9A.4 Dual-Polarization Radar QPE Using Adjusted Specific Attenuation For Different Precipitation Types

Thursday, 17 September 2015: 9:15 AM
University AB (Embassy Suites Hotel and Conference Center )
Yadong Wang, CIMMS/Univ. of Oklahoma, Norman, OK; and P. Zhang, A. V. Ryzhkov, J. Zhang, L. Tang, and S. B. Cocks

A novel Quantitative Precipitation Estimation (QPE) algorithm using the specific attenuation A has been developed recently. As opposed to the conventional R(Z) algorithm, the R(A) estimate is immune to attenuation, radar miscalibration, wet radome, and partial beam blockage. Although the R(A) algorithm is more robust to the variability of drop size distributions compared to other radar rainfall relations, the estimate of A from the ZPHI procedure requires tuning the net ratio α = A/KDP along the radar beam where KDP is the specific differential phase. Our analysis of the R(A) performance on the WSR-88D network indicates that the original R(A) algorithm with default value α = 0.015 dB/deg at S-band performs well for continental rain but tends to underestimate tropical rain.

The new version of the R(A) method incorporates automatic tuning of α based on the rain type and the slope of the Z–ZDR dependency in a particular rain region. According to the new algorithm, precipitation region is first segregated into three rain categories: stratiform, convective, and tropical. For these regions, three values of α are calculated based on the Z-ZDR slopes in each of the three regions. The A field for an entire radar scan is first calculated using the optimal value of α obtained from the largest region. Then the values of A in the other two smaller regions are adjusted with a factor based on the estimated α to match their precipitation types. Rainfall rate for the scan is computed using adjusted A. The modified R(A) algorithm has been tested for light rain, flood, and hurricane precipitation events. Comparison with rain gauge measurements shows that the modified R(A) demonstrates better performance in terms of correlation coefficient, mean bias ratio, and the root mean square error than R(Z) and the original version of R(A) with fixed default value of the parameter α.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner