The AMSU asymmetry could be a combination of several causes: (1) polarization angle alignment; (2) antenna pointing angle; and (3) an intrusion of the solar array. The initial analyses for NOAA-15 and -16 AMSU show that the adjustment of -1.5 degree to the instrument polarization angle is needed in order to eliminate the asymmetry.

To identify the asymmetry along the scan, we simulate the brightness temperatures at 23.8, 31.4, 50.3 and 89 GHz over oceans corresponding to each AMSU-A beam position and compare the simulations with the measurements. Radiative transfer modeling is performed under clear atmospheric conditions using sea surface temperature (SST), wind vector and temperature/moisture profiles obtained from the NCEP global data assimilation system (GDAS) as inputs. The SST is generated using a blended technique of satellite and conventional observations. Surface wind data are analyzed through assimilating available microwave satellite estimates from the SSM/I with buoy measurements. The GDAS global analyses have a resolution of one degree in latitude and longitude and are produced four times a day. The difference between simulated and observed brightness temperatures is illustrated in Fig. 1.

Figure 1. AMSU Brightness Temperature Bias vs. Beam Position

It appears that the bias is asymmetric relative to the nadir. For example, at 31.4 GHz (Fig. 1b), a bias is positive for the beam position from 1 to 14 and negative for the beam position from 15 to 30. While the measurements at 23.8 and 31.4 GHz are obtained from the same AMSU-A2 module, the bias seems to be smaller at 23.8 GHz (Fig. 1a) than that at 31.4 GHz (Fig. 1b). Also, the bias at 50.3 GHz (Fig. 1c) is smaller than that at 89 GHz (Fig. 1d) although both channels are situated on the AMSU-A1 module. The asymmetry appears the worst at 31.4 GHz where the atmosphere is the most transparent. The AMSU radiometer at the window channels is designed to receive the radiance at the vertical polarization at the satellite nadir. As the reflector scans across the track, the electric components at both horizontal and vertical polarization states from the Earth surface are entered into the receiver. The radiance received from any scan angle is a weighted average by

(1a)

where A and B are related to the antenna reflector normal angle (q); the polarization alignment angle (y) and the scan angle (j). Using the AMSU instrument parameters as shown in Fig. 2, we derive:

	(1b)
	(1c)
Where,
	(1d)

       A nominal performance is set by q = 45^o and y= 90^o. Thus, A = sinq and B = cosj

An offset of y to its normal design may result in the asymmetry. Using Eqs. (1a) - (1d), we can simulate the brightness temperature bias from its nominal performance due to the polarization alignment. Figs. 3a~3d display the possible biases at four AMSU window channels by setting y=91^o(solid lines) and y=92^o(dash lines), respectively. The simulations are performed under a clear atmosphere over the ocean with all AMSU instrumental parameters being set to their norms. Apparently, the bias is largest at 31.4 GHz at which the atmosphere is most transparent (see Fig. 3b). At 50.3 GHz, the bias is smaller due to the larger opacity of atmospheric oxygen absorption (see Fig. 3c). At 23.8 and 89 GHz, the biases are similar because the absorption intensities of water vapor are similar. Overall, the simulated biases at four frequencies exhibit similar characteristics to those obtained from the measurements.

Figure 3. Simulated brightness temperature biases at four AMSU channels due to the polarization alignment angle errors. The solid and dashed lines are compared between y=91 and 92 respectively, (a)23.8 GHz; (b)31.4 GHz; (c)50.3 GHz; and (d)89 GHz