Estimating radar reflectivity uncertainties and their impact on retrieved raindrop size distribution parameters

Vertically pointing radars transmit tens-of-thousands of radar pulses in order to generate a single profile of Doppler velocity power spectra. After correcting for the vertical air motion, the Doppler velocity power spectra can be used to estimate the raindrop size distribution (DSD). Assuming that the precipitation is stationary during the observation interval, processing more independent radar pulses results in smaller measurement uncertainties and more accurate reflectivity estimates. The focus of this study is to understand how reflectivity uncertainties propagate into uncertainties in the retrieved DSD parameters including the mass-weighted mean raindrop diameter Dm and the rain rate R.

This study uses a Monte Carlo Simulation to add realistic measurement noise to ideal Doppler velocity power spectra in order to investigate measurement and retrieval uncertainties. Regarding measurement uncertainties, the spectrum moments (i.e. Signal-to-noise ratio (SNR), mean velocity, and spectrum width) are estimated for each simulated noisy spectrum. The bias and uncertainties are estimated from an ensemble of 1000 simulated noisy spectra. Regarding retrieval uncertainties, the best Gamma shaped DSD function using the 3 parameters Nw, Dm, and mu is retrieved for each simulated noisy spectrum. The bias and uncertainties are estimated for the three DSD parameters along with the rain rate R using 1000 simulated noisy spectra.

This study quantifies how the measurement uncertainties are dependent on the Doppler velocity power spectrum signal-to-noise ratio (SNR) with decreasing SNR leading to larger measurement uncertainties as the random noise variance becomes more dominant.