|George R. Kasica
Lessons Learned from Tucson, Arizona March 8, 2007 high temperature forecast error
The synoptic "big picture" for the area around Tucson, Arizona on March 8, 2007 showed us a calm pattern with little indication of any significant weather in the area as you can see on the surface chart from 06Z on March 8, 2007. The areas at the time was under clear skies and winds were light as shown by the station models in the SW US at that time.
Looking at the various forecast models there were several small areas of vorticity predicted to pass over the Arizona area during the forecast period (Upper left panels) but the atmosphere would also stay quite dry as indicated by the low relative humidity in the lower left panels, so clouds and precipitation were unlikely at best. (All the below images were provided courtesy of Pennsylvania State University Department of Meteorology E-Wall - click on each for a full size image)
Given the quiet conditions and clear skies, which were quite similar to the previous day, I assumed that the temperatures would be at least as warm, if not warmer, than they were on the previous day and my forecast therefore was for a high temperature of 87, a low of 53, winds of 12 knots and no precipitation. As it turns out the actual results were somewhat different than the forecast conditions with the actual high temperature only reaching 83, the low almost as forecast at 54, winds nearly as forecast at 14 knots and no precipitation as expected as you can see from the final results page here taken from the Weather Challenge website.
Looking at the conditions at Tucson, Arizona they were actually as expected by the forecast, sunny and relatively light winds as shown by the detailed METARs in this link. You can see that the skies remained clear throughout the 24 hour forecast period and the winds were light with the strongest recorded wind speed of 11 knots. So why did the temperature not reach the expected 87 degree high that was forecast? For that we have to look at the conditions in the atmosphere above the area around Tucson. To begin with we can look at the 850mb temperatures and wind streamline analysis from March 7 at 18 and 21Z. In the larger images you can see that there is a pool of cooler air located to the north of Tucson at the tip of Nevada in the 18 and 21Z image that is noticeably cooler than the atmosphere over Tucson at the time and it is motion is in the direction of the area as shown by the streamlines (highlighted by the red circle). Further, looking at the second row set of images (850mb heights and temperatures from 18Z and 21Z on March 7, 2007 you can see that there is a corresponding decrease in the heights from approximately 1530m to 1500m over the area, another reflection of the influx of cooler air as was shown in the top two images. Lastly, we can look at the results of the cooler air fully arriving in the area on the last set of images in the third row from 18ZX and 21Z on March 8, 2007 where we can see the heights still maintained at 1500m at 18Z and falling still further with cool air arriving to near 1470m by 21Z. The last image shows the highlighted flow of cooler and warmer air over the area (blue and red arrows) for 18Z on March 8, 2007.
So what effect did this cooler air aloft and then mixed down to the surface have on the temperatures in Tucson, Arizona and why didn't I notice it? A good gauge of the effect on the surface temperatures can be seen by looking at the model interpolation grids as used to predict high temperatures using the Delta Method as seen in a previous city, San Antonio, Texas. Looking at the Delta grids below, we can see that the March 7, 2007 12Z ETA forecast model reflects about a three degree temperature drop in the forecast surface high temperature and about a 1C (2F) drop in temperature at 850mb. Similarly the GFS predicted a 1F drop in the surface temperature and a 2C (4F) drop at 850mb. If we apply these changes to the high temperature for March 7, 2007 which was 87F as listed by the Weather Challenge site (green calculations) we can see that the high temperature predicted ranged from 83-86 degrees with several of the predictions quite close to the actual reading of 83F rather than my forecast of 87. In any case all showed a cooling trend that, had I computed the Delta method at the time would have been a tip off that I should look more closely at what was happening and adjust my forecast high temperature downward. Had I taken a consensus number of the 2m temperatures I would have gotten a forecast high of 85 and the 850mb consensus would have yielded 84F, both much closer to the actual reading of 83F than my forecast of 87F.
So as a result of this, what lessons were learned here? In my opinion, I need to be more alert for subtle changes in the atmosphere, not just at the surface, since none of the above was even apparent on the surface charts, but also in the layers above the ground and their eventual effects on the conditions at the surface. What may seem like a calm weather pattern with no obvious changes may in fact as was shown here have subtle activities going on above the surface that make an impact on the surface conditions. To put it in a more direct context, I need to consistently remember that the atmosphere is three dimensional and remember to look at all three dimensions when making a forecast. I see myself too often falling into the pattern of looking at surface charts or one layer of the atmosphere possibly, rather than ALL the layers and getting only a partial picture of what is potentially occurring as a result. In this case, had I looked at the 850mb analysis and or the grid interpolations I would have easily gotten clues that there was something more than a quite weather day like the previous one going on, not at the surface, but above the ground.