The great importance of clouds in understanding atmospheric phenomena is widely recognized, yet faithful representations of cloud and precipitation processes in models at nearly all scales remain elusive. In order to properly constrain model parameters, it is important to obtain reliable observations of cloud properties in varying atmospheric environments. The Temporal Experiment for Storms and Tropical Systems (TEMPEST) mission was proposed to help address this need by deploying a cluster of CubeSats, each containing an identical, five-frequency passive microwave radiometer, into the same orbit. Doing so would allow for the observation of cloud processes at a high temporal resolution and on a global scale.

In order for such a mission to be useful in understanding cloud processes, it is crucial to develop a retrieval algorithm that can distinguish true changes in the atmospheric state from the noise induced by making repeated observations only a few minutes apart at different view angles. To this end, a physical optimal estimation algorithm is developed for the retrieval of water vapor, cloud water, and frozen hydrometeors from cross-track microwave sounders such as the TEMPEST radiometer. The performance of the algorithm is assessed by using high resolution Weather Research and Forecasting (WRF) model output to generate synthetic radiometer observations, while incorporating realistic error estimates, and then comparing the parameters retrieved using the synthetic observations to the actual model parameters.

For rapidly changing clouds, differences in parameters retrieved at various view angles, while not trivial, are small enough that changes in cloud properties can be discerned. This is especially true for view angles near nadir, where the field of view is smaller and changes less rapidly with time. Experiments simulating a cluster of TEMPEST instruments successively observing the same cloud system suggest that using the higher-quality retrievals near nadir to constrain preceding and subsequent observations allows for cloud changes to be observed more clearly. An analysis of the contribution of various forward model errors indicates that incorporating more accurate a-priori information about wind speed, cloud coverage, and cloud heights, perhaps obtained from coincident measurements by other spaceborne instruments, would further constrain the retrieval and mitigate some of the view angle induced biases.