Process-oriented Model Evaluation for the North American Monsoon

The objective of this proposal is to develop process-oriented diagnostics to evaluate global
model representation of the North American monsoon (NAM) and explore the pathways to model
improvements. The NAM is chosen to be the focus of the project because of its significance to the
United States, and also because it serves as an ideal testing ground for model evacuation and
improvement owing to the important roles of many fundamental physical processes and their
interplay with the large-scale monsoon circulation. We will focus three aspects of the NAM, its
moist thermodynamic perspective, the link between the continental monsoon to the subtropical
northeastern Pacific cloud regime, and the multi-scale nature of the NAM. Process-oriented
diagnostics will be developed in the convective quasi-equilibrium framework to evaluate the
seasonality, structure, intensity and variability of the NAM. The simulated convection and cloud
processes will be evaluated using satellite and site-specific data from the ob4MIPs. In particular,
the synergetic analysis of the CloudSat and MODIS will help to link the deficiencies in simulated
cloud processes to uncertain parameters in microphysics schemes. In addition, two bulk metrics,
which link model performance and physics formulation, will be tested and are expected to provide
insights into model improvement. Although we focus on the NAM, the proposed research
addresses some common issues in climate models and will contribute to improvement of the
overall model performance.

The GFDL models (CM4, AM4 and fvGFS) will be employed to assist the development and
testing of the diagnostics and metrics. Perturbed-physics ensembles will be carried out using CM4
and AM4 in the weather forecasting mode, and the high-frequency output will be evaluated to
examine fast-physics error growth and constrain parameter uncertainties based on observations.
Climate simulations will be further carried out to examine slow error growth. In addition, the
fvGFS will be run at the seasonal-prediction mode with a configuration similar to the GFDL fvGFS
experimental 10-day forecasts (i.e., 13-km globally uniform resolution with an interactive, refined
grid of 3-km resolution). These simulations will be used to assess climate model errors, especially
in representing multi-scale processes and weather/climate extremes. The simulations will also help
to explore the capability of the fvGFS in seamless prediction from the synoptic to the seasonal
time scales. The diagnostics and metrics will be developed and tested mainly using the GFDL
model simulations, and further testing of robustness will be carried out using the CMIP6 data, in
particular the CFMIP, GMMIP and HighResMIP.

The proposed research falls right into the focal area of the MAPP’s competition on
“addressing key issues in CMIP6-era earth system models”, and is also highly relevant to the
MAPP’s mission to enhance the Nation’s capability to predict natural variability and changes in
Earth’s climate system.

Climate Risk Area: Water Resources