WPC Home
About the WPC
WPC Product Info
Current Products
Experimental Products
WPC Verification
Model Biases/Verif.
Office and Staff
International Desks
Photo Gallery
What's New
Other Sites

Weather Prediction Center
Camp Springs, MD

Model Biases

This page is no longer maintained. Please visit the page below and update your bookmarks accordingly: /mdlbias/biastext.shtml

Updated 03/31/2000

*** NOTE: Changes have been made to the Eta/AVN/MRF models. Thus biases for these models are no longer displayed. ***

Go To Model:    [NGM] [UKMET] [ECMWF]

The biases and model performance characteristics listed below are based primarily on WPC forecaster perceptions of model performance. Many performance characteristics are undocumented and offered in hopes of stimulating discussion and local model studies. The information is also offered in hopes that they will be useful to others. Comments on any of these biases are welcomed. The WPC will try to keep the homepage up to date and correct any errors. Readers should try to verify the existence of any of these perceived model biases before applying them to a forecast situation.

Short Range Models


  • Overdevelops surface cyclones over land and underdevelops them across the oceans. Can be too slow to fill weakening cyclones.
  • Has a northern bias with respect to storm track prediction. Generally is the farthest north of the three short-range models discussed here with systems moving into the Plains from the Rockies.
  • Tends to overpredict surface pressures for anticyclones across the western USA.
  • Tends to overpredict the strength of sea-breeze convergence zones, and resultant precipitation.
  • During the warm season, the NGM has a high bias, overpredicting the areal coverage of light precipitation (amounts <0.50).
  • Near the Gulf of Mexico, the NGM very often underforecasts 0.50 to 1.00 rains, and GROSSLY underforecasts amounts heavier than that.
  • The NGM has a cold bias with respect to forecasting 1000 to 500-mb thicknesses over surface cyclones, during the winter.
  • Can be slow in deepening surface cyclones across the Gulf Stream during winter.
  • Tends to overdevelop surface cyclones in the lee of model terrain features.
  • Predicts too much stability through too deep of a layer with temperature inversions.
  • The NGM can be the weakest of all models with digging troughs across the western US.
  • Has a bias to underforecast precipitation amounts onto the west coast of the US in the presence of a blocking ridge across the Gulf of Alaska.
  • All models, especially the NGM, forecast too much precipitation across Texas during pre-frontal squall line situations. That is, once the squall line has moved east and mid-level winds have veered to more westerly, the rain event for Texas has usually ended. Often the NGM will still predict rainfall. This bias is most true during spring and fall/early winter.
  • The overall errors in the NGM decrease during the warmer months.
  • Has trouble handling return flow cases after strong cold fronts have pushed into the Gulf of Mexico.
  • Too far north and too strong with systems coming out of the Rockies; fails to consider upstream shortwave and associated jet streak on the west side of the trough.
  • Overforecasts upslope precipitation over the central/southern Rockies.
  • Too slow with Arctic airmasses plunging southward down the plains.
  • Factors affecting NGM QPF:
    • Smoother terrain:
      • NGM overpredict upslope pcpn and have it too far east of Rockies into plains.
      • NGM overpredicts pcpn in the lee of Cascades/Sierras; but not enough pcpn along windward slopes.
      • NGM has little skill with QPF over interior west, partially due to lack of terrain detail.
    • Sea breeze:
      • Sea breeze occurs below horizontal scale of NGM grid points; thus, NGM overpredicts scale of sea breeze and associated boundary layer moisture convergence.
      • NGM also overdevelops lee trough east of southern Appalachians.
      • Result of above is a high bias for Mid-Atlantic and Gulf states during warm season.
    • Convective Parameterization:
      • Kuo convective scheme stabilizes atmosphere at higher levels in vertical than would be done by grid-scale precipitation processes
      • Results: Overprediction of light precipitation and underprediction of heavier precipitation.
    • Evaporation:
      • NGM surface evaporation too high during warm season; likely a factor in overprediction of measurable precipitation
      • NGM surface evaporation over oceans held to half the theoretical value to suppress spurious light pcpn
      • NGM return flow off Gulf too westerly; this (along with restricted ocean evaporation) likely responsible for cool season low bias of heavier amounts across south.

  • NGM QPF performance...biases cool/warm seasons; light amounts (>0.01):
    • Cool season:
      • High bias over Pacific Northwest and lee of Cascades.
      • High bias over portions of interior west and through central/southern Rockies
      • High bias across the deep south; otherwise, generally a reasonable bias east of the Rockies.
    • Warm season:
      • High bias along west coast (central California northward)
      • Low bias over portions of interior west.
      • High bias over portions of upper Mississippi Valley and Midwest
      • High bias across the south
      • Generally a high bias east of the Rockies

  • NGM QPF performance...biases cool/warm seasons; heavier amounts (>0.50)
    • Cool season:
      • High bias lee of Cascades/Sierras; but a low bias along windward slope/coastal sections
      • Low bias over portions of interior west
      • High bias upslope regions of central/southern Rockies
      • High bias over northern/central Plains
      • Very low bias across the south (NGM grossly underpredicts heavy rainfall in the cool season)
    • Warm season:
      • High bias lee of Cascades/Sierras; but again a low bias along windward slopes/coastal sections
      • Low bias over portions of interior west
      • High bias over southern Rockies
      • High bias over Northern Plains/Upper Mississippi Valley
      • High bias across the south (reverse of cool season)
      • Generally a high bias east of Rockies

  • NGM QPF performance...Threat Scores; comparing cool/warm seasons; heavier amounts (>0.50)
    • NGM has higher threat scores in the cool season
    • NGM has higher threat scores along and north of the storm track
    • NGM has its highest threat scores over the Northeast U.S. in the cool season
    • NGM has a band of higher threat scores extending southeast from the Midwest to Southern Appalachians in the warm season (possibly associated with MCS's?)
    • In general, NGM has low threat scores over the interior west in any season
    • NGM underpredicts low level moisture and thus grossly underpredicts rain over the south during the cool season because its Gulf return flow is too westerly.
    • NGM too cold at lowest levels; too warm at and above 850 mb.
    • NGM has trouble simulating the inversion during cold outbreaks

Medium Range Models

General Biases and Comments

  • Models (short and medium range) do not drive arctic fronts fast enough and far enough south through the Great Plains in the winter. ECMWF usually has the best handle on such fronts of the medium range models, but is still usually a bit too slow.
  • The ECMWF and UKMET phase northern and southern stream short wave energy too readily over the entire forecast domain. (This is more true of the UKMET) Correction: Maintain separate streams unless the longwave pattern and trends from the short range models clearly support phasing. The longwave pattern may support phasing on days 4-5 if both the northern and southern stream flows are fairly high in amplitude and fit the mean upper pattern. Don't rely on the speed of an individual short wave. This can be a major correction that often makes you decide the fate of a major storm.  
  • The ECMWF and to a lesser extent the UKMET will attempt to close off unphased systems too often. Correction: Keep higher amplitude but progressive troughs and open short waves moving along except in the presence of a high amplitude upstream ridge or if the trough is bumping into a downstream block.  
  • Closed/closing off upper lows are too fast.
  • All models may incorrectly amplify short waves moving through low amplitude 500 mb flow.
  • The models will often amplify or key in on a particular short wave trough when in reality the next in a series of troughs will be more significant and dampen the first wave out. This often occurs during retrogression of the upper pattern. Correction: Weaken short waves and associated surface development if the model produces a strong solution when wavelength spacing between series of systems is short. 
  • Arctic air will plunge southward at a more rapid pace to the lee of the Rockies and the Appalachians than forecast by medium range models (especially the UKMET) or most short range models.
  • Surface development is overdone in an arctic airmass by all models.
  • Surface cyclogenesis is underdone over the Great Basin ahead of significant upper troughing.
  • Storm motion across the Gulf of Alaska is too slow and too far north (all models).
  • Model surface lows over the oceans deepen too gradually.
  • Model surface lows over land are often too deep.
  • Model surface low development directly underneath a closed upper low is often underdone.
  • The UKMET/ECMWF surface high west of the Rockies is often much too weak.
  • The models underforecast cyclogenesis off the Carolina coasts in the presence of improving upper dynamics.
    Correction: Initiate cyclogenesis as the upper ridge axis reaches the coast.
  • Convective precipitation amounts are usually underdone and overrunning precipitation amounts are usually overdone (all models).
  • All models underestimate precipitation amounts along the west coast due in part to a lack of upstream data and a unrealistic orography depiction.
  • Precipitation is underforecast over mountains by all models due partly to inadequate terrain parameterization.
  • Lake effect snows are underforecast.
  • Model precipitation lingers too long over the high plains even after upslope decreases and upper dynamics shift eastward away from the area.
  • Precipitation produced mainly by warm air advection often begins about 6-12 hours earlier than forecast by the models east of the Rockies assuming the surface pressure field verifies as forecast.
  • The models underestimate precipitation amounts when a subtropical moisture plume is feeding into a weather system.
  • Underforecasts monsoonal precipitation across the west during the warm season.
  • The models often do not have proper initialization over the Pacific Ocean (all models). This can be especially true with tropical storms.
  • The Gulf of Alaska (and elsewhere) can be a real trouble spot in several respects:

    1. All models periodically have some of their largest observed errors in depicting the mean pattern in the Gulf of Alaska.

    2. Superimposed on the errors in the longwave pattern can be additional significant errors in identifying and timing shortwaves. These type errors are at their maximum in faster than normal zonal flow with short spacing between systems.

    3. Models typically handle high moderately high amplitude systems with fairly long spacing between trofs and ridges much better over the Gulf of Alaska ( and everywhere else, for that matter.)

    4. Transitions can be troublesome for all models. Not only are seasonal transitions difficult, but a large scale transition from high to low longwave amplitude, or vice versa, is a major source of errors in all models.

    5. Another potential source of large error comes with extremely high amplitude ridges in the Gulf of Alaska and the forecast supergradient flow crossing to the right of the height contours on the east side of the upper ridge. This usually happens in the colder seasons, often leading to the formation of an upper low somewhere along the West Coast of the US.

    6. Concerning these developments...the ECMWF does fairly well (with occasional big busts) in catching the genesis of cutoff lows along the W coast of the US. The UKMET/NAVY NOGAPS/CANADIAN are typically too progressive with the flow from the Ern Pacific into the Wrn US. Not digging upper energy far enough S, they are prone to miss closing off systems.

    7. The MRF ensemble members can often be helpful in determining when any model, particularly the operational MRF, is an outlier and straying too far from reasonable solutions offered by consensus of all the available models.

    8. A West Pacific typhoon, in the process of becoming extratropical as it enters the westerlies, can significantly alter run-to run continuity in all the models. Poor initialization of a typhoon can lead to large errors in the longwave pattern over the Pacific by Day 3 (and over most of NOAM by Day 5) as well as errors in the shorter wavelength features.


  • Has problems with shallow cold air.
  • Tends to progress shorter wavelength features too quickly.
  • Westerlies are often too far south.
  • The model tends to lower surface pressures too much and too far south and often implies synoptic-scale fronts too far south.
  • Often too low with heights along the southern ends of short wave trofs, resulting in a southward displaced storm track
  • Significant low pressure bias during the warm season over the western U.S., particularly when upper ridge conditions predominate
  • Breaks down amplified long wave patterns too quickly
  • Over the east, tends to be too weak and strung out with surface cyclogenesis along fronts when a significant short wave is digging into the trough position. This results from the forecasted baroclinic zone being too broad and low-level temperature gradients being too weak. (Doesn't take into account latent heat release from the Gulf Stream).
  • Of the medium range models, it has the poorest performance forecasting polar vortices.
  • Often is too flat with the upper and surface patterns over the western Pacific.
  • Eastward forecast bias with upper lows, especially over eastern Canada, possibly due to its tendency to flatten amplified patterns too quickly
  • The UKMET often develops an upper system too far to the south over the Gulf of Mexico and the northwest Caribbean. This is a tropical interaction problem or applies directly to tropical cyclones.


  • Compared to the other 2 operational models described above, the ECMWF does the best in predicting mid/upper tropospheric heights during the colder part of the year(such as October through April). The ECMWF tends to perform quite well in predicting amplitudes of planetary-scale regimes such as the Pacific/North American teleconnection (PNA). This model can also perform outstandingly during low to high planetary-scale wavenumber transition events, and northern hemispheric-scale regime transitions (Berry et al. 1996, CR TM 111).
  • Outperforms the other medium-range forecast models during shallow cold air situations.
  • Tends to overdevelop mid/upper cyclones across the southwestern U.S. Situations arise where this model will be too slow to predict the movement of cyclones from the southwest deserts.
  • Has a slight tendency to forecast mid/upper tropospheric heights and the resultant thickness calculations too high (i.e.; a warm bias).
  • Sometimes, especially during the warmer portion of the annual cycle, this model has too many closed lows. This bias may be related to its high resolution.
  • Tends to overamplify the long wave pattern, resulting in slower than observed progression of systems through the westerlies. This can result in overly weak and northward displaced short waves and associated surface features lifting into the long wave ridge position.
  • Found to have the smallest overall distance errors with springtime closed low forecasts during days four and five.
  • Westward forecast bias of closed cyclones (related to the issue described above)
  • Often too slow moving short wave features in deamplifying or zonal patterns
  • Of the medium range models, the ECMWF performs best with driving Arctic fronts down the east slopes of the Rockies.
  • The ECMWF too often incorrectly digs closed upper lows SWWD then WWD underneath strong upper ridges over the Eastern Pacific.

Questions? Comments? Click here to send mail to the WPC.

Last Modified May 31, 2001