Researchers at Colorado State University have developed a new method for measuring and mapping rainfall using the same radar sensors found in modern cars. These automobile radars can detect rain right at ground level, where it matters most for people and infrastructure. This innovation allows for the creation of detailed rainfall maps at both local and regional scales, especially in urban areas where traditional weather radar systems face limitations. It’s a new, cost-effective way to improve flood forecasting and storm monitoring in urban areas.
Current methods for measuring precipitation—like weather radars and satellites—have significant limitations in urban areas due to blockage by buildings and coarse spatial resolution. Ground-based sensors like rain gauges are accurate but sparse and fixed, often missing hyper-local storm data. To fill these gaps, researchers explored the use of millimeter-wave automobile radars, already installed in vehicles for driver assistance, to detect rainfall at a hyper-local scale. By leveraging the wide availability and mobility of these sensors, especially in cities, this innovation enables a dense, mobile network for real-time precipitation mapping.
This study introduces a novel technique that repurposes automotive radars—typically used for collision avoidance and cruise control—to measure surface-level precipitation. The approach hinges on the radar’s ability to detect signal attenuation caused by raindrops, from which rain rates can be estimated using simulation-based models.
Field experiments validated the concept. During a real storm on CSU’s campus, the radar-derived reflectivity closely matched that of the National Weather Service within a margin of a few decibels. Simulations further demonstrated that a dense network of such mobile radars (mounted on vehicles) can be used to construct accurate, city-scale rainfall maps using spatial interpolation techniques. In simulated urban mapping scenarios over Dallas-Fort Worth, rain estimates showed good spatial agreement with traditional radar networks, especially when using natural neighbor interpolation with 10,000 simulated data points.
Available for Exclusive Licensing
TRL: 4
US Provisional Patent
Chandrasekaran Venkatachalam
Shashank Joshil
Aly Hoeher
Aly.Hoeher@colostate.edu
970-491-7100