Discussion Session 2: What kinds of experimental and observational data can be collected that are of best use to models?

Chair: Murray Roberts. Rapporteur: Brian Gaylord.

A critical issue in both population and ecosystem models is that OA impacts are known for only a subset of the life cycles of most species. Many marine taxa have complex life cycles that include a microscopic larval stage, and information concerning competency periods, dispersal pathways, and patterns of connectivity (for example) are often poorly described.  These weaknesses become especially relevant in the context of spatially explicit models, where issues of source and destination become integral. Emerging work has also revealed the importance of carry-over or legacy effects that link different life stages; such connections suggest that experiments or models that examine only single stages in isolation could miss important impacts.

The vast majority of OA experiments have employed ANOVA-style designs, testing for effects of acidification at only a couple of levels of pCO2. However, there is considerable value in expanding beyond a traditional ANOVA framework. Regression approaches that employ a greater number of levels in the independent variable have the capacity to yield essential insight into the “functional responses” of organisms – defining how plants or animals react to a full spectrum of seawater pCO2 or pH. Such information can assist in the identification of non-linear responses, including threshold effects and tipping points that have the potential to induce dramatic and unexpected outcomes. At present, the likelihood of such nonlinear impacts, and how they might manifest in any number of biological and ecological systems, remains almost entirely unexplored. Regression methods, of course, are only one example of a much broader array of possible statistical and analytical approaches that could be advantageously applied to the subject of OA. Bayesian and AIC-based (Akaike’s information criterion) tools provide just two additional examples that might provide useful avenues for pursuit.

There is a growing awareness that acidification conditions vary strongly in space and time, yet experiments to date have largely omitted such variability. Exceptions include recent in-situ work using FOCE (Free Ocean Carbon Enrichment) methods, and field efforts to characterize finer-scale environmental variability in pH and other carbonate parameters. A related and propagating unknown is how the nature of variability itself will change under future conditions. Although model projections for mean shifts are well-characterized, less attention has focused on variances around the means. In research tackling this latter information gap, it will be crucial to recognize the varied roles of biological processes (e.g. photosynthesis and respiration, but also species interactions) in influencing spatio-temporal variation. Because such biological effects can have positive or negative feedbacks, at times even the direction of impact may be affected, perhaps non-intuitively, and the magnitude of response may depend on location or season.

Issues of species interactions provide a quite general challenge for linking theory and experiments. Many population models focus on single species, and ecosystem models often rely on proper representations of energy flow through multiple trophic levels. However, OA effects on interactions between even single species pairs have been examined in only a few cases. Moreover, the spectrum of possible interactions is large, encompassing predation, competition, grazing, facilitation, as well as parasitic and/or pathogenic relationships. Thus, whether interactions are direct or indirect, appreciable additional effort is required to examine the influence of OA on them. It is also the case that because many species interactions are mediated through behaviour, an expanded focus on OA-induced shifts in behaviour may be warranted, especially since some of the most surprising effects of OA have been realized as behavioural changes. In models that incorporate species interactions, attention to the strengths of connections among taxa (as identified by experiments) could help to simplify what otherwise could become unwieldy analyses of links within given systems.

Considerable current research has been directed at understanding coupled effects of two or more coincident stressors, usually elevated temperature and elevated seawater pCO2. As such studies emerge it will become more feasible to assess the relative importance of seawater acidification compared with other perturbations tied to global environmental change. Such data will also enable models to begin to more accurately represent connections, or lack thereof, among multiple environmental factors that are changing in concert. The capacity for synergistic or antagonistic outcomes, compared to simple additive ones, makes proper accounting of such interactions essential.

Some work has begun to explore the capacity for acclimatization and/or adaptation to OA.This line of inquiry connects implicitly to explorations of patterns of variability in seawater parameters, in that the latter define the environmental context for physiological responses of organisms and their evolution. Mechanistically based physiological experiments may bolster understanding of general themes and strengths of response, which in turn may aid modelling studies that rely on a range of assumptions about the sensitivity of various taxa to OA. Furthermore, such information concerning acclimatization and adaptation will be critical for models attempting to project very far into the future, since these processes have the potential to induce powerful feedbacks on population parameters and the strength and/or character of natural selection.

Although not a main topic of discussion at this ecological workshop there also remain uncertainties and data gaps in our ability to make adequate predictive models of biogeochemical fluxes under OA. Similarly, there remain significant uncertainties in how patterns of primary productivity may change, and in turn how this will cascade through altered pelagic-benthic coupling to impacts upon seabed communities. There are also other, quite general challenges in linking models to data. Communication between experimentalists and theoreticians is essential, and must take place in both directions. Empirical data are necessary for parameterization of models, and model predictions guide the design of effective experiments. Discussions as to common conceptual “currency” will be vital, as will considerations of the kinds of information that will be maximally useful to the broadest array of researcher. Issues of uncertainty and error propagation are also critical concerns, as is the need for sensitivity analyses, especially in ecosystem models. Ideally, empirical OA researchers would collect data that could be used to test models, enabling their refinement and improving the capacity of the models to predict outcomes accurately. Likewise, modellers would generate output that could spark new insights and conceptual advances amenable to experimental testing. In these dual efforts, it should be emphasized that all models and experimental systems are abstractions of the full complexity of nature. Thus, there is no single, best model, nor is there any single, maximally valuable type of data. In this situation, it may be of special value to pay attention to recurrent themes that arise in a broad array of models of different types, as outcomes tied to these predictions may be among the strongest and most robust.