ADAPTIVE CLUSTER SAMPLING OF FRESHWATER MUSSELS IN THE ST. CROIX RIVER
Ben Dickinson, Mark Hove, and Dan Hornbach
Macalester College, Department of Biology
Estimating population size of endangered or threatened species is often difficult due to low density. Typically a large number of samples must be collected to estimate population size with confidence. An alternative approach is adaptive cluster sampling, which increases sampling efficiency for organisms that are spatially clustered, such as freshwater mussels. Adaptive cluster sampling involves selecting sites randomly, but collecting additional samples from neighboring cells when a target species is encountered. We developed a modified adaptive cluster sampling technique to estimate population size of federally endangered Quadrula fragosa and Lampsilis higginsii, and Wisconsin state endangered (Minnesota threatened) Cyclonaias tuberculata in the St. Croix River. Previous research has shown that these species are most commonly found in areas of high mussel density. Therefore we implemented a two-stage stratified sampling design to maximize sampling effort in areas more likely to be inhabited by target species.
In 2005 we randomly selected 20 50-m2 cells from a two kilometer river reach near Franconia, Minnesota. We sampled three 1-m2 quadrats from each cell and if at least 4 mussels were found in one quadrat we labeled the cell “high mussel density.” Of the 20 cells sampled, 10 were high density. We initiated adaptive cluster sampling in these 10 high-density cells. We randomly selected 10 1-m2 quadrats within each high density cell, and collected substrate samples, measured hydrologic parameters, and counted and identified living and dead mussels. When a target species was found we sampled the four adjacent quadrats. Neighboring cells were sampled until no additional target species were observed. We collected 750 mussels at Franconia, including target species from three of 10 50-m2 high-density cells. Of the target species, we found 6 Cyclonaias tuberculata (0.8% of the population), 2 Quadrula fragosa (0.27%) and 3 Lampsilis higginsii (0.4%). Four of these individuals were collected from one adaptive cluster.
In 2006 we applied this sampling technique at Interstate State Park, Minnesota. This area has much higher mussel density than found at Franconia, with 19 mussels per square meter being considered “high density”. Eight of the 20 randomly selected 50-m2 cells were high density, five of which contained Lampsilis higginsii and/or Quadrula fragosa individuals. We collected 3205 total mussels, 194 Cyclonaias tuberculata (6.1% of total population), 6 Lampsilis higginsii (0.19%), and 15 Quadrula fragosa (0.47%). 10 of the 15 Quadrula fragosa were in clusters, as well as four of the six Lampsilis higginsii, indicating that these rare mussels are often spatially clustered, and that adaptive cluster sampling may be an effective way to find these clusters. We had to stop using Cyclonaias tuberculata as a target species because they were so abundant that we triggered adaptive cluster sampling at almost every site. This suggests a population threshold at which adaptive cluster sampling becomes inefficient and simple random sampling is more effective. We are currently processing data from both 2005 and 2006 to estimate population sizes and habitat characteristics for these three mussel species.
