The recently initiated CAPT (CCPP-ARM Parameterization Testbed http://www-pcmdi.llnl.gov/capt ) project seeks to apply numerical weather prediction (NWP) analyses and model development protocols to the verification of physical processes in lower-resolution climate models. The fundamental premise of CAPT is that during a short-range forecast initialized with high quality analyses from modern NWP data assimilation systems, the climate model and observed physics are responding to the same (nearly) meteorology (dynamics). As the difference between the analyzed and observed (“true”) meteorology narrows, the difference between model and observed physics becomes a model vice analysis error.

While the degree of meteorological similarity will firstly depend on the quality of the initial NWP analyses, even with good or even “perfect” analyses, model-observation differences would have to exceed some uncertainty (e.g., due to chaos) threshold to be considered significant in either a statistical or modeling sense. The objective of the work reported in this paper is to estimate noise threshold through model experiments with different high-quality reanalyses. Specifically, we run the NCEP R2 reanalysis model (T62L28) with its own analysis and that of the recently completed ECMWF ERA-40 for two 30+ day intensive observing periods (IOPs) of the Atmospheric Radiation Measurement (ARM) program in the Southern Great Plains (SGP) in 1997 and 2000. These experiments can be considered “perfect” analysis in that the model is initialized with an analysis from its own data assimilation system. We also run the same experiments with the NCAR CAM model to compare how the analysis-dependent difference varies with model.

Two aspects of the differences are emphasized: 1) meteorology – defined using standard NWP scores and analysis-observation departures; and 2) physics – precipitation and surface fluxes over the ARM SGP site. Additionally, we examine how model-observation differences depend on the short-range forecast length, frequency of forecasts and time of day.

This research was performed under the auspices of the U.S. Department of Energy (DOE) by the Lawrence Livermore National Laboratory under Contract W-7405-ENG-48.