VIIRS SDR Algorithm Three-Phase Performance Test Approach for Radiometric Calibration

The Visible Infrared Imager Radiometer Suite (VIIRS) is part of the NPOESS Preparatory Program (NPP). Northrop Grumman Corporation (NGC) is responsible for the calibration and validation of the instrument sensor data records (SDR). NGC took a unique three-phase approach to testing the SDR radiometric performance. The approach was dictated by the progress of the sensor testing and the SDR algorithm development. The synergy of testing the SDR algorithm with both a sensor model and with test data produces an understanding that neither one alone can produce. In the first phase NGC developed a sensor model and produced look-up tables (LUT) consistent with the model. This was necessary because complete characterization was not yet available. It demonstrated consistency between model and SDR Algorithm. In the second phase we used the actual raw data from the sensor testing and characterization as input into the SDR. The LUT were produced from characterization. This had the advantage of being most consistent with actual sensor, but by itself is limited by the limitations of the tests. It also does not involve the model at all. The third phase was to develop sensor model from same characterization data used to produce SDR LUT. By itself this potentially has a problem in that a mistake in both would be masked by a consistent, but wrong result. It did, however, demonstrate consistency between model and SDR Algorithm. This 3 phase process validates to a certain level both the SDR algorithm and the sensor model and guarantees consistency between the theory that the SDR algorithm is based, and the algorithm as programmed. Comparing the results from each phase resolves the limitations inherent in each. The logic is that if as-built sensor is consistent with SDR algorithm, and if SDR algorithm is consistent with as-built sensor model, then as-built sensor is consistent with as-built sensor model. We have completed the 3 phases of test and have shown that the SDR algorithm and sensor model are consistent with the as-built sensor. By comparing with SDR performance predictions based on analysis outside of the actual SDR tests, consistency is shown between how the SDR should theoretically perform, and how the actual algorithm as programmed performs.