NOTE: The nws-changes mailing list is no longer active. The list archives are made available for historical reasons.
On 19 June 2002 a crisis change package was implemented in the Meso Eta modeling system that produces forecast guidance at 00z, 06z, 12z and 18z. The first change involved increasing the frequency of physics calls from every 540 seconds to every 300 seconds. This was found to fix the problem of near surface 2 meter temperatures getting excessively hot during the day. This was observed in (at least) Florida, Texas and California where temperatures exceeded 40 degrees C. An example meteogram from Orlando Florida is attached and shows two periods when the 2 m temperature exceeded 100 degrees F. The soil in this area was very dry and very hot at this time and calling physics more frequently allows the model to make necessary thermal adjustments and avoid the trapping of heat at the lowest levels. COMET's Stephen Jascourt has a website with more examples and detailed descriptions at http://www.emc.ncep.noaa.gov/casestudy/hotFL/heattrap.htm I have also attached an example of the impact on Florida 2 meter temperatures showing a very large area of temperatures in excess of 40 degrees C in the operational run where physics are called every 540 seconds and all of those warm temperatures eliminated in the run calling physics every 300 seconds. The second change involved correcting a small code error in the land- surface model that will increase the amount of evaporation over very dry soil. While the dry soil was connected to the hot temperatures in these cases, this small code error was determined NOT to be a factor. We do expect that 2 m temperatures will remain too warm in the 2-3 degree range due to three remaining factors - all being worked on by EMC. The first involves a slight high bias in the amount of incoming solar radiation (sometimes exacerbated by a low bias in cloud amount), the second involves the lack of a wetlands land use type which causes an underestimation of the green vegetation fraction which yields too little evaporation (and near surface cooling) and the third involves a low bias in the precipitation data (too often based only on radar estimates which can underestimate precipitation intensity) used in the assimilation which leads to too dry soils which again yields too little evaporation (and near surface cooling). These two changes were tested in a 12km parallel from June 3 to June 14. Differences in forecasts were very small except in the areas of unmeteorological hot temperatures. Check out http://www.emc.ncep.noaa.gov/casestudy/timestep/24hfcst.htm for a comparison of precipitation fields. Several more processors have been used in making the forecast to compensate for the increase in work done by calling physics more frequently. This has resulted in NO change in the model completion time. The final change involves a latent defect in our radiation code that up until now had never shown itself. This bug manifested itself in several 06z runs of the Eta Data Assimilation System (EDAS) and caused failures on June 8th, 14th, 15th and 17th. This was tracked down to treatment of very thin clouds that resulted in division by zero. This was fixed by eliminating treatment of very thin clouds and changing the code to eliminate the possibility of ever dividing by zero.
nws-changes archives: