This study therefore seeks to address two main objectives. The first objective is to quantify the amount by which changes in instrument accuracy alias into cloud property trends. For a perfect observing system that has no measurement uncertainty, the accuracy of a climate trend is limited only by the natural variability of the climate variable in question. The accuracy of a climate variable trend as observed by an actual observing system is additionally limited by the measurement uncertainty due to components such as orbital sampling and calibration. Drifts in cloud properties due to calibration changes over time may be disguised as a true climate trend. We are therefore imposing absolute calibration changes to MODIS spectral reflectance used as input to the CERES Cloud Property Retrieval System (CPRS) and running the modified MODIS values through the cloud retrieval algorithms to calculate the resulting changes in cloud properties. We then use these changes to determine the impact of changes in instrument calibration on the ability to detect trends in reflected solar cloud properties.
The second objective is to quantify by how much cloud retrieval algorithm assumptions alias into cloud optical retrieval trends. The initial focus on this objective is how the plane-parallel assumption in reflected solar satellite cloud retrievals aliases into cloud optical thickness trends. In addressing both of these objectives, we are again using the CERES CPRS, which ingests MODIS reflectance. First, we are collecting liquid water cloud fields obtained from Multi-angle Imaging Spectroradiometer (MISR) measurements to construct realistic probability distribution functions (PDFs) of 3D cloud anisotropy (a measure of the degree to which clouds depart from plane-parallel) for different ISCCP cloud types. We also want to know the corresponding distribution of 3D optical thickness bias associated with each of these clouds types, but this is an unknown quantity. A theoretical study with dynamically simulated cloud fields and a 3D radiative transfer model will provide the relationship between 3D cloud anisotropy and 3D optical thickness bias for each cloud type. Combining the results from these first two steps results in distributions of 3D optical thickness bias for each cloud type. Finally, using output from CFMIP2/CMIP5 simulations, we will determine the change in frequency of occurrence of cloud types between two decades (e.g. at the beginning and end or beginning and middle of the 21st century). With the PDFs of 3D optical thickness bias and the change in frequency of occurrence per cloud type, we will then have the information needed to calculate the total change in 3D optical thickness bias between the two decades.
The quantification of aliasing errors from instrument and algorithm sources will be of interest to the satellite retrieval user community. There are two possible outcomes. If significant instrument- and algorithm- based aliases (cloud property trend distortions) are found, the results will then motivate the development and rigorous testing of climate-scale cloud retrieval algorithms. The discovery that current reflected solar cloud retrievals have the accuracy needed for climate studies will help to verify the validity of using these retrievals to detect and attribute changes in Earth's climate. We are working to resolve whether operational reflected solar cloud retrievals are able to discriminate domain-aggregated, long-term climate trends from inherent aliasing errors and recommend the development of climate-specific algorithms as needed. The results from this study can also be used as the requirements of future satellite instruments are determined that will be used to detect changes in climate.