FY 2021

Multiple linear regression models for reconstructing and exploring processes controlling the carbonate system of the northeast US from basic hydrographic data

McGarry, K., S.A. Siedlecki, J. Salisbury, and S.R. Alin

J. Geophys. Res., 126(2), e2020JC016480, doi: 10.1029/2020JC016480, View online (2021)

In the coastal ocean, local carbonate system variability is determined by the interaction between ocean acidification and local processes. Sporadic observations indicate that biological metabolism, river input, and water mass mixing are dominant local processes driving carbonate system variability in northeast US shelf waters. These processes are also reflected in the variability of observed temperature (T), salinity (S), oxygen concentration (O₂), and nitrate concentration (NO₃⁻). Therefore, regionally specific empirical models can be developed, which relate carbonate system parameters to a combination of basic hydrographic parameters. Here, we develop multiple linear regression models that represent the processes that drive carbonate system variability in the Mid‐Atlantic Bight and Gulf of Maine using observations obtained on three hydrographic surveys in summers between 2007 and 2015. The empirical model equations reveal the observation‐based relationships between carbonate parameters and basic hydrographic variables. Unlike other regions where empirical models have been developed, salinity appears in all models. T is the most important parameter for predicting aragonite saturation state (Ω_AR), while S and O₂ are most important for predicting pH on total scale (pH_T). The basic hydrographic variables explain over 98% of the variability in total alkalinity (TA), dissolved inorganic carbon (DIC), and Ω_AR and 89% of the variability in pH_T in the calibration data. We recommend applying models that depend on T, S, O₂, and NO₃⁻ as predictors, which reproduce TA and DIC with R² > 0.97, Ω_AR with R2 > 0.93, and pH_T with R² > 0.77, to reconstruct carbonate system parameters in the region.

Plain Language Summary. Carbon dioxide released to the atmosphere by humans can adversely impact aquatic ecosystems, so it is crucial that we understand the current state of carbon variables and anticipate future conditions. Carbon cycling in the coastal ocean is the result of the interaction of physical and biological processes that occur on multiple time and space scales. Sparse sampling of carbon variables presents challenges to our understanding of carbon cycling in the coastal ocean. Other seawater properties measured more frequently with better spatial coverage, including temperature, salinity, oxygen concentration, and nitrate concentration, can be used in combination to estimate carbon variables. In this study, we rely on measurements from cruises along the northeast US shelf to develop equations to predict carbon variables from other seawater properties. These equations can be used to fill gaps in observations and help incorporate observations into ocean models. The statistical relationships between carbon variables and other seawater properties identified here vary depending on the region, because a balance of different processes is important in each region. On the northeast US shelf, salinity emerges as an important predictor for all explored carbon variables.