Breakout Summary Report
ARM/ASR User and PI Meeting
2 - 6 May 2016
2 May 2016
3:30 PM - 5:30 PM
0
Scott Collis and Shaocheng Xie
Breakout Description
As you arrive keen to share science and excitement at the ASR Science Team Meeting, we will hold a breakout on Monday afternoon on use of radar simulators in achieving ASR science. Speakers have been instructed to specifically focus on the use of radar simulators and how forward-modeled moments can provide insight and build that bridge between the modeling and observing communities we seek. We have invited four speakers to present:3:30 - 3:40 Intro and updates. (Scott Collis)
3:40 - 4:00 Overview of POLARRIS: Motivation, Application, and Architecture (Toshihisa Matsui)
4:00 - 4:20 The Development and Applications of a Polarimetric Radar Forward Operator to Improve Microphysical Parameterizations and Study Deep Convective Storms (Jeff Snyder)
4:20 - 4:40 GCM-SIM: Latest updates and applications (Shaocheng Xie)
4:40 - 5:00 CR-SIM: Latest updates and applications (Pavlos Kollias)
5:00 - 5:30 Discussion
If you wish for me to announce your project at the beginning of the session, please use the attached template. I will spend a minute on each and will accept up to 5 solicitations. Note: As I am not an expert on your work, please make the slides self-explanatory. We welcome funded and unfunded work or even just good ideas.
The session will conclude with 30 minutes of discussion. Since we are organizing this session, the following initial discussion questions are proposed:
1) Once investigators have forward-simulated radar moments, what are some methods for comparing against observations?
2) How to represent and allow for uncertainties (or more correctly lack of information) due to the simplicities of microphysical schemes?
3) To build a smart or a dumb simulator? Should we build in information into the simulator that does not exist in the model microphysical scheme?
4) What is occurring outside of DoE and the USA? Can we learn from other communities?
We are looking forward to your participation,
Scott Collis and Shaocheng Xie
Main Discussion
This session focused on how Radar Simulator packages might be used to achieve ASR and other ARM stakeholder goals. Solicited presenters were asked with specific instructions to focus on usage not technical aspects.
The following talks were given:
3:30 - 3:40 Intro and updates. (Scott Collis)
3:40 - 4:00 Overview of POLARRIS: Motivation, Application, and Architecture (Toshihisa Matsui)
4:00 - 4:20 The Development and Applications of a Polarimetric Radar Forward Operator to Improve Microphysical Parameterizations and Study Deep Convective Storms (Jeff Snyder)
4:20 - 4:40 GCM-SIM: Latest updates and applications (Shaocheng Xie)
4:40 - 5:00 CR-SIM: Latest updates and applications (Pavlos Kollias)
5:00 - 5:30 Discussion
Toshi’s talk focused on a collaboration between NASA and ASR/ARM through Steven Rutledge’s group at CSU. The main thrust of their work is to use the simulator to create synthetic radar data that retrievals can be applied to. Then retrievals are also carried out on observed radar data allowing a more “Apples to apples” comparison. The team is focusing on two retrievals: Vertical Velocity using multi-Doppler and the CSU particle ID (identification of dominant scattering hydrometeor). The project is in its early stages, but some nice examples were shown. The team is developing a Python-based interface and are looking at accelerating the process (which does not use lookup tables; rather, it uses Mueller/T Matrix calculations at each gate) by using MPI.
Jeff’s talk revolved around the simulation of microphysical fingerprints of dynamics in synthetic radar data. This follows from Matt Kumjian’s well-known work detailing columns of high anisotropy (so-called ZDR or differential reflectivity columns) that are associated with supercooled liquid water above the freezing layer associated with strong vertical velocities. New developments in the CIMMS/NSSL simulator code coupled to the Hebrew University Cloud Model (HUCM) microphysics allow the generation of synthetic fully polarimetric radar data (with such exotic moments as Circular Depolarization Ratio, CDR). This allows the team to quickly parameter-scan (?) the model data and document the existence of columns.
Shaocheng gave the audience a scale change to talk about his infrastructure-funded work to implement an ARM KAZR/MMCR simulator into the COSP framework. Shaocheng reported on the latest work introducing more realism into the simulator, including accounting for the change in radar sensitivity with range (due to the beam diameter being greater further from the radar, thus less power per unit volume). Early results and statistical comparisons with the DoE ACME model were shown highlighting some issues with low clouds in the model. There was a healthy exchange between Shaocheng and Toshi on issues with the subcolumn generator (downscaler) and cloud overlap issues.
Finally, Pavlos took us back to convection-resolving models, highlighting the open source (available on the website of the Stony Brook team’s web page) CR-SIM. CR-SIM offers a lookup-table-based forward model (in collaboration with Jothiram Vivekanandan [Vivek] from NCAR) coupled to the WRF framework. Pavlos highlighted the need for support for such packages and that the model of simply handing over code with no help for the modeling community did not work.
We allocated 30 minutes to discussion. We were originally worried this would not be enough but the audience was very subdued and it took great effort to get any clear feedback from the user community. The chair (Scott) forced some conversation along the lines of:
• Can comparisons between forward-modeled measurements and real measurements tell us any more than “the model and obs disagree?” How do you improve a model using forward modeled measurements, which can be a much higher-order moment than that represented in the microphysics of the model?
• With ACME pushing into the grey zone (~13km Regionally Refined Meshes) what does this mean for simulators of radar observables? Could we strike out on our own (away from COSP) with dynamical downscaling? Phil Rasch gave some feedback and interest from the ACME side of things. It is clear going to higher resolutions will require a rethink.
• We asked the question “Should ARM support a simulation framework?” Again, a very muted response from the group. In conversations later with individual PIs an opinion that this should happen on a PI-to-PI level within ASR was voiced.
Key Findings
There are three main uses for radar simulators:1) Direct (statistical) comparison between forward-modeled measurements and ARM measurements.
2) Simulating radar-based microphysical fingerprints from model data in order to understand the phenomenology of such signatures and what microphysical and dynamical processes they map to.
3) The use of simulators to create radar output from model data that can then be used alongside collected radar data in a retrieval framework. This allows the issues with sampling and “indirect-ness”.