Invisible colleges in ecology; levels of organization

Lecture graphics

Ecology involves the study of complex systems spanning many levels of organization

e.g., conventional levels

biosphere

ecosystems

communities

populations

organisms

organs

cells

compounds--organic and inorganic

elements

atomic particles

fundamental particles

Higher levels of organization are characterized by:

increasing complexity of arrangement

decreasing scientific precision (e.g., boundary recognition)

Selection of labels for levels (a human activity) = f(object of study)

e.g., could use guilds or trophic levels instead of communities, depending on phenomena being observed

Furthermore, these labels aren't necessarily size-ordered

So this "tower" model, intended to offer generality, actually accounts for relatively little flow of materials through ecological systems

There are some indicators that ecology is a primarily a "social" activity--if McIntosh's analysis of invisible colleges is correct, ecological theory is largely a function of sociology:

1. Investigator exerts considerable influence over ecological discoveries:

   Selects level of study (e.g., organisms, populations, guilds, communities)

   Selects phenomena of interest (e.g., growth vs. migration)

   Hopefully, these decisions are made w/ an aim toward hypothesis-testing, in the context of Popper and Kuhn

   Hypothesis-testing is a search for mechanisms (why something happens; a hypothesis is a candidate explanation)

2. What is observed is dependent on level of organization of observer [Figs. 1-3, Allen et al. 1984, p. 2]

   
   
   

3. Investigator selects grain and extent of studies, which determine the limits on size and time of observable phenomena

Fine grain required for detecting high-frequency events

Large or long-term studies needed for large or long-term processes

Use of "standard" levels of organization would not be problematic for ecology as a science, except for the difficulty in linking levels

Common phenomena must be used to link levels (e.g., growth; cycling of a particular nutrient)

Terminology of phenomena must also be consistent

--> difficulties (see many articles and resultant confusion)

Take-home messages:

1. Use of traditional terminology ("tower" model) will continue, and is OK if:

   We recognize that these are arbitrary, subjective categories that are not scale-defined

   In fact, traditional terminology is better than development of new terms w/o new concepts (as mentioned by McIntosh)

2. Ecology is in its infancy, and resembles "social" science in many respects

Tools for ecological investigations

Graphics

Diamond (1983, Nature 304:586-587) gave an overview of approaches for studying ecological concepts

classified research into 2 types (while recognizing that these represent regions along a continuum of possibilities):

comparative and experimental studies [right-hand-side of Fig. 4.1 Keddy 1989, p. 83]

in comparative studies ("natural experiments") the researcher exploits naturally occurring perturbations

Diamond divided comparative studies into 2 types:

"natural trajectory experiments" are comparisons of the same community before, during, and after a perturbation

"natural snapshot experiments" are comparisons of communities assumed to be different principally w/ respect to one IV

Diamond also divides experimental studies into 2 types, depending on the location (and, by inference, the level of control exerted by the researcher): laboratory and field experiments

Keddy (1989) adds an additional category: descriptive studies

data sets are collected from a community and are then statistically manipulated to look for patterns which the researcher believes can be attributed to competition

this approach was popularized by animal ecologists in the 1980s

Diamond evaluates the strengths and weaknesses of 4 types of studies relative to 7 criteria:

• Control of independent variables
• Matching of sites
• Maximum temporal scale
• Maximum spatial scale
• Range of possible manipulations
• Realism
• Generality

DESCRIPTIVE STUDIES

Keddy (1989): "generations of plant ecologists have been occupied w/ tallying the contents of quadrats in the summer, and then trying to draw inferences about these observations in the winter."

Many statistical techniques have been developed just to look for patterns in these data sets

The biggest problem w/ this approach can be illustrated w/ a simple example of association analysis using the 2 x 2 contingency table

Data are collected from sample units (usu. quadrats) and the association between any pair of species is calculated using the chi-square test

The null hypothesis is that the species are independently distributed; the alternative hypothesis is that the 2 spp. are either positively or negatively associated

Negative associations are often interpreted as being evidence of competition; actually, at least 4 interpretations are possible:

1. Spp. are restricted to different microhabitats, and so do not interact

2. Spp. are positively associated but the sample unit was so small that only a few indivs. fit in it, thereby obscuring the pattern which occurred at a larger scale

3. Agents such as predators independently control each species and restrict each to a different set of conditions

4. The spp. compete, and competition leads to habitat segregation

It is not possible to distinguish between these causes w/ descriptive data alone

A variation on using association analysis is to choose natural environmental gradients and examine distributional limits of spp. along these gradients in order to infer the existence of competition

Assumption: systems that are structured by competition have different kinds of patterns than those not structured by competition

Problem: it is not clear what type of patterns competition would produce

Nonetheless, statistical tests have been developed to determine distributions. Three alternatives are recognized:

1. Spp. distributional limits are regularly spaced

2. Spp. distributional limits are randomly arranged

3. Spp. distributional limits are clustered along the gradient, producing apparent communities

Again, departures from random patterns do not tell us anything about competition (or any other process)

1. Spp. may have similar distributional limits because of similar physiological tolerance limits

2. Clusters of distributional limits may be attributed to the way the observer divided the gradient

3. Herbivores may stop at a certain point along the gradient, and therefore create discontinuities

4. One or more competitive dominants may set the distributional limits of an entire group of spp.

Only the last hypothesis is consistent w/ competition, and hypotheses 1-3 are very difficult to reject

More graphics

COMPARATIVE STUDIES

Follow directly from descriptive studies, in that observational data are used to describe patterns, and resulting patterns are compared to infer differences in process

Value in comparative studies lies in the spatial and temporal scales which they can consider

Caution:

Since there is not an experiment, researcher is always forced to compare patterns and then invoke mechanisms

Comparative studies are commonly used to infer presence or strength of interactions when experimental studies are difficult to conduct

As w/ descriptive studies, it is tempting to collect data w/o first asking a question

EXPERIMENTAL STUDIES

An experiment requires you to specify a question and a means of answering the question in advance, so experimental studies tend to be better-designed than descriptive or comparative studies

e.g., a common hypothesis proposed for many studies of competition is that competition is structuring a community in a certain manner

a manipulation is then performed to test this hypothesis

dependent and independent variables must be specified

dependent: characteristic that measure the performance of indivs. or pop'ns

independent: abundance of neighbors; a negative relationship is predicted between abundance of neighbors and performance measures

Laboratory experiments

tell what could potentially happen in nature under specific sets of conditions which can not be produced in the field

advantage: researcher can manipulate a wide range of biotic and abiotic variables

disadvantage: limited scope, extreme unrealism; however convincingly phenomena are demonstrated in lab, can not extrapolate from controlled environment to the actual existence of a process in nature

Field experiments

unlike lab studies, there is a reference point: current performance of indivs. or pop'ns of interest [Fig. 4.6 Keddy 1989 p. 94]

Descriptive and comparative studies do not test for an ecological process--they test for a pattern and infer a mechanism

Experiments do not necessarily overcome this problem--they may eliminate the largest number of alternative hypotheses, but manipulations of, for example, 'interference', really test for density dependence

'Apparent' competition (resulting from indirect effects) has been proposed to explain density dependence; examples:

the removed species may have been a host for a pathogen which also damaged the remaining species

the removed species may have attracted a herbivore which also fed upon the remaining species

Parker and Root (1981 Ecology 62:1390-1392) showed that a herbaceous plant species was excluded by some habitats by a grasshopper assoc. w/ a common shrub; removal of the shrub --> incr. in herb, but w/o interaction between the 2 spp.

these examples indicate:

1. mechanisms of ecological interactions may be very complex

2. natural history can not be divorced from experiments

The most important part of ecological research is the choice of the question

Because ecologists usually enjoy field work, it is tempting to rush into data collection w/o first taking the time to think

An appropriate strategy for conducting ecological research is presented by Keddy [Table 4.4 p. 97]

In the absence of generality, ecology is nothing more than natural history. How can we increase the generality of field experiments? Keddy (1989) proposes:

1. Demonstrate that the pattern is a general one

2. Use increased no. of spp.

   tradeoff may be whether to pose complex questions w/ a few spp. or simple ones w/ a lot of spp.

3. Provide a comparative context

   select spp. because they are representative (in some way) of a large group of spp. of interest

   groups can be defined after the experiment is conducted, too (e.g., gap-colonizers vs. shade-tolerant spp.)

   then, present results not based on species responses, but on responses of groups [Table 5.2 Keddy 1989 p. 108]

4. Use general experimental factors

5. Arrange experiments along gradients

   incorporates variation --> expanded generality

   variation is incorporated systematically, so you can tell how trt effects vary along the gradient

   e.g., Gurevitch (1986)

Previous lecture

Next lecture