Severe winter weather poses a major risk to road users due to slippery surfaces, disrupting highway traffic flow through speed reductions, capacity decreases, or complete highway closures. Departments of Transportation rely on Road Weather Information Systems (RWIS) for maintenance planning. However, RWIS has limitations, including downtime, equipment failures, and high installation and maintenance costs, resulting in limited station coverage. To supplement RWIS, some agencies use alternative weather data sources such as the Automated Surface Observing System (ASOS), Next Generation Weather Radar (NEXRAD), Meteorological Assimilation Data Ingest System (MADIS), and Multi-Radar/MultiSensor System (MRMS). While these sources offer valuable insights, most have geographic coverage constraints. In contrast, high-resolution Doppler weather radar systems like NEXRAD provide extensive precipitation estimates and atmospheric wind visualizations. Georeferenced NEXRAD data allows for detailed storm tracking and intensity monitoring. Nonetheless, factors such as beam blockage, radar calibration issues, and atmospheric conditions can impact the accuracy of precipitation estimates. This research aims to enhance NEXRAD's utility by developing and validating advanced calibration techniques using artificial intelligence (AI). Focusing on precipitation and wind attributes, algorithms such as deep learning and Gaussian process regression will be employed, leveraging Ohio's RWIS data as ground truth. The methodology incorporates spatial-temporal dynamics and external factors like radar proximity to improve model accuracy and transferability.
Exhibit D
