Forecasting German crash numbers: The effect of meteorological variables

Area of Science:

• Road safety research
• Traffic accident analysis
• Statistical modeling

Background:

• German Federal Highway Research Institute (BASt) annually publishes road safety statistics.
• Current statistics are incomplete, requiring forecasts for the last 3-4 months.
• Existing forecasting methods lack optimal accuracy.

Purpose of the Study:

• To enhance the accuracy of road safety data forecasts.
• To integrate meteorological conditions into predictive models.
• To provide more reliable data for German policymakers.

Main Methods:

• Application of structural time-series models.
• Inclusion of meteorological variables in the analysis.
• Comparison with previous heuristic forecasting approaches.

Main Results:

• Root mean squared errors reduced by up to 55% compared to the heuristic method.
• Prediction accuracy improved for most of the 27 data series when incorporating meteorological data.
• Only minor increases in prediction errors observed for a few data series.

Conclusions:

• The proposed structural time-series models offer a valid and improved alternative for road safety forecasting.
• Incorporating meteorological data enhances prediction accuracy.
• The findings support evidence-based policymaking in German road safety.