The Accuracy of Radar Estimates of Ice Terminal Fall Speed from Vertically Pointing Doppler Radar Measurements

International audienceDoppler radar measurements at different frequencies (50 MHz and 2835 MHz) are used to characterize the terminal fall speed of hydrometeors and the vertical air motion in tropical ice clouds and evaluate statistical methods for retrieving these two parameters using a single vertically pointing cloud radar. Regarding the observed vertical air motions, it is found that the mean vertical air velocity in ice clouds is small on average, as assumed in terminal fall speed retrieval methods. The mean vertical air motions are slightly negative (downdraft) between the melting layer (5 km height) and 6.3 km height, and positive (updraft) above this altitude, with two peaks of 6 and 7 cm s−1 at 7.7 and 9.7 km height. Regarding the retrieved hydrometeor terminal fall speeds, it is found that the variability of terminal fall speeds within narrow reflectivity ranges is typically within the acceptable uncertainties for using terminal fall speeds in ice cloud microphysical retrievals. This study also evaluates the performance of previously published statistical methods of separating terminal fall speed and vertical air velocity from vertically pointing Doppler radar measurements using the 50 / 2835 MHz radar retrievals as a reference. It is found that the variability of the terminal fall speed - radar reflectivity relationship (Vt - Ze) is large in ice clouds and cannot be parameterized accurately with a single relationship. While a well-defined linear relationship is found between the two coefficients of a power-law Vt - Ze relationship, a more accurate microphysical retrieval is obtained using Doppler velocity measurements to better constrain the Vt - Ze relationship for each cloud. When comparing the different statistical methods to the reference, the distribution of terminal fall speed residual is wide, with most residuals being in the ± 30-40 cm s−1 range about the mean. The typical mean residual ranged from 15 to 20 cm s−1, with different methods having mean residuals less than 10 cm s−1 at some heights, but not at the same heights for all methods. The so-called Vt - Ze technique was the most accurate above 9 km height, while the running-mean technique outperformed the other techniques below 9 km height. Sensitivity tests of the running-mean technique indicate that the 20-minute average is the best trade-off for the type of ice clouds considered in this analysis. A new technique is proposed that incorporates simple averages of Doppler velocity for each (Ze, H) couple in a given cloud. This technique, referred to as DOP - Ze - H, was found to outperform the three other methods at most heights, with a mean terminal fall residual less than 10 cm s−1 at all heights. This error magnitude is compatible with the use of such retrieved terminal fall speeds for the retrieval of microphysical properties