BUFFALO, N.Y. — Advanced computer models can quantify the impacts of climate change and other environmental challenges, providing deep insights into things like streamflow, vegetation, wildlife and even the risk of wildfires.
But what good is all that data if the communities most affected can't act on it?
A University at Buffalo researcher has developed a framework to help scientists incorporate community input into Earth system models, tools that simulate climate as well as chemical and biological processes.
The framework was published in the December issue of AGU Advances and featured in AGU's Eos Research Spotlight . While many studies discuss community engagement in climate research, this study is novel in that it takes a modeler's perspective for incorporating community priorities into complex Earth system models.
"Very often we modelers make decisions about the parameters of our projects behind closed doors, but co-designing a model with the community will ultimately make our work more tangible and relevant," says the study's corresponding author, Yifan Cheng, PhD, assistant professor in the UB Department of Earth Sciences. "We should really want our datasets about a region to not only be accurate and insightful, but also directly benefit the people who live there."
The framework is based on Cheng's experience working with Indigenous communities in Alaska and California as a postdoctoral fellow and later project scientist for the National Center for Atmospheric Research. Both projects were funded by the National Science Foundation.
His collaborators include Nicole Herman-Mercer, a research social scientist with the United States Geological Survey Southwest Climate Adaptation Science Center, who has worked for years with tribes throughout Alaska.
"There are so many decisions that go into the development of Earth systems models. The work that we have done shows that it's possible to include community and end-user priorities when making those decisions," Herman-Mercer says. "Co-designing models with communities and end-users results in a product that not only more readily addresses local concerns than traditional modeling approaches but also increases trust in the model output."
Different approaches for different-sized communities
The Alaska and California projects differed widely in scopes — giving Cheng a range of examples for how co-design can play out.
The Alaska project modeled the climate and hydrology of the entire state, as well as the Yukon Territory and part of the Northwest Territory in Canada. Given this region is home to 229 federally recognized tribes and over 40 First Nations, the project team convened an Indigenous Advisory Council, administered a survey to all Indigenous governments and organizations in the region, and held an in-person community summit to gather feedback from community members and local decision-makers.
Still, the sheer scope of the project — 30 million allocated central processing unit hours — limited how much they could fine-tune the model based on community feedback.
The team's project in the Mid-Klamath region of Northern California was different. There, they collaborated only with the Karuk Tribe and their partner organizations, allowing them to tailor the modeling approach much more directly to tribal questions about watershed runoff, streamflow, salmon populations and cultural and prescribed burning.
They were even able to revise the model mid-project. When the tribe pointed out that the fire model could not simulate dynamic fire spread, the modeling team pivoted to only simulating the effects of static snapshots of different fire practices on the landscape.
And when the model consistently overpredicted streamflow — and tuning could prioritize either summer or winter flows — it was tribe members who chose to prioritize summer flows, given their importance for salmon habitat and spawning.
"These two experiences made me realize there's not a one-size-fits-all approach to co-designing models with a community," Cheng says.
That's why the framework outlines four levels of co-design — elements that modelers can incorporate into a project depending on community needs and available resources. They include:
jointly configuring the model setup with communities;
fine-tuning the model to better capture what communities care about;
incorporating local knowledge and observations directly into the model;
and developing new model functions to represent processes that are important to a community but missing from existing models.
"Modelers can use all or just one of these levels in their project," Cheng says. "Lower-level co-design is not better than higher-level ones — even using just a level 1 co-design can greatly boost the salience and usefulness of the model's output."
Cheng credits his co-authors' preexisting relationships with the Indigenous communities. Herman-Mercer has worked for years with tribes throughout Alaska, while Cleo Woelfle Hazard, fire advisor for the University of California Cooperative Extension, also works as habitat restoration program manager for Karuk Tribe Department of Natural Resources.
"The communication with the communities was smooth because members of our team already had developed these relationships over years," Cheng says. "That kind of trust is very important for co-design."
