Presentation is loading. Please wait.

Presentation is loading. Please wait.

Huaqing Cai, Jim Wilson, James Pinto, Dave Albo and Cindy Mueller

Similar presentations


Presentation on theme: "Huaqing Cai, Jim Wilson, James Pinto, Dave Albo and Cindy Mueller"— Presentation transcript:

1 Developing NIWOT ---- A Regional 1-6 hr Short-Term Thunderstorm Forecast System
Huaqing Cai, Jim Wilson, James Pinto, Dave Albo and Cindy Mueller National Center for Atmospheric Research

2 What is NIWOT NIWOT, which means left-handed, comes from the name of an Arapaho chief who used to live in the Boulder valley area A regional 1-6 hr storm forecast system which combines radar echo extrapolation and NWP model output to produce a better final forecast

3 The Goal of NIWOT Skill Schematic of Extrapolation and NWP

4 NIWOT BLOCK DIAGRAM Injester Pre-Processor Blending

5 NIWOT Major Components
A data injester which grabs all the available data from NCWF, ANC and various models. Depending on running in real time or archive mode, this part can be intergrated into pre-processor A pre-processor which can do a list of tasks such as: converting to common grid and units, filtering, creating ensembles, changing to probability, calculating error-based weights from past performance, feature matching, and scale decomposition, etc A blender which combine an observation-based and a model-based 1-6 hr forecasts according to some weightings (automatic blending with fixed, pre-determinded weights; automatic blending using dynamical weights from past performance; or human-assisted blending) A blending/trending algorithm which grows/decays the extrapolation according model output

6 Problem With Matching: Which One to Match, Extrapolation or Model?
T=1 hr T=3 hr T=6 hr Both intensity and distance are weighted ! Extrapolation Merged Model

7 NIWOT Blender 1.0 We are blending two fields, one is extrapolation from NCWF6, the other is model forecasts from MM5 models (RTFDDA with 5 km resolution) A simple automatic matching/merging technique is used in the real time system The weights are pre-determined according to historic performance and both position and intensity from extrapolation and MM5 have their own set of weights

8 An Example of NIWOT Display
Human can modify the final forecast by growing or decaying inside a polygon

9 Examples of Blended Forecast

10 Examples of Blended Forecast

11 Examples of Blended Forecast

12 Examples of Blended Forecast

13 Examples of Blended Forecast

14 Examples of Blended Forecast

15 Examples of Blended Forecast

16 Examples of Blended Forecast

17 Examples of Blended Forecast

18 Examples of Blended Forecast

19 Examples of Blended Forecast

20 Examples of Blended Forecast

21 Examples of Blended Forecast

22 Examples of Blended Forecast

23 Examples of Blended Forecast

24 Examples of Blended Forecast

25 Examples of Blended Forecast

26 Examples of Blended Forecast

27 Future Development of NIWOT
Object-based feature matching OR cross-correlation/variational feature matching ? Real-time verifications Real-time weights calculations based on real-time verifications What kind of verifications ? Grid-to-grid versus object-based ? Extrapolation technique needs to be refined

28 Difficulties of Nowcasting
No “one size fits all” solution It seems that you will never know for sure which forecast is correct, until you verify it. Although everyone knows “past performance is no guarantee of future results”, everybody is still using it Statistics is the key Basic research is the King (according to Jim)

29 Thanks ! Questions/comments ?


Download ppt "Huaqing Cai, Jim Wilson, James Pinto, Dave Albo and Cindy Mueller"

Similar presentations


Ads by Google