14:15 〜 14:30
[PAE16-03] スパースモデリングによる地球型系外惑星の表面組成の全球マッピング
キーワード：系外惑星、マッピング、スパースモデリング
In the exploration of extraterrestrial life, the characterization of exoplanets has been studied. Direct imaging is one of the fundamental observation methods for characterization. However, even for close terrestrial exoplanets from the solar system, a resolution of a few microarcseconds is required for spatially resolved observations, and even if direct imaging becomes possible, it will only be possible to observe point sources. Kawahara & Fujii (2010, 2011) proposed Spin-Orbit Tomography (SOT) to obtain the two-dimensional spatial distribution of a planet's surface from the temporal variation of observed reflected light without spatial resolution. Recently, Aizawa et al. (2020) obtained highly accurate spatial distribution maps by introducing sparse modeling to SOT.
In addition to the above methods, we focused on Spin-Orbit Unmixing (SOU), which was developed by introducing Spectral Unmixing, a remote sensing method, to SOT. SOU enables us to obtain the spatial distribution of the planet's surface and the reflection spectrum from the luminosity variations at multiple wavelengths (Kawahara 2020). In this study, we introduced sparse modeling to SOU. In SOU, we placed sparsity-inducing constraints and reformulated the solution into an appropriate form to use the proximal gradient method, one of the optimization algorithms. As a result, we obtain highly accurate and more sparse planetary surface distributions and reflection spectra. In this talk, we describe this method and show the results of tests using Earth data.
In addition to the above methods, we focused on Spin-Orbit Unmixing (SOU), which was developed by introducing Spectral Unmixing, a remote sensing method, to SOT. SOU enables us to obtain the spatial distribution of the planet's surface and the reflection spectrum from the luminosity variations at multiple wavelengths (Kawahara 2020). In this study, we introduced sparse modeling to SOU. In SOU, we placed sparsity-inducing constraints and reformulated the solution into an appropriate form to use the proximal gradient method, one of the optimization algorithms. As a result, we obtain highly accurate and more sparse planetary surface distributions and reflection spectra. In this talk, we describe this method and show the results of tests using Earth data.