Towards a Lost-Cost Distributed AWOS: Machine Learning for Camera-Based Ceilometry
Rabin Dhakal, Chad Mourning · 2025
Automated Surface/Weather Observing Systems (ASOS/AWOS) are the surface weather observation stations located throughout the United States. This weather station relies upon sensing instruments that disseminate the weather data and broadcast it over the radio or the ground link in METAR Format. ASOS station accumulates complex weather data and is useful for weather forecasting or climatology applications. On the other hand, AWOS focuses on aviation-specific weather data, catering primarily to the needs of pilots and airport operations. Weather data is comprised of visibility, wind speed, cloud coverage information, temperature and humidity, etc. AWOS depends upon the ceilometer for estimating the cloud coverage and its location from the ground. Accurately estimating cloud-base height typically requires a costly ceilometer system, which uses advanced instrumentation to measure cloud height precisely.This research addresses this challenge by proposing a cost-effective alternative that uses optical cameras, offering a more affordable solution without compromising functionality. Implementing computer vision and synthetic data in a virtual environment, our proposed method achieves an RMSE of 193.91 meters on 442 previously unseen synthetic test samples, demonstrating its potential as a reliable and efficient alternative.