San Francisco Estuary Generalized Additive Model:

A generalized additive model (GAM) is an extension of generalized linear models (GLM) that more accurately model complex relationships between the predictor variables (variables that can affect the response variables, or be used to predict or explain them) and response variables (variables of interest that are measured or observed). This makes them especially powerful in environmental modelling, where the relationships between variables are often non-linear and non-additive. This GAM uses a single spline, the continuous day of the water year (where October 1st is day 1 and September 30th is day 365), to generate the curve. The upper limit for the knots in this model is 12 times the number of years in the selected date range. Knots are points along the spline between which the curve must adhere to a single function. At each knot, the behavior of the curve may change. See Beck and others (2022) for in-depth information about the GAM approach to water quality trend analysis.

This data visualization application contains data from two independent data sources. The first is the United States Geological Survey’s (USGS) Water Quality of San Francisco Bay Research and Monitoring Project that has been measuring changes in water quality for over 50 years (Schraga and Cloern, 2017). The data is publicly available at the San Francisco Bay Water Quality Database. For the past five decades the USGS has conducted ship-based measurements of water quality along a 145-kilometer-deep water transect that spans the length of the entire system from ocean to inland delta. USGS scientists sample the full salinity gradient on monthly cruises, measuring the basic elements of water quality that define the sustainability of the bay as habitat for fish and organisms of the lower trophic levels. Using both oceanographic sensors and discrete water collection, the USGS measures salinity, temperature, light extinction coefficient, chlorophyll-α, dissolved oxygen, suspended particulate matter, nitrate, nitrite, ammonium, silicate, and phosphate.

The second dataset is the product of the Interagency Ecological Program’s Environmental Monitoring Program (EMP), for which the discrete water quality data can be sourced from the Environmental Data Initiative. Since 1975 they have monitored discrete water quality and nutrients in the upper San Francisco Estuary. The objectives of the EMP are to obtain consistent and accurate monthly data at established monitoring stations, provide and document information necessary to achieve compliance with salinity, flow, and dissolved oxygen standards, and to report this information for the purpose of management and conservation of the upper San Francisco Estuary (Battey and Perry, 2024). The EMP dataset is only represented on this generalized additive model page as an option for users to view GAMs at more stations throughout the estuary. The surface, profile, and data exploration pages reflect only the USGS dataset.

This web application is an extension of the original San Francisco Estuary Institute Nutrient Management System's San Francisco Bay Trends webtool.

For the purposes of this application, and only on the data fed directly into the GAM, a simple interpolation is performed for any gap measuring more than 90 days. In this case, the average value for that measurement, station, and month combination is inserted.