Coupling Optical Turbulence to Sensible Heat, Surface Heat Flux and Latent Heat to potentially optimize Machine Learning predictions of Optical Turbulence

Surface layer optical turbulence in the form of C_T² or C_n² can be calculated from surface layer temperature, moisture, wind, and net radiation characteristics that are readily obtainable in many operational numerical weather prediction (NWP) models. The C_T² and C_n² values can be compared to measurements from differential temperature sensors, imaging systems, sonic anemometers, and Sonic Detection and Ranging device (SODAR). A key derived component needed in surface layer turbulence calculations is the sensible heat value. Typically, the sensible heat is calculated using the Bulk Aerodynamic Method, which assumes a certain surface roughness and a friction velocity that approximates the turbulence drag on temperature and moisture mixing from the change in the average surface layer vertical wind velocity. These assumptions generally only apply in free convection conditions. A more applicable approach is to calculate sensible heat flux from incoming solar radiation and ground heat flux by using the Energy Balance Bowen Ratio method. This method applies when free convection conditions are not occurring. The use of the Bowen ratio – the ratio of sensible heat flux to latent heat flux – allows a more direct assessment of the sensible heat that drives the production of surface layer optical turbulence than the Aerodynamic Method. This study compares two methods for estimating surface layer C_T² and C_n² values. These include deriving sensible heat values using the Bulk Aerodynamic and Energy Balance Bowen Ratio methods and collecting measurements from instruments such as sonic anemometers, SODAR, differential temperature sensors, and time-lapse imagery. Additionally, the Energy Balance method often relays negative Bowen Ratio values; this study aims to validate that employing an absolute value of the calculated Bowen Ratio not only makes the method robust to all atmospheric conditions, but also does not produce non-physical results. Applying machine learning (ML) to the underlying physical parameters such as sensible / latent heats and ground fluxes may allow greater site diversity to the ML quantifications that capture, solar, cloud, surface/terrain effects, and allow better NWP-derived optical turbulence quantifications.