D.I. Lekhovytskiy, D.S. Rachkov, A.V. Semeniaka, D.V. Atamanskiy, A.A. Pushkov
We define and compare the accuracy characteristics of both well-known and a number of new algorithms for estimation of Doppler velocities spectrum (DVS) width of meteorological formations (MF) in pulsed Doppler weather radars (PDWR). The main attention is paid to this problem solution under widely used assumption about Gaussian (bell) shape of the DVS, and for different estimates of moduluses of correlation coefficients of samples of reflections from MF out of adjacent intervals of sounding. On the results of extensive theoretical-experimental investigation of different estimation algorithms there are proven the recommendations for their choice and practical implementation.
It is shown, that on a number of criteria both for constant and varying intervals of sounding the estimate of unimodal DVS width based on averaged Burg estimate of correlation coefficient modulus is more preferable than the known and overall used width estimate based on moduluses of two correlation coefficients of samples of reflections from MF separated by the different time intervals.
We evaluate numerically the errors conditioned by distinctions of shape of unimodal spectrum of finite order autoregressive (AR) process, approximating the reflections, from the Gaussian one. It is shown, that the Gaussian approximation of DVS leads to the upper width estimate for such spectrum. The measurement error is greater the greater the correlation coefficient and the lower the order of approximating AR-process. It can be decreased at the expense of complexification of DVS width estimation algorithm and use of additional procedure for order estimation of approximating AR-process.
There are proven the recurrent procedures and schemes for DVS width estimation in specified range resolution cells for two variants of data updating periodicity – in each azimuth direction, and in directions, angular distance between which exceeds the beam width of pencil-beam directional pattern of antenna of all-round looking PDWR. It is shown, that the basic elements of developed schemes are suitable for width estimation of arbitrary-shaped unimodal spectra as well as reflectivity and mean radial velocity of MF.