I recently came back home after a conference biathlon over a period of three weeks across two continents - The Ecological Forecasting Initiative (EFI) 2025 Conference in Blacksburg, Virginia, and the One Ocean Science Conference in Nice, France. Conferences can be overwhelming and also completely exhilarating - the concentration of boundary-pushing ideas in such a small area of space and time makes my own ideas and thinking leap forward more than I could ever accomplish reading papers. These conferences were no exception.
EFI 2025 - It’s always about scale
EFI, the organization on which I serve as an early career and student co-chair, put on tremendously fun small conference at Virginia Tech, nestled in the Blue Ridge Mountains of southwestern Virginia. EFI serves a really forward-looking group of ecologists who think beyond building models of how ecological systems work, and seek to test their predictions in the real world. In my mind, the centerpieces of EFI are the forecasting challenges, which include the NEON and FLARE (Forecasting Lakes and Reservoir Ecosystems) - true examples of ecological prediction at its best, leveraging ensemble modeling created by teams across the world to predict, in real-time, a range of environmental and biological parameters.
These challenges, and the conference overall, really opened my eyes to how the analytical divisions we often slot ourselves into when performing ecological modelling should really be viewed as just a range of techniques for attempting to describe and predict patterns in nature. Time-series analysis methods (ARIMA, lagged regressive techniques) are just another side of the die that also include mechanistic, biophysical/ecophysiological models, black box machine learning, and cutting edge hierarchical Bayesian analysis. Each method has its adopters, and strengths and weaknesses, but ultimately all modelling approaches may be a valid way of describing ecological patterns.
Indeed, some approaches do better at different spatial and temporal resolutions, while abstracting away trends in the data that are well-described by another approach. For example, Dr. Mark Urban highlighted how temporal autocorrelation is responsible for current-state, statistical approaches performing best over short forecast horizons, while process-based models do better at longer time scales. There is no clear horizon where one is clearly better - models have temporal, and likely spatial, “neighborhoods” or domains where inference is high before decreasing (or the inverse!). In some cases, it seems that predictive ability may actually increase over longer time scales, once the ecological processes that actually drive the large-scale patterns we are actually interested in interact at the proper temporal scale.
I really enjoyed listening to a keynote by Dr. Wendy Parker (VT), who works on the philosophy of prediction. Succinctly, good model fit and “mirror fidelity” to an ecological process may not be the best goal for our forecasting approaches - rather, the adequacy for purpose should be considered. Ultimately, we build models to accomplish some goal, and this goal should always be considered when building our models. Once we account for these goals, our models are much better situated for decision support and use by managers and policymakers.
Finally, one of my biggest takeaways is the importance of a robust cyberinfrastructure for iterative forecasting. If one can get continuous data streams, or at least automated packet deliveries, the time lag for model scoring, and by extension development of better models, is much reduced. Having server or cloud-based systems that can automate much of the data cleaning, analysis, and visualization greatly extends the reach of a forecasting system beyond a single group of researchers. There are so many awesome tools now, including Github actions and FAAS (functions as a service) that enable ecologists to speed up the modeling cycle exponentially.
Of course, the next best part about the conference was its location in the Blue Ridge, an area of the country close to my heart. It’s hard to go wrong with a post-conference hike and communal square dance!
French-American Doctoral Exchange (FADEX) Program and One Ocean Science Congress
Within three days of returning from Blacksburg, I got on a flight to Paris to begin the weeklong FADEX program put together by the French Embassy to the United States. I was lucky enough to be selected for this program along with 10 other American PhD students in the ocean sciences to travel to France to tour labs and meet French students and researchers. The program was designed to end in Nice during the One Ocean Science Congress (OOC), itself the immediate scientific entrée to the UN Ocean Conference (UNOC). Given the retreat of the current US regime from funding some of the best institutions and researchers in the world (I am restraining my editorializing), I was particularly interested to see how French and European scientific organs are filling the scientific power vacuum.
After meeting with researchers at the Maison de l’Océan (Prince Albert I of Monaco’s urban temple to the ocean sciences) and Muséum national d’Histoire Naturelle in Paris (and some sightseeing amidst a historic Paris-Milan football match), we took the high-speed train to Brest, where IFREMER and the Université de Bretagne Occidentale have major marine labs. I was struck during this visit by the French system of research neighborhoods, or teams, which may accomodate several PIs within a single umbrella lab. After a quick, scenic 24 hours in Brest, we took high speed rail to Nice to begin the OOS Congress.
OOS was a complete change of pace from the EFI conference - from 100 to over three thousand attendees, and from niche ecological topics to wide-ranging, high-impact ocean science. After battling through an extensive security zone, each day I would enter the congress grounds directly on Port Lympia. Watching my fellow American FADEX attendees give their talks in the big conference tent was a pretty unreal experience!
Some of the other highlights of the OOS were the public-facing ocean exposition called La Baleine (the whale), an impressive display of ocean-themed exhibits in a massive conference center, and the achingly beautiful special art event held at the Nice Opera House. We had the opportunity to meet with the (sadly) small group of US politicos attending OOS/UNOC, including Sen. Sheldon Whitehouse (RI) and the mayor of New Orleans. In better times, I hope to see a much stronger and dedicated US delegation in attendance. As a final part of our FADEX program, we met French doctoral students at the Sorbonne’s Laboratoire Oceanographique de Villefranche-sur-mer.
I am deeply grateful for the Embassy of France in the United States for providing this incredible opportunity. I was heartened and inspired to see a large-scale investment (and trust!) in not only the sciences, but efforts to combat climate change and degradation of the world’s oceans.