Hut Point and Cape Evans-1: The Cape Evans site was equiped with a benthic mooring suspended above the seafloor (~ 30 m depth) by a subsurface buoy. At Hut Point, the mooring was suspended from the surface by a steel cable from a wooden hut through a hole in the sea ice. Bottom depth was > 200 m.

All sensors were deployed at 20 m depth. The moorings were instrumented with a suite of sensors to record time series of temperature, salinity, pH, and tide. Temperature and salinity were measured using a non-pumped conductivity-temperature (CT) MicroCAT sensor (SBE-37 SM; Sea-Bird Electronics) that sampled at 5-min intervals. pH was measured using an autonomous data logger based on a Honeywell Durafet® pH sensor (Martz et al., 2010) and sampled at 1-hr intervals.

Calibration of pH sensors required a discrete water sample collected in situ. This single point calibration approach is justified when the sensor obeys the Nernst equation and the temperature component of the standard potential has been previously characterized; both of which have been repeatedly demonstrated for these sensors (Martz et al. 2010). The water sample was collected adjacent to the sensor by SCUBA divers (Cape Evans) or by lowering a 5 L Niskin sampling bottle from the surface (Hut Point) prior to retrieval. From this sample, a 500 mL water sample was returned to the laboratory for CO2 analysis modified from Standard Operating Procedures (SOP) for spectrophotometric pH (SOP 6b) and Total Alkalinity (TA, SOP 3b) (Dickson et al., 2007) as reported in Fangue et al. (2010). In situ pH was then calculated using CO2calc (Robbins et al., 2010) using the constants of Mehrbach et al. (1973) as refit by Dickson and Millero (1987). Due to the calibration approach used, sensor accuracy depends mostly upon collection of a representative discrete sample. Based on experience, there is an expectation that the data presented here accurately represent pH variability with a finite yet unquantified error in accuracy dominated by sampling errors. Past experience suggests that sampling errors lead to vicarious calibration errors of ~0.01 pH or less. Second order errors due to extending the fit of temperature dependent equilibrium constants in CO2calc and temperature dependent sensor calibration coefficients for the SeaFET sensor, both fit to data above zero, introduces additional unquantified error; yet this error is most likely smaller than the aforementioned discrete sampling error (Martz et al., 2010).

New Harbor: Benthic moorings was suspended above the seafloor (~ 30 m depth) by a subsurface buoy. All sensors were deployed at 20 m depth, with the exception of pressure sensors, which were attached to the mooring anchor. Mooring was instrumented with a suite of sensors to record time series of temperature, salinity, pH, and tide. Temperature and salinity were measured using a non-pumped conductivity-temperature (CT) MicroCAT sensor (SBE-37 SM; Sea-Bird Electronics) that sampled at 5-min intervals. Tidal height was measured using water level loggers (Hobo U20-001-03-Ti; Onset) that recorded water pressure at 10-min intervals.

Jetty and Cape Evans-2: All methods are described in Kapsenberg et al. (2015). Data were collected using an autonomous SeaFET pH sensor containing Honeywell DuraFET electrodes (Martz et al., 2010). Sensor depth was 18 m with ~27 m bottom depth. Sensors sampled on a 2-hour frequency.

Conversion from voltage to pH (on a total scale) was performed using a single discrete calibration sample collected via SCUBA using a 5 L GO-FLO sampling bottle. Sample collection was conducted within the first two weeks of sensor deployment, after sensor conditioning to seawater, in-situ. Samples were preserved with saturated mercuric chloride Standard Operating Procedure (SOP) 1 (Dickson et al., 2007) and analyzed for spectrophotometric pH (total scale, at 25 degrees Celsius) and total alkalinity following SOP 6b and 3b (Dickson et al. 2007). Sample salinity was measured using a calibrated YSI 3100 Conductivity Instrument. In-situ pH was calculated using the program CO2Calc [Version 1.0.1, 2010, U.S. Geological Survey] using SeaFET temperature recorded at the time of sample collection.

The combined standard uncertainty associated with the pH measurement of the calibration sample is ± 0.026 pH units (Kapsenberg et al., 2015). The sources of error are use of unpurified m-cresol dye (0.02, Liu et al., 2011), spatio-temporal mismatch of the calibration sample (± 0.015, Bresnahan et al., 2014), user differences (± 0.006), and calibration of the SeaFET thermistor (± 0.005).

SeaFET thermistors were not individually calibrated resulting in a maximum estimated temperature offset of ~0.3 degrees Celsius.