Date Awarded


Document Type


Degree Name

Doctor of Philosophy (Ph.D.)


Virginia Institute of Marine Science


Herbert M. Austin

Committee Member

David Evans


Long term trends in juvenile recruitment of oyster, bluecrab, and 24 species of finfish in a large temperate estuary (lower Chesapeake Bay, USA) are coherent across the three major tributaries (the Virginia rivers James, York, and Rappahannock). The driving force for these long term trends is geographically large in scale. Anomalous winters in the mid 1970's, with the warmest years on record followed immediately by the coldest, caused a severe perturbation in population dynamics. The extreme conditions caused the system to shift, with recruitment patterns following temperatures by a one year lag. Following this anomalous episode, smoothed mean winter water temperatures have increased steadily from 1979 until 1995 &\rm (3.9{lcub}-{rcub}5.7\sp\circ C,& long-term T = &\rm 4.6\sp\circ C),& closely followed (with zero lag) by the first principal component (PC) from each set of smoothed biological indices. Annual indices of juvenile abundance (means of log-transformed catch per unit effort) were calculated by river for the James, York and Rappahannock Rivers. Two collections of different temporal lengths are analysed, oyster, bluecrab and 14 species of finfish (1965-1995) and 17 species of finfish (1980-1995), with an overlap of seven species of finfish. The indices are smoothed by loess (locally weighted scatterplot smoother), and analyses are performed on the indices, the loess-smoothed indices, and the residuals. Principal components analysis (PCA) on the indices indicates coherence in the population fluctuations by a relatively small number of PC's. Weak relationships are found in the unsmoothed indices and the residuals. Smoothed long-term trends eliminate much of the noise, thus exposing the underlying behavior of populations. PCA on the loess-smoothed indices were remarkably cohesive, with only three or four PC's significant in each of the six treatments, accounting for 93 to 98% of total variance, with 44 to 70% in PC#1. Correlations on the first PC's of the loess-smoothed indices, between rivers, within and between surveys, yielded 87-99% agreement; such coherence indicates the underlying causal factor is geographically broad. Cross correlations and scatterplots of smoothed winter water temperature and PC#1 identify the lag during the perturbation years.



© The Author