A significant bug fix was made for the 12Z run this morning (Tues 20 Jan) in the RUC2 forecast model.

Previously, the RUC2 convection parameterization was set to try to saturate the cloud top to the water-saturation value, even if the temps were very cold. This is the detrainment at cloud top level. However, the RUC2 microphysics understands that saturation at colder temps is with respect to ice, so tremendous amounts of non-physical latent heat release were occurring. We had seen some bad behavior regarding this before, but were spurred on to investigate more throughly by 2 crashes that occurred on Friday 16 Jan that were linked over the weekend to this same problem. Somewhat miraculously (considering the timing), John Brown in our group realized that the convection parameterization could be very simply set to account for ice saturation at colder temperatures (currently set for < -15 deg C), and this seems to solve the problem in a physically consistent manner.

For the case that we reran, precipitation patterns were very similar, but slightly more coherent with the new version. There will be a decrease in spurious warm temps near estimated cloud tops in convective areas, and drier conditions and less ice aloft in the same areas. Wind anomalies were sometimes occurring in the same regions, but not now. Perhaps more importantly, the source of the crashes is removed.

Stan