Weather radar time series simulation: Improving accuracy and performance

Abstract

Time series simulation is an important tool for developing and testing new signal processing algorithms for weather radar. The methods for simulating time series data have not changed much over the last few decades, but recent advances in computing technology call for new methods. It would seem that faster computers would make better-performing simulators less necessary, but improved technology has made comprehensive, multiday simulations feasible. Even a relatively minor performance improvement can significantly shorten the time of one of these multiday simulations. Current simulators can also be inaccurate for some sets of parameters, especially narrow spectrum widths. In this paper, three new modifications to the conventional simulators are introduced to improve accuracy and performance. Two of the modifications use thresholds to optimize both the total number of samples and the number of random variates that need to be simulated. The third modification uses an alternative method for implementing the inverse Fourier transform. These new modifications lead to fast versions of the simulators that accurately match the desired autocorrelation and spectrum for a wide variety of signal parameters. Additional recommendations for using single-precision values and graphical processing units are also suggested.