Evaluating clouds and cloud effects in the AM3 model using atmospheric classification

Parameterizations in models attempt to statistically relate large-scale atmospheric variables such as temperature, humidity, and winds with variables that change on scales too small to resolve, such as cloud properties. We study the observed relationships between such large- and small-scale variables through the creation of a set of atmospheric states and associated distributions of small-scale variables. We then compare these observed relationships to those that exist in the GFDL's AM3 model. We use an iterative clustering technique to define a set of atmospheric states for a region surrounding the ARM Southern Great Plains (SGP) site. Atmospheric state in this context can be thought of as a frequently occurring regional weather pattern. We use 13 years of dynamic and thermodynamic variables from the ERA-Interim reanalysis as the input to a clustering algorithm to define the states, and cloud occurrence data from the vertically pointing millimeter cloud radar at the SGP site to validate the statistical significance of each state. Once the states are defined, we classify the state of the atmosphere every 6 hours for the duration of the study, creating a time series of atmospheric state. This time series is used as a basis for compositing cotemporaneous observations of interest and creating distributions of small-scale variables associated with each of the states. Distributions of both ground-based observations from the SGP site such as cloud occurrence, precipitation, liquid water path, and energy fluxes as well as satellite-derived equivalents are created in this fashion. Snapshots of output from the AM3 model are sorted into the observed atmospheric states, and their associated distributions of cloud, precipitation, and radiative properties are composited to produce modeled distributions of parameterized variables for each atmospheric state. Comparison of the observed and modeled distributions of cloud occurrence, cloud radiative effect, precipitation, and other variables within an individual state provides a test of the model parameterization under a particular set of physical conditions. In contrast, comparing the observed frequency of occurrence of the atmospheric states with their occurrence within AM3 tests how well the model reproduces the dynamic and