The DYCOMS-II field experiment off the California coast in July, 2000 provides an opportunity to test techniques for satellite estimation of cloud parameters.(Stevens et al., 2002). We use radiance measurements made by the MODIS sensor on the Terra satellite to infer optical thickness, cloud droplet effective radius and the cloud temperature and compare these retrieved values against coincident in-cloud aircraft data. Although many different retrieval techniques have been developed during last years, (Arking and Childs, 1985, Rawlins and Foot, 1990, Nakajima and King, 1990, Nakajima and Nakajima, 1995, Perez et al., 2000), most are limited to daytime data, because they rely on the fact that the cloud visible reflectivity is governed by optical thickness and is almost independent of droplet size, while the radiance received in an absorbing band is primarily a function of droplet size. However, during nighttime, the available radiances depend on both parameters. The proposed retrieval method is based on the inversion of a forward radiative transfer model consisting of 3 layers. The top and bottom layers are composed of near-ir absorbing water vapor, while the middle layer is a plane-parallel homogeneous cloud with cloud droplet multiple scattering calculated using the DISORT method. In order to simplify the model, we have selected five MODIS bands affected only by water vapor absorption with low sensor noise.

The inversion is performed by minimizing the differences between the satellite and simulated brightness temperatures, using a numerical evolutionary technique called "Scatter Search" (Glover et al., 2002) and a model state vector that consists of optical thickness, effective radius, cloud temperature and column water vapor path. We have performed retrievals on four MODIS images with coincident DYCOMS aircraft data, and comparisons show good agreement in the retrieved parameters. In the extended abstract we will report further on these four cases and on the sensitivity of the retrieval algorithm to uncertainties in model parameters.