This is the AVN MOS Wind Development Test Results Page

Development was done by using data from 335 stations ; 300 were from the contiguous U.S., 30 from Alaska, and 5 from Puerto Rico and Hawaii. This was a single station development.

TESTING SAMPLES
The full cool season sample was from:

The independent sample consisted of : The dependent sample consisted of: The full warm season sample was from:
  • April 1, 1997, to October 15, 1997
  • March 16, 1998, to October 15, 1998
The independent sample consisted of :
  • the last 15 days of April, May, June, July, August, and September 1998
The dependent sample consisted of:
  • the remainder of the test sample listed above

PREDICTANDS
Speed
  • Valid every 3 hours from 6 to 72 hours after initial model time
u- and v- wind components
  • Valid every 3 hours from 6 to 72 hours after initial model time
PREDICTORS
Model Predictors
  • Forecasts of u- and v- wind components and wind speed at 10m and at isobaric levels in the lower troposphere
  • Thermal stability and relative humidity over layers in the lower troposphere
  • Relative vorticity, divergence, and vertical velocity at isobaric levels in the lower troposphere
Observed Predictors
  • Surface u- and v- wind components and wind speed observed at initial model time +3 hours,i.e., 0300 or 1500 UTC. Observations are potential predictors for the 6- through 15-h forecasts projections.
Geoclimatic Predictors
  • The sine and cosine day of the year and twice day of the year were used
POST-PROCESSING FORECASTS
Wind Speed
  • Wind speed forecasts were inflated for each individual station in order to better forecast higher wind speeds
Wind Direction
  • Wind direction is forecast by using the u- and v- wind components
VERIFICATION MEASURES
Wind Speed
  • Mean Absolute Error of all forecasts
  • Heidke skill score(based on categories of 0-12, 13-17, 18-22, 23-27, 28-32 and >32 knots)
  • Probability of Detection of winds greater than 22 knots(no regional scores are available because the sample sizes are too small for evaluation)
Wind Direction
  • Mean Absolute Error when the observed wind >= 10 knots
  • Cumulative Relative Frequency of forecasts with errors of 30 degrees or less, when the observed wind is >= 10 knots
OVERALL RESULTS
Conclusion
  • The AVN MOS wind speed and direction guidance appeared to improve significantly over the NGM MOS guidance and the direct model output

To view the test stations used click on Description of AVN MOS Test Stations

Please select a product from the menus on the left.