Breakout Summary Report
ARM/ASR User and PI Meeting
10 - 13 June 2019
10 June 2019
1:30 PM - 3:30 PM
50
J.Fan, D.Feldman, K.Lamer, S.Xie, B.Gustafson, R.Jackson, D.Kim, C.Naud, C.Schumacher, M. Ahlgrimm
Breakout Description
While ARM observations can prove invaluable to ensure adequate representation of atmospheric processes in large-scale models, their scale often differs greatly from that of large-scale models such that they may not be representative of grid-box-simulated quantities. This breakout session will, by discussing product development and cross-discipline uncertainty, present current efforts to advance the use of ARM data both for testing large-scale models, such as the earth system model, and for charting a course towards improved fidelity in their representation of atmospheric and land-atmosphere interaction processes. Because the approaches for earth system model development and ARM data analysis and product development differ so substantially, the session will also leave time for an open discussion to strengthen existing paths and establish new ones to work in tandem with earth system model developers to identify where observations can be optimally deployed to reduce model errors and improve model parameterizations.Main Discussion
The speakers presented on a broad range of issues related to challenges in and opportunities for using ARM measurements in the service of global model evaluation and parameterization development. More specifically, the six presentations in the session covered i) the development of model evaluation targets especially for the evaluation of rainfall rate, ii) the development of model evaluation techniques including comparing observed and simulated quantities (e.g., degree of cloud organization) but also trends and relationship between quantities (e.g., relationship between cloud base height and sea air temperature contrast). Talks covered a wide range of cloud conditions from low-level clouds to convective systems and a wide range of techniques from retrievals to forward-simulators.
We had very active discussion including the following topics:
- Pros and cons of using forward simulators and retrieval products
- How to better quantify data uncertainties and communicate data uncertainties from both the observation and modeling sides
- How to build robust communication between observationalists and modelers to span the gap between the languages and approaches that the two groups use
Also, given progress in the last few years, the community expressed a desire to start addressing the scale gap challenge that one faces when comparing ARM high-resolution observations and GCM models. For this reason, a lot of discussion surrounded the topic of subcolumn generators, which handle cloud and precipitation vertical overlap. There was great interest in better understanding and modifying the COSP subcolumn generator to increase its consistency with how GCMs treat overlap.
Key Findings
ARM measurements can be very useful to develop and evaluate model parameterizations. However, the comparison of such data with global or even regional model results is highly non-trivial. The challenges associated with this comparison can be seen most immediately in the discussion of uncertainty: there can be a fundamental mismatch in how observational and modeling uncertainty are represented, and this gap needs to be spanned to understand the extent to which models functionally disagree with ARM observations. To realize the scientific value of ARM data with respect to revealing improvements in process representations within models requires close, sustained collaborations among ARM data developers, observationalists, and modelers. These collaborations are best facilitated through i) individuals with expertise across these disparate disciplines, and ii) mechanisms that enable and actively promote co-development and science. To that end, a consensus was reached that robust communication and mutual education can be much better achieved through projects that require both expertise in ARM observations and expertise in model development.Issues
Despite the fact that the session was hosted Monday when many people were still commuting, the organizers were encouraged by the significant level of attendance at the session. The scope of the session was broad, covering topics from forward simulators to retrieval and from low-level clouds to deep convection. Since the needs and challenges of each of these communities somewhat differ, it could be beneficial to host separate breakout sessions on these topics in the future for more in-depth discussions.Needs
- Need to define, quantify, and communicate uncertainties. The meaning of uncertainty can vary depending on the discipline.
- Need to train each other to have mutual understanding regarding the scales of comparison, uncertainties, and caveats in both sides.
- Need designated projects to build robust communication and mutual education among data developers, observationalists, and modelers.
- Need scientist “liasons” focusing on both observations and modeling. Mechanisms and incentives are needed to promote their work, even though it is not necessarily as high profile as intra-disciplinary research.
- Need to change how we educate students to break down the persistent issue of the categorization of researchers into observation and modeling categories. Likewise, we should continually seek to break down silos in research projects as necessary to prevent us-versus-them mentalities.