A Modified Method of Removing Ground Clutter from Wind Profiler Radar Based on Adaptive Wavelet Threshold

Wind profiler radar (WPR) can be used to retrieve real-time atmospheric wind field data of high resolution. Backscattered echo caused by irregularities of atmospheric refractive index is received by radar antenna and wind velocities is calculated with Doppler frequency shifting speed formula. It is widely used in fields of very short-term weather forecasting, airport operations and public protection, air pollution monitoring, wind field analyses and forecasts of toxic plume trajectories resulting from chemical or nuclear incidents. As a result of being widely used in different situations, WPR is always sited near the city with a large population and complicated geographical environment. Transmission of electromagnetic wave during WPR detecting period is often interfered by various clutters that contaminate WPR data introduce bias in moments and wind velocity estimation. Of all clutters, ground clutter is the primary source because it happens more often than the others. Ground clutter is radar return from more or less stationary targets such as trees, buildings near the cited place. How to eliminate the influence of ground clutter is a most concerned aspect. Ground clutter mainly concentrates around the zero-frequency and it has high amplitude on the power spectrum. The most frequently used methods, such as traditional wavelet threshold processing and zero-frequency elimination of 3 points, both have the ability to separate the meteorology echo from the ground clutter when the turbulent peak is away from the zero-frequency and not covered with ground clutter peak. However, when the near zero-frequency echo is taken into consideration, both of the traditional methods meet their limitation. Based on the wavelet high frequency coefficients, a method of determining threshold adaptively is proposed and the validation of the method is done by using of simulated data and WPR measured data. The corresponding power spectrum before and after self-adapting wavelet threshold processing are compared. Results show that this method performs well even when the signal is close to the zero-frequency and covered completely. Meanwhile, the method has some important features, such as simple theory, small amount of calculation and easy to implement. Cases analysis shows that self-adapting threshold processing can increase the accuracy of peak identification, also provide approach and basis for improving the WPR products.