To date there is no accepted satellite remote sensing method for estimating the liquid water mass fraction in mixed phase clouds.  This study presents such a method appropriate for cold clouds (-40 < T < -15 C) where the liquid water fraction is less than 50%.  It has been long recognized that low levels of liquid water (LW) can dominate the projected area of cloud particles in mixed phase clouds, thus dominating their optical properties.  Based on recent measurements of the ice particle size distribution (PSD) in cirrus clouds, a 6% LW fraction will often decrease the mixed phase effective diameter De by ~ 1/3.  Thus it is important to account for the presence of liquid water in cold clouds for an accurate assessment of the global radiation budget.

To retrieve the LW fraction, cloud emissivity in two window channels and cloud temperature T are retrieved.  This was done by showing cloud emissivities in the MODIS CO2 channels are nearly identical, allowing T to be retrieved from two CO2 channels.  Knowing T, two cloud emissivities can be solved for exactly using the MODIS 12 and 11 micron channels.  From these retrieved emissivities, the effective (corrected for scattering) 12/11 μm absorption optical depth ratio, henceforth β, can be determined and related to T (an advancement unique to this study).

Previous satellite studies have shown that for all-ice clouds, the mean value of β is approximately constant with T.  Our satellite retrievals appear to confirm this for cloud scenes during the TC4 field campaign off Costa Rica.  This observation can be exploited since the incremental presence of LW causes β to increase, as shown in the figure (left panel).  For T < -40 C, clouds are glaciated and the mean β (pink line) is roughly constant, while for T > -40 C, mean β increases with T.  The vertical pink bars indicate the standard deviations in β. This mean increase in β can be used to estimate the temperature dependence of the mean LW fraction in a cloud scene as well as the variance in LW fraction (from the variance in β).  In situ observations show that a given mixed phase cloud tends to be dominated by one phase so that a phase-frequency diagram has maxima for all-ice and all-liquid conditions.  A similar pattern can be inferred from this figure at the warmest temperatures, with fewer points associated with the mean β.

The microphysics algorithm for translating these β changes into changes in LW fraction assumes a mixed phase bimodal PSD, with a temperature-dependent cirrus anvil PSD scheme for the ice portion (given the TC4 cloud scene) and a representative mean diameter and dispersion parameter for the liquid portion or mode.  The algorithm systematically increases the liquid water content (increasing the liquid mode) until the predicted β matches the measured β.  The retrieved LWC/IWC ratio is sensitive to the assumed droplet diameter but less sensitive to other microphysical variables.  The retrieval below (right panel) assumes a mean diameter of 10 microns.  This value can be inferred from the spread of β values.  The MODIS image was filtered to select only pixels classified as single layer cirrus over ocean having cloud emissivities < 0.7.  The pink curve gives the LW fraction corresponding to +1 standard deviation from the mean β, showing considerable variability for T > -30 C.  We are currently attempting to validate this retrieval using two SPARTICUS case studies.