Reflective at GeoMIP
A few weeks ago, myself and other members of the Reflective team — Dakota, Alistair, and Dan — attended the Geoengineering Model Intercomparison Project’s (GeoMIP) 16th Annual Meeting in Tokyo, Japan. The meeting included talks, breakout discussions, and poster sessions, bringing together the community of earth system modelers and climate scientists from around the world who produce and use GeoMIP simulations. At Reflective, we used the meeting to present two of our developing tools, the Uncertainty Database and the Cloud Hub, build connections, and keep up to date with upcoming developments.
GeoMIP, co-chaired by Reflective’s Head of Data Daniele Visioni, coordinates sunlight reflection simulations across modeling centers worldwide. Running the same core set of simulations across models makes results comparable, allowing researchers to estimate model uncertainty in projections of the response to both climate change and sunlight reflection. This meeting was particularly significant given the ongoing preparations for CMIP phase 7 — which will bring new and improved climate models and emissions scenarios — as well as the fast approaching IPCC 7th assessment report.
Much of the GeoMIP experiment planning discussion centered on how to adapt our scenarios in light of these developments, particularly over which of the new ScenarioMIP background scenarios to choose for GeoMIP’s central sunlight reflection simulations. These are important decisions as they will be influential towards how the SAI literature evolves over a long period — both as it pertains to scientific analysis as well as public perception. The world has changed a lot since the last CMIP6-era scenarios were proposed in 2015, and it is necessary to balance these new scenarios with competing priorities from various researchers.
The Uncertainty Database
I presented our Uncertainty Database — now live on our website — which serves as a dynamic, public, scientifically-grounded assessment of the key technical uncertainties in SAI. The community’s interest was encouraging, and the feedback we received will help sharpen the work. One area for improvement that surfaced is ensuring that first impressions of the database align with our intentions. The current matrix on the home page of the database can leave the impression that the areas of research in the “low-low” category are unimportant. Rather, we want to make sure we communicate that every uncertainty on the list would benefit from more research, but we are also trying to achieve a prioritization of the most urgent uncertainties to address.
Additionally, getting people on the same page in terms of quantitative categorization can be challenging, specifically as it’s tempting to air on the side of caution and rate all uncertainties as large. We have tried to overcome this in our work by defining uncertainties in terms of semi-quantifiable statements — or “metrics” — and feedback we received prompted us to consider how our visualisations and first impressions can foreground this aspect.
We also received questions about the ability to provide feedback on the database, leading us to realize that the feedback buttons are not very prominent unless one does a deep dive into the site. We plan on encouraging feedback on the homepage and making the links to contribute more obvious in order to address this issue. Lastly, we received a few comments from people that were unaware we had launched the database, which stressed the importance of doing more outreach with it to ensure it functions as a community resource. On that note, we welcome input on the database’s content — if you disagree with any of our assessments, please use the link on our website to let us know.
The Cloud Hub
Senior scientist Alistair Duffey presented Reflective’s Cloud Hub, a resource for broadening access to earth system model simulation output — particularly of GeoMIP experiments — which provides free compute resources to researchers around the world for running analyses in the cloud. The GeoMIP community are core users of this platform, so it was great to have the opportunity to hear about the amazing diversity of work that is already happening on the Cloud Hub and what features users would need to accelerate their work further. It was highlighted that producing and hosting a set of climate extremes indices (the ETCCDI metrics) across all the recent GeoMIP experiments could save the community substantial duplicated effort, and work is already underway to get those indices calculated and online.
We also surfaced challenges in working with large data, such as model outputs at sub-daily resolution in three spatial dimensions, in the cloud using the models’ native output file formats. This highlighted the need for us to move fast with our ongoing work to produce cloud-optimised (zarr) formatted versions of these recent simulations. These new versions of the G6-1.5K-SAI and -HiLLA simulations are currently in testing and will be shared with the community shortly. This extends work Reflective commissioned last year to produce a cloud optimised store of the previous generation of GeoMIP data.
Other Notes
The Reflective team participated in a breakout session focused on research/analysis coordination. One theme came through repeatedly: many researchers are eager to shift away from what one described as “déjà vu analysis” — research that revisits similar scenarios with updated models — and towards new, usable, policy relevant questions. There was a strong interest in mapping neglected impact areas, development of shared downscaling workflows, and wider use of common data infrastructure (like Reflective’s CloudHub) that accelerates research and reduces friction. Additionally, the advantages of collaborating with researchers outside of the sunlight reflection community doing work in relevant impact areas was made clear, and prompted us to think about ways in which we can support that work.
On a personal note, this was my first GeoMIP meeting. After two and a half packed days, I came away with a much clearer sense of how the community decides which simulations to prioritize, and it was rewarding to share Reflective’s work alongside that. Participating in the breakout session on Tier-2 scenarios (additional scenarios on top of the main few that all modeling centers will run) helped me appreciate the complexity of balancing important, but at times competing, interests, and how these decisions steer the resulting literature and perceptions of SAI. I want to give a big thank you to local hosts Dr. Masahiro Sugiyama and Dr. Shingo Watanabe, and to GeoMIP co-chairs Dr. Alan Robock and Dr. Daniele Visioni for a fantastic meeting.
