Development of Data Screening Method for Wind Lidar under Low S/N Environment

One of wind measuring methods, VAD (Velocity Azimuth Display) is widely used by wind lidar (light detection and ranging). In VAD technique, under the assumption that wind is unique in observation atmosphere, since LOS (line-of-sight) or radial wind velocity form a sinusoidal curve, the curve is detected by sine curve fitting based on the least squares method. An amplitude of the sinusoidal curve is corresponding to the horizontal wind velocity, an initial phase is corresponding to the horizontal wind direction, and an offset is corresponding to the vertical wind velocity respectively. But in low S/N (Signal-to-Noise ratio) conditions such as heavy rain, fog and low density of atmospheric aerosol particles, accuracy of wind velocity and wind direction is degraded. If wind estimation results is degraded, some or all of these three parameters have not been estimated correctly, and it is due to degradation of each LOS velocity. So degraded LOS velocity must be eliminated properly before VAD processing. This paper proposed a data screening method to estimate wind velocity and direction accurately under low S/N environment. The proposed method derive appropriate threshold to remove unnecessary (degraded) LOS velocity utilizing the characteristics of a sinusoidal curve. The threshold is determined as follows. The average of all LOS velocity asymptotic to offset value of sinusoidal curve, that is, the vertical wind velocity component. Next, the amplitude corresponding to the horizontal component is restricted because the horizontal wind velocity range is within at most 30 m/s. Then the threshold is defined as the offset value plus or minus the horizontal velocity range. Invalid LOS velocity is excluded by the threshold, it becomes easier to estimate the remaining two parameters (the amplitude and the initial phase). Under low S/N environment, which is considered the accuracy of LOS velocity itself is deteriorated, the offset value cannot be calculated correctly. But wind lidar has a high range resolution, and variance of vertical wind in a small area is relatively small in general, therefore LOS velocity of adjacent ranges can be utilized to calculate the offset value. In addition, the threshold is derived from the information of a plurality of ranges, it does not deteriorate the original range resolution. We have confirmed the effectiveness of the proposed data screening method applied to simulation data and real data. Figure 1 shows the simulation results of wind velocity and wind direction estimation. In Figure 1, horizontal axis indicates wind velocity and wind direction set value and vertical axis shows estimated value. Fig.1(a) and (b) represent the results before applying the proposed screening method and Fig.1(c) and (d) represent the results after applying the screening method, respectively. Before applying the screening method, estimated wind velocity and wind direction vary widely, that is, estimation accuracy is poor. On the other hand, after applying the proposed screening method estimated value agree well with set value. These results show effectivity of the proposed screening method. This method is able to be applied to Doppler radars and Doppler sodars.