The Impact of Controversial Small Ice Crystals on GCM Simulations

The concentrations of small crystals in ice clouds having length D < 60 µm remains controversial due to (1) difficulties in measuring these crystal sizes and (2) the inability of existing theory to explain their high concentrations (typically 500-5000 L-1). Various instruments used to measure the size distribution (SD) of these small crystals include the FSSP, the CPI/2DC, the CAPS, the 2DS, and others. It is often not clear what percentage of the small crystal concentration is due to shattering of larger ice particles at the instrument inlet.

We have asked the question “if these small ice crystals are real, how will they impact the performance of Global Climate Models (GCMs)?” To answer this question, the GCM must have a realistic treatment of ice particle shapes, the SD shape (including bimodality), ice particle fall velocities (i.e. SD mass sedimentation rates) and ice cloud radiative properties. The GCM experiment described here was not designed with this question in mind, but the results do provide some insight on what the answer might be.

In situ FSSP/2DC measurements indicate that the temperature dependence of the SD in mid-latitude cirrus differs appreciably from that of tropical anvil cirrus clouds. Parameterizations of these measurements have been incorporated into the Community Atmosphere Model (CAM) at NCAR, part of the Community Climate Systems Model (CCSM). Both the tropical and the mid-latitude SD scheme are bimodal, with crystals having D < 100 µm comprising the “small mode”. The amplitude of the small mode in the mid-latitude SD scheme decreases with decreasing temperature, while the small mode of the tropical SD scheme increases with decreasing temperature. The question evaluated in this study was “how does the differing temperature dependence of the small mode in these two SD schemes impact GCM simulations?” Another question evaluated was “does it matter whether one uses a SD scheme for mid-latitude cirrus or tropical anvil cirrus in GCM simulations of climate?”

The treatment of ice clouds in the CAM was modified by implementing the following schemes: (1) SD schemes for tropical anvil and mid-latitude cirrus predicted from cloud temperature and ice water content; (2) the fall velocity treatment of Mitchell and Heymsfield (2005) for accurate ice sedimentation rates; (3) the Modified Anomalous Diffraction Approximation (MADA; parameterized for GCM use) for accurate treatment of ice cloud radiative properties. Ice crystal shape recipes representative of the small and large particle SD modes were based on CPI data. Using realistic SD and particle shape information, ice sedimentation rates and the cloud life cycle were better represented, as well as cloud radiative properties.

One-year CAM simulations were performed using only the tropical SD scheme and only the mid-latitude SD scheme. In the tropical SD simulation, the albedo and emissivity in the upper regions (T < -50 deg.C) of anvil cirrus was strongly governed by the small mode crystals. Due to their higher concentrations in anvil cirrus, the small crystals have a strong impact on bulk ice sedimentation rates and hence the ice water path (IWP), cloud lifetime and cloud coverage. These fall velocity related factors, along with the direct differences in SD bimodality, dramatically increase the shortwave (SW) and longwave (LW) TOA cloud forcing in the tropics (up to -26 and +20 W/m2, respectively, for annual zonal mean) relative to simulations using the mid-latitude SD scheme. Moreover, SW and LW heating rates were greater using the tropical SD scheme due to greater IWP and SD projected area, respectively. This resulted in temperatures in the upper tropical troposphere about 3 deg.C greater than those predicted using the mid-latitude SD scheme. These results suggest that the tropical cold bias predicted by some GCMs in this region may be primarily due to an inadequate treatment of ice microphysics (i.e. small ice crystals and their fall velocities) in tropical anvil clouds.

These findings may also provide clues to how aerosol particles affect cirrus radiative properties. Assuming that aerosol particles first affect the small mode of the SD, mid-latitude cirrus (having a less pronounced small mode) would be most vulnerable to the first indirect aerosol effect. Higher aerosol concentrations would most likely increase the small mode amplitude, resulting in greater heating rates. Recent ECMWF reanalysis and satellite measurements indicate a warming in the upper troposphere in the mid-latitudes but much less in the tropics, contributing to a 200 meter increase in tropopause height. This is consistent with the postulated indirect aerosol effect for cirrus.