WR-TSS: Weather Radar Time-Series Simulation

Functions for accurate and efficient simulation of weather radar time series.

WR-TSS (Weather Radar Time-Series Simulation) is a set of functions for simulating weather radar time-series data using a Gaussian signal model. WR-TSS is focused on simulating a large number of realizations when using the same signal parameters (not for simulating weather profiles). These types of simulations are often used to simulate weather or ground clutter time series for testing signal processing algorithms. The latest version of WR-TSS uses a different approach for simulating time series data. It uses an autocovariance-matrix-based simulator that works directly in the time domain. In general, this simulator should be faster and more accurate than the simulators in version 1, but there could be some cases on some computers (possibly with a large number of samples per realization) where the version 1 simulators could be faster.

In addition to the new simulator, this version of WR-TSS also uses a new approach to speed up simulations that loop through weather signal parameters. This type of simulation is often used to test new radar-variable estimators or ground clutter filters and to produce lookup tables. These simulations often take a lot of time, and reducing simulation runtimes can increase productivity and can decrease the time needed to find errors. The new approach combines the new autocovariance-matrix-based simulator and the reuse of white noise matrices (RWNM) to achieve the speedup. In tests, this approach made the looping simulations up to 16 times faster than conventional simulations and 60 times faster when also utilizing a graphics processing unit (GPU).

Files:

• ts_aim_eig: autocovariance-matrix-based, single-polarization, weather-signal, time-series simulator; utilizes eigendecomposition (or LDL decomposition); uses GPU if available by default.
• ts_aim_eig_dp: autocovariance-matrix-based, dual-polarization, weather-signal, time-series simulator; utilizes eigendecomposition (or LDL decomposition); uses GPU if available by default.
• ts_eim_eig_mat: produces a weather-signal, time-series matrix that’s utilized when reusing white noise matrices; based on eigendecomposition (or LDL decomposition).
• test_ts_am_funcs: simple function to test speed and accuracy of ts_aim_eig and ts_aim_eig_dp simulators, tests with both eigen- and LDL decompositions, and with and without a GPU (will display an error if a GPU is not available).
• sp_looping_sim: single-polarization, weather-signal, time-series simulation that reuses white noise matrices to speed up the simulation, uses GPU if available by default; implements a simple spectrum width estimator simulation that can be modified for other uses.
• dp_looping_sim: dual-polarization, weather-signal, time-series simulation that reuses white noise matrices to speed up the simulation, uses GPU if available by default; implements a simple correlation coefficient estimator simulation that can be modified for other uses.
• test_sp_looping_sims: tests single-polarization conventional and RWNM simulators with and without a GPU to show runtime and speedup figures like the ones in the reference, can be used to find out which simulators are faster for different numbers of realizations.
• test_dp_looping_sims: tests dual-polarization conventional and RWNM simulators with and without a GPU to show runtime and speedup figures like the ones in the reference, can be used to find out which simulators are faster for different numbers of realizations.
• mag2: helper function to compute the magnitude squared of complex numbers

Reference:

Curtis, C. D., 2025: Weather Radar Time Series Simulation: Rapidly Looping through Signal Parameters. J. Atmos. Oceanic Technol., 42, 269-279, https://doi.org/10.1175/JTECH-D-23-0171.1.