FOTL AutoNowCaster (ANC) system was demonstrated at the National Weather Service Dallas/Ft. Worth WFO from 2005-2010.
Goals: Improve the consistency,
reliability, and accuracy of one hour convective storm nowcasts for inclusion in automated aviation weather digital products (NextGen 4-D data cube).
Assess NWS forecaster
involvement in providing value-added enhancements to gridded, automated ANC products for NextGen.
The Federal Aviation Administration NextGen 4-D Weather Cube: Provides a common weather picture for the National Air Space by 2025 Observations Forecasting Numerical Modeling Systems
Statistical Forecasting Systems
4D Wx Cube
4D Wx SAS
Automated Forecast Systems
g Forecast Integration
Soundings S di
Decision S Support t Systems
Custom Graphic G hi Generators
Custom Alphanumeric Al h i Generators
Integration into User Decisions
What is needed to meet the NextGen requirements - and be successful??
High resolution observations Boundaries, winds, stability, gravity waves, capping inversions, elevated dry layers, moisture i t variability i bilit and d gradients di t
NWP - high resolution data assimilation for boundary layer winds and precipitation forecasts - frequent update cycle - at least 1 hr analysis and forecast fields - ability to represent multiple scale interactions - accurate output fields (CAPE, CIN, RH, vertical shear, convergence, vorticity, temperature gradients)
Forecaster - convergence boundary locations - climatology data base - conceptual models and documented weather regimes - QC of final products
Basic B i understanding d t di (most ( t important) i t t) Of elevated convection, secondary convection, outflow characteristics
Operational Role of Forecaster with the ANC system Entering convergence boundaries periodically during shift Nudging storm initiation nowcast fields based on
forecaster understanding of NWP performance, the observations and their experience.
Specification and selection of weather regime
Storm initiation likelihood fields for different synoptic regimes Dry L Line n Regime gm
Airr Mass Regime gm
•7 different regimes (fuzzy-logic rule sets ) available • cold front, warm front, dryline, air mass, MCS, no storms, and mixed (default) regimes Developed in collaboration with NWS forecasters •Developed •User Selectable
RUC NWP Model Forecasts “Too Hot” or “Too Cold” Forecaster has sliding scale toolbar to adjust (nudge) thunderstorm initiation likelihood values over the whole domain
Challenges in Evaluating the Performance of Convection Initiation Nowcasts Difficult to validate forecasts of these “rarer” convective initiation events. Evaluation is hampered by the impacts of large-scale storm characteristics as well
as small-scale errors in timing and location. Difficult to statistically highlight the impact of human forecaster on an
automated nowcast system.
M th d l Methodology ffor E Evaluation l ti 1.
Focus the evaluation on small regions rather than the broader ANC domain.
Stratifying cases into similar forcing categories (i.e., weather regimes).
Examination of time-series performance for specific cases.
Use of contingency table performance diagrams to simultaneously investigate changes in multiple forecast performance attributes. attributes
4 Year Period Evaluated (2006-2009) 44 days analyzed in great detail
ANC Convective Storm Initiation 60 min Nowcast Field
ANC Final 60 min Nowcast Field: Initiation, Growth & Decay
Statistics Over Whole ANC Domain
Convective storm initiation period
Calculating Statistics Over Sub-Domains
Evaluation of nowcasts on the scale of the phenomena being nowcast.
Forecaster Interaction with ANC System 19:21 UTC
4 July 2006 Cold Front Regime 20:21 UTC
Black contours: Location of > 35 dBZ storms at validation time
Time-Series Plots – 4 July 2006 for selected sub-domain boxes
Summary Histograms of Change in Verification Scores With and Without Human Input for all sub-domain boxes
4 July 2006
Performance Diagrams* Diagrams 4 July 2006 *After Paul Roebber, Clive Wilson
Forecaster-entered convergence boundaries have largest impact on convection initiation nowcasts. Examining only those domains that contain boundaries allows us to analyze change in nowcast accuracy due to forecaster involvement.
Performance for 27 August 2006
24 April 2007
12 May 2007
8 October 2007
8 October 2007 Mixed Regime
30 July 2008 MCS Regime
25 March 2009 Warm Front Regime
10 June 2009 Cold Front Regime
Summary PODs of new storms increased by y 20-60% with little or no change g observed in false
alarm ratios. Bias values increased from 0.8-1.0 to 1.0-3.0 with human involvement. Range of CSI values increased from 0.15-0.25 (ANC) to 0.2-0.4 (FOTL-ANC). Values of CSI as large as 0.6 were observed. A true improvement p of the nowcast p performance was achieved through g forecaster
involvement. Forecasters played an important role in improving the quality of automated, high-
resolution, gridded, convective initiation nowcasts for end-users.