With the advent of the Open Radar Data Acquisition (ORDA) system on WSR-88D radars and the introduction of significantly more powerful signal processing hardware comes the opportunity to improve the method used for estimating the spectrum width, a measure of the variability of radial wind velocities within a measurement pulse volume. In addition, the implementation of new operational modes for improved data quality, including SZ phase coding, will involve very different signal processing techniques and hence may require novel methods to meet the WSR-88D specifications. While spectrum width has not been used extensively by radar meteorologists in the past, the new NEXRAD Turbulence Detection Algorithm (NTDA), developed under direction and funding from the FAA's Aviation Weather Research Program, will soon be using the WSR-88D spectrum width as a key input for providing in-cloud turbulence estimates (eddy dissipation rate, EDR) for an operational aviation decision support system. Achieving optimal spectrum width estimator performance would directly benefit the accuracy of the NTDA product.

This paper addresses these issues by evaluating performance characteristics of several spectrum width estimators, including the pulse-pair estimator currently used in the WSR-88D and other correlation and spectral estimators. Evaluations are performed using simulated data representing a variety of scenarios including different signal-to-noise ratios, overlaid power ratios, and spectrum widths. The effect of the spectrum width estimators' bias and variance on the NTDA's EDRs is also examined. Finally, a hybrid algorithm combining several of the best methods is proposed, and it is shown that this algorithm, while somewhat more computationally intensive, is significantly more accurate and robust than the pulse-pair method alone.