Towards operational optical turbulence forecast systems at different scales

TL;DR

This paper demonstrates that an autoregressive method combining mesoscale model forecasts and real-time measurements significantly improves short-term atmospheric parameter predictions for astronomical sites, supporting better adaptive optics operations.

Contribution

The study extends an autoregressive forecasting approach to multiple astroclimatic parameters at the VLT site, achieving unprecedented accuracy over existing methods.

Findings

Unprecedented forecast accuracy for astroclimatic parameters.

Clear improvements over persistence-based predictions.

Preliminary results suggest better accuracy than machine learning approaches.

Abstract

The forecast on a time scale of 1 or 2 hours is crucial for all kind of new generation facilities (ELTs) instrumentation supported by the adaptive optics that will be mainly operated in Service Mode. In a recent study (Masciadri et al. 2020) we have showed that we can forecast the seeing and atmospheric parameters at such short time scales using an autoregressive method achieving unprecedented model accuracies with a substantial gain with respect to forecasts performed the day before (i.e. on longer time scales) obtained with an atmospheric mesoscale model. Equally we showed a gain with respect to the method by persistence using simply real-time measurements in situ on the same short time scale (1-2 hours). The auto-regressive method makes use of the forecasts done with mesoscale atmospheric models and real-time measurements and since 2019 has been implemented in the operational forecast system ALTA Center supporting LBT observations. In this contribution we apply the same approach to the VLT site extending the method to all the main astroclimatic parameters i.e. the seeing, the wavefront coherence time, the isoplanatic angle and the ground layer fraction. We prove that such a method offers unprecedented forecast accuracies for all the astroclimatic parameters with clear gains with respect to the prediction by persistence. Preliminary calculations indicate also better accuracies than those obtained with the machine learning based approach using in situ measurements. We will apply soon such a method in an operational forecast system that we conceived for the VLT.