Using ARM Observations to Evaluate Continental Surface Processes in Atmospheric Climate ModelsPrimary Author: Phillips, Thomas
Additional Authors: CAPT team
Using ARM Observations to Evaluate Continental Surface Processes in Atmospheric Climate Models
Thomas Phillips and CAPT Project Participants
Improving the sub-gridscale parameterizations in global climate models is key to enhancing their simulations of the historical climate and, by implication, their projections of future climate change. Deficiencies in the parameterizations of an atmospheric global climate model(AGCM) at short timescales can be inferred by:
1) initializing the AGCM's state variables realistically from a global weather analysis that is mapped to the model's horizontal and vertical grids,
2) operating the AGCM as a coarse-resolution weather-forecast model, and
3) applying high-frequency observations to identify systematic errors in forecast variables that are strongly impacted by the parameterizations.
Analysis of the systematic forecast errors then can guide efforts to make more realistic the AGCM's parameterized processes. In the experience of the NWP centers, the resulting reduction of forecast errors often correlates with enhanced model performance at climate timescales as well.
The U.S. Department of Energy's 'CAPT' project is applying this NWP-inspired methodology to identify systematic errors in forecasts made with two climate AGCMs--the NCAR CAM3 and the GFDL AM2 models. Central to this investigation are high-frequency field observations supplied by the USDOE's Atmospheric Radiation Measurement (ARM) program at sites representative of diverse climatic regimes (e.g. tropical Pacific, continental U.S., and Arctic climates). Especially noteworthy are hourly atmospheric observations for the entire year 2000 recorded at the ARM site on the U.S. Southern Great Plains (SGP). Once a GCM's forecast variables have been interpolated to the SGP observational grid, it thus is feasible to closely evaluate parameterization-dependent processes over a wide range of continental synoptic conditions.
The present study focuses on systematic errors in year 2000 model forecasts of surface processes at the SGP site. Variables of particular interest include surface radiative fluxes and precipitation, as well as model responses to these forcings in the form of surface turbulent fluxes, temperatures, and humidities. For each season, statistical metrics are used to identify model systematic bias and phase errors in the entire collection of hourly samples of a forecast variable, as well as in its daily averages and mean diurnal cycle. The systematic errors are found to vary substantially according to process, season, and model. In the summer season, for example, SGP precipitation in CAM3 occurs too frequently, due to an overactive convective scheme, while it is too sparse in AM2, possibly because the observed propagation of convective systems across the SGP region is inadequately captured. Other instances of systematic errors that illustrate potential parameterization deficiencies in these AGCMs also will be discussed.
|
