Diego Fustes

The Gaia astrophysical parameters inference system (Apsis) - Pre-launch description

The Gaia satellite will survey the entire celestial sphere down to 20th magnitude, obtaining astrometry, photometry, and low resolution spectrophotometry on one billion astronomical sources, plus radial velocities for over one hundred million stars. Its main objective is to take a census of the stellar content of our Galaxy, with the goal of revealing its formation and evolution. Gaias unique feature is the measurement of parallaxes and proper motions with hitherto unparalleled accuracy for many objects. As a survey, the physical properties of most of these objects are unknown. Here we describe the data analysis system put together by the Gaia consortium to classify these objects and to infer their astrophysical properties using the satellites data. This system covers single stars, (unresolved) binary stars, quasars, and galaxies, all covering a wide parameter space. Multiple methods are used for many types of stars, producing multiple results for the end user according to different models and assumptions. Prior to its application to real Gaia data the accuracy of these methods cannot be assessed definitively. But as an example of the current performance, we can attain internal accuracies (RMS residuals) on F,G,K,M dwarfs and giants at G=15 (V=15-17) for a wide range of metallicites and interstellar extinctions of around 100K in effective temperature (Teff), 0.1mag in extinction (A0), 0.2dex in metallicity ([Fe/H]), and 0.25dex in surface gravity (logg). The accuracy is a strong function of the parameters themselves, varying by a factor of more than two up or down over this parameter range. After its launch in November 2013, Gaia will nominally observe for five years, during which the system we describe will continue to evolve in light of experience with the real data.

Stellar parametrization from Gaia RVS spectra

Among the myriad of data collected by the ESA Gaia satellite, about 150 million spectra will be delivered by the Radial Velocity Spectrometer (RVS) for stars as faint as G_RVS~16. A specific stellar parametrization will be performed for most of these RVS spectra. Some individual chemical abundances will also be estimated for the brightest targets. We describe the different parametrization codes that have been specifically developed or adapted for RVS spectra within the GSP-spec working group of the analysis consortium. The tested codes are based on optimization (FERRE and GAUGUIN), projection (MATISSE) or pattern recognition methods (Artificial Neural Networks). We present and discuss their expected performances in the recovered stellar atmospheric parameters (Teff, log(g), [M/H]) for B- to K- type stars. The performances for the determinations of [alpha/Fe] ratios are also presented for cool stars. For all the considered stellar types, stars brighter than G_RVS~12.5 will be very efficiently parametrized by the GSP-spec pipeline, including solid estimations of [alpha/Fe]. Typical internal errors for FGK metal-rich and metal-intermediate stars are around 40K in Teff , 0.1dex in log(g), 0.04dex in [M/H], and 0.03dex in [alpha/Fe] at G_RVS=10.3. Similar accuracies in Teff and [M/H] are found for A-type stars, while the log(g) derivation is more accurate. For the faintest stars, with G_RVS>13-14, a spectrophotometric Teff input will allow the improvement of the final GSP-spec parametrization. The reported results show that the contribution of the RVS based stellar parameters will be unique in the brighter part of the Gaia survey allowing crucial age estimations, and accurate chemical abundances. This will constitute a unique and precious sample for which many pieces of the Milky Way history puzzle will be available, with unprecedented precision and statistical relevance.

A cloud-integrated web platform for marine monitoring using GIS and remote sensing. Application to oil spill detection through SAR images

Geographic Information Systems (GIS) have gained popularity in recent years because they provide spatial data management and access through the Web. This article gives a detailed description of a tool that offers an integrated framework for the detection and localization of marine spills using remote sensing, GIS, and cloud computing. Advanced segmentation algorithms are presented in order to isolate dark areas in SAR images, including fuzzy clustering and wavelets. In addition, cloud computing is used for scaling up the algorithms and providing communication between users.

SOM ensemble for unsupervised outlier analysis. Application to outlier identification in the Gaia astronomical survey

Gaia is an ESA cornerstone astronomical mission that will observe with unprecedented precision positions, distances, space motions, and many physical properties of more than one billion objects in our Galaxy and beyond. It will observe all objects in the sky in the visible magnitude range from 6 to 20, up to approximately 10^9 sources. An international scientific consortium, the Gaia Data Processing and Analysis Consortium (Gaia DPAC), has organized itself in several coordination units, with the aim, among others, of addressing the work of classifying the observed astronomical sources, using both supervised and unsupervised classification algorithms. This work focuses on the analysis of classification outliers by means of unsupervised classification. We present a novel method to combine SOMs trained with independent features that are calculated from spectrophotometry. The method as described here can help to improve the models used for the supervised classification of astronomical sources. Furthermore, it allows for data exploration and knowledge discovery in huge astronomical databases such as the upcoming Gaia mission.

Mixing numerical and categorical data in a Self-Organizing Map by means of frequency neurons

Graphical abstractDisplay Omitted HighlightsSelf-Organizing Maps (SOMs) are powerful tools with many applications. Nevertheless, they cannot deal directly with categorical variables.In order to present categorical variables to SOMs, they are usually transformed by binarization. This increases dramatically the dataset dimensionality.NCSOM has been presented in order to cope with categorical or mixed data. However, it presents some drawbacks: categorical and numerical variables are not equally balanced and the method is not convergent.A novel SOM variant, called FMSOM, is presented which is able to deal with numerical and categorical variables, giving the same weight to them and ensuring convergence. A scalable implementation of the method is fully described.FMSOM is applied to a benchmark of well known datasets, composed of categorical and mixed data. The results show the potential of the method to analyze this kind of datasets. Even though Self-Organizing Maps (SOMs) constitute a powerful and essential tool for pattern recognition and data mining, the common SOM algorithm is not apt for processing categorical data, which is present in many real datasets. It is for this reason that the categorical values are commonly converted into a binary code, a solution that unfortunately distorts the network training and the posterior analysis. The present work proposes a SOM architecture that directly processes the categorical values, without the need of any previous transformation. This architecture is also capable of properly mixing numerical and categorical data, in such a manner that all the features adopt the same weight. The proposed implementation is scalable and the corresponding learning algorithm is described in detail. Finally, we demonstrate the effectiveness of the presented algorithm by applying it to several well-known datasets.

On the estimation of stellar parameters with uncertainty prediction from Generative Artificial Neural Networks: application to Gaia RVS simulated spectra

Aims. We present an innovative artificial neural network (ANN) architecture, called Generative ANN (GANN), that computes the forward model, that is it learns the function that relates the unknown outputs (stellar atmospheric parameters, in this case) to the given inputs (spectra). Such a model can be integrated in a Bayesian framework to estimate the posterior distribution of the outputs. Methods. The architecture of the GANN follows the same scheme as a normal ANN, but with the inputs and outputs inverted. We train the network with the set of atmospheric parameters (Teff, logg, [Fe/H] and [alpha/Fe]), obtaining the stellar spectra for such inputs. The residuals between the spectra in the grid and the estimated spectra are minimized using a validation dataset to keep solutions as general as possible. Results. The performance of both conventional ANNs and GANNs to estimate the stellar parameters as a function of the star brightness is presented and compared for different Galactic populations. GANNs provide significantly improved parameterizations for early and intermediate spectral types with rich and intermediate metallicities. The behaviour of both algorithms is very similar for our sample of late-type stars, obtaining residuals in the derivation of [Fe/H] and [alpha/Fe] below 0.1dex for stars with Gaia magnitude Grvs<12, which accounts for a number in the order of four million stars to be observed by the Radial Velocity Spectrograph of the Gaia satellite. Conclusions. Uncertainty estimation of computed astrophysical parameters is crucial for the validation of the parameterization itself and for the subsequent exploitation by the astronomical community. GANNs produce not only the parameters for a given spectrum, but a goodness-of-fit between the observed spectrum and the predicted one for a given set of parameters. Moreover, they allow us to obtain the full posterior distribution...

Applications of Cloud Computing and GIS for Ocean Monitoring through Remote Sensing

This chapter focuses on how to monitor marine spills using powerful tools such as remote sensing and Geographic Information Systems (GIS). On the one hand, remote sensing has been widely used as one of the main ways to periodically monitor large areas, as it allows to obtain data under poor weather conditions and in spite of darkness. We particularly center upon a sensor called “Advanced Synthetic Aperture Radar” (ASAR), which is part of the Envisat payload. On the other hand, GIS have emerged in recent years as a set of standards for data organization and representation that allow themanagement of geographic data.We provide a detailed description of the design and implementation of a tool that provides an integrated framework for the detection and localization of marine spills using remote sensing, GIS, and cloud computing. Cloud computing is used because of the enormous amount of data to be processed and the need of communication between users.

Cloud integrated web platform for marine monitoring using GIS and remote sensing: application to oil spill detection through SAR images

Geographic Information Systems (GIS) have gained popularity in recent years because they provide spatial data management and access through the Web. This article gives a detailed description of a tool that offers an integrated framework for the detection and localization of marine spills using remote sensing, GIS, and cloud computing. Advanced segmentation algorithms are presented in order to isolate dark areas in SAR images, including fuzzy clustering and wavelets. In addition, cloud computing is used for scaling up the algorithms and providing communication between users.

Distributed Genetic Algorithm for Feature Selection in Gaia RVS Spectra: Application to ANN Parameterization

This work presents an algorithm that was developed to select the most relevant areas of a stellar spectrum to extract its basic atmospheric parameters. We consider synthetic spectra obtained from models of stellar atmospheres in the spectral region of the radial velocity spectrograph instrument of the European Space Agency’s Gaia space mission. The algorithm that demarcates the areas of the spectra sensitive to each atmospheric parameter (effective temperature and gravity, metallicity, and abundance of alpha elements) is a genetic algorithm, and the parameterization takes place through the learning of artificial neural networks. Due to the high computational cost of processing, we present a distributed implementation in both multiprocessor and multicomputer environments.