8.7 Representing the ice fall speed in climate models using measurements from recent field campaigns

Wednesday, 30 June 2010: 12:00 PM
Cascade Ballroom (DoubleTree by Hilton Portland)
David L. Mitchell, DRI, Reno, NV; and S. Mishra, B. Baker, Q. Mo, and P. Lawson

The life cycle of cirrus clouds depends strongly on their ice sedimentation rates which have been difficult to characterize due to uncertainties in the concentration of small ice crystals (e.g. Mitchell et al. 2008, GRL). Moreover, the ice fall velocity in the U.K. global climate model (GCM) was shown to be second most influential among physical parameters in affecting climate sensitivity (i.e. the equilibrium change in global mean surface temperature resulting from a doubling of CO2), as described in Sanderson et al. (2008, Clim. Dyn.). In spite of its importance, the ice fall velocity in climate models is highly uncertain due in part to its dependence on the ice particle size distribution (PSD), which has been plagued with measurement uncertainties from ice particles shattering on the probe inlet tube. More recently, new probes have been designed to measure the cirrus PSD where particles pass through probe arms instead of an inlet tube, thus reducing ice artifacts from shattering.

To accurately estimate ice mass sedimentation rates, accurate measurements of the PSD regarding ice particle number, projected area and mass are needed. A relatively new probe called the 2D-Stereo (or 2D-S) probe measures these 3 quantities from 10 to 1280 μm, using an ice particle projected area-mass relationship to estimate the size-resolved mass concentrations. Larger sizes (if present)are measured by the 2D-P probe. In addition to having probe arms instead of an inlet tube, the 2D-S measures ice particle inter-arrival times that reveal the spatial proximity of ice particles. Using software processing, clusters of closely-spaced ice particles are assumed to result from shattering and are not counted in the 2D-S PSD calculations. Thus the problem of shattering has been greatly reduced in these processed 2D-S PSD measurements. The 2D-S estimates of ice water content (IWC), based on PSD integrations using the area-mass relationship, generally agree well (within ~ 20%) with CVI measurements of IWC during the TC4 campaign. Hence the size-resolved number, area and mass concentrations of the 2D-S appear reasonable. This study uses 2D-S data from four recent field campaigns: TC4, NAMMA, ISDAC and SPARTICUS.

Ice mass sedimentation rates are the product of the IWC and the mass-weighted fall velocity, Vm. In a measurement context, the definition of Vm is given as:

Vm = Σ v(D)m(D)N(D)ΔD / Σ m(D)N(D)ΔD ,

where v(D) = ice particle fall velocity, m(D) = ice particle mass, N(D) = size distribution, D = ice particle maximum dimension at bin midpoint and ΔD = bin width. Using this equation, the measured PSD of size-resolved number, area and mass concentration can be used to solve for Vm. The ice particle size, area and mass measurements are used to calculate v(D) for each 2D-S size-bin as described in Mitchell and Heymsfield (2005, JAS), where area- and mass-dimension power laws (used to calculate v(D)) are generated from the 2D-S measurements. Standard temperature and pressure were assumed when calculating v(D) and Vm, and a prefactor is applied to Vm to adjust it to the desired temperature and pressure. It was found that correlating Vm with cloud temperature T or IWC accounted for ~ 50% of the variance in each regression. Therefore to estimate Vm when only T and IWC are available, a multiple regression was performed, relating Vm to both T and IWC. This accounted for ~ 60% of the variance in Vm.

The same type of analysis can be done for the PSD effective diameter, De. A regression with T accounts for 71% of the variance in De based on TC4 data.

Most climate models have two-moment prognostic microphysics schemes that predict De. Since the ice particle mass/area ratio is used to calculate both De and Vm, a stronger correlation between these properties is expected. Indeed, the correlation based on TC4 data is 0.961, accounting for 92% of the variance in Vm. Such climate models can therefore estimate Vm rather accurately based on these empirical De-Vm relationships for anvil, midlatitude and Arctic cirrus.

In addition, for the TC4 field campaign, it was found that the PSD number concentration N was about 6.5 times higher on average when T < -40 C relative to N when T > -40 C. Similarly, the PSD mean maximum dimension (Dmean) was about 2.7 times larger on average when T > -40 C relative to Dmean when T < -40 C, with considerable scatter in Dmean for T > -40 C. Moreover, number concentration PSD were bimodal for T > -40 C and monomodal (due to higher concentrations of smaller ice crystals) for T < -40 C. This was true for both anvil cirrus and in situ cirrus. These observations are consistent with theoretical expectations if homogeneous freezing nucleation dominates for T < -40 C. This mechanism would produce ice crystals at a faster rate than would heterogeneous nucleation processes. It is also possible (but arguably less likely) that these observations were due to a process called “size sorting” where larger ice particles fall faster to lower levels while smaller crystals tend to remain at higher levels. Evidence of homogeneous freezing nucleation will be explored in a similar way regarding the other field campaigns and discussed at the meeting.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner