HIRS Pathfinder Intercalibration

After construction of the HIRS Pathfinder clear-sky data set, we began the intersatellite calibration by first looking at the observed Tb differences for all the IR channels. The table below presents time series for each channel and various latitude bands of the monthly mean clear-sky TB from the 27th (center) scan position.
TB Observed time series

The following tables provide values for the intersatellite TB differences between the various satellites and channels. Tables 1 and 2 give the modeled TB difference with NOAA-10 using MODTRAN 3.7 for simulating the HIRS TBs and using TIGR-3 tropical profiles. The figure shows the bias between all the satellites. This figure excludes data for N15 channel 16 and N11-N15 channel 17 data since these channels were moved to a new central frequency. Channel 10 also had changes to the central frequency for N11, N14 and N15 but experienced little change in mean Tb value. Changes in the central frequency for N15 HIRS chn 12 are shown here.
Simulated TB means Table
Simulated TB differences with N10 Table
Simulated TB differences with N10 Figure

The following figures provide values for observed intersatellite bias from the Pathfinder clear-sky data. The observed bias was determined from the monthly mean brightness temperature data for each satellite. The satellite climatology helps remove natural interannual variations from the bias calculation. However, satellites with short lifespans, such as NOAA-8, may only have one monthly mean data point for its climatology. For the following figures, only scan position 27 (nadir) is averaged. January and July climatologies are compared here. The shapes of the bias curves are similar for most channels. Channels 1-6 show negative trends for both months with very similar bias amplitudes for the individual satellites. Channel 7 does show differences for N9,N10,N11 between the two months. It is unknown why Channel 16 shows such a large bias in N9. The observed TBs are 40K warmer, yet the central frequency for the channel is reported in Kidwell to be the same as the other satellites. Channels 8, 18 and 19 have a periodic curve because the daytime observations cause the afternoon satellites to have warmer observations. Channels 18 and 19 are larger in amplitude because the daytime observations have enhanced TBs due to a small contribution from the visible spectrum. Subtropical latitude bands are shown to investigate latitudinal dependencies of the bias. The largest difference from the 10N-10S results is the amplitude of the warm TB signal from the afternoon passes for the surface channels.
Observed Nadir view 10N-10S January TB differences with N10 Figure
Observed Nadir view 10N-10S July TB differences with N10 Figure
Observed Nadir view 10N-30N January TB differences with N10 Figure
Observed Nadir view 10S-30S January TB differences with N10 Figure

So how well do the observed and simulated biases compare? The following histograms compare the data given in the above figures. Each histogram compares one channel for all the satellites. Remember that the observed biases include not only inter-satellite bias due to filter differences but also can include natural variations seen in the atmosphere.
Observed versus Simulated intersatellite bias histograms

A multiple linear regression was developed to predict the Tb differences between satellites. The results below show the how linear regression method can both improve and worsen the intersatellite bias. Results are shown only for channel 1 to 12 since both day and night observations are included here. The linear regression method is not valid for daytime observations for the near IR channels (13-19).
TB Linear Regress correction time series

Given the mixed results from the linear regression method, we decided to remove the intersatellite bias using the EDF method used in Bates et al. (1996). Final results for the EDF adjustment and bias correction are given below.
TB EDF correction time series

Last modified: Tue May 15 19:43:57 MDT 2001