Discussion Session 3: Can we use long-term observations to detect and predict ecosystem change?
Chair: Tim Wootton. Rapporteur: Helen Findlay.
A key challenge in understanding and responding to ocean acidification is predicting how ocean ecosystems will respond to increasing dissolved CO2. Development and analysis of long-term observations will play an important role in meeting this challenge, as they are useful in three fundamental ways.
First, time series data are essential simply to document patterns of change. This is the traditional rationale of long-term observations (i.e. monitoring). Although often viewed as a rather mundane enterprise, and therefore poorly funded, time series are the only direct data sets available, and are therefore the most effective evidence to convince policy makers and the general public that real environmental change is occurring, and they often hint at its implications. Furthermore, the patterns uncovered in monitoring programs often reveal ecological surprises that lead to new lines of investigation and insight into the mechanisms underlying environmental change and its subsequent impacts. Long-term data on ocean acidification (OA) itself are remarkably scarce. Recent funding efforts have led to an expanded network of OA monitoring, which may eventually lead to expanded long-term monitoring activity depending on funding commitment.
Second, a comprehensive program of long-term observations (incorporating both physical/chemical and biological data collection) provides clues about potential ecosystem linkages that may be essential to consider when anticipating effects of global change. Simple visual analyses of contemporary variable trajectories through time by themselves can suggest key linkages to investigate, and they also offer a platform to look for more sophisticated statistical association among variables that offer insight into key interactions using methods such as cross-correlation, structural equation modelling and neural networks. Having a coherent series of data for multiple key ecosystem elements is essential in this enterprise to allow comparisons among linked variables. At present, coordinated data among biological and physical/chemical data series is lacking.
Third, time series data could play a much more important role in understanding and predicting ecosystem response to effects of increasing CO2 and other environmental changes because of their natural linkages to ecological models. It is impractical to manipulate all relevant stressors and ecosystem components in a natural context. Mathematical modelling can help to understand the role of key ecosystem components, to sort out the roles of simultaneously acting processes, and to provide a predictive framework, but requires concrete linkages with empirical data. Most ecological models focus on the dynamics of ecosystems, and use these to scale up to longer-term response. Time series data are the direct empirical description of ecological dynamics (i.e. estimating rates, as well as states), yet ecologists and oceanographers typically work with snapshots of data in time. Long-term observations can potentially play a powerful role in estimating parameters in models to constrain model behavior and make more precise predictions about complex ecological systems. For example, long-term data on the dynamics multiple species and ocean chemistry have been combined using Markov chain models to gain insight into how rocky shore communities respond to ocean acidification. Despite this potential, few data sets are available at present to carry out this activity. Hence, a more concerted effort to develop coordinated datasets documenting long-term dynamics that can readily be applied to modelling frameworks is needed. The general need in ecology for well-parameterized models of multi-species communities and ecosystems provide opportunities to transform the perception of long-term observations from a mundane activity to a critical part of cutting edge ecological science.
Several impediments exist to fully realize the potential of long-term observations utilizing extant data. First, readily accessible information about what long-term data sets exist is broadly lacking. While data sets are traditionally viewed primarily as numerical tables, digital photographs and videos represent a second stream of valuable long-term data that may be available, especially in conjunction with further tool development to efficiently extract data from them. Second, many long-term observations focus on one to a few species or environmental parameters of interest, which constrains their use as a basis for community- and ecosystem-level models. In particular, there is a critical need for concurrent observations of environmental and biological variables at the same site to most effectively describe linked dynamics. Third, most monitoring efforts are generally not designed with application to model parameterization in mind. In particular, many monitoring studies employ randomly placed census sites at each survey time, whereas time series are most effectively applied to parameterize dynamic models when they re-survey the same sites through time, thereby avoiding confounding spatial variability with fluctuating variables. Fourth, there is disparity between sites collecting physical/chemical and biological data, perhaps because the former focuses on sites deemed to reflect as much as possible the conditions typical for the open ocean whereas the latter emphasizes areas with intense biological activity, which can be notoriously variable. Although sometimes frustrating, probing this variability may also yield great dividends as particular sites or periods of time may mimic the broader environmental changes we anticipate in the future. Finally, there is a general lack of funding commitment to undertake long-term observations--most current data sets derive from site-based investigators who manage to piggy-back observations onto other short-term research projects.
Addressing access to extant long-term data can probably be carried out without large investments by:
allocating modest funding to allow investigators to digitize and document extant data
creating well-indexed, web-accessible data repositories and providing stable data archiving resources
developing long-term data clearinghouse websites where investigators and data sites can post relevant metadata describing what datasets are available
implementing strategies to facilitate and encourage use of these resources through mechanisms, such as advertising within the scientific community and data access conditions linked to external funding.
Implementation of some of these activities is beginning. For example, EPOCA in Europe has established a clearing-house for data related to ocean carbon chemistry, and the National Oceanic and Atmospheric Administration (NOAA) now hosts a site for depositing a wide variety of oceanic data through the National Oceanographic Data Center (NODC) in the United States. The Long Term Research in Environmental Biology (LTREB) and Long Term Ecological Research (LTER) programs of the US National Science Foundation (NSF) provides some support for generating data series, and requires funding recipients to make core data from these activities publicly available. Unfortunately, the primary division at NSF that funds marine research, the Ocean Sciences Program, does not contribute funding to these important initiatives to generate longer-term data series, so there is minimal development of marine time series through these programs. In the UK, NERC is supporting Long Term Time Series (LTTS) through National Capability (NC), but like the LTREB program, such endeavors are only likely to be funded if the project has a focused question that happens to require long-term data to answer it. Also in the UK, CEFAS currently generates data series for marine systems, but when funding is cut, long-term time series are the first data to go, rather than being nurtured as valuable resources with added value by virtue of its long-term status.
With limited available funding to develop coordinated marine time series of biological and physical/chemical conditions, creative alternatives should also be considered. Marine field stations offer one such opportunity, because courses in marine ecology and related fields often carry out repeated surveys of the same sites. To be effectively used, however, some degree of continuity in oversight of the data and quality control will be needed. Furthermore, these facilities often lack key environmental monitoring programs to link to biological data sets. Hence, dispersed marine stations may be logical places to expand current monitoring of ocean conditions, including the DIC system, to best integrate these data with biological information, when appropriate quality control can be implemented. Another source of time series data might be citizen scientist programs. These have proven useful to understand bird populations (e.g. Christmas Bird Count, Great Backyard Bird Count, Hummingbird Home) and butterflies (e.g. US Monarch Watch). The REEF organization has been developing programs to improve volunteer observations by divers, including providing a platform for data collation. Partnering with diving centres to collect environmental data on boats and biological data taken from traditional routes used for guided tours could also leverage more underwater eyes to generate informative long-term data series. At present there are no systematic survey sites established, but some tropical and temperate sites are frequently revisited and might be useful resources. Shoreline surveys are also becoming established in some places (e.g. Coastwatch, Marine Conservation Society Beach Watch).Partnering with diving centres to collect environmental data on boats and biological data taken from traditional routes used for guided tours could also leverage more underwater eyes to generate informative long-term data series.
Because of the lack of systematic commitment to developing long-term data series, extant data series are often somewhat esoteric in the ecological conditions they represent. It would be illuminating if such data series were expanded in a coordinated manner to allow comparative dynamics of similar systems to explore common trends and how context may change responses to global change. Are the systematic changes with latitude or between western and eastern ocean basins, for example, that might anticipate some effects of climate change? Although it is challenging to find true replicate ecosystems, sufficient commonality exists in dominant species in coastal ecosystems throughout the world to develop comparative data. For example, ecosystems dominated by mussels and corals are present in many areas, and might be particularly useful to use to develop parallel time series because long-term data is already available at some sites. More geographically coordinated actions and standardization of methods is required to achieve this goal.
Aside from traditional monitoring of unmanipulated sites, developing time series data in the context of experimental manipulations, such as FOCE studies, expand the power of the both the time series and experimental approaches. Traditionally, field experiments are only sampled a few times during their duration, but developing longer-term systematic time series data can both aid the interpretation of how manipulated factors generate treatment differences, and provide novel ecological parameter space to more powerfully estimate parameters in ecosystem models.
One concern with all monitoring programs is that not all possible ecological variables can be measured that may ultimately be useful in applying the data. Strategies for minimizing and dealing with these gaps need to be developed. Opportunities for augmenting time series may be present in historical/paleontological sources, particularly when geochemical information is lacking. A major impediment is identifying records with sufficient resolution to collect time series with a temporal resolution to ecological data, but analysis of skeletons of marine species with identifiable annual growth patterns, such as mussels, coralline algae, and corals, have successfully accomplished this in some locales. These issues could also be overcome by trying to project forward strategically in sampling regimes to take advantage of emerging technologies. For example, developing appropriate protocols for preserving genetic material, along with the resources to allow long-term storage, will allow opportunities to leverage emerging genomics approaches as the technology matures and costs come down. Additionally, more systematic collection and archiving of water samples in coastal areas where biological monitoring is focused could be invaluable in future analyses of retrospective water chemistry and new processes and risks come to light.