OOS 5-5
Optimizing environmental monitoring designs

Monday, August 5, 2013: 2:50 PM
101E, Minneapolis Convention Center
Carrie R. Levine, Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, CA
Ruth D. Yanai, Forest and Natural Resources Management, SUNY College of Environmental Science and Forestry, Syracuse, NY
Gregory Lampman, NYSERDA, Albany, NY
Douglas A. Burns, US Geological Survey, Troy, NY
Charles T. Driscoll, Department of Civil and Environmental Engineering, Syracuse University, Syracuse, NY
Gregory B. Lawrence, U.S. Geological Survey, Troy, NY
Jason A. Lynch, Clean Markets Dvision, US Environmental Protection Agency, Washington, DC
Nina Schoch, Biodiversity Research Institute, Gorham, ME

Uncertainty analyses can be used to improve efficiency in ecosystem monitoring efforts.  We illustrate four methods of data analysis appropriate to four types of monitoring designs, using examples from northeastern North America.

•      To analyze a long-term record from a single site, we applied a general linear model to stream chemistry at Biscuit Brook in the Catskill Mountains. Biscuit Brook has been monitored continuously on a weekly basis since 1999; we simulated reduced sampling efforts and evaluated confidence in the detection of change over time. 

•      To analyze a one-time survey, we applied a detectable difference analysis to loon tissue mercury concentrations, illustrating effect of sampling intensity on statistical power and the selection of a sampling interval likely to detect an expected change over time. 

•      To analyze sampling intensity at one point in time, we subsampled forest inventory plots from the Hubbard Brook Experimental Forest, illustrating the relationship between sampling intensity and uncertainty in the mean biomass or nutrient content.

•      To analyze time-series data from multiple sites, we assessed the optimal number of lakes and the optimal number of samples per year needed to monitor change over time in Adirondack lake chemistry, using a repeated-measures mixed-effects model.


Environmental sampling designs should maximize the information gained relative to the resources expended on data collection and analysis.  Uncertainty analysis can help select a sampling interval likely to detect an expected change over time and evaluate whether effort should be re-allocated in space or time to best reduce uncertainties.