J4.3
DCIT: diagnosing convectively-induced turbulence in near real-time and the challenges of communicating probabilities of rare events

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011: 9:00 AM
DCIT: diagnosing convectively-induced turbulence in near real-time and the challenges of communicating probabilities of rare events
2A (Washington State Convention Center)
Jennifer Abernethy, NCAR/RAL, Boulder, CO; and J. K. Williams

DCIT (Diagnose Convectively-Induced Turbulence (CIT)) is a near-real-time turbulence prediction system developed by the Research Applications Laboratory at the National Center for Atmospheric Research (NCAR/RAL). DCIT incorporates satellite, radar, NTDA and NWP model data to produce high resolution (6km) predictions over the contentintal U.S. of convectively-induced turbulence near convective clouds every 15 minutes.The goal of DCIT is to increase the information available about turbulence around thunderstorms in order to reduce the distances aircraft must fly around them (according to FAA Thunderstorm Avoidance Guidelines). DCIT complements NCAR/RAL's suite of clear-air (GTG, Graphical Turbulence Guidance) and in-cloud (NTDA, Nexrad Turbulence Detection Algorithm) prediction products; it will be an integral input to RAL's Graphical Turbulence Guidance Nowcast (GTGN) product, which will be part of the FAA's NextGen 4D Data Cube.

DCIT uses the statistical learning method called random forests, based on decision trees, to make probabilistic forecasts of CIT. The model training combines satellite, radar, NTDA and NWP model data, using in-situ eddy dissipation rate (edr, a measure of turbulence) observation data from aircraft as turbulence 'truth'. The algorithm output can be mapped to an edr scale, also, to give deterministic predictions of edr. This paper will describe the DCIT algorithm development and training process, handling diverse data sources in a near-real-time system, algorithm verification, and the challenges of effectively communicating risk of rare events with probabilistic forecasts.