Coastal upwellings are common in the Great Lakes but have lacked enumeration and systematic classification of spatial extent, frequency, duration, and magnitude. Near real-time sea surface temperature (SST) images derived from the Advanced Very High Resolution Radiometer (AVHRR) provide indices of upwelling events, but visual inspection of daily images can be tedious. Moreover, the definition of what constitutes an upwelling from AVHRR data is subjective. We developed a semi-automated method to classify upwellings during the period of thermal stratification using daily, cloud-free surface temperature charts from AVHRR SST data. Then we statistically evaluated the location, frequency, magnitude, extent, and duration of upwelling events in Lake Michigan from 1992–2000. Further, we analyzed meteorological data from the National Data Buoy Center buoys in an attempt to improve the reliability of the classification and to provide a means for future forecast of coastal upwelling. Although variable, upwelling events along the western shoreline were preceded by 4 days of southerly and west-to-north-westerly winds, while upwelling events occurring along the eastern shore were preceded by 4 days of northerly winds. Probability of an upwelling event occurring was a function of the direction-weighted wind speed, reaching a 100% probability at direction weighted wind speeds of 11 m s⁻¹ for the western shore. Probability of an upwelling occurrence along the east coast reached 73% at 11 m s⁻¹ and 100% at 13 m s⁻¹. Continuous measurements of wind data with a sufficient temporal resolution are required during the entire upwelling season to improve the predictability of upwellings.