Biologists interested in population viability, and managers concerned with setting harvest quotas are often faced with identifying spatial structure of species in data-limited scenarios (when data exist, they may have many gaps in time, large associated measurement or observation errors). Existing types of data that ecologists have used to identify structure include movement or tagging data between populations, estimates of dispersal distances between sites, genetic methods (e.g. Structure), and time series of survey data (or trends in abundance). This latter method has proved promising, with recent advancements in state-space modeling (both Maximum Likelihood and Bayesian). These multivariate autoregressive state space models (MARSS) have recently become more accessible with new R packages. One of the challenges in using MARSS models is that data other than time series may be informative with respect to spatial structure, yet no framework exists for including these data within MARSS models. Within this talk, we will demonstrate how other data sources (including genetic data) may be included as an additional component of the likelihood.
Results/Conclusions
As a case study, we present an application to Puget Sound Chinook salmon. Chinook have relatively high commercial value, and monitoring programs have resulted in relatively long (> 40 years) of estimates of returning adults. Within the Puget Sound region, there have been 22 previously identified sub-populations. Identifying correlations between the productivity of these subpopulations has been challenging, because of various anthropogenic disturbances that also have spatial correlation. These include, but are not limited to changes in habitat, large scale hatchery programs, straying rates of returning spawners between subpopulations, and fishing. Using the modified MARSS framework, we demonstrate that the information in baseline genetic data may be seen as complementary to the original MARSS approach, that just uses time series data.