Predictive accuracy of temperature‐nitrate relationships for the oceanic mixed layer of the New Zealand region

Abstract

Nitrate concentrations are a major factor controlling phytoplankton growth, hence the recent interest in using remotely sensed sea surface temperature (SST) and chlorophyll concentrations (Chla) to infer nitrate concentrations and substantially improve spatiotemporal estimates of nitrate in the surface ocean. Regression models which predict mixed‐layer nitrate concentrations as a function of temperature and climatological salinity are derived for the subtropical and subantarctic waters of the New Zealand region (30°–50°S, 154°E–160°W). These models are then validated using independent in situ measurements of temperature and nitrate concentrations and remotely sensed SST and Chla. Root mean square (RMS) nitrate prediction errors vary with water mass and exhibit seasonally dependent biases. RMS errors range from 0.8 to 1.8 μM in subtropical waters, 1.6 to 1.9 μM in the Subtropical Front, and 1.4 to 2.5 μM in subantarctic waters, depending on the spatiotemporal sampling characteristics of validation data sets. Prediction errors are correlated with observed chlorophyll concentrations, and a linear chlorophyll correction reduces seasonally dependent prediction biases significantly. Nitrate prediction errors for the New Zealand region are comparable with nitrate prediction errors reported for the North Atlantic and Equatorial and North Pacific, and the regression models give a substantially better description of the seasonal variation of nitrate in the New Zealand region than an existing nitrate climatology. A comparison of predicted nitrate‐depletion temperatures with other published studies highlights the importance of detailed regional validation of temperature‐nitrate regression models.