321 Estimation of the Specific Differential Phase Shift using Kalman Filtering

Thursday, 19 September 2013
Breckenridge Ballroom (Peak 14-17, 1st Floor) / Event Tent (Outside) (Beaver Run Resort and Conference Center)
Jacopo Grazioli, EPFL-LTE, Lausanne, Switzerland; and M. Schneebeli and A. Berne

The specific differential phase shift (KDP) is defined as the range derivative of the differential phase shift on propagation (ФDP). This polarimetric variable depends on the shape, type and concentration of the hydrometeors in a radar resolution volume, and therefore carries valuable information about the characteristics of precipitation. The estimation of KDP from the measured and noisy total differential phase (ψDP) can be affected by significant errors, and especially the high values of KDP can be largely underestimated.

In the present work we propose a Kalman filter (KF) technique to estimate KDP that relies only on the measured ψDP and therefore provides estimate which are fully independent from power-related radar observables. The algorithm aims to provide KDP estimates at the higher possible resolution, in order to capture the fine scale variation of the precipitation. It takes into account the effect of differential shift upon backscattering (d) and the error structure of the polarimetric phase-based variables, through a parametrization based on simulated radar measurements.

The newly developed algorithm is compared with other two standard algorithms of the same type (i.e., that only use phase-information), that are based on linear interpolation and moving window range derivative. The comparison is carried out by means of simulated DSD fields, which allow to compare “true” and retrieved KDP values directly. The KF algorithm outperforms the standard ones in terms of Nash Efficiency, Correlation Coefficient, RMSE and mean normalized bias and it is able to improve the estimation of high KDP values, by reducing their systematic underestimation. It is also shown that the Mean Normalized Absolute Error in the estimation of KDP for all the methods considered can be as large as 25-30% at the range resolution volume scale.

- Indicates paper has been withdrawn from meeting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