Specialty crops, defined here as fruit and vegetable crops, are resource and labor-intensive, high-risk, high-value agricultural products. The yield and quality of specialty crops are highly influenced by local climate conditions. Climate change presents a challenge to specialty crop agriculture and suitable regions for agriculture are expected to shift. Determining the extent of future suitable agricultural regions will be an important aid for future land use management decisions. Machine learning algorithms, such as Random Forest, can utilize a wide range of relevant data, including climate, soil, and topographic features, to predict spatial distributions of suitable agricultural land. This research developed random forests classification models predicting the presence/absence of Southern Ontario specialty crops using contemporary climate variables derived from a ~1 km gridded surface weather dataset and distribution for each specialty crop from Agriculture and Agri-food Canada’s Annual Crop Inventory for an eight-year period. Predictive performance was measured by Matthew’s correlation coefficient and area under the precision-recall curve and showed overall high performance. Trained models were applied to the Ontario Climate Data Portal CMIP5 ensemble mean projections for mid-century (2040-2069) and late-century (2070-2099) periods and two Representative Concentration Pathways (4.5 and 8.5). Results identify regions where future climate presents opportunities and limitations and extents will be overlapped with predicted suitable soil and topography distributions. This work will support land use planning decisions by providing predicted areas of importance for specialty crop production in current and future scenarios in Southern Ontario.