JEFS Status Report Department of Atmospheric Sciences University of Washington

1
JEFS Status ReportDepartment of Atmospheric
SciencesUniversity of Washington

• Cliff Mass, Jeff Baars, David Carey
• JEFS Workshop, August 28-29

2
Who is working on JEFS at the UW?

• Atmospheric Sciences
• Cliff Mass-main faculty lead
• Jeff Baars-research scientist
• David Carey-research scientist
• Some assistance from other staff
• Statistics
• Tilmann Gneiting-main faculty lead.
• McClean Sloughter-grad student working on precip
  and wind bma
• Chris Fraley-staff, software support
• Some assistance from Adrian Raftery-Prof--Mr. BMA

3
The Two Basic Requirements for Post Processing of
JEFS Forecasts are

• Quality controlled observational data for the
  region of interest of all necessary parameters
• Reliable JME grids for several months at least.

4
Acquiring and QCing JEFS Obs Data

• Not in the Statement of Work (SOW), but a
  critical requirement for JEFS.
• Developed a robust QC system at the UW that has
  been tested over the last year and recently
  transferred to AFWA.

5
UW Observation Quality Control System

• Range Check
• A simple gross error check to make sure the
  observation is within a reasonable range.
• Step Check
• Looks for "spikes" in an station's time series
  for a given variable.
• Persistence Check
• Looks for variables that are not changing over
  extended periods, indicative a failed or frozen
  sensor.
• Spatial Check
• Compares each observation to those of surrounding
  stations. Specifically, designed to handle areas
  of complex terrain-like Korea and the NW U.S.

6
Quality Control Display and Warning System

• This system is now working on the UW Northwest
  test bed for a heterogeneous collection of
  roughly 70 different networks.
• View it in real time at
• http//www.atmos.washington.edu/qcreport/

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
JEFS Observations

• Recently began receiving observations for 3-, 6-,
  and 12-hr precipitation.
• in addition to temperature, dew point,
  altimeter, sea level pressure, wind speed and
  direction.
• Now receiving Korean Mesonet Data.
• Format of obs still being adjusted

11
Forecast Verification

• Another major issue is to have forecast
  verification code running operationally, since
  that is essential infrastructure for
  post-processing.
• UW has developed code called verobs which is
  being transferred to AFWA

12
UW-to-AFWA Code Transfers for QC and forecast
verification

• Observations Quality Control Fortran code
  transferred to Scott Rentschler at AFWA, March
  2007.
• AFWA needs to use this software to flag and
  remove observations from the collection we are
  receiving.
• Forecast verification Fortran code, verobs,
  transferred to Scott Rentschler at AFWA, August
  2007.

13
Next Item Bias Correction Software Development
and Transfer to AFWA

• As noted many times in the past, bias correction
  substantially improves probabilistic prediction
  since systematic bias does not represent true
  uncertainty.

14
Bias Correction of T2 and TD2

• Have bias correction running in real-time for 2-m
  temperature and 2-m dew point temperature for the
  UW Mesoscale Ensemble System (UWME).
• Uses optimized settings.
• Have gotten this running on the JME/WRF output as
  well, using older generation JME model data.
• Results shown at JEFS technical meeting,
  Sep-2006.
• Ready to transfer code to AFWA.

15
Bias Correction of PCP12

• Have developed bias correction settings for 12-hr
  precipitation for the UWME.
• As with T2 and TD2, settings were determined
  through an optimization process.
• Removed land use in gridpoint to obs
  matching using bins for precipitation amount
• Included in Weather and Forecasting paper, along
  with T2 and TD2.
• Results look good

16
Results Bias Corrected PCP12, forecast hour 48,
2004-July to 2005-June, using UWME system
Metric Uncorrected Bias Corrected
Mean Absolute Error (inches) 0.060 0.046
Mean Error (inches) 0.018 -0.013
Help Fraction N/A 0.133
Hurt Fraction N/A 0.047
17
Daily Domain Averaged ME for PCP12, f48, December
2004
Daily Domain Averaged MAE for PCP12, f48,
December 2004
18
Daily Domain Averaged ME for PCP12, f48, October
2004
Daily Domain Averaged MAE for PCP12, f48, October
2004
19
Local Bias Correction of DIR10 and SPD10

• Currently working on 10-m winds.
• As with T2 and TD2, a statistically significant
  correlation between model error and land use
  appears to exist, especially for wind speed.

20
Next Issue Transferring BMA to AFWA

• Local BMA for temperature is now working
  operationally over the Northwest and is ready to
  be transferred to AFWA. FORTRAN Code.
• Same approach can be used for dew point
  temperature.
• Global BMA is working for precipitation.
  Currently written in r and must be converted to
  FORTRAN before it can be transferred to AFWA
  (Jeff Baars is doing this now).
• Local BMA for precipitation and BMA for wind has
  been an area of active work between atmos sci and
  statistic and prototypes are working, but still
  being improved and evaluated. The next talk will
  describe this more.

21
We have continued to maintain and improve PROBCAST

• Probcast is our attempt to package uncertainty
  data in a form accessible to laymen and less
  experienced forecasters.
• A version of probcast will be adapted for DOD use
  when the underlying JEFS system is in place.
• Are now extending the verification of Probcast
  over the NW.

22
Before Probcast The BMA Site
23
PROBCAST
24
Verification of Probcast temperatures web tool
25
Next McClean Sloughter, UW Statistics