Studying the effect of weather conditions on daily crash counts using a discrete time-series model

Brijs, Tom; Karlis, Dimitris; Wets, Geert · 2008 · OpenAlex-citations

DOI: 10.1016/j.aap.2008.01.001

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the impact of weather conditions on daily road crash counts, aiming to correct methodological limitations in previous research. Prior studies often relied on highly aggregated monthly or yearly data, which obscures short-term patterns, or failed to account for temporal serial correlation (autocorrelation) in daily data, leading to biased results. The authors introduce an Integer Autoregressive (INAR) model, specifically an INAR(1) Poisson regression, to explicitly model the time-dependency inherent in discrete count data. This approach separates the current crash count into a component dependent on the previous day’s count (reflecting stable infrastructure risks) and an innovation term influenced by daily covariates like weather and traffic exposure. The analysis utilizes daily crash data from 2001 for three major Dutch cities: Utrecht, Dordrecht, and Haarlemmermeer. These sites were selected to ensure varied weather conditions and distinct autocorrelation structures. The dataset includes daily crash counts, traffic exposure metrics (vehicle kilometers derived from loop detectors), and detailed meteorological variables such as temperature, precipitation, wind, sunshine, visibility, and air pressure. The authors employed an Expectation-Maximization (EM) algorithm to estimate model parameters, allowing for site-specific autocorrelation coefficients while assuming constant weather effects across sites. To handle multicollinearity among weather variables, a stepwise selection procedure was applied. The study compares three model specifications: one using day-of-the-week dummies as a proxy for exposure, one using actual traffic exposure data, and a combined model. The results demonstrate that the INAR model effectively captures the serial correlation in crash data, which varies significantly by location (e.g., higher autocorrelation in Haarlemmermeer than Utrecht). The inclusion of covariates substantially reduces overdispersion, rendering more complex negative binomial models unnecessary. Crucially, the findings confirm that several weather conditions significantly influence crash risk. The study validates that day-of-the-week dummies can serve as effective proxies for traffic exposure when actual data is unavailable, producing consistent weather effect estimates. However, the authors emphasize that failing to account for temporal serial correlation leads to inefficient modeling and potentially biased standard errors. The INAR framework provides a parsimonious and interpretable method for analyzing daily accident counts, successfully isolating the direct effects of weather from underlying structural risks and exposure variations.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

StageOutcomeToolModelPromptAttemptsCompleted
discover success OpenAlex-citations 1 2026-06-20
archive success unpaywall 2 2026-06-26
extract success pdftotext 2 2026-06-26
clean success clean 1 2026-06-26
chunk success chunk 1 2026-06-26
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-26
enrich failed 4 2026-06-26
promote success 1 2026-06-20
summarize success llm qwen3.6-27b-prismaquant summ-v5 1 2026-06-26
tag success vector_similarity 6 2026-06-26
verify success 1 2026-06-26

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).