3.1 Powering Aviation Nowcasting Models with New Data Source: Wireless Networks

Monday, 8 January 2018: 12:00 AM
Room 16AB (ACC) (Austin, Texas)
Rei Goffer, ClimaCell Inc., Boston, MA; and D. Rothenberg

Detecting and anticipating transient local weather conditions has long been a difficult challenge in both commercial and military aviation. Traditionally, this problem has been tackled by augmenting existing weather observations with additional instrumentation to peek into blind spots near and around airports, such as placing dedicated doppler radars on site at major air traffic hubs or on board the aircraft which fly through their airspace. But critical operations decisions at airports are still heavily based on manual observation or a limited number (usually not more than one) of instrumentation packages.

These decisions can have a massive influence on the 8.5 million departures which occur annually at over 500 airports in the United States alone. 496,174 hours of delays cost airlines over $30 billion in fuel and operating costs, two-thirds of these delays and resulting financial losses were weather-related.

These losses could be abated with improved detection and anticipation of constantly-changing weather conditions. Towards this end, we present a novel weather observation system leveraging a ubiquitous component of modern infrastructure - the wireless networks which carry global telecommunications. Ambient weather conditions can deleteriously impact the performance of these networks and provide a unique way to passively sense the atmosphere. Furthermore, since this infrastructure is widely deployed in both the developed and developing worlds, it provides a means to improve observations of the atmosphere without investing in new remote sensing or instrumentation platforms.

Using a novel signal processing technology employing this telecommunications data, we present a high resolution (with < 1 minute latency in time and spatial resolution up to 300 meters) precipitation data product with applications to real-time detection, nowcasting, and historical re-analysis. This data product yields strong correlation with rain gauges (up to 90%) and outperforms local radar-derived precipitation estimates in many cases, both in terms of measuring the timing and intensity of rainfall.

Finally, we show how this data can be used in conjunction with traditional observation and forecasting tools to develop improved hyperlocal and nowcasting products. These products could potentially augment decision-making throughout aviation operations, yielding early detection of hazards and allowing early mobilization of suitable responses.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner