How E‑Scooter Trips Vary by Time, Weather and Neighborhood

A recent study examined when and where people use shared electric scooters across Kelowna. Researchers employed a zero‑inflated negative binomial model—a statistical tool designed to handle datasets with many zero entries.

How the Model Works

• Zero‑inflation component: Explains why certain areas see no trips at all.
• Count component: Predicts the number of rides when trips occur.

Other models were tested, but this approach provided the best fit to the data.

Key Findings

• Seasonal and Temporal Patterns
• Usage spikes in the summer.
• Peak activity occurs on weekday afternoons.
• Rainy weather sharply reduces trips.

• Spatial Factors
  Areas with:
• Many cycle paths
• Mixed land uses
• City centers
• Lower ground levels
• Neighborhoods dense with hotels or young residents

experience higher scooter activity.

The zero‑trip component of the model aligns with these spatial patterns, reinforcing confidence in the results.

Model Validation

A separate dataset tested the model’s predictions; it performed well enough for practical application.

Implications for City Planning

These insights enable planners to:

• Optimize scooter placement by targeting high‑use zones.
• Adjust fleet sizes to meet demand accurately.
• Schedule promotional efforts during periods of peak usage.