Optical turbulence simulations at Mt Graham using the Meso-NH mode

The mesoscale model Meso-NH is used to simulate the optical turbulence at Mt Graham (Arizona, USA), site of the Large Binocular Telescope. Measurements of the CN2-profiles obtained with a generalized scidar from 41 nights are used to calibrate and quantify the model’s ability to reconstruct the optical turbulence. The measurements are distributed over different periods of the year, permitting us to study the model’s performance in different seasons. A statistical analysis of the simulations is performed for all the most important astroclimatic parameters: the CN2-profiles, the seeing {\epsilon}, the isoplanatic angle {\theta}0 and the wavefront coherence time {\tau}0. The model shows a general good ability in reconstructing the morphology of the optical turbulence (the shape of the vertical distribution of CN2) as well as the strength of all the integrated astroclimatic parameters. The relative error (with respect to measurements) of the averaged seeing on the whole atmosphere for the whole sample of 41 nights is within 9.0 %. The median value of the relative error night by night is equal to 18.7 %, so that the model still maintains very good performances. Comparable percentages are observed in partial vertical slabs (free atmosphere and boundary layer) and in different seasons (summer and winter). We prove that the most urgent problem, at present, is to increase the ability of the model in reconstructing very weak and very strong turbulence conditions in the high atmosphere. This mainly affects the model’s performances for the isoplanatic angle predictions, for which the median value of the relative error night by night is equal to 35.1 %. No major problems are observed for the other astroclimatic parameters. A variant to the standard calibration method is tested but we find that it does not provide better results, confirming the solid base of the standard method.

💡 Research Summary

This paper presents a comprehensive validation of the non‑hydrostatic mesoscale model Meso‑NH for simulating atmospheric optical turbulence (OT) at the Mt Graham site in southeastern Arizona, home to the Large Binocular Telescope (LBT). The authors used a substantial dataset of 41 nights of CN² vertical profiles measured with a Generalized Scidar (GS) mounted on the VATT telescope. The GS provides turbulence strength from the ground up to 20 km with a typical vertical resolution of about 1 km, making it an ideal reference for model validation.

The Meso‑NH configuration employed a three‑nested‑grid approach: an outer domain (Model 1) covering 800 × 800 km at 10 km horizontal resolution, an intermediate domain (Model 2) covering 160 × 160 km at 2.5 km resolution, and an innermost domain (Model 3) covering 60 × 60 km at 0.5 km resolution. All domains share 54 vertical levels, with the first level 20 m above ground and a logarithmic stretching up to 3.5 km, after which the spacing is roughly constant (~600 m). Simulations were initialized and forced every six hours using ECMWF analyses, run for 12 h, and the first two hours discarded to allow model spin‑up. The model outputs a CN² profile every two minutes for the observatory location.

Calibration of the OT parameterisation follows the MJ01 method (Masciadri & Jabouille 2001). The atmosphere is divided into 5–6 vertical slabs; for each slab a scaling factor aₖ is derived by minimizing the χ² between simulated and measured CN² averages over the night. The minimum kinetic energy parameter E_min is then adjusted per slab as E*ₘᵢₙ,ₖ = E_min · aₖ^{3/2}. This calibrated model is rerun for each night. A variant of the calibration was also tested, but it did not improve the results, confirming the robustness of the standard MJ01 approach.

The validation focuses on four key astro‑climatic quantities derived from CN²: (1) the seeing ε, (2) the isoplanatic angle θ₀, (3) the wave‑front coherence time τ₀, and (4) the CN² vertical distribution itself. For the whole atmosphere, the model reproduces the average seeing with a relative error of 9 % (global mean) and a night‑by‑night median relative error of 18.7 %. Similar performance is observed when the atmosphere is split into the boundary layer (0–1 km) and the free atmosphere (1–20 km), and across the two seasons represented in the dataset (summer and winter).

The isoplanatic angle, however, shows the largest discrepancy: the median night‑by‑night relative error is 35.1 %. This stems from the model’s limited ability to capture very weak or very strong turbulence layers in the high atmosphere (above ~10 km), which dominate θ₀ because the isoplanatic angle scales with the 5/3‑power of altitude. Consequently, while the model’s morphology of CN² (the overall shape of the profile) is satisfactory, the high‑altitude fine structure is overly smooth compared to measurements.

The wave‑front coherence time τ₀, which depends primarily on wind speed profiles, benefits from the previously demonstrated reliability of Meso‑NH wind forecasts. The τ₀ predictions exhibit a relative error around 15 %, indicating that the wind field is sufficiently accurate for this purpose.

Statistical analysis also confirms that the calibrated model does not over‑fit the calibration subset; performance on the independent nights remains comparable, demonstrating the generality of the MJ01 calibration. The attempted calibration variant, which altered the distribution of E_min across slabs, actually degraded the agreement, reinforcing that the original slab‑wise scaling is optimal for this site and dataset.

In summary, the study shows that Meso‑NH, when calibrated with the MJ01 procedure, can reliably predict the average strength of optical turbulence and the associated seeing at Mt Graham, with errors well within operational tolerances for adaptive optics (AO) planning. The main limitation lies in reproducing the extreme high‑altitude turbulence that governs the isoplanatic angle; addressing this will likely require higher vertical resolution, refined turbulence parameterisations, or incorporation of additional observational constraints (e.g., radiosonde or lidar data). Improving the high‑altitude representation would directly benefit AO system design, especially for wide‑field correction where θ₀ is a critical parameter.

Overall, the paper provides a solid benchmark for mesoscale OT modelling, demonstrates the practical utility of a calibrated Meso‑NH system for a major astronomical site, and outlines clear pathways for future enhancements to achieve even higher predictive fidelity.