Carlos Mejia

A neural network approach for modeling nonlinear transfer functions: Application for wind retrieval from spaceborne scatterometer data

The present paper shows that a wide class of complex transfer functions encountered in geophysics can be efficiently modeled using neural networks. Neural networks can approximate numerical and nonnumerical transfer functions. They provide an optimum basis of nonlinear functions allowing a uniform approximation of any continuous function. Neural networks can also realize classification tasks. It is shown that the classifier mode is related to Bayes discriminant functions, which give the minimum error risk classification. This mode is useful for extracting information from an unknown process. These properties are applied to the ERS1 simulated scatterometer data. Compared to other methods, neural network solutions are the most skillful.

Neural network wind retrieval from ERS‐1 scatterometer data

This paper presents a neural network methodology to retrieve wind vectors from ERS-1 scatterometer data. First, a neural network (NN-INVERSE) computes the most probable wind vectors. Probabilities for the estimated wind direction are given. At least 75% of the most probable wind directions are consistent with European Centre for Medium-Range Weather Forecasts winds (at ±20°). Then the remaining ambiguities are resolved by an adapted PRESCAT method that uses the probabilities provided by NN-INVERSE. Several statistical tests are presented to evaluate the skill of the method. The good performance is mainly due to the use of a spatial context and to the probabilistic approach adopted to estimate the wind direction. Comparisons with other methods are also presented. The good performance of the neural network method suggests that a self-consistent wind retrieval from ERS-1 scatterometer is possible.

Determination of the Geophysical Model Function of NSCAT and its corresponding variance by the use of Neural Networks

We have computed two geophysical model functions (one for the vertical and one for the horizontal polarization) for the NASA scatterometer (NSCAT) by using neural networks. These neural network geophysical model functions (NNGMFs) were estimated with NSCAT scatterometer σO measurements collocated with European Centre for Medium-Range Weather Forecasts analyzed wind vectors during the period January 15 to April 15, 1997. We performed a student t test showing that the NNGMFs estimate the NSCAT σO with a confidence level of 95%. Analysis of the results shows that the mean NSCAT signal depends on the incidence angle and the wind speed and presents the classical biharmonic modulation with respect to the wind azimuth. NSCAT σO increases with respect to the wind speed and presents a well-marked change at around 7 m s−1. The upwind-downwind amplitude is higher for the horizontal polarization signal than for vertical polarization, indicating that the use of horizontal polarization can give additional information for wind retrieval. Comparison of the σO computed by the NNGMFs against the NSCAT-measured σO show a quite low rms, except at low wind speeds. We also computed two specific neural networks for estimating the variance associated to these GMFs. The variances are analyzed with respect to geophysical parameters. This led us to compute the geophysical signal-to-noise ratio, i.e., Kp. The Kp values are quite high at low wind speed and decrease at high wind speed. At constant wind speed the highest Kp are at crosswind directions, showing that the crosswind values are the most difficult to estimate. These neural networks can be expressed as analytical functions, and FORTRAN subroutines can be provided.

Inferring the seasonal evolution of phytoplankton groups in the Senegalo-Mauritanian upwelling region from satellite ocean-color spectral measurements

We have investigated the phytoplankton dynamics of the Senegalo-Mauritanian upwelling region, which is a very productive region, by processing a 13 year set of SeaWiFS satellite ocean-color measurements using a PHYSAT-like method. We clustered the spectra of the ocean-color normalized reflectance (reflectance normalized by a reflectance dependent on chlorophyll-a concentration only) into 10 significant spectral classes using a Self-Organized Map (SOM) associated with a hierarchical ascendant classification (HAC). By analyzing a 13 year climatology of these classes, we have been able to outline a coherent scenario describing the Senegalo-Mauritanian upwelling region in terms of spatiotemporal variability of phytoplankton groups: during the onset of the upwelling (December–February), we mainly observed nanoeukaryote-type phytoplankton in the coastal area; in April–May, the period corresponding to the maximum chlorophyll-a concentration, the nanoeukaryote types were replaced by diatom types. This scenario is in agreement with microscope phytoplankton cell observations done during several past cruises.

Sea Surface Salinity Retrieval for the SMOS Mission Using Neural Networks

During the in-flight phase, using neural networks to retrieve the sea surface salinity from the observed Soil Moisture and Ocean Salinity brightness temperatures (TBs) is an empirical approach that offers the possibility of being independent from any theoretical emissivity model. Due to the large variety of incidence angles, several networks are needed, as well as a preprocessing phase to adapt the observed TBs to the inputs of the networks. When using the first Stokes parameter as an input, the retrieved salinity has a good accuracy (with an error of around 0.6 psu). Furthermore, the solutions for improving these performances are discussed.

Neural network wind retrieval from ERS-1 scatterometer data

Abstract This paper presents a neural network methodology to retrieve wind vectors from ERS1 scatterometer data. First, a neural network (NN-INVERSE) computes the most probable wind vectors. Probabilities for the estimated direction are given. At least 75% of the most probable wind directions are consistent with ECMWF winds (at ±20°). Then the remaining ambiguities are solved by an adapted PRESCAT method, which uses the probabilities provided by NN-INVERSE. Several statistical tests are presented to evaluate the accuracy of the method. Its good performance is mainly due to the use of a spatial context and to the probabilistic approach for estimating the direction. Comparisons with other methods are also presented. The good performance of the neural method suggests that self-consistent wind retrieval is possible.

REMOTE SENSING OPERATIONS

We show that Neural Networks can efficiently model multivalued transfer functions. We propose a method related to conditional density approximation p(y/x) and test the validity of the approach on a remote sensing problem.

Empirical transfer function determination by the use of multilayer perceptron

Abstract Neural Networks are relevant statistical methods to extract information from data when physical phenomena are very complicated and cannot be described in terms of theoretical analysis. Scatterometers are active microwave radar which accurately measure the power of the backscatter signal versus incident signal in order to calculate the normalized radar cross section (σ0) of the ocean surface. We use multilayer perceptrons in order to determine the Geophysical Model Function and to estimate the variability of the signal of ERS-1, ERS-2 and NSCAT scatterometers.