Analysis of Cloud Variability and Sampling Errors in Surface and Satellite Measurements

Radiation measurements have been widely employed for evaluating cloud parameterization schemes and model simulation results. As the most comprehensive program aiming to improve cloud parameterization schemes, the Atmospheric Radiation Measurement (ARM) Program has an essential goal to make observations on the scale of a general circulation model gridbox, so as to define the physics underlying some of the important parameterizations in the general circulation models used in climate change studies. While ARM has deployed a network of radiation stations over a domain of about 400 km, extensive radiation and cloud measurements are taken at a single location (the central facility) within a domain of 100 km. An important question is thus raised as to whether these measurements are adequate to represent grid mean values given the high variability of cloud and the surface. In contrast to the coarse-resolution general circulation models, fine-resolution cloud system models (CSMs) are playing an increasing role in revealing the fundamentals of cloud-radiation interactions. As computing power improves, CSMs will have a very high resolution (a few to a few tenths of km), which raises another question: on what scale does modeling need to occur in order to capture the physical properties that drive the system. Answers to both questions hinge on cloud variability and observation density. It is not clear that we can understand cloud behavior well enough from the ARM-deployed instruments nor are we sure which types of measurements and at what spatial density are required for closure of the boundary fluxes in single-column models. To address these questions, we need more sound guidance toward adding additional observation sites and deploying mobile facilities. While it is a formidable task to resolve these questions, we attempt here to analyze related issues by using satellite and ground observations of radiative quantities. Taking advantage of the high spatial and temporal resolution of geostationary operational environmental satellite (GOES) data, we mimic ground-based measurements of varying density and temporal frequency and characterize their observation uncertainties caused by cloud variability at different scales in different seasons. Such scale-dependent statistics of observation uncertainties provide critical constraints on model-observation comparisons, and are thus valuable for improving and validating cloud parameterization schemes.