Predicting bicycle commuting mode share for 864 European cities

Cycling mode share is an important indicator for climate and transport policy, yet consistent measurements are unavailable for many European cities. We develop a machine learning approach to classify commuting cycling mode share as low ( < 5%), moderate (5–15%), or high ( > 15%). The model uses more than 200 predictors covering urban form, the natural environment, socioeconomic conditions, and demographics. We train it on 564 cities and predict cycling levels for another 300, increasing coverage to 864 European cities. Although our results are limited by the quality of the available data, the classifications remained stable across robustness checks. Protected cycling infrastructure is the strongest predictor, followed by terrain and climate. Most European cities are classified as low, whereas high cycling is concentrated in Northwestern Europe. Our approach addresses persistent gaps in coverage and comparability of existing cycling data, supporting planning and research on urban mobility.