One of the primary values of long-term ecological data sets is the crucial background information they can provide for experimental research, especially whole ecosystem manipulation. The North Temperate Lakes (NTL)-LTER project of the University of Wisconsin- Mad ison’s Center for Limnology not only supplies background data for the Rock Lake Acidification Project, but also provides reference data for the development of new statistical techniques for analyzing the often complex results of ecosystem manipulations.
In 1983 the Center’s Trout Lake Station in the Northern Highlands Lake District of Wisconsin was chosen by the US Environmental Protection Agency as a major national site for research on the biological effects of lake acidification. This site was chosen partly because of the existence of the NTL LTER program and the long tradition of limnological research and ecosystem manipulation at Wisconsin. Little Rock researchers have been able to use the sampling and data management schemes developed by the NTL LTER project. In turn, LTER researchers have gained two additional lake basins in their set of lakes. They can also use the insights gained from research on a stressed ecosystem located near the primary LTER lakes.
The Little Rock Lake Project involves the gradual acidification of one half of an hour-glass shaped lake whose two lobes are separated by a chemically resistant, neoprene barrier. The treatment basin is being acidified in steps of 0.5 pH units/2 yr period from a starting pH of 6.1 to a final pH of 4.6, which is roughly the average pH of rain in this region. One half of the lake will be left untouched as a reference lake until 1991. The goals are to document the biological and chemical changes that occur in response to acidification, to identify the direct and indirect mechanisms that regulate responses, and to expand insights to a class of lakes for which responses to acidification were understudied.
The project, which is the only one of its kind in the United States, is a cooperative study involving not only researchers Tom Frost, Tim Kratz, and John Magnuson from the University of Wisconsin-Madison, but also Carl Watras, project site manager, Kathy Webster and Paul Garrison from the Wisconsin Department of Natural Resources, Pat Brezonik and Jim Perry from the University of Minnesota, Bill Swenson from the University of Wisconsin-Superior, Bill Rose from the US Geological Survey, and John Eaton, project officer, from the US Environmental Protection Agency.
The experimental manipulation of whole ecosystems is a powerful analytical approach and, perhaps, the only way to quantify responses in natural systems. A tradition of using natural lakes for experimental studies was started in northern Wisconsin by Chancey Juday in the 1 930s to quantify the relationship between inorganic nutrients, plankton production and fish yield. Arthur D. Hasler greatly refined experimental ecosystem manipulations by splitting and then liming one half of Peter-Paul Lake on the Wisconsin-Michigan boundary.
The advantages of large-scale manipulations are tied to their capacity to
- Incorporate a full range of ecosystem processes
- Encompass all of the life stages of aquatic organisms
- Allow for the proliferation of indirect as well as direct environmental effects
At the same time, however, replication in large-scale manipulations typically is limited by logistic considerations. This lack of replication confounds the interpretation of whole- lake experiments by traditional statistical techniques.
Thus, whole-lake experiments require varied and comprehensive analytical approaches for their interpretation. Attention must be paid to:
- Developing a baseline against which to measure the effects of acid treatment
- Discerning whether a non-random change has occurred in a treated system
- Determining whether observed changes can actually be attributed to the direct or indirect effects of pH change
In the Little Rock project, all three approaches have been combined. Although the primary focus is on comparisons between a treatment and a reference basin, pretreatment data were collected for 1.5 years and the seven LTER lakes nearby serve as secondary references.
To test the applicability of nearby lakes as reference systems, we have developed an analytical approach to evaluate limnological variability patterns (Kratz et al. 1987). We have applied it to a variety of parameters from the North Temperate Lakes LTER site and from other lakes in the region.
Variability patterns for long-term data from a set of lakes within a common region can be expected to fall into three categories:
- Lake-specific parameters, consistent in each lake through several years of sampling but different among lakes
- Year-specific parameters, consistent across lakes in each year but different among years
- Stochastic or complex parameters, consistent neither with lake nor year
A two-way analysis of variance can be applied to long-term, multi-lake data matrix to provide a measure of the total parameter variance that is associated with year or site. Potentially, four variance types may be revealed by this analysis: Type I - with both lake and year effects, Type II - with a lake effect but no year effect, Type Ill - with a year effect but no lake effect, and Type IV - with neither a lake effect nor a year effect.
Data from the seven LTER lakes will permit an analysis of lake and year variability patterns for limnological parameters in the Little Rock Lake region. This analysis, combined with data from the treatment and reference basins of Little Rock Lake, will provide a more general basis for evaluating early responses to experimental acidification. For example, abundance patterns for Daphnia galeata mendotae in the LTER lakes indicate a Type II parameter, consistent year after year in each lake (Thomas Frost, unpublished data). In Little Rock’s treatment basin, substantial increases were observed for Daphnia spp. during the second year at pH 5.6 (Frost and Montz, in press). Although caution must be used in extrapolating from data on one species to the lumped behavior of the genus, the increase in Daphnia would appear to be an unlikely event except in response to a major change in lake conditions, in this case Little Rock’s acidification.
Regardless of the data used to establish baseline conditions, analytical methods are also necessary to test whether observed differences between treatment and reference systems are greater than those that would be expected to occur by chance alone.
Once again the limited replication in whole-system experiments confounds the use of standard statistical procedures. We are pursuing several methods to establish the significance of change in the acidified basin of Little Rock Lake.
In a direct attempt to address the limited replication in environmental assessment work, Stewart-Oaten et al. (1986) suggested an approach involving a comparison before and after a treatment for a reference site and a site in which a change might be expected (termed BACI for a before: after, control impact comparison). Carpenter et al. (unpublished) have developed a modification of the BACI approach that incorporates randomized intervention analysis (AIA) to provide an estimate of the probability that an observed shift in conditions might have occurred by chance.
We are testing the utility of the randomized intervention approach by comparing not only the reference and treatment basins of Little Rock Lake but all of the possible pair- wise comparisons between the two Little Rock basins, the seven LTER study lakes, and three lakes now being used for whole-lake experiments involving shifts of top piscivores at the University of Notre Dame Environmental Research Center located less than 50 km north of Little Rock Lake.
Having detected changes in the acidified basin, it is critical to ascertain whether observed changes can be attributed, either directly or indirectly, to acid treatment. Consideration must be given to alternative explanations for observed changes. We are relying on at least two validation techniques: comparison with the results of previous studies and experimentation on a smaller scale. Before initiating the acidification, a set of predictions was compiled for each target pH, based on reports in the literature and on deterministic models. These predicted responses will provide one benchmark for gauging effects. In addition, smaller experiments are nested within the ecosystem manipulation. These experiments include limnocorrals, in situ bottle incubations, laboratory assays and paleolimnological investigations. Many of the limitations of the whole-system approach can be circumvented by coupling it, hierarchically, with these other experimental scales.
Ecosystem-level experiments are essential in assessing the effects of environmental perturbations like acidification. By coupling the Little Rock Lake acidification to the Northern Lakes LTER project, we can overcome constraints imposed by limited replication and develop powerful analytical tools for the interpretation of complex ecological data.
References:
Frost, T. M. and P. K. Montz. (in press). Early zooplankton response to experimental acidification in Little Rock Lake, Wisconsin, USA. Verh. Verein. Limnol.
Kratz, T. K., T. M. Frost, and J. J. Magnuson. 1987. Inferences from spatial and temporal variability in ecosystems: Analyses from long-term zooplankton data from a set of lakes. Amer. Natur. 129: 830-846.
Stewart-Oaten, A., W. W. Murdoch, and K. R. Parker. 1986. Environmental impact assessment: “pseudo- replication" in time? Ecology 67: 929-940.
For more information, contact Tom Frost or Carl Watras, University of Wisconsin, Trout Lake Station, 10810 County N, Boulder Junction, WI 54512, (715)356- 9494.