Atmospheric Temperature Climate Data Records from Satellite Microwave Sounders - Satellite-based Applications on Climate Change - page 117

by Mears et al. ( 2003 ) are used for the diurnal drift corrections in version 2.0 and
earlier versions of the NOAA MSU/AMSU temperature of mid-atmosphere (TMT)
product. This is a merged product from MSU channel 2 and AMSU-A channel 5
observations. To reduce uncertainties in the diurnal anomaly dataset, a scaling
factor to multiply the anomaly amplitude is introduced. An optimum scaling factor
is obtained by minimizing inter-satellite differences over land. Impact studies
showed that this correction generated a consistent TMT trend between the land and
oceans (Zou and Wang 2010 ), suggesting that the diurnal anomaly is reasonable.
Other products such as temperatures of the upper troposphere (TUT) and lower
stratosphere (TLS) do not include a diurnal drift correction since its effect can be
ignored for these channels (Zou et al. 2009 ).
8.3.4 Residual Inter-satellite Bias Correction
Although inter-satellite biases and the solar heating-induced temperature variability
signals in the radiances were mostly removed from the SNO Level-1c calibration,
small residual errors still exist in the gridded inter-satellite difference time series
(Fig. 8.6 ). These small residual biases need to be completely removed before
merging the satellite data for TCDR generation. Empirical correction algorithms
have been developed by different investigators to remove these biases (Christy et al.
2000 ). It was shown that using the Christy et al. ( 2000 ) approach on top the SNO
calibration yielded stable MSU trends (Zou and Wang 2010 ). In the Christy et al.
( 2000 ) approach, a best fit empirical relationship between the brightness tempera-
ture correction term and the warm target temperature is obtained for the difference
time series as shown in Fig. 8.3 by solving multi-regression equations, and then the
best fit is removed from the unadjusted time series. The correction result on MSU
channel 2 for this method is shown in Fig. 8.8 . As seen, after the residual bias
correction, global ocean-mean inter-satellite differences are nearly zero for all
overlaps with no obvious bias drift.
Currently, global ocean-mean residual biases are corrected using the Christy
et al. ( 2000 ) approach in the NOAA MSU/AMSU TCDR products.
8.3.5 Correction of the Earth-Location-Dependent Biases
Although SNO calibration minimized global mean inter-satellite biases and instru-
ment temperature signals in the Level-1c data, Earth-location-dependent inter-
satellite biases still exist for certain channels on certain satellites in the gridded
time series (Zou et al. 2009 ). This occurred because the nonlinearity of the
radiometer transfer function was assumed to be a quadratic type. It is possible
that higher-order nonlinearities exist for certain channels and these unresolved
nonlinearities in the calibration equation may cause inter-satellite biases to depend