Machine learning-based seeing estimation and prediction using multi-layer meteorological data at Dome A, Antarctica

Xu Hou,Yi Hu,Fujia Du,Michael C. B. Ashley,Chong Pei,Zhaohui Shang,Bin Ma,Erpeng Wang,Kang Huang

2023-04-07

Abstract:Atmospheric seeing is one of the most important parameters for evaluating and monitoring an astronomical site. Moreover, being able to predict the seeing in advance can guide observing decisions and significantly improve the efficiency of telescopes. However, it is not always easy to obtain long-term and continuous seeing measurements from a standard instrument such as differential image motion monitor (DIMM), especially for those unattended observatories with challenging environments such as Dome A, Antarctica. In this paper, we present a novel machine learning-based framework for estimating and predicting seeing at a height of 8 m at Dome A, Antarctica, using only the data from a multi-layer automated weather station (AWS). In comparison with DIMM data, our estimate has a root mean square error (RMSE) of 0.18 arcsec, and the RMSE of predictions 20 minutes in the future is 0.12 arcsec for the seeing range from 0 to 2.2 arcsec. Compared with the persistence, where the forecast is the same as the last data point, our framework reduces the RMSE by 37 percent. Our method predicts the seeing within a second of computing time, making it suitable for real-time telescope scheduling.

Instrumentation and Methods for Astrophysics

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to predict atmospheric seeing at Dome A in Antarctica by using multi - level meteorological station data and machine - learning methods. Specifically, the researchers developed a new framework based on Long Short - Term Memory network (LSTM) and Gaussian Process Regression (GPR) to estimate and predict the atmospheric seeing at a height of 8 meters. This framework aims to overcome the difficulties of long - term continuous seeing measurement using standard instruments such as the Differential Image Motion Monitor (DIMM) in extreme environments like Dome A. In addition, this framework can quickly predict seeing and is suitable for real - time telescope scheduling, thereby significantly improving the observation efficiency of telescopes.

The paper mentions that, by comparing with DIMM data, the root - mean - square error (RMSE) of the seeing estimated by this framework is 0.18 arc - seconds, and the RMSE of the prediction for the next 20 minutes is 0.12 arc - seconds (the seeing range is from 0 to 2.2 arc - seconds), which is a 37% reduction in RMSE compared to the persistence forecast (that is, the forecast value is equal to the last data point). These results show the advantage of this method in prediction accuracy and also prove its feasibility and effectiveness in practical applications.