Given the importance of nitrate in sustaining high primary production and fishery yields in eastern boundary current ecosystems, it is desirable to know the amounts of this nutrient reaching the euphotic zone through the upwelling process. Because such measurements are not routinely available, we developed predictive models of water-column (0–200 m) nitrate based on temperature for a region of the California Current System (30–47°N) within 50 km from the coast. Prediction was done using generalized additive models based on a compilation of 37,607 observations collected over the period 1959–2004 and validated with a separate set of 6430 observations for the period 2005–2011. A temperature-only model had relatively high explanatory power (explained deviance, D2 = 71.6%) but contained important depth, latitudinal, and seasonal biases. A model incorporating salinity in addition to temperature (D2 = 91.2%) corrected for the latitudinal and depth biases but not the seasonal bias. The best model included oxygen, temperature, and salinity (D2 = 96.6%) and adequately predicted nitrate temporal behavior at two widely separated locations (44°39.1′N and 32°54.6′N) with slight or no bias [root-mean-square error (RMSE) = 2.39 and 0.40 µM, respectively). For situations when only temperature is available, a model including depth, month, and latitude as proxy covariates corrects some of the biases, but it had lower predictive skill (RMSE = 2.50 and 5.22 μM, respectively). The results of this study have applications for the proxy derivation of nitrate availability for primary producers (phytoplankton, macroalgae) in upwelling regions and for biogeochemical and ecosystem modeling studies.