Winter-Run Chinook Life-Cycle Modeling for Central Valley Hydromanagement

Winter-run Chinook were listed as endangered under ESA in 1994.  The reasons for winter-run declines included loss of natal spawning habitat, poor survival through the lower Sacramento River and San Francisco delta, poor ocean productivity, and capture in the Sacramento-Klamath Rivers Chinook fishery.   We developed a life-cycle model for winter-run Chinook that operates at a monthly time step to evaluate how these factors have combined to affect the population dynamics from 1980 to 2010.  Because hydrologic management of the Sacramento River and delta are so critical for California water supply, the model is structured spatially to reflect hydromanagement areas.  The model was fitted in a Bayesian framework to provide probabilistic inference on population vital rates as a function of temperature and flow in freshwater and productivity and harvest rates in the ocean. Life-cycle model coefficients may have high correlation due to a lack of observational data between stages, and density dependence functions may furthermore lead to non-linear posterior distributions that are "banana-shaped".  Both of these conditions can cause problems for Markov Chain Monte Carlo (MCMC) samplers. To overcome these limitations, we used an adaptive importance sampling (AIS) algorithm and compared inference under MCMC, AIS, and hybrid AIS-MCMC methods.