Aggregated traffic data can obscure temporal and between-axis variation. Objective: To characterize multiscale patterns in traffic volume, mean speed, and recorded violation rates, and to assess the contribution of temporal dimensions to estimates of unsafe following-distance violation rates. Methods: All hourly records published by Iran’s Road Monitoring System 141 for Razavi Khorasan during the 12 months of Iranian calendar year 1404 were considered. Of 1,166,102 raw records, 1,143,169 observations from 136 directional axes remained after quality control. Violation rates were calculated per 1,000 vehicles as the ratio of aggregated violation counts to aggregated traffic exposure. Temporal differences were assessed using the Friedman test, Kendall’s W, and Holm-adjusted Wilcoxon signed-rank comparisons. Unlike an axis-only or hourly profile, the evaluated estimator jointly stratifies violation rates by axis, hour, day type, and season, reverting to simpler profiles when the corresponding training cell is absent. Its contribution is the incremental evaluation of these temporal distinctions; an additive weighted least-squares (WLS) regression provided an independent benchmark on identical monthly folds and test cells. Blocked leave-one-month-out evaluation within the same province and year used exposure-weighted R², MAE, RMSE, and MSE. Stability checks comprised axis-block bootstrap resampling, minimum-volume sensitivity analysis, and reconciliation of published axis-month totals and reconstruction of daily aggregates from valid hourly records. Results: The hour-of-day effect on the unsafe following-distance violation rate was large (Kendall’s W=0.800). Network exposure-weighted rates were 227.47 per 1,000 vehicles on Fridays and 194.53 on working days; rates in summer and winter were 214.91 and 182.68, respectively. The multiscale estimator achieved MAE=25.13, RMSE=35.09, mean MSE=1410.28, and weighted R²=0.922. Relative to the axis-only profile, MAE and RMSE decreased by 53.35% and 53.83%, respectively. WLS yielded MAE=42.08 and RMSE=56.85; P3 reduced these errors by 40.27% and 38.28%. Equal-axis bootstrap profiles yielded a mean Spearman correlation of 0.997, with a 95% percentile interval of 0.991–1.000. Exposure-weighted profiles yielded 0.990 (0.958–0.999). The exposure-weighted peak remained at 16:00 across the examined minimum-volume thresholds. The directions of the day–night differences persisted across six alternative boundary definitions. Conclusion: The observed patterns identify 15:00–18:00 and weekends as priority periods for monitoring. Temporal stratification improved within-year estimation relative to the axis-only profile. These results do not establish causal effects, prospective forecasting performance, or validity in other years or provinces