**ABSTRACT NOT FOR CITATION WITHOUT AUTHOR PERMISSION. The title, authors, and abstract for this completion report are provided below. For a copy of the full completion report, please contact the author via e-mail at bence@msu.edu. Questions? Contact the GLFC via email at frp@glfc.org or via telephone at 734-662-3209.**



Quantitative Tools for Assessing and Managing Cisco Populations


James R. Bence1, Samuel B. Truesdell1,4, Richard D. Clark1, Nicholas C. Fisch1,5, Jared T. Myers2, and Daniel L. Yule3


1Quantitative Fisheries Center, Department of Fisheries and Wildlife, Michigan State University East Lansing, MI. 48824-1101

2 U.S. Fish and Wildlife Service, Ashland Fish and Wildlife Conservation Office, 2800 Lake Shore Dr. East, Ashland, WI 54806

3 U.S. Geological Survey, Lake Superior Biological Station, 2800 Lake Shore Dr. East, Ashland, WI 54806

4 Gulf of Maine Research Institute, 350 Commercial St, Portland, ME 04101

5 Fisheries and Aquatic Sciences Program, University of Florida, PO Box 110410, Gainsville, FL, 32611-0410


December 2018




We completed stock assessment models for Cisco, Coregonus artedi, in Thunder Bay Ontario, completed management strategy simulations evaluating alternative harvest policies for the Thunder Bay stock of Cisco, evaluated how assessment results would be influenced by reductions in sampling, and conducted technology transfer workshops that have led to development of preliminary Cisco assessment models for other areas of Lake Superior. Stock assessments are a critical to modern fisheries management, supporting the calculation of key reference variables used to make informed management decisions. Given the lack of assessment models for Cisco and considerable uncertainty as to which class of assessment models is appropriate for fishery stocks in general, we developed both a statistical catch-at-age assessment (SCAA) model, and a statistical catch-at-size assessment (SCSA) model for the Thunder Bay stock of Cisco. We addressed: whether a SCAA models actually performs better than SCSA models when age data are available, or is this just an assumption we make in fisheries research and management? Both models were fit using an integrated framework with multiple sources of data including hydroacoustic estimates of spawning stock, fishery dependent and independent age/length compositions, and harvest data. Our results suggest that for Cisco in Thunder Bay, data-limitations related to lack of size-composition data over the size range for which cisco growth is rapid resulted in difficulty estimating relative year-class strength within a SCSA. This led to parameter confounding and ultimately the inability to estimate natural mortality within a SCSA. This hampered the utility of a SCSA model in comparison with a SCAA model when age composition data were available. We used Management Strategy Evaluations (MSEs) to evaluate current and alternative harvest policies for the Thunder Bay stock of Cisco. MSEs can provide information to managers on the relative performance of alternative management policies (strategies) while accounting for uncertainty. Our simulations explicitly accounted for uncertainty in the frequency of strong year classes being produced by Cisco, the stock-recruit relationship, stock abundance, and the sex-specific nature of roe harvest. Assuming future productivity is similar to that observed over the past 30 years, results suggest the current exploitation rate of 10% is sustainable in terms of maintaining spawning biomass above 20% of the unfished level. Variants of constant exploitation rate control rules that included biomass thresholds defining when exploitation rate is to decrease as a function of spawning stock size increased yield, decreased risk, and increased the magnitude of spawning biomass at the end of the simulation period. However, these advantages came at the expense of greater inter-annual variation in yield. Constant catch control rules greatly underperformed constant exploitation rate control rules in terms of magnitude in yield. Constant catch control rules, however, did reduce inter-annual variation in yield compared to constant exploitation rate control rules, and conditional versions of constant catch control rules (i.e., threshold stock sizes below which catch limit was reduced) mitigated some of the increased risks. We evaluated how reductions in the amount of sampling would likely influence stock assessment results by refitting the Thunder Bay Cisco assessment model with only subsets of the full dataset. We adopted this approach because the Thunder Bay assessment had more data available than is often the case. In a first set of analyses, we reduced the number of ages, either by randomly selecting a subset of structures collected per sampled trip, or randomly selecting a subsample from all ages, mimicking a reduction in sampling intensity or just in aging effort. Results suggested that up to a 70% reduction in either sampling effort or aging could be implemented without much loss of information in age compositions or influence on assessment results. A second set of analyses looked jointly at (a) a reduction in the portion of the fishing season for which biological samples were collected (i.e., age compositions based on randomly selected sub-intervals of season) and hydroacoustic surveys done only in a subset of years. Reducing hydroacoustic survey frequency had a larger influence on assessment quality than did a reduction in biological sampling. Reducing biological sampling via random temporal subsets did reduce information content of age compositions, but had little influence on stock assessment results. This is likely a consequence of the strong periodic year classes, characteristic of Cisco in Lake Superior. This dynamic may require less sampling to characterize. Finally we met with state, federal and tribal biologists and explained our assessment model and worked with them to develop preliminary assessment models for several other areas in western Lake Superior.