Ozone (O3) pollution is an increasing barrier to air-quality improvement in arid cities and poses growing public-health risks because of rapid buildup and frequent exceedances. Unlike humid eastern China, arid cities typically experience high temperatures, low humidity, and intense solar radiation, together with rapid motorization, which can lead to complex nonlinear O3 responses. We conducted hourly, synchronized monitoring of traffic flow, meteorological variables, and 11 air pollutants at a roadside site on a major arterial in Yinchuan (Northwest China). We used a random forest regression model with SHAP (Shapley Additive Explanations) to identify drivers of O3 variability. Road traffic showed a clear weekday double-peak commuting pattern. Diurnal patterns differed across pollutants: primary pollutants (e.g., NO and CO) peaked during commuting hours, whereas O3 increased with daytime radiation and peaked at 110.11 mu g center dot m-3 in the afternoon. Model interpretation identified temperature and NO as dominant drivers, with methane and relative humidity (RH) also contributing. Threshold analysis showed that O3 increased rapidly when air temperature exceeded 24.9 degrees C and traffic flow exceeded 386 veh center dot h-1, whereas RH above 52% inhibited O3. SHAP interaction analysis indicated a climate-amplification pattern, with the strongest O3 response under high-temperature, low-humidity conditions. The temperature-NO interaction dominated O3 variability at low-to-moderate NO levels, whereas NO titration suppressed O3 under high-NO conditions. Traffic emissions showed a chemistry-dependent dual role: they promoted O3 accumulation under moderate loads but suppressed O3 via NO titration under severe congestion and high NOx. Overall, we present an interpretable, data-driven framework that links road traffic and meteorology to nonlinear O3 responses and actionable thresholds, supporting time-resolved traffic management and coordinated precursor control in arid cities.