Environmental variability poses a significant challenge to the reliability of vision-based perception systems in autonomous vehicles (AVs), particularly in the context of traffic sign detection. This study evaluates the performance of the YOLOv8 deep learning model under clear, rainy, and snowy conditions. Real-world video data captured in clear weather was augmented using simulation techniques to replicate realistic rain and snow effects, including precipitation, motion blur, lens distortion, and reduced visibility. Detection performance was assessed using detection rate, sharpness, mean saturation, and intensity. Results show a marked decline in accuracy under adverse conditions, with snow causing further issues like occlusion, reduced contrast, and altered reflectivity. Additionally, detection resilience varied among sign classes; high-contrast signs with distinct geometric shapes were detected more reliably than those with fine details or low color contrast. These findings highlight the critical need for robust perception algorithms and the advancement of AV vision systems capable of maintaining accuracy across challenging environmental conditions.