Lazarus dF star
Number of posts : 3337 Registration date : 2008-06-12
| Subject: Classifying planets by colours 24th July 2018, 3:09 pm | |
| Batalha et al. "Color Classification of Extrasolar Giant Planets: Prospects and Cautions" https://arxiv.org/abs/1807.08453From the conclusion: - Quote :
- There is a strong correlation between atmospheric properties and WFIRST-like optical filters (Fig. 6). However, these correlations only truly exist for a population that does not have significant cloud coverage in the visible part of the atmosphere. If a full sample of cloud-free and cloudy planets is considered, there are less strong correlations between atmospheric properties and photometric bins that can be leveraged to classify planets.
- For giant planets, it is only possible to classify planets into physically motivated groups with greater than 90% accuracy if it is known a priori that the planet does not have significant cloud coverage in the visible. However, observations of Solar System and exoplanet giant planets suggest clouds are prevalent in most planetary atmospheres.
- Our machine learning algorithms are unable to classify planets by metallicity with greater than 55% accuracy. However, we are able to classify planets with moderate accuracy ∼70% by classifying by the cloud sedimentation efficiency, fsed. Additionally, binary classification between cloudy and cloud-free populations are successful with an accuracy >90%.
- We find that at least three filters are needed for any kind of classification (cloud, metallicity, etc). We also tested our classification algorithm with another filter set proposed by Krissansen-Totton et al. (2016), but do not find more optimistic results. However, our statistical algorithm is open-source and can be used to determine optimal filters for WFIRST or future mission concepts.
| |
|