Adam Varble — Pacific Northwest National Laboratory
Edward Zipser — University of Utah
Category
Deep convective clouds, including aerosol interactions
Description
An example of amplification of mesoscale circulations (rear inflow and front-to-rear flow) in a WRF simulation of the 20 May 2011 MC3E squall line causing convective updrafts to be sheared off and increasing the size of the high reflectivity convective rain region. The panels show radial velocity (filled) with Rayleigh reflectivity contoured every 10 dBZ in observed (top) and simulated (bottom) longitude-height cross-sections through the squall line. The convective region is on the right moving toward the right and stratiform/anvil regions are trailing to the left.
ARM long-term and field campaign measurements have been instrumental in solidifying robust cloud and precipitation biases in models across every scale. For deep convective systems in particular, predicting system initiation and potential upscale growth into long-lived mesoscale systems as a function of environmental conditions without bias is vitally important to avoiding significant regional temperature and precipitation biases in climate models. Unfortunately, these well known biases exist, and high-resolution models that are looked to for guidance in developing cumulus parameterizations also exhibit biases in convective system cloud and precipitation evolution. Based on many ARM and non-ARM field campaign case studies, there are now several established causes of these high-resolution model biases that will be showcased. These include biased environmental representation, overly intense convective updrafts, and biased parameterizations, notably microphysics parameterizations through insufficient representation of hydrometeor properties or key microphysical processes. From recent analysis of MC3E cases, a new potential cause of bias that will also be highlighted is amplified mesoscale circulations, which can cause over-organization of convective regions.
To constrain the causes of biases requires measurements of processes because flawed process parameterizations or environmental properties are the cause of upscale error growth through nonlinear interactions with other processes. For deep convective systems that evolve quickly, processes can be inferred from frequent temporal sampling of specific observable environmental properties in a confined region. Without such retrievals, modelers are left to tune parameterizations without observational basis in an attempt to match environmental statistics. Potential observing methods using available ARM resources will be discussed that would target some of the identified processes that cause persistent model deep convective biases. Even with such methods, firmly identifying causes of model bias will remain difficult for a number of reasons including limited convective predictability and a shortage of well-observed cases across different convective environments. Further research into stochastic parameterizations, simulation ensembles, and methods for robustly comparing large amounts of model output with small numbers of observations would help.
