Understanding changes in the ocean carbonate system is central to understanding ocean and coastal acidification and the effects these phenomena will have in the future. To create a more complete overview of the recent history of the carbonate system in the nearshore Northeastern United States, several recently published or in-development statistical models have used simple ocean chemistry parameters of salinity, temperature dissolved oxygen, and nitrate, or these variables plus the addition of other input parameters: sea surface temperature, chlorophyll a, sea surface height, bathymetry, and atmospheric pCO2 to generate estimates of dissolved inorganic carbon (DIC) and total alkalinity (TA). Both a Random Forest Regression model and a multiple linear regression model predicting carbonate chemistry parameters was tested for accuracy in predicting fugacity of CO2 (fCO2) by comparing them with the publicly available fCO2 data from the Surface Ocean CO2 Atlas (SOCAT) database. Comparisons revealed a bias by the models to overestimate fCO2, which was also observed when comparing the SOCAT dataset to collocated discrete observations. To resolve these biases in fCO2, a correction was fitted to the modeled datasets. This investigation suggests that models that accurately predict carbonate parameters of DIC and TA, may be limited in their ability to reproduce fCO2 conditions in coastal areas without correction. This study suggests that extrapolating ocean carbonate system models based on parameters outside their intended uses should be considered for their potential limitations.