PREDICTING THE SALINITY HISTORY OF OYSTERS IN DELAWARE BAY USING OBSERVING SYSTEMS DATA AND NONLINEAR REGRESSION
Files
(RESTRICTED ACCESS)
Publication or External Link
Date
Authors
Advisor
Citation
DRUM DOI
Abstract
Salinity is a major environmental factor that influences the population dynamics of fish and shellfish along coasts and estuaries, yet methods for predicting the salinity history at specific sampling stations are not widely available. The specific aim of this research was to predict the history of salinity experienced by juvenile and adult oysters (Crassostrea virginica) collected at sampling stations in Delaware Bay as part of the Selection along Estuarine Gradients in Oysters (SEGO) project. To do so, empirical relationships were created to predict salinity at five oyster bed stations using observing systems data and then applied to construct indices of salinity exposure over an oyster’s lifetime. The desired accuracy was +/- 2 psu. Three independent sources of salinity data were used in conjunction with observing systems data to construct and validate the predictive relationships. Observing systems data from the USGS station at Reedy Island Jetty and continuous near-bottom measurements taken by the U.S. Army Corps of Engineers (ACOE) from 2012-2015 and 2018 were employed to fit nonlinear empirical models at each station. Haskin Shellfish Research Laboratory (Haskin) data were used to evaluate model fit, then ACOE data from 2018 (withheld from model fitting in the validation analysis) and SEGO data from 2021 were used to validate models. The best-fitting models for predicting salinity at the oyster bed stations given the salinity at Reedy Island Jetty were logarithmic in form. The root mean square error (RMSE) of the models ranged from 1.3 to 1.6 psu when model predictions were compared with Haskin data, 0.5 to 1.5 when compared with ACOE data, and 0.6 to 0.8 when compared with SEGO data. All of these models were within the desired accuracy range. Results demonstrate that observing systems data can be used for predicting salinity within +/- 2 psu at oyster bed stations within 39 km in upper Delaware Bay. When these models were applied to estimate low salinity exposure of 2-year-old oysters via the metric of consecutive days below 5 psu, the indices suggested that there could be as much as a 42-day difference in low salinity exposure for oysters at stations 31 km apart. This study helps further our understanding of the salt distribution in Delaware Bay as well as the effect of low-salinity stress on the life cycle and genetic differentiation of oysters. In addition, the approach of using observing systems data to predict salinity could be applied to advance understanding of salt distribution and the effect of low salinity exposure on living resources in other estuaries.