Search

A common task in Earth Sciences is to infer climate information at local and regional scales from global climate models. Dynamical downscaling requires running expensive numerical models at high resolution which can be prohibitive due to long model runtimes. On the other hand, statistical downscaling techniques present an alternative approach for learning links between the large- and local-scale climate in a more efficient way. A large number of deep neural network-based approaches for statistical downscaling have been proposed in recent years, mostly based on convolutional architectures developed for computer vision and super-resolution tasks. This paper presents DL4DS, Deep Learning for empirical DownScaling, a python library that implements a wide variety of state-of-the-art and novel algorithms for downscaling gridded Earth Science data with deep neural networks. DL4DS has been designed with the goal of providing a general framework for training convolutional neural networks with configurable architectures and learning strategies to facilitate the conduction of comparative and ablation studies in a robust way. We showcase the capabilities of DL4DS on air quality CAMS data over the western Mediterranean area. The DL4DS library can be found in this repository: https://github.com/carlos-gg/dl4ds

Direct imaging of exoplanets is a challenging task as it requires to reach a high contrast at very close separation to the star. Today, the main limitation in the high-contrast images is the quasi-static speckles that are created by residual instrumental aberrations. They have the same angular size as planetary companions and are often brighter, hence hindering our capability to detect exoplanets. Dedicated observation strategies and signal processing techniques are necessary to disentangle these speckles from planetary signals. The output of these methods is a detection map in which the value of each pixel is related to a probability of presence of a planetary signal. The detection map found in the literature relies on the assumption that the residual noise is Gaussian. However, this is known to lead to higher false positive rates, especially close to the star. In this paper, we re-visit the notion of detection map by analyzing the speckle noise distribution, namely the Modified Rician distribution. We use non-asymptotic analysis of the sum of random variables to show that the tail of the distribution of the residual noise decays as an exponential distribution, hence explaining the high false detection rate obtained with the Gaussian assumption. From this analysis, we introduce a novel time domain detection map and we demonstrate its capabilities and the relevance of our approach through experiments on real data. We also provide an empirical rule to determine detection threshold providing a good trade off between true positive and false positive rates for exoplanet detection.

We present the most sensitive direct imaging and radial velocity (RV) exploration of $\epsilon$ Eridani to date. $\epsilon$ Eridani is an adolescent planetary system, reminiscent of the early Solar system. It is surrounded by a prominent and complex debris disk which is likely stirred by one or several gas giant exoplanets. The discovery of the RV signature of a giant exoplanet was announced 15 years ago, but has met with scrutiny due to possible confusion with stellar noise. We confirm the planet with a new compilation and analysis of precise RV data spanning 30 years, and combine it with upper limits from our direct imaging search, the most sensitive ever performed. The deep images were taken in the Ms band (4.7$\mu$m) with the vortex coronagraph recently installed in W.M. Keck Observatory's infrared camera NIRC2, which opens a sensitive window for planet searches around nearby adolescent systems. The RV data and direct imaging upper limit maps were combined in an innovative joint Bayesian analysis, providing new constraints on the mass and orbital parameters of the elusive planet. $\epsilon$ Eridani b has a mass of $0.78^{+0.38}_{-0.12}$ $M_{Jup}$ and is orbiting $\epsilon$ Eridani at about $3.48\pm 0.02$ AU with a period of $7.37 \pm 0.07$ years. The eccentricity of $\epsilon$ Eridani b's orbit is $0.07^{+0.06}_{-0.05}$, an order of magnitude smaller than early estimates and consistent with a circular orbit. We discuss our findings from the standpoint of planet-disk interactions and prospects for future detection and characterization with the James Webb Space Telescope., Comment: Accepted to AJ

The NIRC2 vortex coronagraph is an instrument on Keck II designed to directly image exoplanets and circumstellar disks at mid-infrared bands $L^\prime$ (3.4-4.1 $\mu$m) and $M_s$ (4.55-4.8 $\mu$m). We analyze imaging data and corresponding adaptive optics telemetry, observing conditions, and other metadata over a three year time period to characterize the performance of the instrument and predict the detection limits of future observations. We systematically process images from 359 observations of 304 unique stars to subtract residual starlight (i.e., the coronagraphic point spread function) of the target star using two methods: angular differential imaging (ADI) and reference star differential imaging (RDI). We find that for the typical parallactic angle (PA) rotation of our dataset ($\sim$10$^{\circ}$), RDI provides gains over ADI for angular separations smaller than 0.25$^{\prime\prime}$. Furthermore, we find a power-law relation between the angular separation from the host star and the minimum PA rotation required for ADI to outperform RDI, with a power-law index of -1.18$\pm$0.08. Finally, we use random forest models to estimate ADI and RDI post-processed detection limits a priori. These models, which we provide publicly on a website, explain 70%-80% of the variance in ADI detection limits and 30%-50% of the variance in RDI detection limits. Averaged over a range of angular separations, our models predict both ADI and RDI contrast to within a factor of 2. These results illuminate important factors in high-contrast imaging observations with the NIRC2 vortex coronagraph, help improve observing strategies, and inform future upgrades to the hardware., Comment: Accepted to AJ

Post-processing algorithms play a key role in pushing the detection limits of high-contrast imaging (HCI) instruments. State-of-the-art image processing approaches for HCI enable the production of science-ready images relying on unsupervised learning techniques, such as low-rank approximations, for generating a model PSF and subtracting the residual starlight and speckle noise. In order to maximize the detection rate of HCI instruments and survey campaigns, advanced algorithms with higher sensitivities to faint companions are needed, especially for the speckle-dominated innermost region of the images. We propose a reformulation of the exoplanet detection task (for ADI sequences) that builds on well-established machine learning techniques to take HCI post-processing from an unsupervised to a supervised learning context. In this new framework, we present algorithmic solutions using two different discriminative models: SODIRF (random forests) and SODINN (neural networks). We test these algorithms on real ADI datasets from VLT/NACO and VLT/SPHERE HCI instruments. We then assess their performances by injecting fake companions and using receiver operating characteristic analysis. This is done in comparison with state-of-the-art ADI algorithms, such as ADI-PCA. This study shows the improved sensitivity vs specificity trade-off of the proposed approach. At the diffraction limit, SODINN improves the true positive rate by a factor ranging from ~2 to ~10 (depending on the dataset and angular separation) with respect to ADI-PCA when working at the same false positive level. The proposed supervised detection framework outperforms state-of-the-art techniques in the task of discriminating planet signal from speckles. In addition, it offers the possibility of re-processing existing HCI databases to maximize their scientific return and potentially improve the demographics of directly imaged exoplanets., Comment: Accepted for publication in Astronomy & Astrophysics

As planning for the next generation of high contrast imaging instruments (e.g. WFIRST, HabEx, and LUVOIR, TMT-PFI, EELT-EPICS) matures, and second-generation ground-based extreme adaptive optics facilities (e.g. VLT-SPHERE, Gemini-GPI) are halfway through their principal surveys, it is imperative that the performance of different designs, post-processing algorithms, observing strategies, and survey results be compared in a consistent, statistically robust framework. In this paper, we argue that the current industry standard for such comparisons -- the contrast curve -- falls short of this mandate. We propose a new figure of merit, the "performance map," that incorporates three fundamental concepts in signal detection theory: the true positive fraction (TPF), false positive fraction (FPF), and detection threshold. By supplying a theoretical basis and recipe for generating the performance map, we hope to encourage the widespread adoption of this new metric across subfields in exoplanet imaging., Comment: 11 pages, 8 figures, accepted to AJ

HD 141569 A is a pre-main sequence B9.5 Ve star surrounded by a prominent and complex circumstellar disk, likely still in a transition stage from protoplanetary to debris disk phase. Here, we present a new image of the third inner disk component of HD 141569 A made in the L' band (3.8 micron) during the commissioning of the vector vortex coronagraph recently installed in the near-infrared imager and spectrograph NIRC2 behind the W.M. Keck Observatory Keck II adaptive optics system. We used reference point spread function subtraction, which reveals the innermost disk component from the inner working distance of $\simeq 23$ AU and up to $\simeq 70$ AU. The spatial scale of our detection roughly corresponds to the optical and near-infrared scattered light, thermal Q, N and 8.6 micron PAH emission reported earlier. We also see an outward progression in dust location from the L'-band to the H-band (VLT/SPHERE image) to the visible (HST/STIS image), likely indicative of dust blowout. The warm disk component is nested deep inside the two outer belts imaged by HST NICMOS in 1999 (respectively at 406 and 245 AU). We fit our new L'-band image and spectral energy distribution of HD 141569 A with the radiative transfer code MCFOST. Our best-fit models favor pure olivine grains, and are consistent with the composition of the outer belts. While our image shows a putative very-faint point-like clump or source embedded in the inner disk, we did not detect any true companion within the gap between the inner disk and the first outer ring, at a sensitivity of a few Jupiter masses., Comment: 12 pages, 6 figures

Uncovering the ingredients and the architecture of planetary systems is a very active field of research that has fuelled many new theories on giant planet formation, migration, composition, and interaction with the circumstellar environment. We aim at discovering and studying new such systems, to further expand our knowledge of how low-mass companions form and evolve. We obtained high-contrast H-band images of the circumstellar environment of the F5V star HD206893, known to host a debris disc never detected in scattered light. These observations are part of the SPHERE High Angular Resolution Debris Disc Survey (SHARDDS) using the InfraRed Dual-band Imager and Spectrograph (IRDIS) installed on VLT/SPHERE. We report the detection of a source with a contrast of 3.6x10^{-5} in the H-band, orbiting at a projected separation of 270 milliarcsecond or 10 au, corresponding to a mass in the range 24 to 73 Mjup for an age of the system in the range 0.2 to 2 Gyr. The detection was confirmed ten months later with VLT/NaCo, ruling out a background object with no proper motion. A faint extended emission compatible with the disc scattered light signal is also observed. The detection of a low-mass companion inside a massive debris disc makes this system an analog of other young planetary systems such as beta Pictoris, HR8799 or HD95086 and requires now further characterisation of both components to understand their interactions., Comment: 7 pages, 6 figures

HR8799 is orbited by at least four giant planets, making it a prime target for the recently commissioned Spectro-Polarimetric High-contrast Exoplanet REsearch (VLT/SPHERE). As such, it was observed on five consecutive nights during the SPHERE science verification in December 2014. We aim to take full advantage of the SPHERE capabilities to derive accurate astrometric measurements based on H-band images acquired with the Infra-Red Dual-band Imaging and Spectroscopy (IRDIS) subsystem, and to explore the ultimate astrometric performance of SPHERE in this observing mode. We also aim to present a detailed analysis of the orbital parameters for the four planets. We report the astrometric positions for epoch 2014.93 with an accuracy down to 2.0 mas, mainly limited by the astrometric calibration of IRDIS. For each planet, we derive the posterior probability density functions for the six Keplerian elements and identify sets of highly probable orbits. For planet d, there is clear evidence for nonzero eccentricity ($e \simeq 0.35$), without completely excluding solutions with smaller eccentricities. The three other planets are consistent with circular orbits, although their probability distributions spread beyond $e = 0.2$, and show a peak at $e \simeq 0.1$ for planet e. The four planets have consistent inclinations of about $30\deg$ with respect to the sky plane, but the confidence intervals for the longitude of ascending node are disjoint for planets b and c, and we find tentative evidence for non-coplanarity between planets b and c at the $2 \sigma$ level., Comment: 23 pages, 14 figures

For several years, we have been developing vortex phase masks based on sub-wavelength gratings, known as Annular Groove Phase Masks. Etched onto diamond substrates, these AGPMs are currently designed to be used in the thermal infrared (ranging from 3 to 13 {\mu}m). Our AGPMs were first installed on VLT/NACO and VLT/VISIR in 2012, followed by LBT/LMIRCam in 2013 and Keck/NIRC2 in 2015. In this paper, we review the development, commissioning, on-sky performance, and early scientific results of these new coronagraphic modes and report on the lessons learned. We conclude with perspectives for future developments and applications., Comment: To appear in SPIE proceedings vol. 9908

We present the first scattered-light images of the debris disk around 49 Ceti, a approximately 40 Myr A1 main-sequence star at 59 pc, famous for hosting two massive dust belts as well as large quantities of atomic and molecular gas. The outer disk is revealed in reprocessed archival Hubble Space Telescope NICMOS-F110W images, as well as new coronagraphic H-band images from the Very Large Telescope SPHERE instrument. The disk extends from 1."1 (65 au) to 4." 6 (250 au) and is seen at an inclination of 73 deg, which refines previous measurements at lower angular resolution. We also report no companion detection larger than 3 MJup at projected separations beyond 20 au from the star (0." 34). Comparison between the F110W and H-band images is consistent with a gray color of 49 Ceti's dust, indicating grains larger than approximately greater than 2 micrometers. Our photometric measurements indicate a scattering efficiency/infrared excess ratio of 0.2-0.4, relatively low compared to other characterized debris disks. We find that 49 Ceti presents morphological and scattering properties very similar to the gas-rich HD 131835 system. From our constraint on the disk inclination we find that the atomic gas previously detected in absorption must extend to the inner disk, and that the latter must be depleted of CO gas. Building on previous studies, we propose a schematic view of the system describing the dust and gas structure around 49 Ceti and hypothetical scenarios for the gas nature and origin.