Community ecology

1
Community ecology

• Aaron M. Ellison

2
Ecology

• Ecology is a broad discipline comprised of many
  sub-disciplines. A common, broad classification,
  moving from lowest to highest complexity, where
  complexity is defined as the number of entities
  and processes in the system under study, is
• Physiological Ecology (or ecophysiology) and
  Behavioral ecology examine adaptations of the
  individual to its environment.
• Population ecology (or autecology) studies the
  dynamics of populations of a single species.
• Community ecology (or synecology) focuses on the
  interactions between species within an ecological
  community.
• Ecosystem ecology studies the flows of energy and
  matter through the biotic and abiotic components
  of ecosystems.
• Landscape ecology examines processes and
  relationship across multiple ecosystems or very
  large geographic areas.
• http//en.wikipedia.org/wiki/Ecologist

3
Statistics

• Inferential statistics is used to model patterns
  in the data, accounting for randomness and
  drawing inferences about the larger population.
  These inferences may take the form of answers to
  yes/no questions (hypothesis testing), estimates
  of numerical characteristics (estimation),
  prediction of future observations, descriptions
  of association (correlation), or modeling of
  relationships (regression). Other modeling
  techniques include ANOVA, time series, and data
  mining.
• http//en.wikipedia.org/wiki/Statistician

4
What do community ecologists do?

• Identify patterns
• Describe of patterns of distribution and
  abundance of gt1 species
• Determine processes
• Identify and quantify interactions among species
• Relating patterns to processes
• Are observed patterns consistent with
  hypothesized mechanistic processes (what is
  P(data model))?
• Attempt to synthesize and generalize across
  systems
• And

5
Care for and curate ANOVA tables
Data compiled by Jessica Butler from the 2005
issues of Ecology (294 articles), Ecological
Monographs (26 articles) , and Ecological
Applications (161 articles)
6
Four questions to frame the discussion

• Are data from community ecology different from
  all other data? Hierarchical modeling has been
  applied routinely to population-level,
  observational data, but rarely (ever?) applied to
  community-level, experimental data. What are
  appropriate (hierarchical) models for addressing
  datasets in community and ecosystem ecology?
• Are community ecologists destined to live
  embittered lives in a theoretical backwater? The
  agony of community ecology is self-inflicted and
  reflects unending political and emotional
  conflicts rather than debates using rational
  criteria (Evan Weiher Paul Keddy, 1999)
• Why should we dip twice? Can hierarchical models
  add value to the analysis of experimentally-derive
  d data of community ecologists, or is ANOVA
  enough?
• Should we take a rest from teaching classical
  statistics to community ecologists?

7
Alternative community states, regime shifts, and
state-and-transition models
8
An open, and interesting, question

• Are ecological communities characterized by
  alternative (stable) states?
• The hallmark of alternative stable states is the
  presence of multiple basins of attraction across
  a range of parameter values. A system with
  alternative stable states must contain at least
  two separate basins of attraction for at least
  one point in the parameter space. (After P. S.
  Petraitis S. R. Dudgeon 2004 J. exp. Mar.
  Biol. Ecol. 300 343-371)

9
Some related (and statistical) questions

• A community ecologist asks
• Are two observed communities consistent with a
  model of alternative (stable) states?
• An ecosystem ecologist asks
• Can we detect a regime shift in multiple noisy
  time-series datasets?
• A manager/restoration ecologist asks
• What (statistical) criteria indicate the success
  of a management effort when using a
  state-and-transition model of community or
  ecosystem development?
• And how do we educate the one(s) who do not even
  know how to frame the question?

10
Some important differences

• Alternative stable states (the community
  perspective)
• Parameters are constant
• State variables change
• Different community states result from
• Differential recruitment or initial conditions
  (history matters)
• An acute perturbation sufficiently large

• Regime shifts (the ecosystem perspective)
• State variables are constant
• Parameters change through time
• Different regimes are caused by
• New interactions among state variables
• A chronic (directional)

perturbation to the system changes the landscape
of the attractor(s)
to move the system to a new (local) basin of
attraction
Figure from Beisner et al. (2003) Frontiers Ecol.
Env. 1 376-382
11
A clash of (statistical) cultures

• Community ecologists come from a design-based
  tradition
• Experimental tests of the existence of
  alternative stable states must fulfill three
  conditions (definitions are consistent among
  studies)
• States must be shown to occur in the same
  environment
• Experimental manipulation must be a pulse
  (one-time) perturbation
• Experiments and observations must be carried out
  over a sufficiently long time and over a large
  enough area to ensure that the alternative states
  are self-sustaining
• Standard analytical framework is AN(C)OVA,
  perhaps in a BACI framework
• P-values are common are data consistent with the
  model?

12
A clash of statistical cultures

• Ecosystem ecologists come from a model-based
  tradition
• Identification of a regime shift (definitions
  vary among studies)
• Thresholds, not gradual changes, characterize the
  response of ecosystems to chronic environmental
  changes (a community ecologists press
  perturbation)
• These thresholds are identified by either sudden
  and dramatic or slow and gradual changes in
  parameters. But, the key characteristic is that
  the time-scale for the change between regimes is
  much shorter than the time within alternate
  regimes.
• Available ecological time series are often too
  short for robust analysis (in contrast with
  longer oceanographic records)
• Statistical framework is time-series analysis
  (occasionally dynamic linear models, PCA, or
  Fisher Information see Manuta 2004)
• Statistical detection of shifts in time series is
  not reliable evidence for underlying non-linear
  processes leading to multiple stable states.
• P-values are rare (but see Solow Beet 2005)
  What is the likelihood of a regime shift, given
  the data?

13
Three datasets of interest

• Alternative community states species composition
  in the Gulf of Maine (experimental)
• Regime shifts time-series of phosphorus change
  in a eutrophied lake (simulation)
• States-and-transitions Vegetation on
  rehabilitated mine sites (observational and
  experimental)

14
Alternative community states in the Gulf of Maine
15
Alternative community states in the Gulf of Maine

• Petraitis and Dudgeon (2005 J. exp. mar. Biol.
  Ecol. 326 14-26) examined changes in species
  composition following experimental clearings
  (1-8m diameter) at 12 sites in 4 bays on Swans
  Island, Maine.
• Data annual (some years twice) censuses
  1996-2002 of numbers of mussels, snails,
  barnacles, and fucoid algae percent cover of
  mussels, barnacles, fucoid algae.
• Analyses presented in paper
• Univariate mixed-model nested ANOVAs on each
  taxon Y f(date, size, bay, site(bay), and all
  interactions)
• Multivariate Procrustes analysis (MDS) to examine
  successional trajectories of different size
  clearings in multivariate space.

16
Time-series of phosphorus change in a eutrophied
lake

• Carpenter and Brock (2006, Ecol. Lett. 9
  311-318) simulated phosphorus dynamics of lakes
  subject to eutrophication.
• Data time series (t 300 years) of P input
  and P output in water, soil sediment.
• Analysis presented in paper
• Dynamic linear model examined changes in
  within-year variance (SD) associated with DLM
  predictions. Increased changes in SD observed
  prior to regime shift.

Photo by Brett Johnson, NTL-LTER
17
Vegetation on rehabilitated mine sites
18
Vegetation on rehabilitated mine sites

• Norman et al. (2006, Rest. Ecol. in press)
  describe rehabilitation of abandoned bauxite
  mines in Western Australia. They examined
  vegetation succession at 9 replicate sites (1 ha
  each) within 4 mines. Each site had 3 seed
  treatments 2 fertilizer treatments.
• Data vegetation composition (density, percent
  cover) 1, 2, 5, 9, and 14 years after
  establishment of treatments. of vegetation in 20
  22-m quadrats along transects within each mine
  site treatment fertilizer combination. 5
  sites have adjacent controls for reference
  vegetation. (Data used with permission of
  Alcoa-Australia)

Photo credit http//www.osmre.gov/

• Analysis presented in paper
• univariate ANOVAs
• similarity indices
• ordinations (deleted from final version of ms.
  because of inconclusive results)

19
Open questions

• Are there common, underlying models that can
  describe the community ecologists alternative
  stable state and the ecosystem ecologists
  regime shift?
• Are there common analytical tools that can use
  community ecologys datasets on alternative
  stable states (designed, ANOVA-type data) and
  ecosystem ecologys datasets on regime shifts
  (observational, time-series data) in such a way
  that both types of data can provide tests for a
  single theory?
• Can parameters for (bi-stable) models of regime
  shifts or alternative stable states be estimated
  from short time-series?
• Can these analyses be used to provide forecasts
  or benchmarks for restoration ecologists who are
  managing for alternative community states?