2020MNRAS.497.4910T

In this article, we present the first results ever obtained by applying the autoregressive (AR) technique to precipitable water vapour (PWV). The study is performed at the Very Large Telescope (VLT). The AR technique was proposed recently to provide forecasts of atmospheric and astroclimatic parameters on short time-scales (up to a few hours) by achieving much better performance with respect to the 'standard forecasts' provided in early afternoon for the coming night. The AR method uses real-time measurements of the parameter of interest to improve the forecasts performed with atmospherical models. Here, we used measurements provided by the Low Humidity And Temperature PROfiling microwave radiometer (LHATPRO), a radiometer measuring the PWV at the VLT continuously. When comparing the AR forecast at 1h with the standard forecast, we observe a gain factor of ∼8 (i.e. ∼800 per cent) in terms of forecast accuracy. In the PWV <= 1 mm range, which is extremely critical for infrared astronomical applications, the RMSE of the predictions is of the order of just a few hundredth of millimetres (0.04 mm). We therefore proved that the AR technique provides an important benefit to VLT science operations for all instruments sensitive to PWV. Also, we show how such an ability to predict PWV can also be useful to predict the sky background in the infrared range [extremely appealing for Mid-infrared ELT Imager and Spectrograph (METIS)]. We quantify such an ability by applying this method to the New Earth in the Alpha Cen region (NEAR) project supported by the European Southern Observatory (ESO) and Breakthrough Initiatives.