Proceedings of SPIE
A mathematical model for statistical estimate of the bias and noise in satellite retrievals of atmospheric profiles and a case study are presented. The model allows accurate validation of actual performance of the remote sensing system while in orbit by comparing its measurements to correlative data sets, e. g. radiosonde network. The model accounts for the following factors: (i) The satellite and validating systems sample volumes of the atmosphere at times and locations that are not exactly co-located. (ii) The validated and validating systems have different characteristics, e. g. different vertical resolution and noise level. All the above factors cause apparent difference between the data to be compared. The presented model makes the comparison accurate by allowing for the differences. To demonstrate its practicability we present the case study that involves the radiosonde data from three stations: ARM Tropical Western Pacific (0.5O S, 167O E), ARM Southern Great Planes (37O N, 98O W), and Lindenberg (52O N, 14O E). For each station we considered temperature profile validation scenario and estimated associated errors. The model can be used for interpretation of the validation results when the above mentioned sources of discrepancies are significant, as well as for evaluation of validation data sources, e.g. GRUAN (GCOS Reference Upper-Air Network).
Pougatchev, Nikita; Bingham, Gail; Seidel, Dian; and Berger, Franz, "Statistical Approach to Validation of Satellite Atmospheric Retrievals" (2007). Space Dynamics Lab Publications. Paper 108.