Search for: All records

Research in both ecology and AI strives for predictive understanding of complex systems, where nonlinearities arise from multidimensional interactions and feedbacks across multiple scales. After a century of independent, asynchronous advances in computational and ecological research, we foresee a critical need for intentional synergy to meet current societal challenges against the backdrop of global change. These challenges include understanding the unpredictability of systems-level phenomena and resilience dynamics on a rapidly changing planet. Here, we spotlight both the promise and the urgency of a convergence research paradigm between ecology and AI. Ecological systems are a challenge to fully and holistically model, even using the most prominent AI technique today: deep neural networks. Moreover, ecological systems have emergent and resilient behaviors that may inspire new, robust AI architectures and methodologies. We share examples of how challenges in ecological systems modeling would benefit from advances in AI techniques that are themselves inspired by the systems they seek to model. Both fields have inspired each other, albeit indirectly, in an evolution toward this convergence. We emphasize the need for more purposeful synergy to accelerate the understanding of ecological resilience whilst building the resilience currently lacking in modern AI systems, which have been shown to fail at times because of poor generalization in different contexts. Persistent epistemic barriers would benefit from attention in both disciplines. The implications of a successful convergence go beyond advancing ecological disciplines or achieving an artificial general intelligence—they are critical for both persisting and thriving in an uncertain future. 

Freshwater ecosystems provide vital services, yet are facing increasing risks from global change. In particular, lake thermal dynamics have been altered around the world as a result of climate change, necessitating a predictive understanding of how climate will continue to alter lakes in the future as well as the associated uncertainty in these predictions. Numerous sources of uncertainty affect projections of future lake conditions but few are quantified, limiting the use of lake modeling projections as management tools. To quantify and evaluate the effects of two potentially important sources of uncertainty, lake model selection uncertainty and climate model selection uncertainty, we developed ensemble projections of lake thermal dynamics for a dimictic lake in New Hampshire, USA (Lake Sunapee). Our ensemble projections used four different climate models as inputs to five vertical one-dimensional (1-D) hydrodynamic lake models under three different climate change scenarios to simulate thermal metrics from 2006 to 2099. We found that almost all the lake thermal metrics modeled (surface water temperature, bottom water temperature, Schmidt stability, stratification duration, and ice cover, but not thermocline depth) are projected to change over the next century. Importantly, we found that the dominant source of uncertainty varied among the thermal metrics, as thermal metrics associated with the surface waters (surface water temperature, total ice duration) were driven primarily by climate model selection uncertainty, while metrics associated with deeper depths (bottom water temperature, stratification duration) were dominated by lake model selection uncertainty. Consequently, our results indicate that researchers generating projections of lake bottom water metrics should prioritize including multiple lake models for best capturing projection uncertainty, while those focusing on lake surface metrics should prioritize including multiple climate models. Overall, our ensemble modeling study reveals important information on how climate change will affect lake thermal properties, and also provides some of the first analyses on how climate model selection uncertainty and lake model selection uncertainty interact to affect projections of future lake dynamics. 

Abstract Lake ecosystems, as integrators of watershed and climate stressors, are sentinels of change. However, there is an inherent time-lag between stressors and whole-lake response. Aquatic metabolism, including gross primary production (GPP) and respiration (R), of stream–lake transitional zones may bridge the time-lag of lake response to allochthonous inputs. In this study, we used high-frequency dissolved oxygen data and inverse modeling to estimate daily rates of summer epilimnetic GPP and R in a nutrient-limited oligotrophic lake at two littoral sites located near different major inflows and at a pelagic site. We examined the relative importance of stream variables in comparison to meteorological and in-lake predictors of GPP and R. One of the inflow streams was substantially warmer than the other and primarily entered the lake’s epilimnion, whereas the colder stream primarily mixed into the metalimnion or hypolimnion. Maximum GPP and R rates were 0.2–2.5 mg O 2 L −1  day −1 (9–670%) higher at littoral sites than the pelagic site. Ensemble machine learning analyses revealed that > 30% of variability in daily littoral zone GPP and R was attributable to stream depth and stream–lake transitional zone mixing metrics. The warm-stream inflow likely stimulated littoral GPP and R, while the cold-stream inflow only stimulated littoral zone GPP and R when mixing with the epilimnion. The higher GPP and R observed near inflows in our study may provide a sentinel-of-the-sentinel signal, bridging the time-lag between stream inputs and in-lake processing, enabling an earlier indication of whole-lake response to upstream stressors.