Predicting Reliable Dust Temperature Maps from Molecular Line Data Using Machine Learning

Tenta Dougome  (Kagoshima University)

To estimate the mass of the molecular cloud, the dust temperature is required. Furthermore, it is crucial to understand the physical condition of the star formation. To estimate dust temperature, it is necessary to observe the continuum with multi-wavelength. However, the areas observed with multi-wavelength and spatial resolution required for star formation studies are limited. Therefore, to estimate the dust temperature in area where the dust temperature estimated from the Herschel data is not available, we constructed a model predicting dust temperature from molecular line data using a machine-learning method called Extra Tree Regressor. we used observational data of dust temperature in the Orion A molecular cloud obtained by the Herschel Space Observatory as teacher data. The molecular line data of 12CO(1-0), 13CO(1-0), and C18O obtained by the Nobeyama 45-meter radio telescope were used as feature data. Applying the model constructed from the features and teacher data to the entire Orion A molecular cloud, we obtained the predicted dust temperature. Our prediction model achieved an accuracy of 20.85±4.53 K compared to the observed value of 20.25±4.61 K, demonstrating high accuracy. The results of this study indicate that machine learning techniques are effective in predicting dust temperature from molecular line data.