## Using Cosmic Data to Identify AMSU Biases Over Different Orbits and Locations

The comparisons between COSMIC-simulated AMSU Ch9 Tbs and collocated AMSU Ch9 Tbs from N15, N16, and N18 within 15 minutes and 50 km are shown in Fig. 2a, b, and c, respectively. The figures depict that COSMIC synthetic AMSU Tbs for N15 (TbœSMiC_Ni5), N16 (TbœSMiC_Ni6), and N18 (TbœSMiC_Ni8) are highly correlated with TbAMsU_N15 (correlation coefficient = 0.99), TbAMSU^16 (correlation coefficient = 0.99), and TbAMsU_N18 (correlation coefficient = 0.99), respectively, and with small standard deviations to their means of COSMIC-N15 (1.20 K), COSMIC-N16 (1.11 K), and COSMIC-N18 (1.22 K) pairs. This demonstrates the usefulness of COSMIC RO data for inter-calibrating AMSU Tbs, and is consistent with the results of COSMIC RO and AMSU Tbs comparisons for September 2006 (Ho et al. 2009). Because GPS RO data are not affected by orbital change and have no calibration issues, the small differences of the slopes and offsets for COSMIC-N15, COSMIC-N16, and COSMIC-N18 pairs are mainly caused by calibration, orbital drift errors, and measurement anomalies for AMSU Tbs for different missions.

Fig. 2 Comparison of COSMIC-simulated AMSU Ch9 Tbs. (a) N15 AMSU Ch9 Tbs, (b) N16 AMSU Ch9 Tbs, and (c) N18 AMSU Ch9 Tbs for July 2007. Pixels in gray triangle are from the 60°N to 90°N zone, pixels in dark dot are from the 60°N to 60°S zone, and pixels in dark cross are from the 60° S to 90° S zone. The best fit is in dash line. Diagonal one-to-one fit is in solid line

Fig. 2 Comparison of COSMIC-simulated AMSU Ch9 Tbs. (a) N15 AMSU Ch9 Tbs, (b) N16 AMSU Ch9 Tbs, and (c) N18 AMSU Ch9 Tbs for July 2007. Pixels in gray triangle are from the 60°N to 90°N zone, pixels in dark dot are from the 60°N to 60°S zone, and pixels in dark cross are from the 60° S to 90° S zone. The best fit is in dash line. Diagonal one-to-one fit is in solid line

To further quantify COSMIC-AMSU Tb biases at different locations for different missions, we use different symbols in Fig. 2 to represent COSMIC-NOAA pairs at the 60°N to 90°N zone (in gray triangle), the 60° S to 60°N zone (in dark dot), and the 60°S to 90°S zone (in dark cross) and list the mean AMSU Ch9 Tb biases for N15-COSMIC, N16-COSMIC, and N18-COSMIC pairs over the 60°N to 90°N and the 60°S to 90°S zones in Table 1. It is obvious that the absolute NOAA-COSMIC mean Tb biases are smaller in the Northern Hemisphere and are larger in the Southern Hemisphere. The absolute biases for N18-COSMIC pairs are larger than that for N15-COSMIC and N16-COSMIC pairs (Table 1 and Fig. 2).

60°N to 90°N |
60° S to 90° S | |

N15-COSMIC |
- 0.05 K |
- 0.73 K |

N16-COSMIC |
- 0.22 K |
- 0.83 K |

N18-COSMIC |
- 0.55 K |
- 1.50 K |

N15-N16 |
0.03 K (0.17 K) |
0.09 K (0.10 K) |

N16-N18 |
0.47 K (0.33 K) |
0.57 K (0.67 K) |

N15-N18 |
0.50 K (0.50 K) |
0.69 K (0.77 K) |

Fig. 3 Comparison of (a) N15 AMSU Ch9 Tbs and N16 AMSU Ch9 Tbs, (b) N16 AMSU Ch9 Tbs and N18 AMSU Ch9 Tbs, and (c) N15 AMSU Ch9 Tbs and N18 AMSU Ch9 Tbs for July 2007. Pixels in gray triangle are from the 60°N to 90°N zone, and pixels in dark cross are from the 60° S to 90° S zone. The best fit is in dash line. Diagonal one-to-one fit is in solid line

Fig. 3 Comparison of (a) N15 AMSU Ch9 Tbs and N16 AMSU Ch9 Tbs, (b) N16 AMSU Ch9 Tbs and N18 AMSU Ch9 Tbs, and (c) N15 AMSU Ch9 Tbs and N18 AMSU Ch9 Tbs for July 2007. Pixels in gray triangle are from the 60°N to 90°N zone, and pixels in dark cross are from the 60° S to 90° S zone. The best fit is in dash line. Diagonal one-to-one fit is in solid line

To see if the NOAA-COSMIC biases are consistent with AMSU inter-satellite biases from different missions, AMSU Ch9 Tbs for N15-N16, N16-N18, and N15-N18 pairs are plotted in Fig. 3a-c, respectively. In Fig. 3, AMSU pairs in the 60°N to 90°N zone and the 60°S to 90°S zone are in gray triangle and in dark cross, respectively. The mean Ch9 Tb biases for N15-N16, N16-N18, and N15-N18 pairs for the 60°N to 90°N and the 60°S to 90°S zones are also listed in Table 1. For NOAA polar satellites, the collocated N15-N16, N16-N18, and N15-N18 pairs within 15 minutes and 50 km all occurred only over polar regions. It is shown in Table 1 that the mean AMSU inter-satellite biases are consistent with those for COSMIC-AMSU pairs where the mean NOAA-COSMIC Tb biases for different missions are all negative. The mean N15-N16, N16-N18, and N15-N18 Tbs at both the 60° S to 90°S and 60°N to 90°N regions can almost always be reproduced using Tb biases of NOAA-COSMIC pairs. For example, the mean N15-N18 Tb bias in the 60°N to 90°N zone (0.5 K) is equal to Tb bias of N15-COSMIC (-0.05 K) minus that of N18-COSMIC (-0.55 K). The reason that some mean N15-N16, N16-N18, and N15-N18 Tbs cannot be exactly reproduced using Tb biases of NOAA-COSMIC pairs (although their Tb differences are small) is mainly due to sample differences among NOAA-COSMIC and NOAA-NOAA pairs.

To find the cause of NOAA-COSMIC Tb biases at the different locations shown in Table 1 and Fig. 2, we plot the binned N15-COSMIC AMSU Ch9 Tb differences for each local time and binned latitude variation for N15 orbit for each local time in Fig. 4. In general, the variation of N15-COSMIC AMSU Tb biases is highly

Fig. 4 Binned N15-COSMIC AMSU Ch9 Tb differences for each local time and binned latitude variation for N15 orbit for each local time. The gray solid line is for N15-COSMIC AMSU Ch9 Tb = 0 for different local times

Fig. 4 Binned N15-COSMIC AMSU Ch9 Tb differences for each local time and binned latitude variation for N15 orbit for each local time. The gray solid line is for N15-COSMIC AMSU Ch9 Tb = 0 for different local times

coherent with the N15 orbit change with local time, where N15 satellite may be warmed up during the day and cooled down during the night. It is shown in Fig. 4 that N15-COSMIC Tbs are in general lower during the southern hemispheric winter where N15 is under the shadow of the Earth (solar zenith angle is larger than 80 degrees) and are higher in the Northern Hemisphere (ranges from 7 to 17 local times). Because GPS RO data are not affected by the temperature variation of the satellite component, the mean N15-COSMIC AMSU Tb biases are mainly from AMSU Tb anomalies due to the heating or cooling of the satellite component. Note that because both NOAA satellites are cooled or warmed in the same manner when collocated NOAA pairs are collected, the effect of AMSU temperature anomalies due to satellite heating or cooling is canceled out in AMSU inter-satellite bias comparisons (comparing Figs. 2 and 3). Due to lack of an absolute reference, only relative inter-satellite biases are corrected (absolute AMSU inter-satellite biases are all smaller than that for NOAA-COSMIC pairs), which may lead to uncertainties for climate trend analysis.

## Post a comment