Water quality measurements in San Francisco Bay by the U.S. Geological Survey, 1969–2015

The U.S. Geological Survey (USGS) maintains a place-based research program in San Francisco Bay (USA) that began in 1969 and continues, providing one of the longest records of water-quality measurements in a North American estuary. Constituents include salinity, temperature, light extinction coefficient, and concentrations of chlorophyll-a, dissolved oxygen, suspended particulate matter, nitrate, nitrite, ammonium, silicate, and phosphate. We describe the sampling program, analytical methods, structure of the data record, and how to access all measurements made from 1969 through 2015. We provide a summary of how these data have been used by USGS and other researchers to deepen understanding of how estuaries are structured and function differently from the river and ocean ecosystems they bridge.

The U.S. Geological Survey (USGS) maintains a place-based research program in San Francisco Bay (USA) that began in 1969 and continues, providing one of the longest records of water-quality measurements in a North American estuary. Constituents include salinity, temperature, light extinction coefficient, and concentrations of chlorophyll-a, dissolved oxygen, suspended particulate matter, nitrate, nitrite, ammonium, silicate, and phosphate. We describe the sampling program, analytical methods, structure of the data record, and how to access all measurements made from 1969 through 2015. We provide a summary of how these data have been used by USGS and other researchers to deepen understanding of how estuaries are structured and function differently from the river and ocean ecosystems they bridge.

Background & Summary
On April 10 and 11, 1969 oceanographers from the U.S. Geological Survey (USGS) conducted the first hydrographic research cruise along the salinity gradient of San Francisco Bay (SFB)-one of the largest estuaries on the west coast of the Americas. Although it was not the researchers' original intention, that survey launched an observational program that continues and expanded into a program of long-term ecosystem research that has contributed to the development of estuarine oceanography as a scientific discipline. In that era little was known about how estuaries function as transitional ecosystems between land and sea, where seawater and fresh water meet. Early USGS studies focused on: estuarine circulation where surface waters flow seaward over a landward-flowing bottom layer 1 ; sediment accumulation in an estuarine turbidity maximum 2 ; geomorphology 3 ; marsh vegetation and land forms 4 ; biogeochemistry of nutrients, oxygen and carbon 5 ; benthic invertebrate communities 6 ; urban pollution 7 , and its flushing by river inflows 8 . Over time the research expanded into new domains to measure, model and understand: tidal circulation and transport processes [9][10][11] ; human modifications of sediment supply 12 and geomorphology 13 ; sedimentwater nutrient exchanges 14,15 ; microbial biogeochemistry 16,17 ; bioaccumulation and cycling of contaminants including petroleum hydrocarbons 18 , metals 19 , mercury 20 , PCBs 21 , and selenium [22][23][24] ; disturbance by introduced species 25,26 ; ecosystem metabolism 27 ; phytoplankton communities 28 , productivity 29 , and regulating processes 30,31 ; zooplankton ecology 32 ; responses to climate variability 33,34 and climate change 35 . Central to this research is a core set of measurements repeated over time at a network of sampling sites (Fig. 1, Table 1) spaced along the estuarine salinity gradient. The San Francisco Bay system has been a useful place for studying estuarine dynamics because it includes two different estuary types. South Bay (stations 21-36) is an urbanized marine lagoon, and North Bay (stations 15-657) is the estuary of California's two largest rivers, the Sacramento and San Joaquin. Central Bay connects South and North Bays to each other and to the coastal Pacific Ocean (Fig. 1). Thus, one goal of USGS research has been to compare two different estuary types 36 . The data set includes measurements of salinity, temperature, suspended particulate matter, light penetration, dissolved oxygen, chlorophyll-a as an indicator of phytoplankton biomass, and concentrations of dissolved inorganic N, P and Si. The sampling program maps longitudinal and vertical distributions of these estuarine properties and captures their variability at seasonal, annual and decadal time scales.
The data described here were collected for one research purpose-to measure and understand how an estuarine ecosystem changes in response to human activities and the climate system. However, we recognized from the beginning of this effort that the data have value beyond this one purpose. We have encouraged and supported use of these data by others, and the diversity of applications of this data set has been both surprising and gratifying. We illustrate this diversity with examples of scientific articles ( Table 2 (available online only)) that used the data for purposes we could not have imagined, ranging across disciplines of archaeology, geochemistry, hydrodynamics, ecotoxicology, conservation biology, sediment dynamics, and biology of organisms from microbes to seabirds. Some of these publications were collaborations with visiting scientists, postdocs and graduate students. Others were done independently of USGS research. The collective knowledge accumulated from this research over decades has contributed to the global understanding of estuaries as ecosystems situated where land, ocean, atmosphere and people converge. Our purpose here is to widen accessibility of these data so their value continues to grow.

Methods
USGS water-quality studies in San Francisco Bay include two types of measurements: (1) laboratory analyses of discrete water samples collected aboard ship (chlorophyll-a, dissolved oxygen, suspended particulate matter, dissolved inorganic nutrients), and (2) shipboard or submersible sensors to measure salinity, temperature, chlorophyll fluorescence, dissolved oxygen, turbidity, and light attenuation. The analyses of discrete water samples were used to calibrate the chlorophyll fluorescence, dissolved oxygen, and turbidity sensors, with individual calibrations for each sampling cruise, and often separate calibrations for each bay region. Therefore, the data record includes both discrete measurements (e.g., Discrete_Chlorophyll-a) and sensor-based in-situ measurements (e.g., Calculated_Chlorophyll-a).
From 1969 through March 1987 the discrete water samples were collected by submersible pump that delivered bay water to a shipboard fluorometer, nephelometer, thermistor, and conductivity sensor 37 . Vertical profiles were obtained by lowering the pump to prescribed depths, typically 0, 2, 5, 10, 20 m. Since April 1987 the discrete water samples have been collected near surface (~1.5 m) by pump and~1 m above bottom with a Niskin bottle, and vertical profiles of salinity and temperature have been obtained with a Sea-Bird Electronics SBE-9 CTD. In 1988 turbidity, chlorophyll-a fluorescence, and photosynthetically active radiation (PAR) sensors were added to the CTD package, and in 1993 a dissolved oxygen sensor was added. In 2002 we started using a Sea-Bird Electronics SBE-9plus CTD. The individual sensors on the CTDs changed over time as new technologies emerged (see below). The CTD system is lowered through the water column at a rate o1 m s − 1 , collecting >24 samples/meter for Seabird sensors and >5 samples/meter for third party sensors. The values we report are averages over 1-m depth bins centered at the depth reported (i.e., CTD values are means of all measurements made 0.5 m above and 0.5 m below the reported depth).

Sampling design
This data set was acquired as a component of a research program whose goals evolved over time, so the frequency, spatial coverage, and makeup of water-quality measurements varied from year to year. We characterize sampling effort for five constituents, measured as the number of samples binned by station, month, and year (Fig. 2). Sampling effort was greatest in South Bay and during March and April, reflecting a key research objective to follow dynamics and ecological and biogeochemical consequences of the spring phytoplankton bloom 30 . The record reveals multi-year gaps in measurements of SPM and dissolved oxygen; that chlorophyll-a measurements first began in 1977; and that nutrient (e.g., phosphate) concentrations were measured less frequently than other constituents (Fig. 2). Sampling became more regular, and all constituents were measured each cruise starting in 1993 when this program became incorporated into the Regional Monitoring Program for Water Quality in San Francisco Bay (http://www.sfei.org/rmp).  Table 1.

Discrete_Chlorophyll-a
Chlorophyll-a measurements began in 1977, and methods changed as new instrumentation and widelyaccepted standard methods emerged. Samples were collected onto Gelman GFF (glass fiber) filters and pigments were extracted with 90% acetone. The absorbance of the extracts was measured with a Varian 635D spectrophotometer following Strickland and Parsons 38 . Chlorophyll-a concentrations were calculated using the SCOR-UNESCO trichromatic equations 39 . Beginning in 1983, we used Lorenzen's 40 spectrophotometric equations. In 1992 we began using a Hewlett Packard 8452A diode array spectrophotometer. In 1999 we began measuring chlorophyll-a concentrations fluorometrically using the acidification method on a Turner Designs TD-700 fluorometer calibrated with chlorophyll-a standard 41,42 . Since 2011 we have used a Turner Designs Trilogy fluorometer. After each method change we compared results of the older and newer approach on replicate samples across a range of chlorophylla concentrations to verify that bias was not introduced as new instruments and methods were used.

Calculated_Chlorophyll-a
Vertical profiles of chlorophyll-a were derived from calculated concentrations based on calibrations of an in-vivo fluorometer done each cruise to account for variability of phytoplankton species assemblages and the relationship between chlorophyll-a and fluorescence 43

Discrete_DO
Water samples for dissolved oxygen measurement were collected into 300-ml BOD bottles that were filled from the bottom and allowed to overflow at least 3 times their volume. Winkler reagents 38 were added immediately and bottles were stored capped with water in their cap-wells. In the laboratory, 100.2 ml of acidified sample was titrated manually following Carpenter 44 . Beginning in 1993, the samples were analyzed with a Metrohm 686 titroprocessor autotitrator 38 using the potentiometric titration method of Granéli and Granéli 45 . Potassium iodate standardization of the sodium thiosulfate was conducted (Knapp et al., 1991). In 2007 the autotitrator was replaced with a Metrohm Titrino 798.

Calculated_DO
In 1993 we added a Sea-Bird Electronics SBE-13 sensor to the CTD package to obtain vertical profiles of dissolved oxygen. The sensor was calibrated prior to each cruise with 100% and zero saturation endpoints, and additionally with Discrete_DO measurements (above) each cruise. In 2002 we began using a Sea-Bird Electronics SBE-43 oxygen sensor calibrated each cruise with Discrete_DO measurements.

Data record 1
The dataset includes 210,826 records, each representing a water sample from a unique date, station, and depth. All measurements made between 4/10/69 and 12/16/15 are available in one csv file (SanFranciscoBayWaterQualityData1969-2015v3.csv) uploaded to the USGS ScienceBase repository (Data Citation 1). An xml-formatted metadata file is also available at that repository.

Technical Validation
Results from each sampling cruise were examined carefully by at least two members of the research team to ensure that all values fell within expected ranges, to verify that calibration regressions were an acceptable basis for computing quantities from shipboard sensor measurements, to ensure completeness of each cruise data report, and to verify that values transcribed from field notes were accurate. The complete 1969-2015 data set was validated with three steps: (1) range tests to ensure that the measured values fell within ranges that are plausible and consistent with knowledge of San Francisco Bay and other estuaries; (2) pattern tests of time series of all measurements to ensure they followed plausible and understandable patterns of variability over time; (3) pattern tests of all measurements by sampling station to ensure they followed plausible and understandable spatial patterns. Sea-Bird Electronics sensors were calibrated annually by the manufacturer and have initial accuracies of: temperature = ± 0.001°C, conductivity = ± 0.0003 mS m − 1 , pressure = ± 0.015% of full range, dissolved oxygen = ± 2% of saturation (http://www.seabird.com). Li-Cor LI192 sensors were calibrated by the manufacturer and sensitivity is typically 4 μA per 1,000 μmol m − 2 s − 1 (https://www.licor.com). Cruise-specific calibrations of shipboard fluorometers, nephelometer/optical backscatter, and oxygen sensors yielded highly significant (P o10 − 16 ) linear relationships between all discrete and calculated concentrations of chlorophyll-a, SPM and DO (Fig. 3). Median absolute deviations between discrete and calculated concentrations were: 0.40 μg l − 1 for chlorophyll-a; 2.10 mg l − 1 for SPM; 0.10 mg l − 1 for DO. Linear regressions yielded residual standard errors between discrete and calculated concentrations of: 1.36 μg l − 1 for chlorophyll-a; 8.2 mg l − 1 for SPM; 0.16 mg l − 1 for DO (Fig. 3).
Discrete chlorophyll-a values are mean concentrations in replicate (2, 3, or 4) aliquots from each sample. If the replicate results differed by more than 10% of their mean the results were not included in the data set. The mean coefficient of variation between replicate aliquots from 3,564 chlorophyll-a samples collected between 2005 and 2013 was 2.4%. Agreement between all replicates was within the recommended guideline for the method: >90% of the coefficients of variation (CV) between samples are o5% (ref. 42). Discrete suspended particulate matter precision was 1%-10%. Analytical precision of the potentiometric DO method is o0.3% (ref. 45).  (Fig. 4). Ammonium was analysed using different methodologies, but replicate samples analysed in each laboratory confirmed consistency between them (Fig. 4). The SFSU laboratory reported detection limits as 0.05 μM for ammonium, nitrite, nitrate+nitrite and phosphate, and 0.1 μM for silicate.
As a preliminary step in the 2014 transition from SFSU to the USGS National Water Quality Laboratory (USGS-NWQL), we collected triplicate water samples along the salinity gradient of San Francisco Bay to compare analyses by SFSU, USGS-NWQL, and the Chesapeake Biological Laboratory (CBL) as an independent laboratory. We continued analysis of duplicate samples by USGS-NWQL and CBL through 2015. We compare results of the three laboratories in Fig. 5. The USGS-NWQL has the following minimum reporting levels: 0.7 μM ammonium, 0.1 μM nitrite and phosphate, 0.7 μM nitrate+nitrite when totalo10 μM, 2.9 μM when total >10 μM, and 1.0 μM silicate. Replicate samples are intermittently analysed by USGS-NWQL to measure precision. Replicates have mean coefficients of variationo5% for all nutrients: nitrite = 3.1%, nitrate+nitrite = 2.3%, ammonium = 4.6%, phosphate = 1.6%, and silicate = 0.01%.
Although nutrient methods changed over time, routine analyses of blanks and standards confirmed that methods changes did not reduce analytical precision or accuracy.

Usage Notes
This Data Descriptor identifies a csv file that contains the complete record of USGS water-quality measurements made in San Francisco Bay from 1969-2015. Users may prefer to access the data from our project web page that includes a database from which queries can be made to select and download subsets of the full data record (https://sfbay.wr.usgs.gov/access/wqdata/index.html ). This web page also provides visual displays of water-quality spatial variability for each sampling cruise, and more detail about the research project and team members.