512 Assimilating Remotely Sensed Satellite Derived Data with Land Surface Model Derived Data: Diagnosing and Reconciling Differences

Wednesday, 9 January 2013
Exhibit Hall 3 (Austin Convention Center)
Rosalyn F. MacCracken, George Mason University, Fairfax, VA; and P. Houser, E. Wood, J. Sheffield, D. Lettenmaier, R. T. Pinker, C. D. Kummerow, M. Pan, H. Gao, A. Sahoo, and J. Bytheway

A new balanced global terrestrial water cycle dataset is being created for the NASAs' Making Earth Science Data Records for use in Research Environments (MEaSURE) project. This dataset will be comprised of multiple remotely-sensed datasets and model generated data, and will be merged into a single unified multi-decade, high spatial resolution, climate consistent Earth Science Data Record (ESDR). Understanding where sources of temporal and spatial differences occur between datasets, and how to reconcile these differences, are some of the major challenges in the creation of this dataset. Some possible sources of differences in the satellite data are varying retrieval methods, radiative transfer models and processing algorithms. Other sources of differences come from model generated data which include varying initializations, parameterizations, and interpolation of output data. Throughout the creation of this dataset, it was necessary to address issues of inconsistencies between the input datasets. These inconsistencies included biases and uncertainties between different variables, and inconsistencies in measurements and reporting time periods, as well as other differences.

This poster summarizes the methodology used to diagnose the differences between the multiple datasets that were merged to form the final ESDR as well as methodologies used to reconcile these datasets. Monthly and seasonal averages will be evaluated between datasets to determine how these averages vary in time. Statistical methods, such as correlation and trend analysis will be also used to identify the magnitude and geographical location of the differences. Calibration will be used to reconcile these differences, which will be done by using a weighted adjustment from the model/observation errors. By performing this adjustment, the model generated data will match the observed statistics of the variables in the water cycle. >

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner