Florida Lake Management Society Annual Conference, Naples, Florida, June 4 – 7, 2007
AN EXPERIMENTAL ANALYSIS OF PROBABILITY-BASED SAMPLING REGIMES
Todd Tietjen Department of Wildlife and Fisheries, Mississippi State University Mississippi State, MS
Comprehensive sampling of water quality in regions with large numbers of aquatic systems can be extremely difficult given logistical and time constraints involved in sampling numerous, widespread and diverse bodies of water. One approach that attempts to overcome these difficulties is probability-based sampling. This sampling regime uses a randomly selected subset of the total population of lakes in order to develop a statistically based assessment of the water quality for an entire region/population. In natural systems it is impossible to assess the validity of the probabilistic approach to monitoring, as a complete set of data from the entire “population” of lakes cannot be obtained. Using the >70 aquaculture ponds on the Mississippi State University campus, I have developed an experimental test of this monitoring approach in order to test its performance in a population of lakes that by design are similar in size, shape, water source and geology. The water quality parameters temperature, dissolved oxygen, pH, specific conductance, in-vivo chlorophyll fluorescence and in-vivo phycocyanin were measured in each pond at daily to weekly intervals for 1 year. I will explore the variability encountered in these controlled systems in order to explore the level of confidence that can be obtained using different sample sizes and sampling intervals in order to facilitate better application of the approach to natural lakes.
As monitoring efforts continue to be curtailed in order to control costs of environmental monitoring one approach that is growing in importance involves probability-based sampling. This approach derives from the basic statistical concept that a random sample can be used to assess the central tendency (mean) of a given parameter for a given population. A simple example would be a “region” with 100 lakes to be sampled for temperature. Sampling constraints dictate that no more than 20 of the lakes will actually be sampled. By randomly selecting 20 lakes to be sampled a mean temperature along with a measure of variability can be calculated to describe the overall population of lakes. The problem with this approach is that it is difficult to assess accuracy and precision of the sample. By using a collection of 70 ponds with similar geology, soils, environmental exposure, size, depth, and water source this study addresses the questions of how large of a sample is required, what level of precision can be obtained, and evaluates the overall approach.
Ponds were sampled from the ~2 m from shoreline, ~0.5 m below the surface between the hours of 10:00 and 14:00 at weekly intervals between May and December of 2006. Temperature, dissolved oxygen, pH, turbidity, and specific conductance were measured with a calibrated
Session 4 – Page 5