Exploring Variability in Population Dynamics and the Influence of Environmental Factors on Recruitment of an Estuarine Fish

The focus of this dissertation was to explore variability in population dynamics and environmental factors influential to recruitment of spotted seatrout in Florida with the end goal of testing the incorporation of select environmental variables into the current regional stock assessment models for spotted seatrout. In Chapter 2, I compared the age and size structure of six estuary populations of this species and determined whether there was significant spatial covariation in recruitment. The results of this chapter indicated that the dynamics of each local estuary population are governed more likely by environmental factors than genetic similarities. Further, they suggest that the geographical management regions are incongruous with spatial stock dynamics and that it may be more appropriate to conduct assessments on an estuary scale as opposed to lumping estuaries together in ambiguous management regions. In Chapter 3, I evaluated the importance of a suite of environmental predictor variables to recruitment of spotted seatrout using a novel machine learning algorithm that inherently models higher-level interactions. This was important because co-linearity among environmental factors can frequently result in spurious relationships. Because Chapter 2 indicated that there is little homogeneity in the dynamics of each estuary, the importance of the environmental predictors was evaluated on an estuary-by-estuary basis. The results of this chapter indicated that, generally, salinity, water temperature, river flow, precipitation and drought are the top five most important predictor variables with the exception that river flow may be exceedingly important in Cedar Key and northeast Florida and, moreover, that there is relatively little latitudinal difference in variable importance among estuaries. This algorithm was shown to be more accurate and less biased than traditional statistical methods like generalized linear models. Finally, in Chapter 4, the top five predictor variables were incorporated into regional quantitative assessment models for spotted seatrout via two methods which adjust predicted annual recruitment. This was done because any residual, unexplained model variation, may be reduced by the presence of an environmental variable. Such inclusion may result in more precise reference points and more accurate estimates of forecasted recruitment. While precipitation significantly improved model plausibility for the northeast and northwest regions, the southeast and southwest regions were unaffected by these variables likely because these models contain many records of age and length data which corroborate the variation in young-of-the-year abundance (proxy for recruitment). It seems, therefore, that models with few length or age observations may be improved by an environmental index but this should be evaluated on a case-by-case basis and is cause for further investigation. In contrast, management reference points were unaffected and this is likely because natural mortality and growth parameters were not estimated by the model. Future research should explore 1) alternative methods for inclusion and 2) the overall benefit of an environmentally-explicit assessment for management purposes. Nevertheless, mid-21st century precipitation was used to adjust forecasted recruitment in the northeast and northwest regions. Under both precipitation scenarios and strict assumptions regarding constant fishing effort and selectivity, recruitment will be highly variable. Therefore, management strategies that ensure a well-developed age-structure as well as anthropogenic land and water usage that preserve valuable estuary habitat will promote resilience of the spotted seatrout stock in a rapidly changing climate