wrf ensemble model performance during atmospheric river events in california
DESCRIPTION
AGU 2010. WRF Ensemble Model Performance during Atmospheric River Events in California. Edward Tollerud 1,5 , Tara Jensen 2,5 , Huiling Yuan 1,3 , John Halley Gotway 2,5 , Paul Oldenburg 2,5 , Isidora Jankov 1 , Wally Clark 4 , Ellen Sukovich 4 , Gary Wick 4 , and Randy Bullock 2,5 - PowerPoint PPT PresentationTRANSCRIPT
WRF Ensemble Model Performance during Atmospheric River Events in
California
Edward Tollerud1,5, Tara Jensen2,5, Huiling Yuan1,3, John Halley Gotway2,5, Paul Oldenburg2,5, Isidora Jankov1, Wally Clark4, Ellen Sukovich4, Gary Wick4, and Randy Bullock2,5
1 ESRL/GSD, Boulder, CO 2 NCAR/RAL, Boulder, CO3 CIRES, Boulder, CO 4 ESRL/PSD, Boulder, CO
5Developmental Testbed Center
Acknowledgments to the USWRP for funding
AGU 2010
Research Objectives: HMT and DTC
AGU 2010
DTC – Evaluate present and future and EMC operational models
HMT – Evaluate forecast values of regional ensemble forecast system
Principal focus on QPF
Investigate impact of verification data choices
DTC - Build demonstration realtime web display and examine standard and state-of-the-art verification methods
Precipitation during 1200 UTC 20 January – 21 January 2010
(Note poor observing network in Nevada)
Four days of heavy rainfall: 1/17-1/21
Cumulative Rainfall at ATA, 1200 UTC 17 Jan – 21 Jan
WRF ensemble member forecasts
The Forecast Prospect: Dealing with uncertainty
AGU 2010
WRF ensemble mean
GFS
Obs at ATA – 240 mm
The Forecast Prospect: Selecting NWP Guidance
AGU 2010
FY 2010 HMT-West: Demonstration Website
Basin-specific and RFC-specific verification domains installed
Assessment of ensemble member QPF possible in near real time; verification dataset options available
30-day boxplots provide statistical summary of model QPF performance
AGU 2010
ETS for January
Large day-to-day variability related to rainfall amount; extensive rain means better scores
RMSE for January
Large rainfall means larger errors
AGU 2010
AGU 2010Full-season statistics – Gilbert Skill Score – All events
30-day summary scoring for January, ETS
GFS degradation at higher thresholds
‘Quirky’ scores for verification using Stage IV analysis: Nevada impact
AGU 2010Score Idiosyncracies and Data Impacts
GSS for QPF Threshold > 0 inches Aggregated for January
for lead times 6-114 h
FAR for QPF Threshold >0.1 inches Aggregated for January
for lead times 6-114 h
AGU 2010
MODE/MET objects: Spatial Verification for ensemble QPF fields
In the pipeline: timeseries plots of quantitative attributes scoresSee Clark poster in afternoon session of more MODE applications
AGU 2010
Profiler Winds at Bodega Bay
Diurnal cycling of winds, from southerly to westerly (upslope); how well to models perform?
January 18
January 19
Full explanation and diagnosis could use wind field verification as well as QPF verification
Summary and Future DRTC /HMT Plans
• General Assessment from one field season: WRF ensemble mean at higher resolution than GFS performs better with most scoring metrics (not statistically scrutinized yet…..)
•This season’s real-time web display will include probabilistic scoring and other operational models (NMMB, RRR, SREF,…)
•HMT hypotheses for DTC testing: Best model for guidance, value of ‘hotstart’ analyses, value of ensemble forecast methods
• New direction for 2011: Assess operational and research model microphysical forecasts with HMT-West research observations
Final Thought: UNCERTAINTY, UNCERTAINTY, UNCERTAINTY (forecasts) (observations) (verification)
AGU 2010