Making bikes count for climate and health

However, cycling volume data is often scarce, as most cities rely on a few isolated counting stations. New research demonstrates how machine learning and existing data can be leveraged to estimate city-wide cycling volumes.

A recent study, “From Counting Stations to City-Wide Estimates: Data-Driven Bicycle Volume Extrapolation”, published in Environmental Data Science by Silke Kaiser (PhD candidate, Hertie School), Prof. Nadja Klein (Karlsruhe Institute of Technology), and Prof. Lynn Kaack (Hertie School), explores how limited bicycle traffic data can be extrapolated to predict street-level cycling volumes across entire cities.

Despite the growing importance of traffic data for urban planning, bicycle volume extrapolation remains underexplored. This study tackles the gap by applying machine learning to diverse data sources—including counting station data, newly available data such as crowdsourced cycling, and bike-sharing data, as well as traditional indicators such as infrastructure and weather—to estimate bicycle volumes across Berlin. Silke Kaiser explains, “For locations without direct cycling counts, combining machine learning models with multi-source data allows us to predict cycling volumes with a very reasonable error.” The study also finds that short-term sample counts can be beneficial for even more precise estimates: “With just ten days of sample counts per location, we can further halve the error.”

These estimates support evidence-based decisions on infrastructure improvements, enabling policymakers to prioritize high-traffic areas and justify investments in cycling networks. While the research focuses on Berlin, its insights are applicable to cities worldwide. As urban areas strive for sustainability, data-driven approaches like these can guide planning, ensuring resources are allocated where they have the most significant impact.

You can read the full paper here

The Publication received funding from the European Union’s Horizon Europe research and innovation programme under Grant Agreement number: 101057131 — CATALYSE — HORIZON-HLTH-2021-ENVHLTH-02