Auto-Nowcast System Tom Saxen (July) Huaqing Cai (Aug) National Center for Atmospheric Research

Size: px

Start display at page:

Download "Auto-Nowcast System Tom Saxen (July) Huaqing Cai (Aug) National Center for Atmospheric Research"

Kristina Patrick
5 years ago
Views:

1 Auto-Nowcast System Tom Saxen (July) Huaqing Cai (Aug) National Center for Atmospheric Research Summer 2006 ATEC Forecaster Conference Photo courtesy of Greg Thompson

2 Overview: Introductory comments on the ANC system. Discussion on how the ANC system works. Role of the human in the ANC system. Future plans.

3 What is the NCAR Auto-nowcast System? It s an automated, data fusion system that weights and combines all available data sets to produce short-term 0-1 hr nowcasts of thunderstorm initiation, growth and decay. Primary focus of this presentation will be on the initiation forecasting capabilities since the hope is to provide this sort of capabilities at all ranges (except CRTC, sorry Craig) in the next few years.

4 System Forecast Output Initiation Forecasts Extrapolation And Trending Initiation Likelihood Forecast Combined Forecast Blue Regions - Little chance of storm development Green Regions - Moderate likelihood Red Regions - Areas of forecast initiation

5 What is the difference between the ANC system and the MRRD (Multiple Regional Radar Display)/AN-Lite? The MRRD/AN-Lite system is primarily a radar data display system with limited forecasting capabilities, providing simply 60 minute extrapolation forecast (both TREC motion vectors and TITAN storm motions). The MRRD/AN-Lite system also utilizes NIDS data as it s primary source of radar data, which is less than ideal. The ANC system utilizes the full beam to beam (or Level II radar data).

6 What ranges have what system? ATC: CRTC: DPG: EPG: RTTC: WSMR: YPG: MRRD/AN-Lite (with Level II radar data) None MRRD/AN-Lite (with Level II radar data) MRRD/AN-Lite MRRD/AN-Lite Full ANC system MRRD/AN-Lite

7 What makes the Auto-nowcast System different from other thunderstorm nowcasting systems? Most operational systems for generating analyses and shortperiod nowcasts of rainfall and other weather parameters rely on the presence of already existing storms. I.e. they are primarily extrapolation based systems.

8 Flow Chart for the Auto-Nowcaster System Forecaster Input Data Sets Radar WSR-88D Satellite Mesonet Profiler Sounding Numerical Model Lightning Analysis Algorithms Predictor Fields Final Prediction Fuzzy Logic Algorithm - Membership functions - weights - Combined likelihood field

9 Forecast Logic Cumulus development Separate sets of forecast logic for different types of convective regimes. Boundary characteristics Satellite Cloud Typing B-L characteristics Large-Scale Environment Initiation Component Boundary characteristics Storm Motion and Trends Large-Scale Environment Growth and Decay Component

Predictor Fields used for Initiation Likelihood Forecast Environmental conditions (RUC or RT- FDDA) Frontal likelihood Layered stability CAPE (max between 900 and 700 mb) CIN (mean value between 975

10 Predictor Fields used for Initiation Likelihood Forecast Environmental conditions (RUC or RT- FDDA) Frontal likelihood Layered stability CAPE (max between 900 and 700 mb) CIN (mean value between 975 and 900 mb) Mean 875 to 725 mb Relative Humidity Boundary-layer Convergence LI (based on METARS) Vertical velocity along boundary (maxw) Boundary-relative steering flow New storm development along boundary Clouds Clear or Cumulus Vertical develop as observed by drop in IR temps

11 Flow Chart for the Auto-Nowcaster System Forecaster Input Data Sets Radar WSR-88D Satellite Mesonet Profiler Sounding Numerical Model Lightning Analysis Algorithms Predictor Fields Final Prediction Fuzzy Logic Algorithm - Membership functions - weights - Combined likelihood field

12 Example of fuzzy logic Predictor Field 1 Convergence line Lifting Zone Likelihood Membership Function Yes.5 No 0 Lifting Zone Likelihood.5

13 Predictor Field 2 Membership Function Convergence Likelihood Likelihood Convergence

14 Predictor Field 3 Cumulus clouds Likelihood 1 Membership Function Likelihood Cumulus cloud type

15 Likelihood 1 Likelihood 2 Likelihood 3 Weight 1 Weight 2 Weight 3 Final combined likelihood of initiation

16 L i f t e d I n d e x C a p e V e r t _ S u m In te J Kg-1 Wt: 0.17 In te Count of "unstable" 25mb levels Wt: m b M e a n R e la t i v e H u m i d i t y Wt: 0.20 In te In te % Wt: ok F r o n t a l L i k e li h o o d C o n v e r g e n c e In te In te Interest Wt: s-1 Wt: 0.08

17 S a t _ C u B o u n d a r y R e la t i v e S t e e r i n g F lo w R a te O f C h a n g e ( R O C ) IR T e m p R a te Wt: In te In te m s-1 Interest Wt: M a x W In te In te Wt: Wt: 0.15 m s-1 Interest I n i t i a t i o n a lo n g B o u n d a r y In te Interest Wt: 0.25

18 Boundary Collision: Wt: 0.12 Sat_Clear: Wt: 0.40 Lake: Wt: 0.10 Initiation Levels: 0.70 => Init => Init => Init 3

Blue Regions - Little chance of storm development Green

19 System Forecast Output Initiation Forecasts Extrapolation And Trending Initiation Likelihood Forecast Combined Forecast Blue Regions - Little chance of storm development Green Regions - Moderate likelihood Red Regions - Areas of forecast initiation

20 Why do we need a forecaster in the loop?? (or rather over the loop) Forecasters see the larger picture Conceptual Models Ignore bad data points Understand limitations of NWP and observations Set current convective regime so appropriate forecast logic is utilized Input significant surface convergence features (cold fronts, dry lines, gust fronts, etc.)

21 Draw Tool

22 Draw Tool

23 Draw Tool Entering a convergence boundary in real time is as simple as this demonstration!

24 Forecaster Tools: Boundary Entry

25 Future Plans: Transition to Probabilistic Forecast Product Idea is to transition the ANC forecast products to a probabilistic product. Goal is to implement this within the WSMR ANC system for next summer (2007).

expected Greens: Regions of moderate likelihood Blues: No initiation

26 Transitioning ANC Forecasts to Probabilities Current Initiation Interest Field Current Final Forecast Field Reds: Initiation expected Greens: Regions of moderate likelihood Blues: No initiation is expected Blue areas: Initiation regions Yellow/Orange/Reds: Growth/Decay Fcst

27 Transitioning ANC Forecasts to Probabilities Example possible probabilistic final forecast field Verification reflectivities (> 35 dbz) are overlaid. Includes both initiation and growth/decay information. Warmer colors represent higher probabilities of convection. Still need to finalize these methods and also calibrate the probabilities.

28 Future Plans: Testing Enhancements at Other Ranges Would like to test the ANC-like capabilities at a range or possibly two. Idea is to utilize model data with other local data sets to produce an initiation likelihood field. Timeline for this is still not clear.

29 Future Plans: 0-6 hour Convective Weather Forecast Product New effort in underway in the convective weather group to develop a 0-6 hour convective weather product. Idea is to merge ANC-like capabilities with numerical model output to produce these forecasts. The Army test ranges would be ideal candidates for these systems due to access to high quality meso-scale models and ANC-like capabilities.

Probabilities Fcst Valid Times: 1600 2200 UTC by 120 mi

30 1-6 hr Probability Forecasts Extrapolation Fcst Valid Times: UTC by 60 min RUC Convective Probabilities Fcst Valid Times: UTC by 120 mi Radar Data UTC by 30 min WSI - Validation Forecast Time

1-6 hr Merged Probability Forecast Linear summation WSI: 1400-1600 UTC by

31 1-6 hr Merged Probability Forecast Linear summation WSI: UTC by 30 min; Fcst: UTC by 60 min WSI - Validation Forecast Time

32 Questions/Discussion..

P5.11 TACKLING THE CHALLENGE OF NOWCASTING ELEVATED CONVECTION

P5.11 TACKLING THE CHALLENGE OF NOWCASTING ELEVATED CONVECTION Huaqing Cai*, Rita Roberts, Dan Megenhardt, Eric Nelson and Matthias Steiner National Center for Atmospheric Research, Boulder, CO, 80307,