Here is an example of a trend for the Eta run.  The most recent run is in yellow (12Z Eta) and the older run is in blue (00Z Eta).   500 mb comparison is on the left and PMSL is on the right.

Note the newer Eta run is slower and deeper with the Pacific system and is indicating more emphasis on the east coast with the surface low.

Here is the trend loop for the GFS run.  This is a more consistent model solution in both areas of interest as compared to the Eta, but even so it too indicates a slight tendency to shift the east coast surface low toward the coast.

The above gave a run to run trend comparison for only 2 runs - the 00Z and 12Z runs.  Typically you should choose cycles 12 hours apart to better discern a trend signal.  Viewing a trend across model cycles less than 12 hours apart may not provide a strong enough signal.

In addition to a trend loop, you can view a trend at a specific valid time.  This is particularly useful for discerning trends for a given feature at a given time.

For example, the loop below is from 4 consecutive in sequence model runs focusing on the east coast low:

• 18Z Eta from July 14 2004 valid 06Z July 18 2004 at (fhr 84)
• 00Z Eta from July 15 2004 valid 06Z July 18 2004 at (fhr 78)
• 06Z Eta from July 15 2004 valid 06Z July 18 2004 at (fhr 72)
• 12Z Eta from July 15 2004 valid 06Z July 18 2004 at (fhr 66)

Note that the 06Z Eta seems to be an outlier.  Loosely, the trend of the Eta is to put more emphasis on the coastal low.

A similar trend from the GFS a similar trend as the Eta, but more subtle:

• 12Z GFS from July 14 2004 valid 06Z July 18 2004 at (fhr 90)
• 00Z GFS from July 15 2004 valid 06Z July 18 2004 at (fhr 78)
• 06Z GFS from July 15 2004 valid 06Z July 18 2004 at (fhr 72)
• 12Z GFS from July 15 2004 valid 06Z July 18 2004 at (fhr 66)

At times the trend can offer a subtle signal (like this case above).

Below is a trend with a strong signal (for a different event) - focus on OH/IN...

Note the consistent westward trend for both the QPF max and vertical motion max in the OH valley.  Also note the cooling trend in KS.

Another way of viewing trends are in multipanel formats.  Below is an example of a QPF trend over the CONUS.  The Eta trend is on top and GFS trend on the bottom with the most recent on the left and oldest on the right.

Note the trend in QPF in northern WY for both models.  On this scale its very hard to discern but you can see the GFS trend toward the Big Horn mountain range in WY the closer we get to verification time.   BE CAREFUL - subtle differences can have huge implications to a WFO.

View this image below.  First note the difference in model solution between the Eta (red) and the GFS (green) in the upper left corner of the image.

Then note the trend for the GFS lower left and the Eta lower right (following the above convention of older run in blue and newest run in yellow).

The mid level flow at 700 mb over UT between the Eta and GFS is tremendously different and therefore needs to be considered.

Just based off of this information the trend of the Eta is to forecast a stronger and therefore slower system like the more consistent GFS.  Since the Eta is trending to a GFS like solution.. the GFS is favored.

You may have heard the saying "The trend is your friend".  This is NOT always true.  Trends noted in longer lead times can reverse as you drawer nearer to verification time.  The best trends are those that are in the short time frame (earlier than fhr 36) and comprised of at least 4 model consecutive model cycles.

Back to the case at hand - we do not note a strong trend signal, but one that did indicate a low formation and track closer to the coastline.

For the east then we have the following pieces of information:

1. No significant errors noted in the initialization
2. A weak trend toward a coastal solution for the low
3. Model differences were confined below 500 mb to the surface
4. Eta can resolve terrain features better than the GFS

(continue)