We compared satellite and ground-based observations of tropospheric NO₂ to test whether satellite observations could successfully detect the behavior of tropospheric NO₂. The satellite data were tropospheric NO₂ vertical column density (VCD) derived from Global Ozone Monitoring Experiment (GOME) spectrometer measurements (hereafter GOME-NO₂), and the ground-based data were surface NO₂ volume mixing ratio (VMR) observed by the network of air-quality monitoring stations in Japan. The analysis was performed in the Tokyo region (the Kanto Plain) from January 1996 to June 2003. A strong correlation between GOME-NO₂ and the surface VMR was observed, with the two quantities showing similar seasonal variation of maximum in winter, minimum in summer. This provided initial evidence that GOME was successful in observing the behavior of NO₂ near the surface level in the Tokyo region. We performed a more rigorous comparison in which the surface NO₂ VMR was scaled to the tropospheric VCD using vertical NO₂ VMR profiles, which were calculated using the chemical transport model CMAQ/REAS. This second comparison indicated that the GOME observations represent the behavior of NO₂ more closely at the relatively unpolluted stations than at the highly polluted stations in the network of air-quality monitoring. This tendency was thought to result from the horizontal heterogeneity within a GOME footprint. Comparison with a previous study in northern Italy showed that the GOME-NO₂ measurements over Tokyo tended to be smaller than those over northern Italy. Because Tokyo is located in a coastal land region with a gulf, areas of ocean intruding into the GOME pixels could lower the observed GOME-NO₂. The pollution in Tokyo is so spatially concentrated that the rural regions contaminating GOME pixels could also reduce the observed NO₂ concentration from its true spatially resolved value.