A method is presented for determining the concentrations of open water and of several ice types using multichannel satellite passive microwave data. The method uses the Kalman filter and provides the ''best fit'' to a time series of data. A crucial element of the procedure is a physical model of how the concentrations of ice types change with time in response to freezing, melting, aging of one ice type to another, and creation of open water by divergence of the ice cover. A measurement model relates the state of the ice cover to the multivariate microwave data. The procedure offers three distinct advantages over algorithms that interpret separately data from each instant in time: it provides a framework for incorporating additional data into the diagnosis of ice concentrations, it takes into account the known uncertainty in the microwave observations and the pure type signatures, and it allows the resolution of ice types with ambiguous signatures. Two examples are presented which make use of scanning multichannel microwave radiometer data and surface temperature and ice velocity data from drifting buoys to estimate the concentrations of open water, first-year, second-year, and older multiyear ice for a Langrangian region of ice. Two other examples include melt ponds in place of second-year ice. Some of the parameters in the physical model (melt rates) and in the measurement model (signature of second-year ice or of frozen melt ponds) are unknown. Reasonable, but arbitrary, values of the unknown parameters are used in the examples. ¿ American Geophysical Union 1989