EMC: Eta Model NTIM Jan 96

WD2200 W/NP22

Here is some information (extended NTIM) on the recent changes to the early Eta Model run.

Geoff DiMego, Ken Mitchell and Zavisa Janjic

CHANGES TO THE OPERATIONAL Early' ETA MODEL

At 1200 UTC 31 January 1996 the forecast model code and post-processor used for the operational early Eta model run were changed to be identical with the codes now running the Mesoscale Eta model. The main reason for this change was to avoid the occasional non-reproducable failures that were occurring with the original early Eta version. Since NCEP's available resources are limited and the prospect of quickly finding the source of the error were slim, the decision was made to implement the Meso Eta code which has been running without failure since July. The configuration of the early Eta remains 48 km and 38 levels while the Meso Eta runs at 29 km and 50 levels. The content and timeliness of current Eta Model products available on the OSO server and the NOAA Information Center anonymous ftp server ( nic.fb4.noaa.gov or 140.90.50.22) remains unchanged.

As a result of this change, two areas of the Eta Model's parameterized physics have been upgraded as described below.

1) LAND-SURFACE PHYSICS: A two-layer soil model with a vegetation canopy replaced the former single-layer bucket model. Soil moisture and soil temperature are explicity forecast in the two soil layers, along with a surface skin temperature based on the surface energy balance equation. Evaporation is comprised of 3 components including a) direct evaporation from the soil surface, b) direct evaporation from wet vegetation, and c) transpiration through the vegetation canopy. To support this new treatment, the Eta model includes N. American gridded fields of 8 soil types, 12 vegetation types, monthly vegetation greeness fraction (interpolated to day of year) to capture the seasonal cycle of vegetation and quarterly surface albedo fields. Soil moisture and soil temperature are currently initialized from the global model data assimilation system (GDAS), which carries a similar land-surface physics package. The snow cover is still updated daily with the operational snow cover analysis from the Air Force. Near future plans include replacing the GDAS-based initial soil moisture with soil moisture from a continuously cycled Eta Data Assimilation System that includes assimilation of an hourly precipitation analysis. (See upcoming article by F. Chen, Ken Mitchell et al.; 1996 Journal of Geophysical Research)

2) TURBULENT VERTICAL TRANSPORTS: In the surface layer formulation of the Eta Model, a viscous sublayer scheme has been implemented to accommodate the new land surface scheme. With this method, the ratio of roughness heights for temperature and momentum is a function of the Reynolds number. A correction is also applied to avoid the singularity in the case of free convection. A fraction of the surface buoyancy flux is converted into kinetic energy of unorganized flow near the surface so that the friction velocity, and, therefore, the Monin-Obukhov length remain nonzero. The integral similarity functions have been replaced by more suitable ones. Above the surface layer, the application of Mellor- Yamada Level 2.5 scheme for turbulent kinetic energy (TKE) was reviewed in order to (a) identify minimum constraints that guarantee satisfactory performance of the scheme for the full range of atmospheric forcing ("realizability"), and (b) develop a robust, accurate and affordable computational procedure. In the new formulation only one constraint is retained and it is applied to a diagnostically computed "master length scale". The order of computations has been rearranged, the equation for the TKE production/dissipation is solved iteratively, and the empirical constants have been revised. Results obtained with the new scheme remain in reasonable agreement with field-study observations, even in the case of very strong instability.(see upcoming paper by Z. Janjic, or here for more details)

Based on model case studies, retrospective parallel testing, and concurrent parallel testing, the following changes should be expected in the early Eta forecasts: