Monte Carlo techniques are used to examine a potential error in the interpretation of satellite measurements for broken cloud fields. Solar radiation reflected from clouds and scattered by the atmosphere will enhance observed visible radiances, when clouds lie on the antisolar side of a cloud-free satellite field of view (FOV). Results of the simulation show that the assumption of a plane-parallel, cloud-free atmosphere in atmospheric correction in a common retrieval algorithm introduces a bias in the inferred surface reflectivity and biological concentrations in the ocean from space. The biases arise because the radiation scattered from the adjacent cloud into the FOV is not correctly determined by calculations with a one-dimensional radiative transfer model assuming a cloud-free atmosphere. It is a function of the optical properties of the atmosphere, the distance of the adjacent cloud from fields of view, the solar zenith angle, etc. Optically thicker and closer clouds to the fields of view lead to a larger enhanced radiance. The bias is up to more than 10% at a nadir viewing angle for a cloud optical thickness of 9, a solar zenith angle of 30¿, and a surface albedo of 0.1. The enhanced radiance overestimates retrieved surface albedo by only 0.01, which may be negligible for a study of the surface radiation budget. The bias is, however, significant in the estimation of biological pigment concentration in the ocean. Because water-leaving radiance is only a small part of radiance detected at a spaceborne sensor, at most 10--20%, a small error in measured radiance causes a large error in retrieved pigment concentrations. Accurate atmospheric correction therefore is needed for the retrieval of phytoplankton concentrations from space. Assuming an improvement of satellite sensor technology in the near future, the issue shown in the present study will become significant because real clouds are often broken. ¿ 2000 American Geophysical Union