Adaptive niche-genetic algorithm based on backpropagation neural network for atmospheric turbulence forecasting

Area of Science:

• Atmospheric optics
• Optical turbulence
• Remote sensing

Background:

• Direct measurements of the refractive index structure constant ($C_n^2$) are limited across various climates and seasons.
• Accurate forecasting of optical turbulence is crucial for applications like astronomy and telecommunications.

Purpose of the Study:

• To develop an indirect method for forecasting optical turbulence by estimating $C_n^2$.
• To validate the proposed model against field campaign data and compare it with existing numerical weather prediction models.

Main Methods:

• Utilized a backpropagation neural network optimized by an adaptive niche-genetic algorithm to estimate $C_n^2$.
• Validated the model using six-day $C_n^2$ data from the 30th Chinese National Antarctic Research Expedition.
• Compared performance metrics (correlation coefficient, RMSE, bias) against the Weather Research and Forecasting (WRF) model.

Main Results:

• The developed model demonstrated a better correlation with observed $C_n^2$ data.
• The indirect estimation method reliably predicted optical turbulence parameters.
• The proposed model outperformed the WRF model in estimating $C_n^2$.

Conclusions:

• The indirect method using a genetic algorithm-optimized neural network is effective for forecasting optical turbulence.
• This approach offers a viable solution for estimating $C_n^2$ in data-scarce regions or seasons.
• The findings have implications for improving the accuracy of optical systems affected by atmospheric turbulence.