Using meteorological forecasts to predict astronomical ‘seeing’

The main limitations affecting visible-light and IR astronomy are related to visibility, atmospheric humidity, and spatial resolution. High-resolution imaging is fundamentally limited by optical turbulence (OT) in the atmosphere. OT is induced by dynamical turbulence developing within a stratified temperature field and depends on weather conditions. The resulting degraded telescope resolution is called ‘seeing.’ At the best ground-based observatories, characterized by a seeing of order 0.5 arcseconds, the resolution of a telescope with an aperture diameter of 10m is no better than 20cm (at ground level) at optical wavelengths.

At the present time, when large and expensive telescope projects are increasingly becoming the norm, observatory managers rely on a proper assessment and understanding of weather conditions for optimal scheduling of scientific operations. Flexible rotas (i.e., planning optimal use of the available instrumentation) are now frequently used at the largest astronomical observatories. Site testing in preparation for new observatory developments is another important challenge. For site characterization for the European 42m (diameter) Extremely Large Telescope (E-ELT)¹ or the Antarctic ‘Dome C’ project,² we designed the DIMM (differential image-motion monitor) and SSS (single-star scidar)³ optical monitors. Since monitoring requires the availability of specific instruments and logistics, and it is limited in place and time, we are investigating other ways to retrieve atmospheric OT conditions from meteorological measurements or forecasts.

Atmospheric conditions are defined by the speed, temperature, and composition of the constituent elements. The corresponding state of the atmosphere is governed by the Navier-Stokes equations prevalent above and around the site. For astronomical seeing prediction, one must obtain the local (microscopic) atmospheric conditions on the metric OT scale. Several models can be used, but most are computationally time consuming, unable to provide a global analysis for a given site, or use the meteorological forecast directly to predict the temporal OT-strength C_n² profiles, where C_n² is the OT index.

We analyzed⁴ the correlation between the C_n² and macroscopic mean profiles of the horizontal wind-speed component and temperature, with a vertical resolution of 100m to 1km. We developed and validated a parametric model on the basis of our analysis of a large C_n² and meteorological profile database.⁵ This model can be implemented easily in a forecasting scheme and used for astronomical site-management purposes (see Figure 1). Any mesoscale model can be used to provide hourly forecasts over 24 hours and deliver the main meteorological parameters of interest. The typical horizontal grid step is 1km, while we simultaneously use more than 30 vertical levels. Higher horizontal resolution may be achieved, depending on the terrain model. Our parametric model is initialized using global data provided by meteorological agencies, complemented by local measurements. The OT model is then applied to the meteorological forecast, thus providing predictions for the seeing, altitude, and intensity of the main OT-producing layers.

Figure 1. Seeing forecasting scheme. OT: Optical turbulence. Cn²(C_n²): OT index. P.T.U.: Pressure, temperature, humidity.

Here we present results obtained with the Weather Research and Forecasting Model and applied to La Palma (Canary Islands, Spain), where the Observatorio del Roque de los Muchachos (ORM) is sited on a crater rim in the northern part of the island. We ran the simulations for one month, with hourly sampling. The model was initialized using data from the Global Forecast System. Figure 2 shows a typical forecast, displaying the island's seeing conditions with 1km resolution. Figure 3 represents the temporal evolution of the C_n² profiles at the ORM over the course of one month.

Figure 2. Seeing map of La Palma (Canary Islands, Spain). The color bar represents the prevailing seeing conditions, in arcseconds. See video⁶ for an animation. UT: Universal time.

Figure 3. Temporal evolution of the OT-strength C_n²profiles (as a function of altitude, h) at the Observatorio del Roque de los Muchachos provided by the forecast. The one-day forecast was computed every hour over the course of one month. The color bar represents the prevailing C_n² values on a logarithmic scale.

Meteorological studies and predictions are crucial to assess astronomical observing conditions. Seeing forecasts are of major importance since they enable planning for optimal use of observing time. They also allow tuning of adaptive-optics systems as well as site characterization. Conceptual tools are now available to calculate model results. Our next step is to compare our model results with measurements to validate the models as forecasting products. Once this has been achieved, this approach could be implemented and routinely used at astronomical observatories.