Upper Mississippi River Restoration Program Long Term Resource Monitoring

Introduction/Background:

Freshwater mussels are the largest group of endangered animals in North America. Over the past 50 years, about 20 species have been lost from the Upper Mississippi River basin. Mussels are threatened by changes in flow patterns within rivers caused by dams, dikes, and levees; by increases in sediment loads; and by invasive species, such as zebra mussels and Asian carps that compete with mussels for food. Conservation of mussels is of great concern to the States, the Fish and Wildlife Service, and the Corps of Engineers and considerable effort is being expended to assess the effects of proposed habitat rehabilitation projects (HREPs) on mussels. For example, estimates of mussel mortality in 2005 as a result of the drawdown of Pool 5 range from about 100,000 to upwards of several million (WDNR et al. 2005). Similarly, crude estimates of the number of endangered Lampsilis higginsi that might be adversely affected by the construction of islands in Pool 11 exceed the incidental take level (500 individuals) identified by the U.S. Fish and Wildlife Service. However, there are two uncertainties in evaluating the potential effects of HREPs on mussels. First, if a given HREP kills some number of mussels, it is usually unknown what proportion of mussels in the pool this represents. For example, if the Pool 5 drawdown killed 100,000 mussels, but there were three million mussels in the pool, then this loss rate may be acceptable, given that this represented only 3% of the total mussel population. Second, it is unclear if the short-term loss of mussels may be mitigated by a longer term gain in improved habitat.

Unfortunately, we do not currently have the ability to rigorously assess the importance of the loss of mussels as a result of HREPs. Most surveys of mussels done associated with HREPs involved sampling in the general vicinity of an HREP (e.g., around the footprint of a proposed island or in the dewatered area of a drawdown) and generate data on species composition and abundance at the spatial scale of the HREP. These estimates cannot be used to make inferences at the pool scale because the HREP area cannot typically be presumed representative of the pool. This scale discrepancy is critical because the pool scale represents the effect scale for water-level drawdowns and the scale typically presumed for mussel population analysis in the UMR. Estimates of mussel abundances at both island and pool scales are urgently needed to accurately assess the effects of HREP efforts. Estimates of abundances at the island scale are needed by managers to get high resolution data at the impact site to accurately assess the potential loss of mussels as a result of island construction. Estimates of abundances at the pool scale are needed to put the HREP-level data in perspective with the rest of the pool.

Recently, substantial progress has been made in evaluating sampling designs for mussels in small to medium sized rivers. In these smaller systems, research has suggested that systematic sampling may be appropriate, because it is easy to implement in the field, provides good spatial coverage, and is the preferred design for sampling rare, spatially clustered populations in the absence of prior information on distributions (Christman 2000, Strayer and Smith 2003). Multiple random starts allow valid statistical inference in systematic designs. For example, a study of mussel populations on the Cacapon River, WV, showed that systematic designs were consistently more efficient than stratified random designs, when distances between sampling units were chosen appropriately and > 1 random start was used (Pooler and Smith 2005). Because freshwater mussel populations have patchy distributions at multiple spatial scales, more sampling effort should be allocated in the locations where the organism occurs than where it does not occur (Villella and Smith 2005). Thus, adaptive designs, which allow sampling effort to increase where mussels are found, are promising designs for mussel populations. Examples of adaptive designs are adaptive cluster sampling, sequential sampling, and double sampling for stratification (Salehi and Seber 1997, Strayer and Smith 2003, Salehi and Smith 2005, Villella and Smith 2005).

High cost and difficulty with collecting mussels in large rivers are the central issues separating designs for small to medium sized rivers versus large rivers. The natural approach when a large area must be sampled is to sample in stages. First, divide the large area into medium sized sections or primary sampling units. Second, select primary sampling units using a probability sampling technique. Third, sample within the primary unit. Much of what has been learned about sampling in small to medium sized rivers is directly applicable to the third step, sampling within a primary unit. Thus, the essential challenges to designing surveys for large rivers are 1) how should primary units be defined (i.e., size and shape), 2) how should primary units be selected, and 3) how should sampling effort be allocated within and between primary units?

Relevance of research to UMRS/LTRMP:

Over the past 10 years, several attempts have been made to include mussels as a component of the LTRMP, however, the potential costs and the lack of a statistically-rigorous sampling design have generally stalled these efforts. We argue that costs are really unknown because no one has estimated the costs for a statistically-sound sampling program in a large river to get pool-wide estimates of mussel abundances. Under some of the new, more efficient sampling designs, costs may be considerably less than previously anticipated. The proposed research is relevant to the UMRS and the LTRMP because designs for sampling mussels at various scales are urgently needed to provide river managers with tools to rigorously assess the effects of HREPs on mussels. In the longer-term, evaluation of the mussel data generated as a result of the HREPs may be used to refine a sampling plan for the long-term monitoring of mussels in the UMRS.

Methods:

We propose to evaluate multiple sampling designs for suitability in obtaining relative density and population estimates at island and pool scales. Most of the accumulated mussel sampling wisdom has been developed for sampling streams. Unfortunately, designs preferred for streams may not be transferable to large rivers because the cost of sampling may be prohibitive, given the large areas to be covered. This work will be done at two spatial scales: (1) the island-scale (the effect scale associated with the footprint of an island HREP) and (2) the pool-wide scale (i.e., spatial scale of most drawdowns). We will use variance estimates from existing quantitative data from studies that employed random sampling (Pool 10, Holland-Bartels et al. 1990; Pool 9, Sietman 2006) and newly acquired estimates from systematic sampling in Pool 5 (to be completed in 2006) to simulate data. Simulated data will be run through various sampling designs (e.g., simple random sampling, systematic sampling with single and multiple starts, probability proportional to size designs, adaptive cluster sampling, two-stage conventional sampling, two-stage adaptive cluster sampling, two-stage sequential sampling, and stratified random sampling) to determine which designs maximize relative efficiency and minimize sampling effort and cost. We will also estimate the costs and precision of estimating mussel abundances within 5, 10 and 20% of the true population.

The proposed work substantially elaborates on existing work by UMESC staff. For example, we were recently asked by the Pool 5 drawdown team to develop a sampling approach for mussels to evaluate the effects of the 2006 drawdown on mussel populations. However, this work on Pool 5 is preliminary in that the sampling design was drafted within a short period (about 6 weeks from initial consultation to the sample site estimation) and, consequently, limited data were used to develop the design. The work proposed here will build upon the Pool 5 work in that inferences from the Pool 5 sampling will be including in selecting designs suitable for use throughout the UMR. Thus, the designs proposed in this APE can be used to assess the effects of future HREPs on mussels.

Literature Cited:

Christman, M.C. 2000. A review of quadrat-based sampling of rare, geographically clustered populations. Journal of Agricultural, Biological, and Environmental Statistics 5:168-201.
Holland-Bartels, L.E. 1990. Physical factors and their influence on the mussel fauna of a main channel border habitat of the upper Mississippi River. Journal of the North American Benthological Society 9:327-335.

Pooler, P.S. and D.R. Smith. 2005. Optimal sampling design for estimating spatial distribution and abundance of a freshwater mussel population. Journal of the North American Benthological Society 24:525-537.

Salehi, M.M., and G.A.F. Seber. 1997. Two-stage adaptive sampling. Biometrics 53:959-970.

Salehi, M.M., and D.R. Smith. 2005. Two-stage sequential sampling: a neighborhood-free adaptive sampling procedure. Journal of Agricultural, Biological, and Environmental Statistics 10:84-103.

Sietman, B. 2006. Abundance of unionid mussels in shallow areas of Pool 9 of the Upper Mississippi River. Final report submitted to the U.S. Army Corps of Engineers, St. Paul, MN. 63 pp.

Strayer, D.L. and D.R. Smith. 2003. A guide to sampling freshwater mussel populations. American Fisheries Society Monograph 8, Betheseda, MD.

Villella, R.F. and D.R. Smith. 2005. Two-phase sampling to estimate river-wide populations of freshwater mussels. Journal of the North American Benthological Society 24:357-368.

WDNR et al. (Wisconsin Department of Natural Resources, Minnesota Department of Natural Resources, and the U.S. Army Corp of Engineers). 2006. Preliminary report on the effects of the 2005 Pool 5, Mississippi River drawdown on shallow-water unionids. Wisconsin Department of Natural Resources, La Crosse, WI. 24 pp.

Principal Investigators:

Teresa Newton (tnewton@usgs.gov), Brian Gray (brgray@usgs.gov), Jim Rogala (jrogala@usgs.gov), David Smith (drsmith@usgs.gov) and Steve Zigler (szigler@usgs.gov)