Intrinsic and Extrinsic Variability in Community Dynamics Working Group

First Meeting 7-9 October 1996
Meeting Summary


On 7 and 8 October:

Introductory Comments

After a brief welcome to NCEAS by its director, Jim Reichman, and several of its staff, each member of the Community Dynamics Working Group (CDWG) provided a short, biographic introduction.

Long-term datasets.

These presentations were followed by summaries of the various data sets that would be available for analyses by the CDWG. These included:

Analytical Approaches.

Following these presentations, other group participants described some of the analytical techniques that they have used for ecological data including:

Tony Ives - Sloppy techniques for sloppy data.

Brian Dennis - Statistical approaches to time series data, and multivariate autoregressive models.

Alan Hastings - Non-linear dynamics in the evaluation of extrinsic and intrinsic variability.

Tim Wootton - Markovian community models and path analysis when evaluating experimental manipulations.

Peter McEvoy - Lessons from investigations of biological weed control.

Kathy Cottingham - Dynamic linear statistical models.

Brainstorming Session

A series of basic questions were posed by our group. These were generally related to our community dynamics theme and stimulated by our earlier discussions. These included:

1. How to evaluate the effects of multiple stressors of lake systems?

2. How to determine the intrinsic sources of population variability, particularly in cross-lake comparisons?

3. What questions are unanswerable? Is there an ecological uncertainty principle? What are predictors and what are data requirements? What can we get from large-scale experiments? What can we get from bag experiments?

4. What is the relationships between the number of species in an ecosystem and its resistance/resilience to stress?

5. What details are relevant in evaluating system responses to stress?

6. Do different lakes show similar patterns of interactions among species? Do different perturbations cause similar responses in the same lake?

7. How can one examine processes that operate on different time scales, particularly in approaching stationarity?

8. What are the limits to linear and non-linear models, particularly in terms of their predictive abilities and statistical inference?

9. What would be ideal experimental designs for future whole-lake manipulations?

10. What can you predict about other stressors from the results shown for one stressor?

11. Can we develop some specific, mechanistic models of interactions?

12. What are the effects of different space or measurement errors? What are some of the specific inferences that can be made with different kinds of error? Can existing experiments be considered as replicates?

13. Are there common responses to the same and different perturbations across sites? How does recovery following stress compare among sites? How can this be quantified?

14. To what extent do we need to worry about species? How can we compare different systems even if they don't have the same complement of species?

15. How long do experiments need to run in investigating community dynamics? What time scales are appropriate?

16. Can subsets of a community be analyzed separately?

17. How can we predict the effects of introduced novel species?

18. How can time-series data for one or a few lakes be compared with snapshot data from a group of lakes? (e.g., presence/absence through time vs. space, longitudinal vs. horizontal surveys, community succession)

19. Are there life-history correlates of population variability?

From this group of questions we discussed a smaller set of research themes that could be the focus of initial, to-do efforts by our working group. These included:

One practical, shorter-term question that can be addressed would be to compare results under AR(1) models across lots of lakes.

On 9 October:

Identification of Target Questions and Working Sub-groups

Can different years be considered as relatively different events? To what extent can it be determined that different starting conditions among years generate subsequently different conditions within the years?
Sub-group = Hastings, Fee, Frost.
Initial work should target Lake 239 at ELA . Consider zooplankton community structure elements as response variables. Drivers to consider include spring conditions (mixing or not) and initial community structure at ice out.
Will test for stochastic vs. nonlinear phenomena with nonlinear responses being fairly steady through time and stochastic ones being relatively independent through time.

Can nonlinear models be developed for the relatively simple communities in Mono Lake? a) Can these then be extended to other datasets? b) Is there a divergence of small vs. large systems (e.g., state variable in mesocosms - could be dependent upon the type of perturbation being evaluated).
Sub-group = Melack, Hastings, Dennis, Ives, Cottingham, Fee, Wootton, plus other collaborators who are not working group participants.
Consider nutrients, phytoplankton and zooplankton.
Consider measurements of all dependent variables and test for Kalman filter applications.
Mono Lake data includes measures of spatial, horizontal heterogeneity may include a consideration of a physical model.

Examination of patterns of species vs. various types of groupings of species based on trophic interactions, functional behavior, taxonomy, etc. across lakes and research areas.
Sub-group = Wootton, Yan, Cottingham, Ives, Fee, Carpenter, Frost, Klug, Fischer.
Which lakes to be evaluated? Consider reference lakes first then treatment lakes to see what experiments can tell you. Longitudinal comparisons. Start with ELA 239, follow with LRL, Paul Lake (TCL), Red Chalk and Harp Lakes (DRA).
Examine patterns for zooplankton, phytoplankton and fish data.
Eventually may consider external drivers for these variables.

Are there substantial state shifts in communities that can be detected due to invasions or strong year classes of native species
Sub-group = Yan, Cottingham.
Study systems: Harp Lake (DRA) where Bythotrephes, a predatory cladoceran, has invaded the zooplankton community and Paul Lake (CTL)where the lake has been strongly influenced by a strong fish year class.

To what extent can the different lake datasets be examined with AR(1) models?
Sub-group = Ives, Dennis.
Study Systems: ELA 239 and 223, Paul Lake, LRL.
Work will initially focus on zooplankton data. An AR(1) will be fit and then residuals will be tested for autocorrelation. Biota are expected to equilibrate rapidly to manipulations but the equilibrium should move within the year with seasonality.
Brian Dennis is expecting to spend 4 days in our January 97 meeting developing MatLab code for these non-standard AR(1) models.

How resilient are the various reference lakes (can be considered as fitting a pseudo-AR(1) model)? How much variance can you explain in a dataset using a simple linear model across lakes? What is resilience (see 5)? What does the analysis tell you? What are other techniques?

To what extent can differences in the responses of different lakes to stress be explained by the diversity of the lakes and other drivers?
Sub-group = Klug, Micheli, Cottingham, Fischer.
This will be pursued from a whole-lake perspective.
And from the perspective of different scales of experiments.
And from a comparison with other types of systems such as oceans and estuaries.

What do mesocosms tell us about whole-lake processes?
Sub-group = Klug, Micheli, Cottingham, Fischer, McEvoy, Carpenter, Hastings, Wootton, Melack, Frost, Fee and Ives.
This will involve a literature search and analysis to compare mesocosm and whole-lake experiments.

Discussions of these questions were followed by practical considerations of how to make data available throughout our group. This included a short meeting with some of the NCEAS data management staff. We identified a data steward for each of the data sites listed above:

LRL and NTL - Frost

ML - Melack

DRA - Yan

ELA and NOLSS - Fee

TCL - Cottingham.

In addition, Tom Frost will serve as a liaison with the NCEAS staff on facilitating the distribution of these data.

We also had a brief planning discussion for our January 1997 meeting. Tony Ives and Tom Frost plan to be in Santa Barbara for two weeks starting 4 January. We hope that all of the group will be present for some extended period around then but not everyone should feel obligated to spend the whole two weeks. We anticipate that folks will be moving through for different portions of the period. We will correspond about the details of the schedule as we get a little more firm with our plans.

Finally, please keep in mind that none of the working sub-groups listed above is at all exclusive and that not everyone has been identified with a group. Please be sure to let us know of any group that you would like to participate with. Likewise, we certainly haven't exhausted all of the questions that we could address and we would be very interested in supporting efforts on questions that haven't been listed here.