Climate change is already affecting the Canadian climate with increases in air temperature, evapotranspiration rate and risk of rainstorms and drought. Potato, a major crop in Prince Edward Island (PEI), Canada, contributed 10.8% of the provinces GDP in 2018. However, potato yields are very sensitive to these changes in weather and an increase of 1?1.4�C could reduce yields by 18-32%. The ability to predict yield before harvest is invaluable information for both decision makers in charge of national food security strategies, and farmers who need to plan their on-farm actions to maximize yield under changing conditions. Process based models and machine learning (ML) are being increasingly applied for yield prediction. The process based model STICS (Simulateur mulTIdisciplinaire pour les Cultures Standard) version 9.2 is tested to predict yield, total biomass, leaf area index, and total and plant nitrogen using multi-year data from research farms in Eastern Canada. The calibrated model will be tested to predict yield in a commercial farm with multiple field years in Prince County, PEI. The commercial farm data includes detailed management practices and gridded yield map. This data set includes around 15000 data points and will be used to train an ML model for yield prediction. We compare ML and the STICS model performance in predicting yield and provide some insight on the conditions in which each model performs reliably.