Investigations of the Uncertainties Associated with HID Algorithms and Guiding Input to a Novel, Synthetic Polarimetric Radar Simulator

November 29, 2017

Julie Barnum

Hosted by Steven Rutledge (Advisor), Michael Bell, Brenda Dolan, Steven Reising (Electrical and Computer Engineering)

Download Video

Abstract

With the advent of polarimetric radars, the need to produce simulated radar observables from model has also become apparent, in order to directly compare the same quantities between observations and models (e.g. rain rate calculations, hydrometeor identification - HID). To the end of evaluating model performance, for both a spectral bin microphysics (SBM) scheme and bulk microphysics scheme (BMS), a novel, synthetic polarimetric radar simulator created by Matsui et al. (2017) was implemented in this study: POLArimetric Radar Retrieval and Instrument Simulator (POLARRIS). POLARRIS takes in model data and simulates polarimetric radar variables in the forward component (POLARRIS-f), and then the inverse component of POLARRIS (iPOLARRIS) utilizes retrieval algorithms that are also employed in observations to make direct 1-to-1 comparisons between model simulations and observations. This inverse component is novel in its ability to help bridge the gap between model output and observations due to the fact that model output and observations without this framework are not directly comparable.

The simulation of ice hydrometeors is not straightforward, and several assumptions are required to create polarimetric data for these species, such as the assumption of the size distribution, particle densities, particle melting, the input axis ratio, and canting angle assumptions. The last two variables are notoriously difficult to pin down for ice hydrometeors. This work aims to narrow down the appropriate inputs for axis ratio and canting angle assumptions that create the most comparable results with observations for three ice hydrometeors: aggregates, ice crystals, and graupel for two different meteorological regimes (mid-latitude supercell and tropical, monsoon MCS). Rain was also carried through as a check on model output. Through various sensitivity tests, it was concluded that, when run through the range of potential values, changes in axis ratio had a larger impact on the resulting polarimetric data than did changes in the canting angle assumptions.

With this in mind, the 22 Z integrated hour mid-latitude supercell from the 23 May 2011 MC3E case was simulated to help determine, for each hydrometeor type, the most appropriate axis ratio value(s) and canting angle assumptions that produced comparable results with observations. It was found using co-variance plots that, for the BMS, the use of a singular axis ratio, mean canting angle, and degree of particle tumbling often produced differential reflectivity and specific differential phase values that converged to one value. While these values were within the observed values, they did not manage to simulate the breadth of observed values. Reflectivity values were also much too low compared to observations. SBM results, regardless of the type of input assumptions, tended to produce broader ranges for these variables, and also managed to better capture the reflectivity range seen in observations than was the case for the BMS. However, the reflectivity ranges seen in SBM were at times too expansive. The differences between SBM output and BMS output is likely due to the differing inherent assumptions in each microphysical scheme. The sensitivity of the simulated hydrometeors’ polarimetric data was also probed against changing axis ratio and canting angle input assumptions. It was found that, in particular, BMS differential reflectivity values were quite sensitive to changes in input assumptions, regardless of the regime (tropical MCS vs. mid-latitude supercell).

HID was found to be the most effective method to evaluate the performance of the two different model microphysical schemes (SBM vs. BMS) with respect to observations. Input assumptions that produced the most comparable results with respect to observations for each hydrometeor were compared using HID stacked frequency by altitude (SFAD) diagrams for convective and stratiform precipitation. This analysis found that although the co-variance plots revealed many model shortcomings, the HID proved to be fairly robust, especially for MC3E. The sensitivity of the HID retrieval itself was also investigated with respect to changing inputs (i.e. the membership beta functions) to the HID algorithm. The resulting HID was fairly sensitive to changes in the inputs to HID, particularly for model simulations. Observations seemed less responsive to changes in these input assumptions to HID. Longer simulation time frames, the potential inclusion of simulated melting hydrometeors, and investigation of other radar wavelengths are all suggested to help further utilize this methodology for evaluating model microphysical schemes’ abilities to accurately simulate polarimetric data and HID retrievals with respect to observations.