分类： 天文学 >> 天文学 提交时间： 2023-02-19
A. Zavagno
F. -X. Dupé
S. Bensaid
E. Schisano
G. Li Causi
M. Gray
S. Molinari
D. Elia
J. -C. Lambert
M. Brescia
D. Arzoumanian
D. Russeil
G. Riccio
S. Cavuoti
摘要： Context. Filaments are ubiquitous in the Galaxy, and they host star formation. Detecting them in a reliable way is therefore key towards our understanding of the star formation process. Aims. We explore whether supervised machine learning can identify filamentary structures on the whole Galactic plane. Methods. We used two versions of UNet-based networks for image segmentation.We used H2 column density images of the Galactic plane obtained with Herschel Hi-GAL data as input data. We trained the UNet-based networks with skeletons (spine plus branches) of filaments that were extracted from these images, together with background and missing data masks that we produced. We tested eight training scenarios to determine the best scenario for our astrophysical purpose of classifying pixels as filaments. Results. The training of the UNets allows us to create a new image of the Galactic plane by segmentation in which pixels belonging to filamentary structures are identified. With this new method, we classify more pixels (more by a factor of 2 to 7, depending on the classification threshold used) as belonging to filaments than the spine plus branches structures we used as input. New structures are revealed, which are mainly low-contrast filaments that were not detected before.We use standard metrics to evaluate the performances of the different training scenarios. This allows us to demonstrate the robustness of the method and to determine an optimal threshold value that maximizes the recovery of the input labelled pixel classification. Conclusions. This proof-of-concept study shows that supervised machine learning can reveal filamentary structures that are present throughout the Galactic plane. The detection of these structures, including low-density and low-contrast structures that have never been seen before, offers important perspectives for the study of these filaments.
分类： 天文学 >> 天文学 提交时间： 2023-02-19
A. Gkogkou
M. Béthermin
G. Lagache
M. Van Cuyck
E. Jullo
M. Aravena
A. Beelen
A. Benoit
J. Bounmy
M. Calvo
A. Catalano
S. Cora
D. Croton
S. de la Torre
A. Fasano
A. Ferrara
J. Goupy
C. Hoarau
W. Hu
T. Ishiyama
摘要： In the submm regime, spectral line scans and line intensity mapping (LIM) are new promising probes for the cold gas content and star formation rate of galaxies across cosmic time. However, both of these two measurements suffer from field-to-field variance. We study the effect of field-to-field variance on the predicted CO and [CII] power spectra from future LIM experiments such as CONCERTO, as well as on the line luminosity functions (LFs) and the cosmic molecular gas mass density that are currently derived from spectral line scans. We combined a 117 $\rm deg^2$ dark matter lightcone from the Uchuu cosmological simulation with the simulated infrared dusty extragalactic sky (SIDES) approach. We find that in order to constrain the CO LF with an uncertainty below 20%, we need survey sizes of at least 0.1 $\rm deg^2$. Furthermore, accounting for the field-to-field variance using only the Poisson variance can underestimate the total variance by up to 80%. The lower the luminosity is and the larger the survey size is, the higher the level of underestimate. At $z$3$ the variance decreases more slowly with survey size and for example drops below 10% for 1 deg$^2$ fields. Finally, we find that the CO and [CII] LIM power spectra can vary by up to 50% in $\rm 1 deg^2$ fields. This limits the accuracy of the constraints provided by the first 1 deg$^2$ surveys. The level of the shot noise power is always dominated by the sources that are just below the detection thresholds. We provide an analytical formula to estimate the field-to-field variance of current or future LIM experiments. The code and the full SIDES-Uchuu products (catalogs, cubes, and maps) are publicly available.
点击量
待评议
点击量
下载量
评论
