As climate models move to finer resolution, they can be evaluated against observations using new metrics. We propose to use and extend a set of measures developed from observations, on the scales that high-resolution global climate models are now reaching, to evaluate a targeted set of processes in current climate models. These will be evaluated for a set of models across a range of resolutions, including the higher resolutionmodels fromthe CoupledModel Intercomparison Project 5 (CMIP 5), and various higher-resolution models from specific modeling groups. An example of a moderately high resolution model (Community Atmosphere Model at 0.5 degree resolution) is used to show that a model with parameterized convection can qualitatively capture several aspects of the categories below, but there is considerable sensitivity to ill-constrained factors such as entrainment.
The analysis will provide assessment of model suitability for evaluation of changes in these statistics under climate change and provide feedback for model development, drawing on tools developed under previous NOAA funding. Specifically, we focus on four categories of related features of precipitation, water vapor and temperature characteristics, on a set of statistics to quantify these, and on the underlying mechanisms producing these features.
1. Onset of deep convection, its water vapor-temperature dependence, and relation to entrainment assumptions. A set of convective onset statistics from remote sensing and in situ data provide a quantification of recent developments on the dependence of convection on water vapor in the lower free troposphere. There are several indications from other groups and from NOAA-supported prior work, of a strong sensitivity of climate models to errors in this process.
2. Excursions to high water vapor and strong precipitation regime. Prior work has provided evidence of long tails in the probability distribution (PDF) of water vapor, with a Gaussian core surrounded by approximately exponential tails, characteristic of an advection interacting with a forcing that maintains a gradient. Such tails imply much more frequent excursions into the high-water vapor regime associated with intense precipitation than would occur with Gaussian statistics.
3. Quantification of similar long-tail behavior for surface temperature PDFs, seen in preliminary work in many locations. The presence of such tails implies a rate of increase of extreme events under a shift of the distribution under global warming very different from that of a Gaussian.
4. Interactions at the margins of convective zones: the inflow air mass transported into a convective region is modified along its trajectory until conditions for convective onset are reached. If the onset condition evaluated in 1 is incorrect, the margins of the convection zones can have errors of hundreds of kilometers, creating large errors at the regional scale. Under global change, differences among model representations of this condition can yield large differences in the predicted regional change.
Coordinated with a proposal from Rutgers University, diagnostics from internal variability will help constrain models in this category, and the role of these mechanisms will be evaluated for regional climate change in the models.