![]() |
|
|
Calibration/Validation for Solar Reflectance ChannelsOperational Calibration of the AVHRR Solar ChannelsNOAA's operational calibration of AVHRR solar channels is based on continuing observations of the selected calibration target, the Libyan Desert. Every month, new data were collected, combined with previous data, and analyzed to yield current calibration. NOAA-16The orbit of NOAA-16 has drifted such that no data with solar zenith angle less than the threshold value (60º) have been collected over the Libyan Desert since Oct. 18, 2006. This situation is expected to continue until March 2007. In addition to the loss of data, the quality of the collected data is also in question since NOAA-16, being six years in operation, has exceeded its designed lifetime. Noisy data occur more frequently lately than earlier in its life. On the other hand, instrument degradation has stabilized to the point that data collected in the past seems adequate to characterize the instrument in future. In the operational calibration algorithm, the measurements available at the time are fitted to an empirical model: Equation 1:
where t is time; y is estimate by regression; R0 is a regression parameter representing the instrument measurement of target reflectance at t=0; d is a regression parameter representing the linear degradation rate; A is a regression parameter representing amplitude of seasonal variation of target reflectance, a manifestation of local bidirectional reflectance distribution function (BRDF); ß is a regression parameter representing phase angle at t=0, and ? is frequency (2p/365). If the model is proper and the data contain only random errors, the four regression parameters should become nearly constant over time. Figure 1 is a summary of instrument performance monitoring for the AVHRR solar channels on NOAA-16. The four panels are time series of instrument degradation rate d; required adjustment to compensate for the accumulative instrument degradation y; ratio of R0/Rr where Rr is reference target reflectance (37.8%, 42.6%, and 66.9% for Channels 1, 2, and 3, respectively); and the amplitude of seasonal variation of target reflectance A. One obvious feature in all panels and for all channels is that the data are noisy at the beginning, especially before the year 2002. This is understandable, as NOAA-16 was launched in September 2000 and it takes at least one year to model a phenomenon with annual variation as Eq. 1 implies. Another feature is that the time series for Channel 3A is short, since this channel was permanently deactivated on May 1, 2003. Figure 1. Instrument monitoring for NOAA-16 AVHRR solar channels. The NOAA-16 was launched on Sept. 21, 2000. Instrument characterization in the first year after launch was not possible because of the annual variation of surface bidirectional reflectance distribution. Later, the sensor’s performance becomes more predictable, especially for Channel 1. ![]() Focusing on the plots for Channel 1 in Fig. 1, one notices that the degradation rate and ratio of R0/Rr have been stable most of the time since 2002, indicating skill of the model and quality of the input data. The total adjustment increases steadily over time, also as expected. The amplitude of seasonal variation of target reflectance increased very slightly over time. This may be real, as BRDF tends to increase with increased solar zenith angle (which is the case when NOAA-16 orbit drifted). Much the same can be said for Channel 2, except that there are more long term and short term variations on all these parameters. This is probably because Channel 2 measurements are affected by the water vapor amount in the atmosphere, which has its own seasonal variation and even inter-annual variation that make Eq. 1 less a valid model. Nevertheless, the model based on Eq. 1 captures the major part of the variation even for Channel 2. Based on these recent developments and our analysis of instrument monitoring, it is recommended to stop collecting new data for NOAA-16 AVHRR solar channel calibration. Instead, monthly update of calibration coefficients should be based on previously collected data and extrapolation into future. NOAA-18NOAA-18 was launched on May 20, 2005. As we have learned from Fig. 1, it is not possible to calibrate its AVHRR solar channels using the Libyan Desert shortly after launch. Instead of waiting for 18-24 months, as were the cases for NOAA-16/17, it was decided to use part of the NOAA-16 model for NOAA-18 to account for the BRDF. This is based on the assumption that the BRDF for NOAA-16 and NOAA-18 should be similar since they are on similar orbit. In terms of Eq. 1, the A and ß as derived from NOAA-16 data were used, together with NOAA-18 data, to derive R0 and d for NOAA-18. Recent results (Fig. 2) suggest that adequate data may have been collected by NOAA-18 over the Libyan Desert to calibrate its AVHRR solar channels, independent of NOAA-16. It just so happened that NOAA-16 may no longer be able to consistently collect new and high quality data in future. It is therefore recommended to make the operational calibration of NOAA-18 AVHRR solar channels independent of NOAA-16. Since we are not yet fully confident about the length of NOAA-18 AVHRR data and we do have the option of using NOAA-16 AVHRR model, we will monitor NOAA-18 more closely in the near future (next six months or so). Figure 2: As Figure 1, but for NOAA-18. ![]() NOAA-17Figure 3: As Figure 1, but for NOAA-17. ![]() |