Dataset: Delaware Bay ACOE Salinity Data

Data curation was conducted on the ACOE datasets to prepare them for analysis (Howlader 2022). All null values were removed and time stamps were converted to the same time zone (UTC/GMT). Comparison datasets used include Station USGS 01482800 Delaware River at Reedy Island Jetty, New Castle County, station USGS 01463500 Delaware River at Trenton, NJ, and Brandywine Shoal Light, NOAA Station ID: 8555889.

Methods description from Howlader (2022):

"Potential instrument malfunctions and fouling problems were investigated in the ACOE dataset. Graphs of daily average salinity at an oyster bed station versus daily average salinity at Reedy were created and visually inspected to identify erroneous measurements caused by sensor malfunction or fouling. Visual inspection revealed that several hundred data points (0.8% of total data set) were substantially lower than most of the data and fell below the 95% prediction interval when a nonlinear model was fit to the data. To determine if these data points were the result of instrument malfunctions or fouling, these data points were plotted against time. Plots (not shown) showed sudden changes in salinity or gradual, linear increases in salinity measurements near the end of a deployment indicative of sensor fouling. To ensure that the anomalous data were due to instrument malfunctions and sensor fouling and not the result of sudden changes in freshwater flow or wind, both freshwater (from station USGS 01463500 Delaware River at Trenton, NJ) and wind (Brandywine Shoal Light, NOAA Station ID: 8555889) data were used. The data were formatted, null values were removed, and data was plotted on the same graph as salinity data to confirm that there were no significant high discharge events coincident with the anomalous salinity observations. Hourly wind data from June-December of 2012 2013, 2014, 2015, and 2018, were likewise plotted with salinity to confirm that there were no strong wind events during the times of the anomalous salinity observations. These analyses demonstrated that the salinity anomalies were most likely a result of sensor fouling and those observations were therefore removed from the dataset."

After removing anomalous data, the data shifted according to the tidal basis using the R package HelpersMG (Girondot, 2015) (from Howlader 2022): 

"Each of the ACOE and Haskin time series were shifted by the tidal phase lag and aligned in time to prepare the time series for analysis. Because the Reedy, Hope Creek, Arnolds, Cohansey, Shell Rock, and New Beds stations were located along a transect from upper to lower bay, and the tidal wave progresses from lower to upper bay, the peak in salinity during a tidal period occurred at different times at each station. Shifting the time series by the tidal phase lag corrected this difference. First, the "tide.info" function of the HelpersMG package in R was used to predict the tidal heights (m) at Reedy and the oyster bed stations. The time series of tidal heights were then graphed, and the time difference between peaks (i.e., the tidal phase lag) was calculated. This difference was used to adjust the time series of salinity at the oyster bed stations."