In situ measurements and neural network analysis of the profiles of optical turbulence over the Tibetan Plateau

Abstract

ABSTRACT The vertical profile of $C_n^2$ is the main factor for accurate astronomical observation and laser communication, however, hardware-based instruments and associated data are not widely available due to logistical and financial issues. In this article, we developed an indirect method, a hybrid network structure which is a combination of the backpropagation neural network and the simulated annealing algorithm, to fit the vertical profile of $C_n^2$. Radiosonde measurements from a field campaign over the Tibetan Plateau at Dachaidan (37.7○N, 95.3○E, 3180 m ASL) were performed in 2020 August to estimate the accuracy of our model, during which a balloon-borne portable turbulence meteorological radiosonde was used to measure the atmospheric optical profiles. Besides, the integrated astronomical parameters (the coherence length r0, seeing ε0, isoplanatic angle θ0, and the wavefront coherence time τ0), derived from $C_n^2$ and wind-speed vertical profiles, are investigated for astronomical applications using the proposed model. In addition, quantitative evaluations such as the correlation coefficient, the root mean squared error, and the systematic bias are used to quantify the performance of our model. More interesting, this model is found to outperform a widely used external scale model for the prevalent atmospheric conditions and shows better correlation and reliable estimates.