GESTIÓN DE RESIDUOS ÍNDICE
GESTOR DE RESIDUOS DE CONSTRUCCIÓN Y DEMOLICIÓN.
332.03 — Machine Learning and Big Data for Ex- oplanets and Astrobiology: Results from NASA Frontier Development Lab
Daniel Angerhausen1,2
1 Center for Space and Habitability, Bern University (Bern, Switzerland)
2 Blue Marble Space Institute of Science (Seattle, Washington, United States)
We present results from NASA’s Frontier Develop- ment Lab 2018, an Artificial Intelligence/Machine Learning incubator tackling challenges in various fields of space sciences. Herw we focus on the results of the Exoplanet and Astrobiology teams: planet candidate classification in survey data and modeling and retrieval of exoplanet atmospheres and spectra in the context of life detection. A particular focus will be on two data sets produced: a set of 3 million exoplanet spectra calculated with the GSFC Plane- tary Spectrum Generator (PSG) and a set of 150.000 exoplanet atmospheres computed withATMOS. The Exoplanet team used state-of-the-art deep learning models to automatically classify Kepler and TESS transit signals as either exoplanets or false positives (Ansdell et al. 2018, Osborn et al. 2019). TheirAs- tronetcode expanded upon work of Shallue & Van- derburg 2018 by including additional scientific do- main knowledge into the models to significantly in- crease overall performance . The Astrobiology team 1 project demonstrated how cloud computing capa- bilities can accelerate existing technologies and map out previously neglected parameter spaces (Bell et al., 2019). They succeeded in modelling tens-of- thousands of atmospheres over a few days, using the software ATMOS that was originally intended for use in single run applications. In Soboczenski et al., 2018 and Cobb/Himes at al., 2019 the Astrobi- ology 2 team presented a ML-based retrieval frame- work calledINARAthat consists of the first Bayesian deep learning model for retrieval and a data set of 3, million synthetic rocky exoplanetary spectra gener- ated usingPSG(Zorzan et al, in prep.; Himes et al., in prep.). References: Bell, A., et al. 2018, NIPS 2018 CiML workshop; Soboczenski, F., et al. 2018, NeurIPS Workshop on Bayesian Deep Learning, arXiv:1811.03390; Ansdell, M., et al. 2018, ApJL, 869 (1), L7; Shallue, C. J.,
& Vanderburg, A. 2018, AJ, 155, 94 Cobb A., Himes M. et al. 2019, AJ, in press, arXiv:1905.10659 ; Osborn H. et al. 2019, A&A, in press, arXiv:1902.08544
332.04 — Magnetic field effects on the motion of charged dust in rings and discs, motivated by Sat- urn’s spokes.
Mia Mace1
1 School of Physics, University of Bristol (Bristol, United Kingdom) Spokes are a transient feature in Saturn’s rings. Widely accepted to be charged dust levitated away from the main ring plane, the exact mechanisms and physical environment leading to their formation and propagation is still shrouded in mystery. Sev- eral theories have been proposed, including micro- meteoroid impacts and lightning. However, none can fully explain the appearance of spokes.
This study is an ongoing exploration into charged dust motions in rotating tilted dipole magnetic fields. A systematic investigation of the orbital dynam- ics and resonances in Saturn’s rings has been per- formed with simulations. The numerical method used is flexible and allows a comparison to the other non-diffuse ring systems of our solar system, where spokes are not observed. The aim is a greater un- derstanding of why spokes are only observed in Sat- urn’s B ring. Extending the work beyond planetary rings, predictions can be made about the existence of spoke-like features in more extreme solar systems, which are analogous in structure — a central body and disc, such as white dwarfs.
332.05 — Persistent Homology of Flows on Extraso- lar Planets
Jack William Skinner1; James Cho2; Heidar Thrastarson3
1 Astronomy Unit, Queen Mary University of London (London, United Kingdom)
2 Flatiron Institute (New York, New York, United States) 3 Jet Propulsion Laboratory, California Institute of Technology (Pasadena, California, United States)
High-resolution simulations and observations gen- erate copious amounts of high dimensional, large volume, heterogeneous datasets, which are increas- ingly difficult (if not prohibitive) for analysis by tra- ditional (statistical, spectral, or graphical) methods alone. Persistent homology is a novel computa- tional method for practically ascertaining the topo- logical ‘shape’ of such data. Here the shape is characterized by tallying the number of connected elements and n-dimensional ‘holes’ (e.g., closed loops, three and higher dimensional voids, etc.), as
well as ‘coves’ (depressions or protrusions on the holes), in the data. An example is the recent high- resolution, long-duration simulations of hot-Jupiter atmospheres that produce highly complex flow and temperature fields, containing up to many thou- sands of storms across a wide range of spatial and temporal scales. To clearly demonstrate the efficacy of the homology analysis method, we use it to ana- lyze an idealized vortex model of these storms, fo- cusing on the nonlinear evolution of such storms at the extremely high Reynolds number associated with planetary flows. Features, such as the num- ber of storms and filaments around their periphery, their ‘tubular’ or ‘blobby’ morphologies, and peri- odic bursts of instability are captured and quantified. Understanding such features is crucial for validating theory and numerical models, as well as for inter- preting and guiding observations. Broadly, homo- logical analysis is a widely applicable tool that can help to directly address the large data problem faced in many areas.
332.06 — Extreme Solar Systems and the Fermi paradox : limits to growth?
Aurelien Crida1,2
1 Lagrange, Université Cote d’Azur (Nice, France) 2 Institut Universitaire de France (Paris, France)
The amazing data we now have on exoplanets allows us to estimate that our Galaxy hosts about 50 billion terrestrial exoplanets. On the other hand, on our own planet, life appeared very quickly, then devel- oped slowly, until a remarquable acceleration in the last 500 million years that lead to mankind. In the developement of our civilisation, a similar exponen- tial acceleration is observed, with the distance fled by man made objects multiplied roughly by 10 every ten years since the first plane flew accross a field at the dawn of the XXth century. At this pace, we should conquer the whole Milky Way by the end of this cen- tury. But although our planet formed nine billion years after the first stars, none of the many exoplan- ets seems to be just a century ahead of us : they have not sent their flying saucers to Earth.
This paradox, first coined by E. Fermi, shows that something must be wrong in the above reasoning. In fact, it is obvious to every participant of this confer- ence that the laws of physics forbid to cross a hun- dred thousand light-years in just a hundred years, so that exploring the whole Galaxy can not be done so fast. This illustrates that any exponential process reaches its limits quickly : the distance we explore can not keep being multiplied by ten every ten years. This is also true for the economic growth, which is
also exponential, and based on the use of ressources and the emission of pollution in a finite world ...
The CO2 emissions already change the Earth’s cli- mate, in such a way that they represent a threat for our future developement. Have all the other extrater- restrial lives destroyed their environment by trying to conquer the Galaxy? Have they spontaneously de- cided to stop their exponential growth in order to preserve their planet? The fact that none of them succeeded to come here forces us to think about our own behaviour. In particular, this conference will certainly be fantastic, and I will be enjoying it a lot. But is it reasonable that we all fly here? Should our community start thinking about how to do science with less greenhouse gasses emission?
332.07 — Formation and evolution of short-period planets around magnetized host stars
Douglas NC Lin1,2
1 Astronomy and Astrophysics, University of California, Santa Cruz (Santa Cruz, California, United States)
2 Institute of Advanced Studies, Tsinghua University (Beijing, China)
During the formation and early evolution of short- period planets, kiloguass fields persist on their host stars. Their interaction with planets’ evolving natal disks determines not only the amount of building block grains and the orbital destiny of protoplanets, but also emerging stars’ spin rate. After the disk de- pletion, the relative motion between the stellar spin and the planets’ orbit leads to unipolar induction and Lorentz force which can cause significant orbital evo- lution, Ohmic dissipation, and low-frequency radio emission. I show how this effect may enable remote sensing of super Earths’ surface composition. I also show how to generate analogous magnetic field in proto-Jupiters and how these process may have de- termined the spin rate of Jupiter and the orbits of its Galilean moons.
332.08 — Forget Limb Darkening Laws: Transit Modeling Using Stellar Atmosphere Intensities
Jerome Orosz1; Donald R. Short1; William F. Welsh1;
Gur Windmiller1
1 Astronomy, San Diego State University (San Diego, California, United States)
Limb darkening (LD) laws are ubiquitous in their use in exoplanet transit modeling. They provide a fast and easy way to parameterize the changing intensi- ties across the disk of the star, allowing transit mod- els to be quickly computed and compared with ob- servations. However, we know that LD laws are a
poor representation of the stellar intensities, partic- ularly at the limb where the intensities drop off ex- tremely quickly. This is especially important when considering subtle effects such as planet oblateness, since these effects are most pronounced at the limb. In spite of the shortcomings, researchers often do not use the LD laws that match stellar models. Instead, they fit for the coefficients of a LD law. This fur- ther removes the transit modeling from actual stel- lar physics. While the fit to the transit may be better when solving for LD coefficients, it is no more than a consequence of (i) allowing more freedom in the models; (ii) a re-statement that LD laws are not real- istic representations of the stellar intensities.
It would be far more advantageous to use an ac- tual model atmosphere to give the specific intensi- ties, and we provide a fast method to do so, us- ing tabulated model intensities from ATLAS mod- els to compute the transit. The code is only slightly slower than the one using ad hoc LD laws. By us- ing actual model atmosphere intensities, there are no free parameters to solve for (given the stellar prop- erties). Comparison of this method with the tradi- tional method using ad hoc LD laws shows the de- rived planet radius is systematically wrong by 0.1% or more, depending on the impact parameter. In ad- dition to improving the accuracy (rather than pre- cision) of transit modeling, we note an area where our method may be extremely fruitful: transit spec- troscopy. The wavelength-dependent transit depth is a function of both the planet’s atmosphere and the stellar LD variation. By eliminating the use of param- eterized LD laws we (i) completely remove any de- generacy between transit depth and LD coefficients; (ii) include astrophysical knowledge of the star’s in- tensity distribution, which varies strongly as a func- tion of both position and wavelength.