WMOD00318 (2) - Laserfiche WebLink

Home Browse Search

838 JOURNAL OF APPLIED' METEOROLOGY VOLUME 34 step. Calculation of F scav follows Young ( 1974a) for the parcel model simulations. It thus depends in time on the net collection kernel for Brownian, phoretic, and aerodynamic transport of aerosol to cloud droplets. <;:alculation of F scav in RAMS follows the simplification of Young's equations given by Cotton et al. ( 1986). This equation was taken to be valid for S; > 1.058 and for T < 269.2 K. For all of the equations, extension below about 253 K implies extrapolation from experimental results, although values appear physically reasonable to as low as 248 K. In all cases as well, F is not allowed to exceed a maximum of 1.0. Extraordinary circumstances would be required to achieve such a result. b. Initialization of seeded aerosol concentrations For initializing seeded aerosol mass and concentra- tion in the numerical simulations, the published aerosol mass release rates, aircraft speed, and horizontal and vertical dispersion rates were combined with the num- bers of particles per gram released based on the PSD to determine the initial seeded volume aerosol con- centration. This volume concentration was determined at a time of 6 min after seeding was initiated. At this time, about two-thirds of a 37-km seeding line had been dispensed, and the initial plume would have filled more than 10% of a 1000-m horizontal and 100-m vertical grid box used in the 3D numerical simulations with RAMS. In the seeding simulation, the entire grid was filled with the nuclei concentration thus deter- mined. Therefore, the cloud was overseeded by several times from a total mass standpoint. This method was considered to be a reasonable compromise to overcome the current inability to accurately represent dispersion of seeded mass and concentrations at subgrid scales in mesoscale models. To accurately predict plume con- centrations in the model immediately after seeding would have required excessively overseeding a model grid volume and thus greatly overestimating seeding effects on cloud microphysics and precipitation. Con- versely, assuming the initial concentrations to be equivalent to those present at the time after the grid volumes were filled (30-40 min) would have greatly underestimated subsequent plume concentrations and missed much of the ice formation occurring by contact- freezing nucleation after the onset of seeding. The published ice nucleus aerosol mass generation rate was 0.4 g km-'. The speed of the cloud seeding aircraft was 80 m S-I. Horizontal and vertical disper- sion rates were estimated as 1.0 and 0.1 m s -1, re- spectively. Integrating the PSDs shown in Fig. 2 gives I X 10 15 and 4 X 10 15 particles per gram nucleant generated, respectively, for the "ground" and "air- borne" cases. Using all of the above values, initial seeded concentrations at 6 min after release were de- termined as 40 000 llnd 160000 L -I, respectively, for the "ground" and .,irborne" :>C;;Ds. The latter result was used for RAMS simulauons. c. Parameterization tests Preliminary to the mesoscale model simulations, some simulations were done with a microphysical par- cel model. The purposes of this exercise were to in- vestigate the importance of knowledge of the artificially generated aerosol size distribution, to demonstrate the validity of the ice nucleation parameterizations, and to determine the expected relative importance of the different ice formation processes for conditions relevant to the 18 December 1986 case study. The parcel model used was a modified version of the model described by Rokicki and Young (1978); a derivative of the 2D microphysical model described in detail by Young ( 197 4b ). The version employed is very nearly the same as used and described by Stith et al. ( 1994), Rogers et al. ( 1994), and DeMott ( 1994) for various investiga- tions of ice initiation due to natural and man-made aerosols. In broad terms, the model simulates the three major processes of nucleation, diffusion, and collection that are key to precipitation formation. Precipitation does not occur from the parcel, and, except for a spec- ification of vertical motion, cloud dynamics is not con- sidered. For the parcel model simulations, artificial ice ini- tiation by deposition, condensation-freezing, contact- freezing, and immersion-freezing nucleation followed either the detailed results as a function of AgI-AgCI particle size summarized by DeMott ( 1994) or the pa- rameterizations given in Eqs. (3.2)-(3.5). The Fscav realized in any time step was determined by the col- lection rates of ice nuclei by cloud droplets due to the combined effects of Brownian collection, thermopho- resis, diffusiophoresis, and aerodynamic capture. Cal- culations of collection rates followed the equations from Young (1974a) and the specification of the par- ticle size distribution. Freezing occurred at the droplet temperature that is explicitly calculated in the model. In the simulations presented, a constant updraft was imposed on a seeded cloud parcel since we were only interested in demonstrating results to be expected in the mesoscale simulations. A representative 0.2 m S-1 updraft was chosen for a parcel initialized 0.5% super- saturated with respect to water at -6.3 oC and 700 mb. CCN coefficients were adjusted to cause 130 cm-3 droplets to form initially, consistent with observations. Fifteen discrete size bins were used to specify each of the two different particle size distributions used. Figure 3 shows parcel thermodynamic and micro- physical evolution with time, as well as the cumulative production of ice crystal concentrations by the four nucleation mechanisms quantified for three simula- tions. Two simulations demonstrate the effect of the presumed aerosol size distribution on ice formation in the parcel model. Dashed curves are for the presumed aircraft generator particle size distribution. The dash- dot curves presume the size distribution given by DeMott et al. (1983) for a ground-based generator.