OTProf estimates high-resolution optical turbulence profiles

The paper introduces OTProf, a deep-learning pipeline that estimates high-resolution vertical profiles of optical turbulence, C_n^2, from coarse-resolution ERA5 reanalysis input. The authors evaluate OTProf against in-situ reference profiles in the Netherlands and the standard Hufnagel-Valley analytic model, reporting better reproduction of vertical structure and improved estimates of integrated metrics such as the Fried parameter (r0) and the scintillation index (sigma_I^2).

OTProf is a supervised learning model trained to map ERA5 fields to high-resolution C_n^2 profiles. The model preserves large-scale meteorological dependence and topographic effects embedded in reanalysis while upscaling to finer vertical resolution. Key practitioner takeaways:

• •Training uses paired reanalysis and observational profile data, with loss terms that emphasize vertical structure and integrated metrics.
• •OTProf reduces gross errors from analytic parametrizations and mesoscale-model interpolation while remaining orders of magnitude cheaper than running high-resolution mesoscale simulations.
• •A noted limitation is prediction smoothing, which attenuates rare, strong turbulence peaks and can produce overly optimistic integrated values.

Accurate C_n^2 profiles are critical for ground-based optical astronomy, adaptive optics design, and free-space optical communications. Existing approaches trade accuracy, cost, and coverage: analytic models like Hufnagel-Valley are simple but crude; mesoscale models are accurate but computationally expensive; and measurements require specialized instrumentation. OTProf sits between these extremes: it leverages widely available ERA5 reanalysis to deliver more physically consistent, higher-resolution profiles with much lower operational cost. For observatory site characterization, instrument scheduling, and link-performance forecasting, this method can expand access to realistic turbulence profiles at continental scales.

Validate OTProf across diverse climates and complex terrain, and quantify how smoothing affects adaptive-optics performance metrics. Follow-up work should explore uncertainty quantification and hybrid workflows that fuse sparse profile measurements with OTProf outputs for operational forecasting.