A21E-0188
Diagnosing Cloud Biases in Climate Models by Comparing Forecast-Mode Simulations With Satellite Observations
Tuesday, 15 December 2015
Poster Hall (Moscone South)
Christopher R Jones1, Christopher Stephen Bretherton1, Jongil Han2, Ruiyu Sun2 and Ming Zhao3, (1)University of Washington Seattle Campus, Seattle, WA, United States, (2)Environmental Modeling Center, IMSG, College Park, MD, United States, (3)NOAA Geophysical Fluid Dynamis Laboratory (GFDL), Princeton, NJ, United States
Abstract:
Accurately simulating marine clouds is a persistent challenge for weather and climate models. Assessing and interpreting the root of systematic cloud biases is exacerbated by the interplay of a wide range of physical and dynamical processes. The goal of this study is to use forecast-mode global simulations to analyze cloud biases that develop in short-term simulations in which the large scale dynamics are still constrained by the initial conditions. We use multiple configurations of the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) and Geophysical Fluid Dynamics Laboratory (GFDL) prototype AM4 models to produce 3 day forecasts starting from each day of July 2013, initialized with NCEP Reanalysis. Comparing the top of atmosphere (TOA) outgoing longwave radiation (OLR) and reflected shortwave radiation (RSW) from each model against Clouds and the Earth's Radiant Energy System (CERES) observations for the same period, we find the models have different regional bias patterns that do not vary substantially with forecast lead, and which are surprisingly consistent across every day of July 2013. Relative to CERES, we find the GFS models broadly simulate too little low cloud across a wide swath of the globe, with an offshore region in the southeast Pacific with too much cloud, contributing to a net TOA radiation bias on the order of 10 W m-2. The GFDL models tend to simulate too much high cloud in the Inter Tropical Convergence Zone and too little coastal stratocumulus. Using the TOA radiation biases as a guide, we identify two regions to further compare vertically resolved cloud fields: the GPCI transect, and the mid-latitude NE Pacific. The mid-latitudes in particular are a region where the GFS and GFDL models show opposite OLR and RSW biases from each other when compared against CERES. Our next step is to use these cloud biases diagnosed in forecast-mode simulations to guide model development.
