Long-term evaluation of cloud fraction simulated by seven SCMs against the ARM observations at the SGP site
Poster PDF
Authors
—
Wuyin Lin — Brookhaven National Laboratory
Yanluan Lin — Tsinghua University
Satoshi Endo — Brookhaven National Laboratory
Leo Donner — Geophysical Fluid Dynamics Laboratory
Audrey B. Wolf — NASA - Goddard Institute for Space Studies
Anthony D. Del Genio — National Aeronautics and Space Administration
Roel Neggers — University of Cologne
Yangang Liu — Brookhaven National Laboratory
Category
Cloud Properties
Description
This study evaluates the performances of seven single-column models (SCMs) (European Centre for Medium-Range Weather Forecasting [ECMWF] IFS; Geophysical Fluid Dynamics Laboratory [GFDL] AM2, 3; Goddard Institute for Space Studies [GISS] model E2, and Community Atmosphere Model [CAM] 3, 4, 5) by comparing simulated cloud fraction with observations at the Atmospheric Radiation Measurement Climate Research Facility’s Southern Great Plains (SGP) site from January 1999 to December 2001. Compared with the three-year averaged Active Remote Sensing of Clouds (ARSCL) cloud fraction, the ECMWF and GISS SCMs underestimate cloud fraction at all levels. The striking feature for the GFDL SCMs is the underestimation of low-level cloud fraction but overestimation of high-level cloud fraction. The three single-column CAMs (SCAMs) overestimate high-level cloud fraction but have low-level cloud fraction similar to the observation, due to the compensation between the biases in convective and non-convective cloud fractions. The frequency distribution of cloud fraction shows a large discrepancy between the observation and SCMs. In the observation, it is a distinct U-shaped distribution of cloud occurrence, heavily concentrating on near-clear (<5%) and near-overcast (>95%) conditions; in contrast, in the SCMs, cloud events occur much more frequently between the two extremes. The three SCAMs overestimate mid-to-low level cloud events with moderate cloud fraction ranging from 10%–55%, which can be attributed primarily to the convective cloud fraction diagnosed based on convective mass flux. Further analysis of the non-convective cloudy events reveals different relationships between cloud fraction and relative humidity in the models and observation. Results imply that the cloud schemes, especially the diagnostic cloud schemes that involve the use of threshold relative humidity, need further improvement.
