Breakout Summary Report
ARM/ASR User and PI Meeting
19 - 23 March 2018
End-to-end Forward Simulators
19 March 2018
4:00 PM - 6:00 PM
30
Katia Lamer, Mariko Oue, Robert Schrom, Zhiyan Jiang, Brenda Dolan, Jeffrey Snyder, and Shaocheng Xi
19 March 2018
4:00 PM - 6:00 PM
30
Katia Lamer, Mariko Oue, Robert Schrom, Zhiyan Jiang, Brenda Dolan, Jeffrey Snyder, and Shaocheng Xi
Breakout Description
While some of ARM’s sensors semi-directly measure geophysical quantities such as those produced by atmospheric models (e.g., liquid water path), most measure indirect quantities that require interpretation. Forward-simulators offer an alternative path by transforming model quantities to observables. They rely on our knowledge of instrument characteristics and particle scattering to produce virtual observations from model output.Main Discussion
Presentations highlighted that a lot of the uncertainty in forward space can be attributed to ice particles (e.g., ice crystals, aggregates, graupel, hail), both because models do not explicitly evolve and report their properties and because they are poorly documented observationally. For forward-simulations, important ice crystal properties include but are not limited to hydrometeor canting angle, aspect ratio, maximum dimension, water fraction, and ice particle. Performing forward-simulation ensembles based on different assumptions was proposed as a way to address uncertainty in forward-simulators. Statistics constructed using all ensemble members have been used in an attempt to quantify uncertainty; Alternatively, the forward-ensemble member closest to observations has been used for model evaluation.The complexity of forward-simulators is ever increasing. Forward-simulators for multiple sensors (e.g., radar and lidar) compatible with models of multiple scales (e.g., LES and GCM) are being developed. Beyond returned power, virtual observation of quantities relevant to cloud phase identification are starting to emerge (e.g., polarimetric and Doppler quantities). Presentations showed how forward-simulators can be used to mimic observations including instrument sampling limitations (sensitivity, attenuation, and sampling), and how observation-based retrievals can be performed on these virtual observations just like on ARM observations in an end-to-end process (e.g., hydrometeor phase identification and 3-D VAR wind retrievals). Presentations highlighted the versatility of forward-simulators, which can be used to evaluate retrieval, optimize sensor placement, and sampling strategy as well as to evaluate numerical models.
This session helped address the ongoing need for communication between forward-simulator developers, modelers, and users.
Needs
There is a need for increased communication between forward-simulator developers and modelers. It seems like lessons learned while coupling each new model to a forward-simulator are not captured and shared. As such, the process of model to forward-simulator coupling remains time consuming. We propose to complement the release of any static user guide with an evolving Frequently Asked Question (FAQ)-type resource.
For an objective model evaluation using observations, forward-simulators must reflect instrument capabilities.
This requires consistent and transparent sensor characterization (e.g., minimum detectable signal).
Large-scale models and high-resolution ARM observations differ in scales.
There is a need to develop a tool to account for these scale differences.
There is a need to evaluate the few approaches that have been proposed (e.g., model sub-column generation, creation of CFADs or observation resampling through the cloud vertical structure approach).
Forward-simulations contain uncertainty especially owing to our limited knowledge of ice crystal properties both in models and observations.
We suggest continuing work related to ice crystal characterization, especially aggregates (e.g., using MASC data for size and velocity and canting angle).
We suggest continuing work related to expanding scattering databases.
We suggest continuing to explore the possibility to evolve detailed information about frozen hydrometeors in models, but only to the point where this information improves simulations and not only to reproduce observations.