The Columbia River plume (CRP) is a buoyant plume that influences the Oregon and Washington shelf with the delivery of freshwater, silicic acid, trace metals, and particulate and dissolved organic matter. The highly dynamic plume contains submesoscale features that have an impact on the chemistry, biology, and transport of water and material offshore. Bio-optical classification of the larger plume water mass has confirmed seasonal and annual flow patterns but has not described the internal structure of the plume in a biogeochemically relevant way, as there were no in situ data to validate classification. The objectives of this study were to define water types statistically within the CRP using in situ measurements of biogeochemically and bio-optically relevant variables, to build a training data set from these water types, and to apply this training data set to 250 m resolution Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua imagery from an oceanographically downwelling and upwelling period to predictively discriminate water masses within the plume. This study’s classification technique was effective at predicting water types in the CRP. The three-variable input matrix (temperature, salinity, and chlorophyll a fluorescence) performed better than the two-variable input matrix (temperature and salinity) at distinguishing fine-scale structure within the plume at the river mouth. Retentive features such as the plume bulge and eddies were observed at the river mouth and on the Washington shelf. This classification approach was limited to the available continuous variables measured by shipboard, mooring, and satellite sensors. Two new classification methods are proposed that build on the framework of the classifier described here.