Parameterizing subgrid-scale hydrometeor transport in deep convection using a PDF-based closure
Authors
Mikhail Ovchinnikov — Pacific Northwest National Laboratory
Minghuai Wang — Nanjing University
May Wong — National Center for Atmospheric Research (NCAR)
Category
Vertical Velocity
Description
Figure 1. (a-c) Joint density distribution of rain mixing ratio (g kg-1) and vertical velocity (m s-1) at height 3.5 km in a well-developed deep convection event. (d-f) Relative flux contribution (%) to the total flux at that altitude; solid black line indicates 0.1%. Vertical dashed line marks zero vertical velocity. Horizontal dashed line marks mean mixing ratio at that height. White crosses and circles in (a) indicate quadrant means and scaled means, respectively. Updraft regions are within 3 km radius of updrafts stronger than 5 m/s. The majority of the upward rain flux is carried by strong but infrequent active convective towers (quadrant I in b and e), while stratiform regions and collapsing convection are responsible for downward advection of rain (c and f).
To improve representations of convection in GCMs and to enable them to account for the impact of aerosols on all type of clouds, a growing number of convection parameterizations are being augmented with microphysical representations of cloud and precipitation. Because neither convective updrafts nor spatial cloud structure are explicitly resolved by these schemes, the convective (subgrid-scale) vertical transport of hydrometeors needs to be parameterized.
In this study we examine potential ways of parameterizing subgrid-scale vertical fluxes of hydrometeors using a high-resolution simulation of a deep convection case during the ARM 1997 Intensive Observation Period. A cloud-resolving model uses a horizontal grid spacing of 250 m and employs the Morrison microphysics scheme, which treats prognostically mass mixing ratios for four hydrometeor types (rain, graupel, snow, and ice). Three different ways of parameterizing the turbulent vertical fluxes of these hydrometeors are evaluated against the benchmark simulation: an eddy-flux approximation, a quadri-modal decomposition, and a scaling method that accounts for within-quadrant (subplume) correlations. The benchmark simulation indicates that convection tends to transport hydrometeors from below the level of their maximum mass concentration to aloft. In contrast, the down-gradient nature of the eddy-flux approximation leads to transport of hydrometeor mass away from concentrated regions. Unlike the eddy-flux approach, the quadri-modal decomposition is able to correctly capture the signs of the fluxes, but underestimates the magnitudes. The third approach that accounts for within-quadrant (subplume) correlations by scaling the quadri-modal fluxes improves the results for the vertical transport of all hydrometeors except snow. Analysis using diagnostic joint PDFs of vertical velocity and hydrometeors (e.g., see attached figure), and vertical profiles of the hydrometeor budgets will be shown. These results indicate the importance of accounting for the correlation between transport and the representation of subgrid-scale microphysics. Based on this study, a subgrid-scale transport scheme for hydrometeors is proposed for an assumed probability density function (PDF) parameterization of convection, such as CLUBB.
