IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 56, NO. 3, MARCH 2018 1507
Intercalibration of Polar-Orbiting Spectral
Radiometers Without Simultaneous
Observations
Terhikki Manninen , Aku Riihelä, Andrew Heidinger, Crystal Schaaf, Alessio Lattanzio, and Jeffrey Key
Abstract—A new intercalibration method for two polar-
orbiting satellite instruments or two instrument constellations’
Fundamental Climate Data Records (FCDRs) is presented. It is
based on statistical fitting of reflectance data from the two
instruments covering the same area during the same period,
but not simultaneously. A Deming regression with iterative
weights is used. The accuracy of the intercalibration method
itself was better than 0.5% for the Moderate Resolution Imaging
Spectroradiometer (MODIS) versus MODIS and Advanced Very
High Resolution Radiometer (AVHRR) versus AVHRR test data
sets. The intercalibration of an AVHRR FCDR generated by
NOAA versus a combined MODIS Terra and Aqua data set
of red and near-infrared (NIR) channels was carried out and
showed a difference in the reflectance values of about 2% (red)
and 6% (NIR). The presented intercalibration method can be
used for checking the calibration of two instruments or FCDRs
in all viewing angles used separately.
Index Terms— Calibration, remote sensing, statistics.
I. I NTRODUCTION
A
NECESSARY condition for high-quality long-term
Climate Data Records (CDRs) derived from satellite
remote sensing is accurate intercalibration of instruments over
long timescales [1]. For CDRs, this most often involves the
same type of sensor on a series of satellites, e.g., the Advanced
Very High Resolution Radiometer (AVHRR) on the series
of NOAA polar-orbiting satellites. It may also be useful,
however, to combine observations from different sensors and
satellites. For example, surface albedo is one of the essential
climate variables and a key parameter for the energy balance of
the earth [2]. Albedo retrievals are usually performed utilizing
only a single instrument or an instrument “family” (i.e.,
AVHRR on multiple satellites). This limits the total number
of available observations per time period per terrestrial scene.
Combining multiplatform observations can yield improve-
Manuscript received March 7, 2017; revised June 14, 2017 and
August 29, 2017; accepted September 25, 2017. Date of publication
November 7, 2017; date of current version February 27, 2018. This work
was supported in part by EUMETSAT through the Project CM SAF and
in part by the National Aeronautics and Space Administration for support
of the MODIS work under Grant NNX14AI73G. (Corresponding author:
Terhikki Manninen.)
T. Manninen and A. Riihelä are with the Finnish Meteorological
Institute, FI-00101 Helsinki, Finland (e-mail: terhikki.manninen@fmi.fi;
aku.riihela@fmi.fi).
A. Heidinger and J. Key are with NOAA/NESDIS, Madison, WI 53706 USA
(e-mail: andrew.heidinger@noaa.gov; jeff.key@noaa.gov).
C. Schaaf is with the University of Massachusetts Boston, Boston,
MA 02125-3393 USA (e-mail: crystal.schaaf@umb.edu).
A. Lattanzio is with EUMETSAT, 64295 Darmstadt, Germany (e-mail:
alessio.lattanzio@eumetsat.int).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TGRS.2017.2764627
ments in both the accuracy and temporal resolutions of surface
albedo retrievals. This is especially important for optical
remote sensing in frequently cloudy areas, such as the Arctic.
Various spaceborne optical imager families, such as
AVHRR or Moderate Resolution Imaging Spectroradiome-
ter (MODIS), have different imaging channel wavebands.
Also, the individual imagers within instrument families have
different spectral responses. The intrafamily differences can
and recently have been compensated for by intercalibration,
leading to the creation of Fundamental CDRs (FCDRs) spe-
cific to a single instrument family [7]. When intercalibrating
observations from two different instrument families, even
though both are FCDR quality, due attention must still be paid
to the differences in spectral coverage between the instrument
imaging channels. In this paper, we seek to perform such
an intercalibration between MODIS and AVHRR data sets.
In this case, the intercalibration is extremely important for
the near-infrared (NIR) channel, the spectral width of which
is markedly larger for AVHRR than for MODIS. Although
the motivation of this paper is to compensate for the spectral
difference of MODIS and AVHRR, the presented intercalibra-
tion method can also be applied to normal intercalibration of
different individual instruments of the same family.
The basic premise of intercalibration is that two similar
instruments should produce the same reflectance value when
they view the same target simultaneously with identical view-
ing geometry. In reality, this requirement cannot be rigorously
fulfilled in the calibration of earth-observing instruments.
An established intercalibration method for two satellite instru-
ments is the use of simultaneous nadir observations (SNOs),
which has been proved to be very effective over a wide spectral
range [1]. Another approach is statistical intercomparison,
which is the most frequently used for low-resolution data.
Pseudoinvariant calibration sites (PICSs), such as the Libyan
Desert, have been utilized for both the SNO-based absolute
calibration and the statistical intercomparison of two or more
satellites [2]–[4]. Several approaches have been investigated
for the intercalibration of the AVHRR instrument series with
MODIS data over the PICS [5]–[8]. These methods require
reliable atmospheric correction and either SNOs or reliable
bidirectional reflectance distribution factor estimates. The
method presented here does not need any information about
the surface or atmosphere properties concerning the images
used.
The goal of this paper is to derive a general concept
of intercalibration applicable to any collection of optical
polar-orbiting satellite imagers. Hence, one must consider the
0196-2892 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.