Out with the Old, in with the New: McICA to Replace Traditional Cloud Overlap Assumptions

Pincus, R., NOAA - CIRES Climate Diagnostics Center

Atmospheric Thermodynamics and Vertical Structures

Cloud Modeling

Pincus, R., R. Hemler, and S.A. Klein, 2006: Using Stochastically Generated Subcolumns to Represent Cloud Structure in a Large-Scale Model. Mon. Wea. Rev., 134, 3644–3656.


As shown by the difference between the two panels, the standard way (AM2, top panel) of mixing solar reflection and transmission differs systematically from the Independent Column Approximation approach.


As shown by the difference between the two panels, the standard way (AM2, top panel) of mixing solar reflection and transmission differs systematically from the Independent Column Approximation approach.

Because cloud-radiation interactions depend critically on the vertical amount of clouds, different assumptions about how this alignment occurs lead to differences in climate model results. As reported in the Journal of Geophysical Research in 2003 and summarized on the DOE Atmospheric Radiation Measurement (ARM) Program website, a new radiative scheme called the Monte Carlo Independent Column Approximation (McICA) was developed to improve the treatment of cloud variability in climate models. In collaboration with the National Oceanic and Aerospace Administration's Geophysical Fluid Dynamics Laboratory (GFDL), ARM researchers recently incorporated their new McICA scheme into the GFDL's atmospheric climate model (AM2) in order to gauge its effectiveness in improving the accuracy of cloud-radiation interactions. The McICA code demonstrated a significant improvement in solar radiation at the top of atmosphere of about 4 W/m2.

Most climate models predict only the average amount of ice and liquid water contained in a 250x250 km atmospheric column. This value is then used to compute the flow of radiation energy through the atmosphere. However, the models only predict the amount of cloud cover in each vertical layer (typically about 25 in total) but don't include information on vertical alignment of clouds. Therefore, traditional climate models (such as AM2) use a "random overlap" assumption to account for cloud vertical alignment. The McICA code, however, uses statistical techniques to represent greater flexibly in the vertical overlap assumptions, creating an inherently more accurate solution.

In this new code, each grid box in a climate model is subdivided into a large set of subcolumns, each containing a different vertical alignment of thick and thin clouds. These alignments are chosen to have the same statistical properties as occur in nature and observed by ARM field instruments. Radiation transport in each subcolumn is calculated independently (the so-called Independent Column Approximation, or ICA), producing a more realistic calculation of actual cloud-radiation interactions. The computational overhead for this process, however, was deemed too expensive for implementation in climate models. So, the researchers again used the statistics, but this time in the form of a Monte Carlo, or random walk, technique (thus, the moniker McICA). This method sped up the calculations enough for implementing in the AM2 model, revealing the improvement over the traditional method.

The AM2 climate prediction model is one of the two most widely used climate models in the United States (the other being the Community Atmospheric Model, by NCAR). However, its assumptions for cloud distribution are inaccurate when compared to direct observational data. The McICA provides a flexible and more accurate way to compare model output with satellite data, and, because of the inherent cloud variability it introduces, allows modelers to avoid the traditional and time consuming step of tuning cloud optical properties. With ARM funding, the GFDL is building a new cloud scheme to deal with small scale cloud variability; this will benefit greatly from the flexibility afforded by the McICA.