Remote sensing of ocean colour allows to analyze spatial and temporal variability of chlorophyll-a concentration (Chl) in a subsurface layer and estimate primary production of phytoplankton (PP) at regional and global scales. However, improvement of remote sensing algorithms requires systematical validation of satellite data by in situ observations. In this work, we compare satellite-derived and field estimates of Chl and modeled PP in the western part of the Japan/East Sea (35-440 N, 130-1370 E) in spring 2004 and autumn 2005 and 2011. As satellite PP we use its VGPM estimates from Ocean Productivity database. Field PP is calculated on the base of in situ observations of Chl, assimilation value, diffuse attenuation coefficient and sea surface temperature using two models: VGPM and a model developed in V.I. Il'ichev Pacific Oceanological Institute (POI). Comparison shows fair correlation between satellite-derived and field data: r = 0.56-0.63 for modeled PP and r = 0.58 for Chl. Their averaged estimates for all stations of observation have small difference, which, however, may be about 1.5-1.7 times in some periods. On average, satellite Chl overestimates in situ observations, while satellite estimates of PP are less than PP calculated with the use of the model developed by POI but greater than PP calculated using VGPM. The difference between modeled satellite and field PP estimates is due to features of vertical distribution of Chl within the euphotic zone and the difference in values of assimilation number used in the models.