Forecasting seasonal peaks in roadkill patterns for improving road management

Area of Science:

• Wildlife ecology
• Conservation biology
• Road ecology

Background:

• Roadkill is a diffuse issue for many species, making fixed mitigation structures like passages less effective.
• Temporary mitigation, such as traffic calming, requires understanding seasonal roadkill patterns.

Purpose of the Study:

• To assess the feasibility of predicting seasonal roadkill peaks for non-flying vertebrates.
• To evaluate the predictive performance of time-series models for roadkill data.

Main Methods:

• Utilized 3 years of systematic roadkill monitoring data (ca. 6400 records) for eight non-flying vertebrate species in Brazil.
• Employed generalized additive mixed models (GAMMs) to model roadkill time-series at the state level.
• Used the first 2 years of data for training and the third year for testing model predictions.

Main Results:

• Models performed reasonably well in tracking roadkill variations over time for species with abundant data.
• Prediction accuracy for the exact number of collisions was limited.
• Models showed poorer goodness-of-fit and prediction ability for species with fewer roadkill records.

Conclusions:

• Seasonal roadkill peaks can be forecasted for species with higher roadkill rates.
• Predictive models offer valuable insights for implementing timely, seasonal wildlife management actions.
• This approach can complement existing mitigation strategies for diffuse roadkill patterns.