Derivation of Uncertainties for GPM Microwave Radiometer Inter-Calibration

Abstract:

The Global Precipitation Measurement (GPM) mission utilizes several microwave radiometers on individual platforms to measure global precipitation. In order to derive consistent precipitation measurements among the radiometers in the constellation, the radiometers need to be inter-calibrated. This is achieved by adjusting the brightness temperatures (TBs) of the instruments to agree with a common reference using the double difference method. The GPM X-Cal team has derived inter-calibration offsets for the radiometers in GPM, and it has been shown that inter-calibrating the radiometers has a significant positive impact on obtaining consistent precipitation estimates. One important aspect of the inter-calibration that is currently not incorporated in the precipitation algorithm is the uncertainty. The inter-calibration uncertainty may have a significant impact on the precipitation retrievals, so the X-Cal team has recently started to study this in more detail. This presentation will analyze the various contributors to the inter-calibration uncertainty and show how uncertainties can be calculated for the microwave imager channels from 10 to 91 GHz.

Two main contributors to the inter-calibration uncertainty are the radiative transfer model (RTM) and the ancillary inputs to the RTM. The RTM is used to simulate the TBs that a radiometer would observe. This is a necessary component of the inter-calibration algorithm, used to account for differences in frequency, earth incidence angle, and orbits between the radiometers being compared. Since the same RTM is used for all radiometers being inter-calibrated, any errors that the RTM introduces by improperly modeling the TBs should be removed to first order. However, if there are errors that are not linear with respect to frequency or EIA, this will cause sensitivity in the inter-calibration to the RTM. A preliminary study showed sensitivity for those frequencies around the water vapor absorption line (21 to 24 GHz) and for the higher frequencies near 90 GHz, which means that the double difference method is not able to completely account for errors in the RTM. This presentation will expand on that study by quantifying these errors for GPM inter-calibration.