Calibrating a System of Satellite Instruments - Satellite-based Applications on Climate Change - page 24

differences and uncertainties have to be resolved in the recalibration. This type of
problem is significantly reduced with hyperspectral sounders such as AIRS and
IASI, which are both spectrally resolved and contributed to the excellent agreement
between them (Wang 2009, personal communications).
The second feature from Fig. 2.5 is that the largest bias occurred between
NOAA-14 and NOAA-15 HIRS which oscillates seasonally. This was further
investigated in a study (Cao et al. 2008b ) which demonstrated that spectral uncer-
tainty in the HIRS on either NOAA-14 or NOAA-15 is likely the culprit, and a
spectral shift is necessary to reproduce this large bias based on studies with IASI
spectra. This also suggests that prelaunch characterization of spectral response
functions is critical for postlaunch instrument performance.
Finally, it is evident that for both channels 4 and 6, the inter-satellite biases
between HIRS on NOAA-18 and other satellites have large variability. This is due
to the fact that NOAA-18 HIRS has a loose lens problem which introduces large
noise for all long-wave channels. Apparently, instrument noise increases the uncer-
tainty in SNO time series analysis.
On the other hand, given the relatively low noise for most of the operational
instruments, in general, noise is not the major constraint in the inter-satellite analysis.
Studies have shown that the SNO method can resolve inter-satellite biases on the
order of 0.1 K in the sounding channels of the microwave and infrared instruments
and 1% in the visible/near-infrared imagers (Cao et al. 2005a , b , 2008a ). In both
cases, the uncertainty is larger than the instrument noise levels.
The SNO method has its limitations. First, since the SNOs occur at many different
locations in the polar regions, the spectral characteristics of the SNO sites are not yet
well quantified and introduce uncertainties in the intercalibration of window channels.
Second, while the SNO method works well for the sounding channels in the micro-
wave and infrared, larger uncertainties are found for low-resolution surface channels
where surface inhomogeneity and pointing accuracy become the limiting factors (Cao
et al. 2005a , b , 2009 ; Zou et al. 2006 ). Third, for the infrared window channels (and
some infrared sounding channels), the temperature at the SNO (below 280 K) is
limited to a narrow range which does not cover the full range of the global surface
temperature. Fourth, for the infrared instruments, inter-satellite biases at the SNO
points may not be representative of the biases over an orbit due to orbital variations of
calibration accuracy in response to fluctuations in instrument temperature and stray
light in certain parts of the orbit (Cao et al. 2001 , 2004 ; Trishchenko et al. 2002 ).
Finally, the SNO method is very sensitive to geo-location and sampling errors, for
example, the AVHRR 4 KM GAC (Global Area Coverage) data does not have an
optimal match with the MODIS 1KM data due to the sampling scheme used in
AVHRR, which introduces uncertainties in intercalibrating AVHRR and MODIS.
Significant progress has been made intercalibrating GOES and POES radiometers
in recent years under theWorldMeteorological Organization's (WMO) Global Space-
based Inter-calibration System (GSICS) program, extending the existing work in this
area (Gunshor et al. 2004 ). Conceptually, since the GOES nadir is fixed at a given
location and the POES satellites pass the GOES nadir point regularly, it would seem to
be an ideal configuration for intercalibration. However, there are several challenges.