Ecological Society of America

♦  Science Resources  »  Science Office  »  Tools and Publications

Sustainable Biosphere Initiative Project Office Workshop Recommendations Ecological Resource Monitoring: Change and Trend Detection

The Sustainable Biosphere Initiative, a project of the Ecological Society of America (ESA), hosted a workshop, "Ecological Resource Monitoring: Change and Trend Detection," at the Patuxent Wildlife Research Center. This joint effort of the Statistical Ecology Section of ESA, the Section on Statistics and the Environment of the American Statistical Association (ASA), and the Environmental Protection Agency's Environmental Monitoring and Assessment Program (EMAP), provided a forum for interaction between the statistical and ecological communities. The workshop sought to assess the current scientific knowledge and science gaps in change and trend detection methodology used in ecological resource monitoring.

Monitoring is a multidisciplinary activity common to all environments, but activities must be tailored to each habitat. The workshop drew upon experience from different scientific disciplines, monitoring methods and habitats. Twelve invited papers covered four types of monitoring programs: intensive or sentinel sites, networks of sites, surveys, and complete coverage, as well as two environments: aquatic and terrestrial. These presentations provided stimulus for the lively and extended discussions of the thirty­five scientists in attendance (see attached list). The general themes that emerged from this workshop comprise a set of findings and policy recommendations for EMAP and the National Environmental Monitoring Framework (Committee on Environment and Natural Resources, CENR), an effort to increase cooperation, coordination and our understanding of the state of environmental resources.

Findings and Recommendations

Recommendations from the workshop are particularly timely, as federal environmental monitoring programs such as EMAP, CENR, National Research Council, and the Office of Science and Technology Policy, currently are being evaluated.

I. Environmental Monitoring Guidelines

• An environmental monitoring framework must recognize the different roles of intensive (sentinel) sites, networks of sites, surveys, and complete coverage in assessing trends in ecological resources. Each type of monitoring program yields unique or specific information. Intensive sites provide process understanding, (e.g., nutrient fluxes, recruitment) but are not applicable at the regional or national level; networks are generally used for monitoring of large-scale resources such as air or soil and may (e.g., atmospheric deposition) or may not (e.g., lake acidification) be useful for characterizing regional scale phenomena; and surveys, which are often more policy oriented, provide information that can be statistically generalized to regional or national scales.

• Intensive sites are often ecologically unique, may not be representative of subsets of populations, and should not be treated like statistical samples from a region. For continental or regional scales, networks of sites and surveys (such as EMAP) are generally more effective in detecting trends than intensive sites because they provide information on cause-and-effect relationships. Trends detected at intensive sites within an area may be different than trends revealed through survey monitoring. Networks of sites provide data about the functioning of a regional ecosystem and the impacts of large scale environmental stressors.

• Intensive sites collect more information per site than monitoring networks, which collect more information per site than surveys. Monitoring networks are often most successful in following stressors per se than responses to stressors. Monitoring sites may or may not be representative of regional phenomena (depending on siting criteria and what is measured). Survey sites are generally designed to be representative of habitats.

• Because ecological and statistical significance are not the same, it is better to report confidence intervals for measured attributes rather than simply stating a result was (or was not) statistically significant. It is important to clearly state assumptions made in data analysis so that the limitations of inferences can be determined.

• The term "trend" has many uses. Researchers must clearly identify what is meant by "trend". "Trend" usually infers general tendency or unidirectional change. However, the design and analysis of data from a sampling or monitoring program may be jeopardized by an unclear definition of trend. For example, different designs are required to obtain information on changes in the ambient mean level of an attribute, versus in the mean level of that attribute as a response to anthropogenic stressors. Using the correct design has implications for all aspects of a program.

• Choice of sampling methodology for detecting trends also depends upon the question being asked. From a statistical standpoint, it is important to formulate the question, and determine an appropriate statistical method or approach, sampling frequency and duration.

• Intensive measurements at a few sites may provide detailed mechanistic information for those sites, but this information may not be generally applicable to the region. Annual index-period measurements at survey sites provide regional trend information that can not be obtained from intensive sites, but the trend information at any one survey site is less precise than the types of data collected at an intensive site.

• Survey designs need to include protocols for long-term estimation of net change, descriptions of components of change, and indicators of trends. Monitoring designs must be inherently flexible to accommodate changes in objectives over time. A process should be established to determine how to add new monitoring schemes, maintain existing ones of continued importance, and drop obsolete ones. Wherever possible, changes in sampling methodology should be accompanied by periods of collecting data using both methods so that the consistency of long­term databases is preserved.

II. Environmental Stressors and Effects

• Early detection of changes in ecological resources through stressor-oriented monitoring is more effective than effects-oriented monitoring. If possible, it is better to detect the cause of an effect than to detect the resulting multivariate effects. However, effects­oriented monitoring is essential to assess trends in the state of natural resources. Empirical information relating stressors to effects must be available to predict probable changes in the biotic components of the system.

• Interaction of anthropogenic and natural disturbances may be a major driver of change. Look for interactions and synergistic effects.

• There may be no pristine areas available to establish baseline data. We must recognize that observed changes may actually reflect or be superimposed on larger or longer term responses to previous disturbances.

III. Spatial and Temporal Scales

• Choosing an appropriate spatial and temporal scale for measurement depends on the level of understanding of the ecological processes to be monitored and the indicators or responses to be measured. Reporting monitoring results at multiple spatial and temporal scales of aggregation is also necessary to detect trends at regional­or continental­scales.

• Ecologists and statisticians must work together on methods for extending findings from intensive sites to surveys.

Trend studies should focus on the description of trends with associated estimates of uncertainty rather than hypothesis testing.

Data must be evaluated carefully particularly if models are postulated: sometimes models are misspecified. The change of interest could be in the variance not just the means. Issues such as observer bias, interaction between anthropogenic and natural disturbances, and scale of aggregation may need to be addressed.

Interpretation of observed trends is often based on process understanding. It is important to ensure that such understanding is used to determine what attribute of the system is monitored, where monitoring sites are located, and how often sites are sampled. Such temporal and spatial scale considerations may not be independent. For example, multiple stressors affect the composition of kelp bed communities: El-Nio events which are global and interannual; storms, which are regional and episodic; and herbivory, which is local and possibly continuous.

IV. Ecology and Statistics

• The usefulness of information on changes and trends in the condition of an ecological resource increases when accompanied by information on changes and trends in one or more stressors. Patterns of change in a measured attribute may appear to be noise but may actually be the signal of a stressor acting on a different temporal or spatial scale. Statistical means are not always indicative of trends, and statistical variance may be a better indicator of change.

V. Policy Implications

• Data from diverse sources gathered at several scales must be synthesized to provide an overall assessment of ecological change. There should be a balance of effort expended on data collection and data analysis and a commitment to a structure for long­term data archiving and curation. Incompatibility of historical data, changes in design, and retirement of investigators are problematic to maintenance of long-term databases. If the framework is to succeed, it is recommended that roughly 20% of resources be allotted to archiving collected data and a process for data sharing should be established.

• Survey studies address policy questions requiring descriptions of the state of natural resources, while intensive sites address policy questions requiring scientific understanding of ecological processes. A process should be established to decide how to add new monitoring schemes, maintain existing ones of continued importance, and drop obsolete ones.

Statistical and ecological experts can aid in bringing agencies together to deal with diverse interests in a network of monitoring schemes. Joint research should be encouraged between ecologists and statisticians on integration of findings from intensive sites and surveys. This is a formidable challenge. Improved collaboration of ecologists and statisticians can contribute an interdisciplinary focus to the National Monitoring Framework thus providing reliable detection of change in ecological resources, and ultimately, increasing our understanding of novel and complex ecological phenomena.

---

WORKSHOP PAPERS

Design and Analysis Approaches for Trends: RESEARCH SITES

Chair: Aaron M. Ellison, Mount Holyoke College

Ghost communities and the problem of reduced expectation in kelp forests. Paul K. Dayton, M.J. Tegner, P.B. Edwards, and K.L. Riser. Scripps Institution of Oceanography.

Ecotone displacement trends on a highly dynamic barrier island: Hog Island, Virginia. Michael Fenster, Dept. of Environmental Sciences, University of Virginia.

Long-term environmental monitoring: Some perspectives from lakes. ¹Craig A. Stow, S. R. Carpenter, , K.E. Webster, and T.M. Frost. ¹Center for Limnology, University of Wisconsin, Madison.

Detecting trends in species composition. Thomas E. Philippi, Savannah River Ecology Laboratory.

Design and Analysis Approaches for Trends: NETWORKS OF SITES

Chair: Sylvia R. Esterby, Canadian Wildlife Science Division.

Vegetation index time series image data and the analysis of ecosystems dynamics. J. Ronald Eastman, Dept. of Geography, Clark University.

Improving trend detection in river solute fluxes. R. J. Hooper, U.S. Geological Survey, Atlanta, GA.

Modeling and estimation of population change for the North American Breeding Bird Survey. William Link and J. Sauer, Biometrics Research, Patuxent Wildlife Research Center.

Interannual variability at three inland water sites: Implications sentinel ecosystems. Alan Jassby, Division of Environmental Studies, University of California, Davis.

Design and Analysis Approaches for Trends: SAMPLE SURVEYS

Chair: Bruce P. Hayden, University of Virginia.

Design and estimation for examining the dynamics of natural resources. Sarah M. Nusser, F. J. Breidt, and W. A. Fuller., Department of Statistics, Iowa State University.

Can site-specific trends be extrapolated to the regional level? A lake acidification example for the northeastern U.S. John Stoddard, ManTech Environmental Research Service Corporation, C.T. Driscoll, S. Kahl, and J. Kellog.

Sampling methods for estimating change in forest resources. Charles Scott, U.S. Department of Agriculture.

Detecting regional-level trends in ecological responses. N. Scott Urquhart, Oregon State University.

Copies of this report are available from: