Copyright 1998, Patricia S. Muir

The following section of notes gives general background about air quality impacts on ecosystems, and discusses the criteria required, in an ideal world, for establishing whether an air pollutant is causing an adverse response.

For information on air pollutant regulation in the US, and in-depth discussion of tropospheric ozone pollution, click on tropospheric ozone, here. This will take you to another set of documents within this BI 301 web site.

If you are interested in air quality for a particular region of the US, you can click on score card, which will take you to a site that provides relatively local-scale data on emissions of a variety of air pollutants . Within that site, toxics will take you to information on air toxics (as opposed to criteria pollutants) in the air near your area of interest.


Prior to 1970's (and passage of the Clean Air Act in 1970 and its amendments in 1977) people were usually most concerned about problems associated with large point sources of pollutants. Industries and electricity-generating facilities often had a deadly combination: short stacks and huge emissions of pollutants. Some of you may be familiar with some of the infamous legacies associated with these levels of pollution. For example, if you have been in Sudbury, Ontario, Columbia Fall, Montana, or Kellog-Wallace Idaho, you have probably seen the barren zones nearest to the stacks that are (or were) associated with the large metal smelting facilities there..

In these cases, extreme local effects did occur. Emissions of air pollutants were only weakly regulated until as recently as the 1960's, if at all. In extreme cases like these, the ground was often totally bare for some distance from the stacks. There would typically be a gradient away from the source, with increasing diversity and complexity of the ecosystems with increasing distance, and finally at some distance, the system would be indistinguishable from the undisturbed, equivalent type of system (within the limits of natural variability).

While it was difficult even then to establish causation in courts of law (where these cases usually ended up) it was relatively easy to study cause and effect when scientists were dealing with clear point source problems such as these. In comparison, it is far more difficult to assess cause and effect relationships for most current problems, for the following reasons:


I'll begin by describing the conventional methods for assessing causation in clear cut, point source situations. Doing so will give you some sense about what scientific context people were starting from when they began to approach more regional problems. I believe that these insights will help you to understand why it has been so difficult to deal with regional -- much less global problems.

What kinds of evidence would one like to be able to muster to establish whether or not pollutants are responsible for observed ecosystem effects? That is, what are criteria of proof?

1. Pollutant concentration:

It should be demonstrated that the pollutant or pollutants are present at concentrations that are high enough to produce the observed effects. This was typically determined by: (1) monitoring of air quality using instruments and (2) knowledge about effects likely to be caused by exposure to given concentrations, based on experiments in controlled environments. (See NCLAN for an example of these experiments involving tropospheric ozone pollution.)

2. Spatial consistency

The pattern of effects (decreased growth, losses of certain species, etc.) should correspond with measured pollutant concentrations, being more intense where concentrations are higher, and diminishing as concentrations decrease. Often, injury will be worst closest to the source, but in some cases pollutant concentrations are highest at some distance from the source. Concentration with distance from source is a function of stack height, topography, prevailing wind direction, and so forth.

The spatial pattern of injury is often determined by sampling air quality and vegetation (or whatever response parameter is of interest) in plots located along transects (radii) out from the source of pollution. In selecting sampling sites, it is important to control for site similarity. It is important to control for similarity in features such as elevation, soil type, topographic position, and so forth, to minimize chances that differences between plots are really due to a factor other than pollution that differs between the sites. For the same reason, it is also important to attempt to collect all samples at the same season of the year, again to avoid confounding (confusing) seasonal effects with those attributable to the pollutant.

Given these cautions, then, one would establish sampling plots, and record data on the ecosystem attribute of interest. Air quality studies often rely on plants as the potential indicators. Why use plants as indicators?

(1) They don't move around so are relatively easy to sample.

(2) Many plant species are sensitive directly to gaseous pollutants, while effects on animals are often (but not always!) more indirect. Animal communities often change or decline as a result of pollutant exposure because their food supply changes or diminishes, rather than as a result of direct effects of the pollutants. That is, animals are often secondarily or indirectly affected by pollutants.

What might be observed along a gradient of distance and pollutant concentration?

· Mortality of some species in plots where pollutant concentration is highest. Different species have different sensitivities, of course. This is known from controlled fumigations in the laboratory as well as from field observations. For example, many lichens are notoriously sensitive to air pollutants. Surveys for air pollutant injury often rely particularly on assessments of those plants known to be highly sensitive to the pollutant or pollutants in question. These species are called bioindicators.

· Visible and characteristic symptoms on known bioindicators in areas where pollutant concentrations are relatively high. Some pollutants do produce such symptoms, and books are published showing characteristic injury symptoms on characteristic species for various air pollutants. Caution must be used here, however, in that similar symptoms can be produced by other agents, in some cases. In addition, visible symptoms are not always correlated with real damage to plant function; some plants can look heavily symptomatic, but be growing just fine, while the converse can also occur; plants that look fine but are struggling physiologically.

· Changes in plant communities. In many cases the pollutant concentration isn't sufficient to cause a great deal of actual mortality, but it may tend to weaken some species, affording a competitive advantage to species that are more tolerant of the pollutant. In such cases, the composition of the community might shift along the gradient

· Reduced cover by vegetation where concentrations are high, cover increasing as air quality improves with distance. ("Cover" is simply the percentage of the ground surface that is covered by vegetation. It can be sampled in terms of cover by individual species, by life form (shrubs versus herbs, for example), or in total.

· Reduced productivity in sites with higher pollutant concentrations compared to those with cleaner air. This can be measured by biomass harvesting or by using measurements of photosynthesis or of growth rates.

These data are then analyzed in relation to pollutant concentration or distance/direction relative to the source area, seeking to determine whether a pattern in injury corresponds to the pattern in pollutant concentration.

(3) Temporal consistency

This criterion involves assessing the time of onset of injury compared to the time of onset of air pollution in the area. That is, did injury begin when the pollution began or shortly thereafter, or are they unrelated in time?

This criterion is one of the more difficult to achieve, because the ideal case in which we have "before" and "after" data on the response parameter(s) is rarely met. Not having anticipated a pollution problem coming to an area, there was no reason for anyone to sample before the problem began!

We do have some ways of trying to go back in time to make this assessment, but their utility is limited. One approach involves using pollen preserved in layered lake or swamp sediments. This science is called "palynology." Given that the pollen reflects to some extent the composition of the vegetation in the area, changes in pollen may reflect what the plant communities were like before the onset of pollution, tracing though the present. I think you can imagine how rarely pollutant problems and good places for pollen preservation coincide, however.

Another long term record that can be helpful in getting "pretreatment" information is tree rings (the science of dendrochronology). Because the width of tree rings indicates the growing conditions in a given year, changes in ring width can indicate change potentially attributable to air pollutants. The science can be useful, but tree rings are difficult to interpret. For example, growth is sensitive to climate as well as to air quality, and sometimes climate actually influences air quality (see tropospheric ozone for an example). Thus, it is very complicated to pull out extraneous "signals" attributable to agents other than air pollution, leaving the air pollutant signal intact.

Establishing temporal consistency also requires understanding about what lags may be involved between the onset of pollutant exposure and demonstration of effects. In some cases, pollutant effects don't manifest themselves immediately. For example, most effects of acidic deposition on terrestrial ecosystems, where they occur, are probably mediated by effects on soils, which change only slowly. In some cases, decades of exposure may be required for effects to show up in soils and then in plants.

Thus, in practice, it is often very difficult to establish temporal consistency.

(4) Match between field and laboratory observations.

Symptoms and sensitivities observed in the field should, ideally, match those demonstrated to occur in controlled environments (e.g., laboratory experiments) and researchers should be able to be reproduce symptoms that were observed in the field using realistic doses in controlled environments.

Meeting this criterion can be problematic, but is often insisted on. Why might it be problematic?

Basically because the environment in the lab is not same as the environment in the field! Factors in the field can affect sensitivity and symptoms, such that plants respond differently in the lab than in the field. Other biotic and abiotic stresses in the field can interact with pollutant exposure to modify symptoms or sensitivities from those that would be observed in the controlled environment. For example, if plants are stressed by lack of water in the field they may respond differently to pollutant than they would under well-watered conditions in the lab. Thus, both the symptoms and the dose at which they occur may be different in the lab than in the field.

(5) The pollutant or its reaction product(s) may be measurable in tissues and in soils

Some pollutants leave residues, or are incorporated into tissues such that concentrations in or on living materials or soils can be measured. Concentrations can then be compared to pollutant concentrations as measured in the air. This, combined with knowledge about levels required for toxicity can provide useful information.

Not all pollutants accumulate, however, so this criterion cannot be achieved for all pollutants. Pollutants that can accumulate include some sulfur compounds, fluoride, and various metals.

(6) A known mechanism whereby the pollutant could produce the observed effect should be known.

For example, if growth is slowed in areas of high pollutant concentration, one should be able to demonstrate a physiological mechanism through which the specific pollutant of interest could affect growth (e.g., disruption of photosynthesis). Alternatively, the mechanism may involve ecological interactions, as is the case of death of ponderosa and Jeffrey pines in southern California.

In summary, it is more difficult to establish "proof" of causation than one might think even in relatively simple cases, associated with a clear point source. You can probably imagine the difficulties in studies involving pollutants that are elevated in concentration over regional (or larger) scales, such as tropospheric ozone. As an example, in regional studies, the gradients in concentration are often so long that it is impossible to sample sites along the gradient without other environmental factors changing along the same gradient as well. That is, it can be nearly impossible to find control sites that are sufficiently similar to the affected area that reasonable comparisons can be made.


In practice, a combination of observational (descriptive) and experimental (manipulative) approaches are needed to establish a particular source as causal for a problem. This applies both to point source/local problems and those operating over larger geographic scales. The two kinds of approaches are complementary and provide different kinds of information about the system, as listed below. Each has strengths and weaknesses, and either alone is generally insufficient, as described below.

Observational approach

The scientist observes the system as it exists in field. This means that the investigator doesn't manipulate the system. Classically, the observer contrasts various measures of ecosystem health (as described above) between the polluted area and a control area, or along a pollution gradient. As mentioned previously, the control area or areas along the gradient are chosen to be as similar to the polluted area as possible in all ways other than pollutant concentration. That is, sites should be on similar aspects, elevations, soils, topographic positions, and with similar vegetation. Thus, by careful choice of sample locations, the scientist has controlled as much as possible for unaccounted for sources of differences between the two areas. However, such differences are likely to be greater in an observational study than they would be in the laboratory or other controlled environment.

Response measures assessed are then contrasted between areas and the scientist may suggest that any differences observed are likely to be related to differences in pollution exposure.


The investigator hasn't simplified or altered the system by manipulating it.

The investigator is trying to assess interactions at a realistic level of complexity. That is, many variables are retained in the system.

Weaknesses or limitations:

These basically lie in lack of control over the system. Even when the investigator is as careful as possible, a gradient in injury away from the pollutant source could be caused by some other factor than happens to co-vary along the same gradient. For example, a decline in tree growth that began at the same time as pollution began could have been caused by a climatic change that happened at the same time.

Observational studies basically allow one to demonstrate correlations. (Correlations are relationships between two variables in which one factor changes in parallel with another such that there is correspondence between two sets of data, describable as a line.)

For example, if tree growth rates were increasing along a Y axis, while distance from the pollutant source was increasing along the X axis, one might be tempted to infer that the pollutant is responsible for the lowered tree growth rates nearer the pollutant source. Can one safely assume that this is so? NO! This apparent relationship between tree growth and pollutant concentration could be caused instead by another, co-varying factor; maybe the trees nearer to the source are older and hence growing more slowly, for example.

It is, unfortunately, common to confuse correlation with causation. I'm sure that you can see the danger in this; correlation does not equal causation!

The example involving tree growth given above is an example of "spurious correlation" -- suggesting a relationship that doesn't have the assumed cause -- a false causal inference.

While biological reasoning may suggest a mechanism for the correlation – in this case, that the pollutant slowed tree growth -- correlational data haven't established the connection rigorously! Data like this won't hold up in court or by scientific criteria as "proof" of causation.

However, establishing correlations like this, based on observational data, is important as a step towards determining cause and effect. Correlations have an important role in generating hypotheses for subsequent tests. If hypotheses generated in the field are then borne out by experiments in the lab, there is important confirmation that there is likely cause and effect -- with evidence from both the field and the lab.

An important caveat: failure to duplicate effects observed in the field under laboratory conditions may be misleading. Failure to duplicate these effects doesn't necessarily mean that inferences based on observations in the field were wrong. How might this be? Remember, important factors operating in the field to modify responses to pollutant may be lacking in the laboratory environment. For example, maybe water stress or other stresses in the field interacted with the pollutant, exacerbating plant responses. By contrast, when the plants are maintained under well-watered conditions in lab, they may be less (or in some cases, more!) sensitive.

Experimental approach

The role of the scientist in the experimental approach is to manipulate one or two variables, and observe the effect on some component of the ecosystem (e.g., a tree or an herb) under controlled environmental conditions. For example, the investigator would provide different levels of pollutant and observe plant responses to those various levels, again, working in the laboratory, greenhouse, or other controlled environmental situation in which all other variables can be held "constant."


This approach allows one to isolate the effects of one or two variables, while holding others constant -- that is, it allows more control over the system. Thus, effects attributable to the pollutant can be observed more clearly in this system with its reduced complexity. The investigator has control over unknown sources of variation to a much greater degree than does the field investigator

Weaknesses or limitations

There is question about the degree to which the experimental situation mimics the real world. For example, some of those "extraneous" sources of variation present in the field may be important modifiers of pollutant response, yet would be lacking in the controlled environment.

It is also difficult (nearly impossible) to do experiments at the ecosystem level. One really can't transport entire ecosytems into the lab to manipulate (although microcosms,as at the local EPA Lab on 35th Street, come close to allowing this). The difficulty of doing ecosystem-level work is a concern because many effects observed in the field probably arise through complex interactions involving several ecosystem components. For example, pollutant exposure may result in altered relationships with pests and pathogens, or beneficial symbionts (such as mycorrhizal fungi). These altered relationships may then mediate pollutant effects in the field, yet would be missed in the laboratory.

Finally, experiments are usually short-term, largely because funding agencies usually provide only short-term funding. As mentioned above, some pollutant effects may take years to become apparent (e.g., when effects are cumulative), and these would be missed in short-term lab studies.

Thus, experiments are often (usually) constrained to short time windows and simple systems.

Thus, in practice, it is usually necessary to use a combination of the two types of approaches.

To move to the next section in these notes, which deal with tropospheric ozone pollution, click "Ozone.". To return to the master Table of Contents for this BI 301 home page, click "Contents," and for reminders on how to navigate within and among these documents, click "Navigate."

This page is maintained by Patricia S. Muir at Oregon State University. Page last updated Otober 14, 2009.