Breakout Summary Report

ARM/ASR User and PI Meeting

2 - 6 May 2016

Session Title: New Tools to Bridge the Gap between Models and Observations
Session Date: 5 May 2016
Session Time: 8:30 AM - 10:30 AM
Number of Attendees: 0
Summary Authors: Scott Collis, Shaocheng Xie, and Laura Riihimaki

Breakout Description

The goal of the session is to address one of the big challenges that ARM and ASR face: how to bridge the gap between models and observations? We will discuss a few new tools that are being developed by ARM to support ASR cloud modeling studies and GCM parameterization developments, as well as ARM data processing and discovery. They include ARM diagnostic packages and radar simulators for both LES and GCMs, the ARM Data Integrator (ADI) for data processing, and new data discovery tools. We will have a few invited talks that cover these areas to initiate conversation and leave plenty of time for general discussion. We will also discuss priorities on new data products and tools identified by the recent DOE CESD workshops and the recent BER data informatics workshop as well as the ARM user community.

Main Discussion

The session was very well attended with a nice mix of ARM infrastructure and the modelling community. It started with questions listed by the session Chairs on challenges that ARM/ASR face in bridging scale gaps between data and models and how can we facilitate use of ARM data with various data and modeling tools developed within the programs. Next came a few short solicited talks on each of the main topics, deferring the bulk of conversation until the end of the session. The session was seeking for feedback from the following areas:
• Comments on ARM/ASR efforts to develop and use these modeling and data-processing/discovery tools
• How to effectively use these tools in support of ASR studies
• Is it a good idea to have a designed area for distributing the tools and associated ARM data to the ASR/ARM community? Is there any support or maintenance that we should provide?
• Should we build a test-case library particularly for the CESD modeling testbeds?
• Effort to address data uncertainty quantification?
• Any missing areas that we need to make an effort to address?

Testbeds: Shaocheng Xie introduced CESD modelling testbeds including FASTER, CAPT, Regionally Refined Meshes, and the Aerosol Modelling Testbed. He also showed how some of the testbeds such as CAPT are being efficiently used in ACME model tuning and developments.

Tools for merging/analyzing data: Laura Riihimaki covered ADI as well as the work by Bhargavi Krishna at ORNL towards rich parallel databases describing ARM data. There was good audience engagement along the lines of how, exactly, ADI could be used to automate some of the data-gathering process. Scott Collis covered infrastructure efforts on the Python-ARM Radar Toolkit (Py-ART) specifically focusing on the community aspect of the project. Specific questions to Scott covered the variety of formats that could be read using Py-ART and also a question on how, specifically, Py-ART can support the modelling community. Scott advertised the upcoming Py-ART roadmap survey and was looking for feedback from the modelling community to better meet their needs.

ARM diagnostic packages for GCMs and fine-scale models: Chengzhu Zhang from LLNL detailed the ARM diagnostic package for GCMs, which automates a key set of statistical comparisons between GCM output and ARM (and other) observations. Discussion included how the package would be made available to the modeling community. One planned path is through PCMDI. Andy Vogelmann (BNL) from the LASSO team introduced some of the very detailed diagnostic work using ARM observations to determine the accuracy of the LASSO Data Bundles. Various metrics were presented, including equitable threat scores. Good discussion followed including how to hold back data from assimilation in order to aid model assessment.

Instrument simulator packages: Yuying Zhang of LLNL detailed progress on the ARM GCM simulator for MMCR/KAZR and the underlying methodologies. Pavlos Kollias of Stony Book University detailed work on CR-SIM, including an upcoming new version that will allow scanning radars and work being carried out toward a LIDAR simulator--which, in turn, will allow a simulated ARSCL product from cloud resolving model output. This session garnered a lot of feedback. Two specific areas were: With increasing resolution, what does a scanning radar simulator that goes across grid cells look like in a GCM (from Shaocheng Xie and Andrew Gettelman) and how much sophistication do we put into a simulator as against correcting for “quality-reducing” phenomena (attenuation, non-Rayleigh, partial beam filling, etc.) in the measured data?

Cloud classification and other indices of atmospheric state: Laura Riihimaki described the cloud classification product that will allow the subsetting of data by cloud type supporting the LASSO project, and discussed requests to extend the code to other sites or indices. Suggestions included making the code available so stakeholders could tailor the retrieval to their needs. Maike Algrihmm requested a large-scale classification of weather state. Mark Miller said the ENA Site Scientist team is working on an algorithm to classify weather state at the Azores. Initial data has only been run during the CAP-MBL period but will be expanded to an operational product when necessary ENA instrumentation is running. Laura encouraged Mark to submit this as a PI product and will follow up with Mark about the best way to make this available to ARM data users. This also led to a conversation later between the translators that we need to better know what product work is being carried out by Site Scientist Teams.

General discussion: Discussion ensued after the talks, with a variety of feedback including questions on how best to make code and tools available. Scott Collis, as usual, recommended GitHub and moving there earlier rather than later. The fact ARM has a GitHub account that would permit users to have a private repository before going public was raised and interested parties were directed to contact Sherman for more information. A final discussion point was on data visualization. Minghua Zhang expressed a desire to see flow vectors overlaid on moisture fields and for these images to be animated. This was raised as a desired capability and the feedback could be generalized to a need for more tools to do such visualizations. Both Scott Collis and Bill Gustafson have experimented with ParaView for 3D visualizations but the tool required a lot of customization for geophysical use. Further discussion with Minghua clarified that a set of quicklook movies of LASSO output would be a sufficient first step if interactive visualization is difficult. Andrew Gettelman also said he is working on subsetting detailed GCM output over ARM sites and is willing to provide that to the ARM archive as a PI product if desired by the community.

Needs

• Better visualization for gridded model (and other?) data, specifically for the generation of animations.
• Need to work closely with the ASR science teams. Specifically, opportunity exists to strengthen connections between site scientist teams and ARM infrastructure.
• Better coordination with similar activities conducted in other communities outside ARM/ASR such as the ARM metrics/diagnostic work.