Subpixel heterogeneity remains a key issue in the estimation of land parameters using satellite passive microwave sensors; the scales of spatial variability on land are typically much smaller than sensor footprints (tens of km). Disaggregation is a necessary component of any successful assimilation or retrieval scheme attempting to exploit satellite passive microwave observations to estimate parameters at the local scale. This paper quantifies the similarity between ground-based brightness and satellite brightness observations at 19 and 37 GHz for Arctic tundra on the North Slope of Alaska, identifying and quantifying sources of the differences. To the extent that this very homogeneous area represents a limiting case, the impact of subpixel heterogeneity in less homogeneous areas may be gauged. The ground-based radiobrightness observations were collected during the Radiobrightness Energy Balance Experiment 3 (REBEX-3) conducted on the North Slope in 1994--1995. A comparison was made of 381 days of brightness observations from a tower-mounted Special Sensor Microwave Imager (SSM/I) simulator representing the full range of annual conditions with coincident satellite SSM/I observations. Issues such as instrument stability, the effects of atmospheric radiative transfer, and consistency of satellite pixel locations are considered. Linear correlations between tower-based and SSM/I brightness observations of 0.93, 0.94, 0.93, and 0.92 were observed for the 19V, 19H, 37V, and 37H channels, respectively. Footprint sizes were 2 ¿ 4 m for the tower-based observations and 43 ¿ 69 km for the resampled SSM/I observations. Atmospheric, topographic, and time-of-observation effects can account for the differences between the best fit lines and the 1:1 lines, with calibration errors accounting for the residual differences.