform and Neural networks. Three more types of neural networks were used. The selected neural networks include the Time Delay Neural Networks (TDNN), Radial Basis Functions network and Neural Network Adaptive Wavelet. All networks were implemented in neural network simulator SNNS. The Neural Network Adaptive Wavelet was implemented by changing the standard sigmoidal nonlinearities to wavelet nonlinearities in the neurons. We compare the results obtained with unfiltered and filtered data. Using data obtained by filtering the wavelet transform coefficients significantly improved the results for all networks. The combination of TDNN with wavelet filtered data gave the best results.