Experiments in 1-6 h Forecasting of Convective Storms Using Radar Extrapolation and Numerical Weather Prediction Acknowledgements Mei Xu - MM5 Morris Weisman – WRF James Pinto –WRF, NCWF-6, computer support Steve Weygandt - RUC Tom Saxen – NCWF-6, Extrapolation Cindy Mueller – NCWF-6, Extrapolation, management Jenny Sun – Forecast VDRAS Dan Megenhardt – computer support Rita Roberts – Scientific advise Frank Hage – Display support
Overarching Goal Blend Numerical Forecasting Methods and Observational methods To improve 1-6 h nowcasting
Predictability Forecast Length Extrapolation NWP Forecast Skill Blended Best
Challenge - How to blend extrapolation and model nowcast Extrapolation Forecast Numerical Model Forecast
8 methods that produce 1-6h forecasts 4 numerical and 4 observational Forecasts evaluated with the objective of developing ideas for blending numerical and observational To meet this challenge – NCAR conducted a forecast extravaganza this summer
Study area June 2005
Example Initiation case
Extrapolation Probabilities Extrapolation plus smart trending (synoptic situation and time of day) Observational Techniques Examined
Probabilities 20 km grid 3 h forecast cycle ACARS, VAD, profiler, GOES precip water) NWP Techniques Examined nested grid 3h forecast cycle observational nudging radar data assimilation (conus mosaic of reflectivity) 4 km grid 24h forecast cycle initialized with 40km ETA The point is- State of the art techniques were available
Subjective evaluation of forecast quality
1 – forecast and observed almost perfect overlap. 2 – majority of observed and forecast echoes overlap or offsets <50 km. 3- forecast and observed look similar but there are a number of echo offsets and several areas maybe missing or extra. 4 – the forecasts and observed are significantly different with very little overlap; but some features are suggestive of what actually occurred. 5- There is no resemblance to forecast and observed. Forecast Quality Definitions Wilson subjective categories
Forecast Observed Quality = 2.0Quality = 3.0 Quality = 4.0 Quality = 5.0 Examples of Forecast Quality
1.Quality of forecasts for echo Existing at forecast time. 2. Quality of NWP forecasts of initiation 3. Quality of NWP forecasts of change in area size
1. Echo present at forecast time Forecast Quality Extrapolation NWP Best
Quality = 4.0 Forecast observed
Forecast Length, hours Accuracy of Rainfall Nowcasts >1 mm/h GRID MESH 20 km Jun-Oct 2002 Courtesy of Shingo Yamada JMA Extrapolation NWP Critical Success Index (CSI) Cross over region
Best NWP Results 2-hour forecast 4-hour forecast 6-hour forecast Initiation (number cases) 17 Initiations fx correct (percent) Forecast quality (category) Offset median (hours) False alarms (number) 5 2. Initiation Forecasts
2, 4 and 6 hr forecasts of trend in area size 3. Area Size Trend Forecasts g+ large growth g medium growth g- small growth nc no change d- small dissipation d medium dissipation d+ large dissipation 7 Trend Categories forecast observed Error 2 categories
2, 4 and 6 hr forecasts of trend in area size 3. Area Size Trend Forecasts g+ large growth g medium growth g- small growth nc no change d- small dissipation d medium dissipation d+ large dissipation 7 Trend Categories forecast observed Error 2 categories
2, 4 and 6 hr forecasts of trend in area size 3. Area Size Trend Forecasts g+ large growth g medium growth g- small growth nc no change d- small dissipation d medium dissipation d+ large dissipation 7 Trend Categories forecast observed Error 6 categories
6 h Best NWP results 3. Area Size Trend Forecasts BestWorse
Overarching Goal Blend Numerical Forecasting Methods and Observational methods To improve 1-6 h nowcasting Summary
1. Model – frequent cycling (3hr), assimilate radar reflectivity 2. Initiation – Give full weight to model 3. Existing storms – Extrapolate and trend area size based on model trend (more weight for dissipation trend) Unfinished – examine model and extrapolation predictability stratified by precipitation organization, synoptic situation and time of day.
Thank You
2-hour trend 4-hour trend g+ large growth g medium growth g- small growth nc no change d- small dissipation d medium dissipation d+ large dissipation 6-hour trend Trend Category Number cases Area Size Trends
Forecast Observed Quality = 1.5
Example Initiation case