Dataset: Computed surface partial pressure and air-water flux of carbon dioxide

pCO2 and the air–water CO2 flux were computed at monthly resolution from 1998 to 2018 from tidal fresh to polyhaline waters.

Detailed description of processing of pH, temperature, salinity, and dissolved oxygen data

Data from the beginning of the regular monitoring program in 1985 through 2018 were downloaded from the Chesapeake Bay Program’s Data Hub (see related datasets), with the original intention of producing a 33-year analysis of the air–water CO2 flux. However, a close examination of the pH data led us to question some measurements in the early part of the record. We found a dramatic difference in long-term trends between stations measured by institutions in the state of Virginia and stations measured by the state of Maryland, particularly from late spring to early fall. The boundary between the station groups runs east–west within the mesohaline portion of the bay in segment 5, where the Potomac River estuary intersects the mainstem bay. The boundary separates strong negative linear trends to the south (Virginia stations) from neutral and weakly positive linear trends to the north (Maryland stations). A closer examination of summer data at two stations straddling the boundary between the states (CB5.3 and CB5.4) revealed a distinct separation of the data before 1998 and convergence thereafter. The Virginia measurement protocols changed around this time (Chesapeake Bay Program, 2012). Specifically, Old Dominion University (ODU) took over sampling from the Virginia Institute of Marine Science (VIMS) in 1996. VIMS measured pH by first collecting a water sample and then inserting a pH probe into that sample whereas ODU and Maryland measured pH with an in situ probe. In 1997, ODU switched from using a Hydrolab Surveyor II sonde to a YSI 6000 sonde. It seems likely that the data quality of the Virginia measurements improved greatly after 1997 and we thus chose to limit our analysis to 1998–2018. For this shorter time period, the four parameters were measured exclusively with in situ probes from YSI® and Hydrolab®, and pH was reported on the NBS scale. The specific models used are described in Chesapeake Bay Program (2012) and probe accuracies can be found in the respective user manuals. As an example, the user manual for the YSI 6-series multiparameter water quality sondes provide accuracies of 0.15 ºC for temperature, 1% or 0.1 part per thousand (whichever is greater) for salinity, 0.2 for pH, and 2% or 6.25 mmol m–3 (whichever is greater) for dissolved oxygen measured with a polarographic sensor and half these values for an optical sensor.

Surface time series of pH, temperature, salinity, and dissolved oxygen were subset from the full Chesapeake Bay Program database depth profiles as the shallowest available measurement in the upper 2 m (36% at 0.5 m, 50% at 1 m, and 14% at 2 m). The native temporal resolution of the measurements was preserved in the surface subset. All data flagged with “Problem” by the Chesapeake Bay Program were removed (<0.1% of data for temperature and salinity, 0.6% for pH, and 0.4% for oxygen). The oxygen supersaturation was computed as the oxygen concentration minus the saturation concentration (computed from temperature and salinity, Garcia & Gordon, 1992). Surface values of pH, temperature, salinity, and dissolved oxygen supersaturation were quality controlled by removing gross outliers in three steps. First, pH was converted to hydrogen ion concentration. Second, the interquartile range for each variable for the whole 1998­­–­2018 data set was computed. Third, all data 20 interquartile ranges above or below the 75th and 25thpercentiles, respectively, were removed. pH was the only variable affected by this filtering, resulting in removal of 0.26% of the pH data. pH was subjected to a second round of quality control following the same procedure, but this time on a station-by-station basis using 10 interquartile ranges as the threshold, leading to removal of 0.5 to 3.5% of the pH data from 18 of the 33 stations. 

Because pH is a non-linear function of hydrogen ion, any time the pH data were manipulated in the analyses that follow (e.g., calculation of monthly averages), pH was converted to hydrogen ion concentration, the hydrogen ion data were processed, and the results were converted back to pH.