Temperature profoundly affects the physical, chemical, and biological attributes of lakes, and is influenced by several abiotic factors. Lake temperature modelling permits regional estimates of seasonal fish thermal habitat availability; however, this requires models that are accurate across large spatial scales. To address this, we fit a semi-mechanistic seasonal temperature-profile model (STM) to 369 morphometrically diverse North American lakes with data spanning 1971-2016. STM with a fixed-depth thermocline formula accurately modelled lake temperature (median pseudo <i>R</i><sup>2</sup>: 0.95, median lake-year-specific RMSE: 1.13 ºC). We used random forests to select candidate predictors, then used linear mixed-effects modelling, based on these predictors, to create empirical equations to predict STM parameters from lake-specific morphometric and climate measures. We tested the accuracy of our equations by predicting thermal profiles in 776 Ontario lakes, finding good agreement between predicted and observed temperatures (median lake-year-specific RMSE: 2.38 ºC) and stratification occurrence (91.7%). These findings enhance our understanding of the factors that influence lake temperatures and can be used to identify lake types and regions that may be especially susceptible to climate change.