The success of the Updateable Model Output Statistics (UMOS) system developed and implemented by the Meteorological Research Branch and the Canadian Meteorological Centre of the Meteorological Service of Canada has inspired the adaptation of this system to the forecast of surface winds on the Great Lakes. Application of a MOS technique to this forecast problem is hampered by the removal every fall of the moored buoys supplying the marine forecaster with the in-situ observations most representative of the off-shore marine environment. UMOS accomodates the limited observational database by quickly adjusting to new model implementations. The design of the system also facilitates the pooling of the samples from all of the buoys in a given lake, in the expectation that the buoy responses in the off-shore marine environment would be much less idiosyncratic than terrestrial observing sites. This amounts to a sacrifice of the resolution of some finer scale mesoscale features for a more stable solution in the coarser mesoscale and synoptic regimes. The progress of this adaptation is the subject of this paper, in which UMOS forecasts for the marine winds on the Great Lakes will be compared with the model output, as well as other operational guidance available to the marine forecaster, for the spring and first half of the summer of this year. In addition some of the strengths and limitations of the system as a development tool, which became apparent during the course of this investigation, will be discussed.