Historically, the BMJ convective parameterization in the operational Eta has been the most aggressive of all convective parameterizations at intercepting water and preventing the grid-scale scheme from blowing up. In contrast, in the AVN, the grid-scale scheme drops most of the water when model precipitation exceeds two inches, even during summer. One consequence shown here is that the Eta seldom predicted 24-hour amounts greater than two inches while the AVN did so too often.
For more detail about how these maps were constructed and an extensive suite of model and observed precipitation climatology maps, consult the COMET Climatology of Forecast and Observed Precipitation case.
With this history, the precipitation "bombs" occurring in the Eta since July 2001 have come as a surprise. As shown on the first page of this case, amounts in excess of three inches occurred in just six hours in two separate episodes in one model run. Amounts over four or five inches in twelve hours are common in similar episodes on other days. The Eta is not alone — similar or worse episodes are typically seen the same days in the AVN and the NSSL version of the Eta running with the Kain-Fritsch convective parameterization. These events did not happen with the 32-km Eta in summer of 2000 nor with the 10-km "nest in the west." Since then, the diffusion in the Eta model was reduced, which has usually resulted in better forecasts. But the excessive diffusion used previously may have helped thwart the development of bull's-eyes of vertical motion and other fields. Also, resolution may have contributed, and as we move to finer resolution models, we may expect to see more of these kinds of "features."
These events in the Eta model have often occurred in environments favoring MCS development, and MCSs with heavy rain have actually occurred on some of these occasions — albeit not usually where the model predicted. For the case we are looking at here, heavy precipitation forecast for South Dakota the first night actually occurred in North Dakota and over a smaller region in Nebraska, while the second night Kansas event did actually occur, though heaviest in two separate areas on either side of the model's bull's-eye. Click the Display Graphic button below to see these forecast and observed 24-hour amounts. Since convection can produce small-scale maxima between stations while remote sensing estimates have various other limitations, both the CPC analysis based on gauge data and the RFC mosaic based largely on radar estimates are shown.
