Breakout Summary Report
ARM/ASR User and PI Meeting
Pairing SCM/LES and GCM/ESM for observation-guided model development
25 June 2020
11:00 AM - 1:00 PM
120
Ann Fridlind with edits from all co-conveners
25 June 2020
11:00 AM - 1:00 PM
120
Ann Fridlind with edits from all co-conveners
Breakout Description
The performance of a climate model's single-column model (SCM) physics may provide aquite faithful representation of that climate model's regional cloud properties. For instance, the SCM
representation of subtropical marine low clouds along boundary layer Lagrangian trajectories is expected
to resemble the parent model's regional physics along similar trajectories (Neggers JAMES 2015). An
actionable implication is that suitable SCM case studies and long-term observations can be combined to
effectively guide climate model physics improvements and confirm their impacts in relevant long-term
statistics. This session invites discussion of current and potential future community projects that are
motivated by specific climate model development objectives, and that employ a combination of SCM/LES
and GCM/ESM approaches using ARM observations. A leading example is provided by the GASS Diurnal
Cycle of Precipitation Project (https://portal.nersc.gov/cfs/capt/diurnal/). A recently derived Lagrangian
SCM/LES case study of highly supercooled drizzling stratus over Antarctica and development of an open
source radar-lidar forward simulator for statistical evaluation of GCM/ESM cloud phase offer a pairing that
is focused on polar cloud phase.
Planned objectives of the breakout discussion:
(1) establish degree of interest in this general approach, or modifications thereon
(2) collect input on known climate model biases or ECS-relevant processes of interest
(3) discuss possible coupled LES-SCM/GCM-obs exercises to address those
Short introductory talks will be limited to briefly outlining the concept, the DCP project example, the role
for a ground-based simulator component, and a climate modeling perspective.
Main Discussion
The first hour was spent on three presentations outlining the concepts being proposed (Ann), the currently ongoing GASS Diurnal Cycle of Precipitation (DCP) project (Shaocheng), and an open source ground-based lidar-radar simulator tool to enable ESM evaluation using long-term data sets (Israel and Bobby). Questions were asked about the DCP project results thus far and the capabilities of the simulator—whether scattering libraries could be expanded with an ensemble approach and whether microphysics PDF assumptions could be altered. Bobby and Israel answered that those capabilities could be readily added in an open source and community code manner, by design.The second hour was spent in open discussing re the following leading questions:
● Is there a need for more organized modeling activities within ASR?
● Should a model-centric focus group be reinstituted?
● Should the focus be expanded beyond IOPs to harness long-term data?
● Could such a group support multiple, diverse group activities with varying group sizes?
A wide range of participants from both domestic and international institutions shared concise and thoughtful reactions. We have appended at the end a rough transcript of all discussion period contributions as a numbered list (view Full Breakout Session Report https://asr.science.energy.gov/meetings/stm/presentations/2020/1119.pdf).
Key Findings
There was unanimous support for some kind of more organized modeling activities but some diversity of opinion on how best to pursue that. We refer to the appendix list by number in the following summary findings:● Wary of the enormous overhead associated with past intercomparisons, there was wide support for a more limited bulletin board (web page) approach to stimulating smaller group activities (2,4,6,8). But some were concerned that that would be "another place to forget to check" (3) or felt that that would be insufficient to develop the scientific excitement and sustained momentum of direct community interaction such as occurred around past GCSS case studies (3,6,17).
● Others favored an explicit group activity that would be capable to span modeling activities from LES to ESM (5,14), noting for instance that it is possible to intentionally focus observational constraint studies on the physics to which ESMs are actually established to be most sensitive (14). This bears similarity to the DCP project example, spanning modeling frameworks and focused on the physics behind a well-known climate model bias.
● Some fully supported the value of new and revisited intercomparison activities to study model physics (7,14,17), and also welcomed the LES/SCM and ESM pairing to address the pitfalls of tuning to limited cases or specific sites that have been problematic (7,16). In the case of intercomparison activities, there was support for low-overhead participation options (1,5).
● There was wide and unopposed support for creating a readily accessible and easy-to-use LES/SCM case study library (4,5,9,11,12) and for using a standardized format like DEPHY (4,5). Efforts are also already being made towards that goal (10,13,20).
● A two-pronged approach was suggested to in parallel pursue the value of getting a wide community pointed in the same direction for intercomparison activities, in addition to supporting a "one stop shop" for time-pressed modelers to be able to quickly access existing case studies off the shelf and readily use them (11).
● The importance of large-scale forcing uncertainty led to discussion of an ensemble forcing approach, existing and past efforts to use an ensemble approach (LASSO and VARANAL examples), and the potential value of a covariance matrix approach to assessing uncertainty in large-scale forcing (18-20,23).
Issues
Some wish to meet and discuss across modeling expertise and pursue intercomparison studies and paired physics work whereas others feel the overhead is problematic. We propose that maintaining an entirely optional participation is desirable. It may also be desirable to maintain periodic virtual meetings to support international and non-PI participation.Needs
Program support is needed for the case study library. It would also be highly desirable to offer limited computing environment support where participants can upload results and plot against specific case study observations and other uploaded model results. Aside, we note that DEPHY workshop participants running SCMs desired to be able to download LES results for cases without having to generate that part themselves.Decisions
We conclude that there is sufficient interest and perceived value among a wide enough group that further discussion is warranted. However, concerns about the amount of overhead on giant omnibus studies are also well warranted. We recommend to develop a low-overhead plan with the following modular elements:● A modeling focus group with participants spanning LES, ESM and observational expertise and entirely optional participation. Such a group could operate in parallel to and entirely deprecated to existing working groups. Discussions could consider development of new LES/SCM cases with any number of participants, paired analyses of ESMs versus long-term data, considerations relevant to the case study library, etc. The group could also serve as a clearinghouse to testing and coordination of the case study library. Joint activities could involve any number of participants but all main work would be done via offline coordination (similar to GCSS serving as an umbrella for many case studies simultaneously in the past).
● A readily accessible case study library using a standardized format such as DEPHY being developed under French leadership (with widespread U.S. participation) for any SCM and LES input forcing and outputs. Development could include ingesting E3SM's existing case study library, and testing the work flow in the course of group activities (e.g., revisiting M-PACE LES/SCM, perhaps paired with an AWARE case and long-term data analysis as proposed).
● A bulletin board (web page) to track activities and any other community communications.
Group activities should be as low-overhead, high-impact, and optional as possible. Tools should be as easy to use as possible. Open source and community codes should be used to build the input files (allowing for ease of tweaking any case elements), plot SCM or LES results versus observations, etc.