| Title |
Association between Road Geometric Characteristics and Traffic Accident Risk Areas Using the National Standard Node-Link Network: A Case Study of Nationally Supported Provincial and Provincial Roads in Gyeonggi Province |
| Authors |
송영선(Song, YeongSun) ; 윤공현(Yun, Konghyun) |
| DOI |
https://doi.org/10.12652/ksce.2026.46.5.0525 |
| Keywords |
도로 기하특성; 교통사고 위험지역; 링크 형상 기반 곡률지표; 표준노드링크; 조건부 로지스틱 회귀 Road geometric characteristics; Traffic accident risk area; Link-geometry-based curvature index; Standard node-link; Conditional logistic regression |
| Abstract |
This study examined the association between road geometric characteristics and traffic accident risk-area designation on nationally supported provincial and provincial roads in Gyeonggi Province. Link-based traffic accident risk areas, the national standard node-link network, annual average daily traffic (AADT), and road-ledger data for 2020-2024 were integrated. For each representative risk-area link, non-risk links with similar AADT and link length on the same route and in the same year were matched, yielding 364 matched strata. The link-geometry-based curvature index was significantly correlated with the road-ledger density of curves with radii below 300 m (Spearman rho=0.657, p<0.001). In the primary conditional logistic model, the odds ratio (OR) per one-standard-deviation increase in the ln(1+C)-transformed and standardized curvature index was 1.217 (95 % CI 1.054-1.404, p=0.007), while the OR for number of lanes was 1.718 (p<0.001), indicating higher relative odds of risk-area designation among road sections with comparable traffic exposure and link length. However, a 2,000-resample cluster bootstrap weakened the evidence for the curvature index (percentile bootstrap 95 % CI 0.993-1.497; empirical p=0.062). The two representative-link assignment rules selected the same link for only 138 of 471 risk areas (29.3 %), and both curvature-index and lane estimates attenuated in the strictly common sample of 257 risk areas. These findings suggest a positive association between road geometry and risk-area designation while also demonstrating that the estimated magnitude and statistical precision are sensitive to spatial-linkage rules and sample composition. |