## Using the Calibrated AMSU Data to Calibrate Other Overlapped AMSU Measurements

From the previous section, we know the best fit for N16 Tbs and COSMIC data for one particular month (in this case, July 2007, e.g., Fig. 2b ), which is

and the best fit for N15-N16 pairs (Fig. 3a ) is

To use the COSMIC-calibrated N16 Tbs to calibrate N15 Tbs is to construct the calibration coefficients for COSMIC-N15 pairs using Eq. (1) and Eq. (2), and we have

The robustness of the slope and offset in Eq. (3) can be estimated by comparing TbCOSMIC^15^16 to TbCOSMIC^15 constructed from COSMIC-N15 pairs (Fig. 2a), where the best fit for COSMIC-N15 pairs is

By applying TbAMSU^15 in Fig. 2a to Eqs. (3) and (4), respectively, we have the computed TbCOSMIC^15 and TbCOSMIC^15^16 (Fig. 5a). Figure 5a depicts that the correlation coefficient of TbCOSMIC_N15_N16 and TbCOSMIC_N15 is equal to 1.0 and

their mean bias is very close to zero (~0.06 K). The very tight fit of TbCOSMIC^15 and TbCOSMIC_N15.N16 (the standard deviation is about 0.04 K) demonstrates the consistency between the slope and offset (calibration coefficients) found in the N15-COSMIC pairs and those constructed from both N16-COSMIC pairs and N15-N16 pairs. The uncertainty of the N16-calibrated N15 Tbs is about 0.06 K when compared to the COSMIC-calibrated N15 Tbs. Similarly, we can also calibrate N18 Tbs using calibration coefficients for N16-COSMIC (Eq. 1) and N16-N18 pairs from Fig. 3b. The best fit for N16-N18 pairs is

The calibration coefficients for new COSMIC-N18 pairs reconstructed using Eqs. (1) and (5) are

TbCOSMIC_N18_N16 = 0-981 ■ TbAMSU_N18 + 4.91, (6)

where the best fit for COSMIC-N18 pairs (Fig. 2c) is

TbCOSMIC_N18_N16 computed from Eq. (6) are plotted together with TbCOSMIC_N18 computed from Eq. (7) in Fig. 5b. Figure 5 b shows that the correlation coefficient of TbCosMIC_N18_N16 and TbCOSMIC^18 is equal to 1.0 and their mean bias is also very close to zero (~-0.01 K).

## Post a comment