A casual glance at the sky on a partly cloudy day confirms that clouds are patchy or clustered in space. We perhaps imagine, however, that when viewed up close, say on meter scales and below, clouds are fairly uniform. In fact, more quantitatively, most treatments of cloud processes such as particle collision rates and radiative transfer in clouds, are implicitly based on just such an assumption: that particles are distributed in a perfectly random manner at small scales. Is this assumption valid and when it is not, are the deviations something we should worry about? Theory and computational work suggest that even in the absence of mixing, cloud particles are indeed spatially correlated on small scales if a cloud is sufficiently turbulent. Recent atmospheric measurements tend to confirm this picture, but the experiments are challenging and fraught with instrumental artifacts, and therefore are not yet satisfactory.

We have tackled this problem directly by building an instrument capable of measuring the 3-dimensional distribution of cloud particles. The instrument, called HOLODEC, digitally records optical holograms of a small volume of cloud. Subsequently, we reconstruct the holograms by performing the computational equivalent of propagating an electromagnetic wave through the hologram, resulting in a real image. Using this technique we obtain all three spatial coordinates and the shape and size of particles in the measurement volume. HOLODEC took its first trip into the clouds aboard the National Center for Atmospheric Research C-130 aircraft late last year. Our early data analysis confirms that we can see cloud droplets with radii 10 micrometers and above, and that we can reconstruct more complex shapes such as ice crystals. As we develop automated routines for droplet detection we will be well on the way toward quantifying particle spatial correlations in clouds.