Accurate forecasts of the number of larval sea lamprey (Petromyzon marinus) within a stream that will enter into metamorphosis are critical to currently used methods for allocating lampricide treatments among streams in the Great Lakes basin. To improve our ability to predict metamorphosis we used a mark-recapture technique, involving the marking of individual larval lamprey with sequentially coded wire tags, to combine information regarding individual and stream level parameters collected in year t, with direct observations of metamorphic outcome of lamprey recaptured in year t 1. We used these data to fit predictive models of metamorphosis. The best model demonstrated excellent predictive capabilities and highlighted the importance of weight, age, larval density, stream temperature and geographic location in determining when individual lamprey are likely to transform. While this model was informative, it required data whose measures are not practical to obtain routinely during the larval sea lamprey assessment program. A second model, limited to data inputs that can be easily obtained, was developed and included length of larvae the fall prior to metamorphosis, stream latitude and longitude, drainage area, average larval density in type-2 habitat, and stream lamprey production category (a measure of the regularity with which treatments are required). This model accurately predicted metamorphosis 20% more often than current models of metamorphosis; however, we recommend further validation on an independent set of streams before adoption by the Great Lakes Fishery Commission for ranking streams.