Weather Radar Time-Series Simulators: Improving Accuracy and Performance

Simulation of time series data is an important tool for testing weather radar signal processing algorithms. With the advent of faster computers, extensive simulations are often employed, but this can lead to multi-hour or multi-day simulation times. One approach to significantly reducing these long simulation times is to improve simulator performance. Current radar simulators can also be inaccurate when utilizing narrow spectrum widths, leading to simulation biases. Improving the accuracy of time-series simulators can keep these biases from affecting radar variable estimates and from generating possibly misleading conclusions. A couple of simulator improvements will be introduced that address issues of both accuracy and performance. These improvements can be applied to time-series simulators that are based on reproducing either a desired spectrum or a desired autocorrelation function. After the improvements are described, the simulations are tested for increased accuracy at narrow spectrum widths and for improved performance with a wide range of weather signal parameters. Some advice on taking advantage of graphics processing units is also provided.