(back)

Initialization Assessment Techniques - Moisture
A comparison of RH to WV imagery and observed PWs to initialized PW can be conducted In lieu of comparing RH against observations at multiple levels.  Therefore, its good practice to compare the 400 mb RH to WV imagery.

As you can see below the Eta compares well for the most part to WV imagery.  Notable discrepancies include the moisture plume extending from northwest CA to southwest over the Pacific and northwest of the remnants of the tropical system (Blas) offshore of Baha.  It also seems to have been underdone a bit from western Oregon and into ID.

The GFS initialized the moisture better with Blas, but like the Eta seems a bit underdone in the moisture plume from CA and points offshore.. and also a tad dry in southwest WA and west ID.  But more notable is the orientation of the GFS RH with the strong short wave in the Pacific.  This reinforces the assessment that both the GFS and Eta seem to have under initialized the strength of this feature.

A comparison of the observed PW values over North America compared to those initialized by the model will help bolster the moisture assessment.

The Eta's PW initialization is contoured in blue and shaded with the observed values in yellow.  By sampling the model initialization the Eta has over initialized the PWs in a band from LZK to JAN and BMX.  This is in a region of a surface front and may have a bearing on QPF amounts predicted by the Eta.

Further west the Eta is not comparing well with the observations from CA to ID.

Meanwhile the GFS in the west seems to have an unsubstantiated minimum over northern CA and southern OR.  Both models are struggling with PWs in this region.

In the east the GFS has done a better job in the region of the front than the Eta, but it too is over initialized by at least 10%.  At this point it is time to relay this information to the Day 1 QPF forecaster.

Wind assessment
Now that you have made an assessment of circulation and moisture, we can proceed to wind assessment.  At a minimum you should assess upper and lower jet levels.

A good place to start is at jet level.  Below is an image of the wind observations on a North American Scale over the GFS (shaded) initialization of wind at 250 mb.  Note the different types of wind observations available for via AWIPS to utilize in your assessment:

As with PWs, you can run your cursor over a gridded field and "sample" the output.  This is an easy way to compare the value of the grid vs the value of the observations.  Pay special attention to discrepancies in regions of sharp gradients.

There seems to be a jet axis running from just offshore of CA through southern OR to southeast WA and southern Alberta.  Because this axis looks to be small in scale, it apparently has not been picked up by the model.  It is possible the resolution of the display is unable to depict this feature.  In which case, this discrepancy can be ignored.

Further east the Eta has done a good job resolving the jet structure over the upper plains and northeast, although it was a bit high with the jet maxima near ND and MA.

The GFS in the west seems to have done a slightly better job than the Eta...

..and did better with the jet maxima in ND...

.. and just offshore of the homeland (MA).

Similarly, a comparison can be conducted for mid and low level jets at 700 and 850 mb.

Height and Temperature Assessments Aloft
A quick comparison of Eta and GFS heights and temperatures at standard levels to observations reveals no significant discrepancies.

CAUTION !  AWIPS only shows height info from RAOBs...and there is a lot more data that goes into an model initialization than raob data.
The only operational tool known allowing a user to compare model initialized fields to the observations is found at http://www.hpc.ncep.noaa.gov/moc/moc.html

However, it usually lags the current cycle by about 6 hours.

Below is the display showing the f00 GFS 500mb contours and the raw observations that were ingested in the GFS initialization scheme.   The color dots represent the observation type (legend provided on the bottom right of the image).  In the region of interest off the Pacific Northwest Coast most of the observations are satellite based...but there are a few aircraft observations.

If we choose to view how these observations compare to the analysis, most of the obs are lower than what was initialized by the model (cool colors indicate observation values lower than f00 and vice versa for warm colors - according to legend in bottom right).


 

This further substantiates our initial impression that the models have under initialized the strength of the Pacific shortwave.

Surface Temperature and PMSL Assessments
You can compare the model initialized PMSL field to the HPC surface analysis valid at the same time.  You should also compare the observed temperatures to the models' initialization.

In the east the Eta PMSL in red compares quite well to the manual analysis in blue.

However some discrepancies are seen in the central plains where the Eta has over initialized a weak surface high.  Since the discrepancy is only 1 mb, it can be ignored.  In the intermountain region, the Eta does not match very well to the observations with some differences of about 2 mb.  Given the variability of terrain in this region, a 2 mb difference is acceptable.

Similarly the GFS has captured the PMSL pattern well in the east...

..and offers a better depiction of the surface pattern in the central plains and the intermountain region than the Eta, but not quite matching the manual analysis.

Finally a look at surface temperatures near the front in the east (the only real region of thermal contrast observed) reveals both models provided a good depiction of the thermal structure in the front's vicinity.  Although the Eta is a bit warm from just north of the surface front in MS and AL.

So too is the GFS.

At this point you have looked at enough data to make a crude assessment of model initialization.  Remember, ideally you would do this over the entire model domain in 3 dimensions. Realistically, limitations in time and technique prevent assessment to that level of detail from being made.

4. Implications of Initialization Errors
Over the course of your assessment you should have found some discrepancies (it is rare that you do not).   In general describe discrepancies in the discussion if they:

Very often you will find discrepancies at one level and area incongruous with other discrepancies.  They may not represent a difference in phase or amplitude, but are noted near regions of gradients or where significant weather will evolve from.  The only course of action in this case is to just state the discrepancies in the discussion.

In general you should try to use your meteorological background as an aid in anticipating the impact of discrepancies.  For example, an upper level jet that is weakly initialized by the model may imply the model's secondary circulations may be weak and hence a-geostrophic adjustment at the surface may be under predicted.

There are no guarantees your assessment of discrepancies will always lead you in the right direction in terms of model choice.  However, you will find that over time, your choices will lead to using the output more knowledgeably as compared to using it verbatim.

You should coordinate significant initialization discrepancies with your fellow staff.

5. Short term validation
Strictly speaking this is not part of initialization assessment, but by the time you begin your initialization assessment, you will likely have model output available out through fhr 03.  You should look at the short term performance of the model compared with the observations.  The idea is to see if the model is going awry in the short time frame.  If it is doing so significantly, it is more than likely it will not recover and get back on track.  This kind of discrepancy should be noted in the discussion.

(continue)