Resumen:

The detrimental effects of light pollution on astronomy, ecology, and human health are intimately linked to the development, and subsequent misuse, of the rapidly evolving lighting technology. Because of the so-called LED revolution, in the course of the last decade, the Spectral Power Distribution (SPD) of man-made light pollution has been shifting from one containing a few emission lines to one containing dozens - if not hundreds - of lines, along with a significantly increased continuum level in the blue-green spectral range. Because of this SPD reshaping, there is ever growing evidence that most of the photometric technologies and techniques that have so far been used to monitor light pollution worldwide are not able to cope with the massive and fast spread of LED sources. Therefore, the only way to make a significant leap forward in the quantitative characterization of this phenomenon is through spectroscopy, a technique capable not only of disentangling the different contributions to the Night Sky Brightness (NSB), but also of following its complex, multiperiodic variability. In this talk, I will introduce our interdisciplinary project aimed at developing a portable optical spectrograph for the first-ever spectral monitoring of the Chilean night sky. This device will be capable of capturing low-resolution flux-calibrated spectra of large sections of the celestial sphere at once. With its lightweight, modular, and versatile design, it will enable a comprehensive spectral characterization of both natural and artificial sources of NSB. We will eventually gain an accurate understanding of how light pollution affects the upper half of the Chilean horizon, a natural and cultural heritage that is our scientific, social, and moral obligation to protect and preserve.

Resumen:

During the last decade several networks for automatic detection of fireballs have been deployed, with the main scientific goal of enabling a rapid recovery of meteorites and determination of its parent's pre-atmospheric orbit. The geographical distribution of these networks is heavily biased towards the Northern hemisphere. The Bocosur network is a contribution to the global deployment of automated fireball networks, and particularly to the unbiasing of their geographical distribution, since it is located in Uruguay, South America (Lat: -30° to -35°). Its main scientific goal is the detection of fireballs of asteroidal origin, massive enough to produce meteorites, and also to inspire secondary-level students and teachers through their involvement in this citizen-science oriented project. The deployment of this network started in 2019, and was completed in March, 2023, when we installed 20 stations separated ~120 km, covering an area of ~ 180,000km2. During this period of time, one major technological upgrade was made when we migrated from a well-known camera to a higher-resolution, more sensitive system. We were able to build a completely autonomous system at an affordable cost that can be replicated in all the stations. A comparison between the astrometric and photometric performance of these two detection systems is reported. Also, a photometric methodology for estimating the brightness of very bright fireballs is presented and validated against the known magnitudes of Jupiter and the full Moon. We obtain mean residuals of the astrometric reduction of ~5', and the discrepancy between the obtained brightness of Jupiter and the Moon average to 0.18 and 1.0 magnitudes, respectively.

Resumen:

The transient objects known as supernovae constitute one of the most exciting laboratories of study in various areas of astronomy because of their involvement in different physical processes. The large influx of current data leads us to search for new ways to correctly classify these objects in the shortest amount of time possible, even when reliable labels are scarce. In this paper, we present a novel approach that extends the Minimax Entropy semi-supervised learning algorithm to the classification of multiband time series data and tests it on observational light curves of different supernova categories obtained by forced photometry from the Zwicky Transient Facility (ZTF). We quantify the performance of our classifier according to the number of real supernova light curves delivered during the learning period as well as the evolution of its performance based on the number of observations per supernova available when predicting with new data. Our results show that the classification accuracy improves over the lifetime of the transient as more photometric data become available. We demonstrate the ability of our model to provide early classifications of observed transients from the ZTF data stream compared to fully supervised classification models.