George R. Kasica
Lessons Learned from International Falls, Minnesota February 23, 2007 low temperature forecast error

The "big picture" for International Falls, Minnesota on Friday February 23, 2007 was one of an area under the influence of an arctic high pressure area and a corresponding 500mb ridge as shown in the February 22, 2007 0Z ETA model graphic that is valid for midnight (6Z) on Friday February 23, 2007. However this was forecast to move off to the east of the area during the forecast period and allow clouds to enter into the picture as you can see by clicking on the image below and looking at the animation of the model output.


(February 22, 2007 0Z ETA model output at 30 hours valid at 06Z February 23, 2007 Courtesy of the Pennsylvania State University e-Wall showing the high pressure area and 500mb ridge over the International Falls, Minnesota area - Click image for 30-54 hour animation sequence)

For a better look at the possibility of clouds in the area we can look at the AVN and WRF model cloud forecasts for February 23rd below. In the image we can see that at 12Z (6am) on the 23rd it is predicted that there will be a lid layer of clouds over the area by the WRF model (two left hand panels) and at best partly cloudy skies as International Falls is on the edge of the cloudiness as predicted by the AVN model (two right hand panels).



(February 23, 2007 12Z cloud forecasts are courtesy of the Pennsylvania State University e-Wall)

As a further indication of the cloud potential for the area during the forecast period we can look at the February 22, 2007 12Z Model Output Statistics (MOS) data (shown in full detail here) and by looking at the red highlighted areas we can see that all three models are predicting that the majority of the day will have significant cloudiness by the listings of BK and OV which indicate broken or overcast skies throughout the period.

As a result of the indications for a cloudy day, I had expected the low temperature to remain relatively close to the consensus (or average) of the MOS predictions of a high of 26 degrees and a low of 3 degrees. I chose to modify this slightly and went with a forecast of 24 for a high temperature and a low of 5 degrees to further account for the moderating effects of the cloud cover that was expected.

The actual results for the day were very close (actually exactly correct) for the high temperature at 24F, however, I missed the low temperature forecast by a large margin of 10 degrees as it was actually -5F as shown here in the final day 8 results from the Weather Challenge.

What was the cause of the low temperature forecast failure? Looking at the meteogram below it is fairly obvious that the expected cloud cover that I had counted on to moderate the low temperature early in the period never materialized as shown by the clear circles in the middle image. In addition corresponding light winds indicated by the circle around a circle iallowed the temperatures to fall quite rapidly as shown by the purple temperature trace in the top image below as it rapidly fell from 3 above at about midnight to -5 by shortly before 3am (9Z) that morning.


(February 23, 2007 International Falls, Minnesota meteogram showing the unanticipated clear skies and corresponding rapid temperature fall early in the forecast period. This resulted in the lower than expected minimum temperatures. Meteogram courtesy of the University of Wyoming)

So what did I learn as a result of the missed low temperature forecast error? In my opinion, I need to clearly look at and consider more sources of data than what I had in making this forecast. A good example of this is that the above cloud forecast made by the AVN model (right hand panels) that showed the area around International Falls to be in a possibly clear sky condition. Had I paid more attention to this possibility and as a result investigated it further possibly by looking at current observations (click for detail) from earlier at International Falls as shown below I would have seen that the skies had already cleared out (red highlighted CLR indications) and that the winds had been light to calm (bold green highlights) for some time, therefore my expectation for clouds werefalse. As a result of that I could have adjusted my forecast to account for the clear skies and increased cooling effects that the clear skies and light winds would bring.

Main Page Previous 1  2  3  4  6  Next