COS 11-7 - Using Virtual Population Analysis (VPA) to estimate under-sampled recruits to improve population dynamics models

Monday, August 8, 2016: 3:40 PM
Floridian Blrm A, Ft Lauderdale Convention Center
John V. Gatto, Biology, Florida International University, Miami, FL and Joel Trexler, Department of Biological Sciences, Florida International University, Miami, FL
Background/Question/Methods

The regulation of population dynamics is often determined by the most vulnerable life stage experiencing density-dependent mortality, the biotic and abiotic drivers of juvenile life stages have long interested ecologists.  This is of particular interest within hydrologically dynamic wetlands dominated by seasonal changes in hydrology that largely drive fish population dynamics.  However, our ability to adequately sample the smallest individuals of populations has hindered our ability to fully understand cyclical patterns of species recruitment.  Several statistical methods have been developed to estimate and correct for missing cohorts, notably the Virtual Population Analysis (VPA).  We used a 17-year time series from 19 small fish populations in the Florida Everglades to determine spatial-temporal changes in fish recruitment.  Each species was separated into 1-mm length cohorts and we estimated instantaneous mortality rates (Z) by fitting generalized linear models to adult cohorts for each site and sample year.  These mortality estimates were then used to perform a VPA on each species to incorporate missing individuals into our time series.  Spatial-temporal variation on fish recruitment was then analyzed across sites to document seasonal changes in recruitment as driven by hydrology.

Results/Conclusions

Our analysis revealed that larval and juvenile cohorts (<18mm SL) were largely underrepresented in our time series.  Therefore, the spatial-temporal distribution of juvenile recruits was most effected by the VPA.  In contrast to juveniles, adult recruits were adequately sampled and very few changes in adult recruit abundance were revealed.  Furthermore, the incorporation of missing juveniles revealed seasonal patterns of fish recruitment that were not apparent in the original time series.  Recruitment of juveniles began in April, earlier than originally predicted, with very little difference in population age-structure towards the beginning of the dry season (February).  The extent of this deviation between the original and reconstructed time series varied among species, sites, and wateryear.  For example, the pattern of recruitment changed for golden topminnows; however, the magnitude of this change had little impact on the estimated pattern of recruitment.  In contrast, the peak in recruitment for the bluefin killifish was shifted two months with a greater increase in juvenile density than originally predicted (>50% in some cases).  Our analysis demonstrated the importance of understanding the limitation of sampling gear and its implications in interpreting cyclical patterns within a time series.