IEEE Geoscience and Remote Sensing Magazine - June 2015 - 31

However, the presence of sand dunes, having a specific
North South orientation creating a spatial asymmetry,
makes the site probably less reliable for very high resolution data due to its sensitivity to directional effects. There is
no current paper on this topic.
There are many examples of re-calibration procedures
applied to visible and near infrared sensors performed by
using observation from bright sites [27]. The re-calibration
procedures are performed in order to maintain the long
term stability of radiometric calibration during the mission
lifetime and even program lifetime [28] [29]. The re-calibration is also required to maintain consistency between
instruments of an identical design, operating in a constellation mode, as with the DMC satellites and the RapidEye
Black Bridge for instance.
In the Copernicus world, the services are using data
from different missions, it is therefore important to
validate the inter-calibration between instruments. Furthermore, multi-temporal data are being ingested; highlighting the importance of validating stability of the radiometric calibration.
The sensitivity of inter-calibration methods, when applied to medium resolution data and by using desert sites,
is largely discussed in [22] and [23]. It shows that inter-comparison is accurate with doublet selection of concomitant
identical and reciprocal observations. The measure to compare geometry of observation is given in [23]. In the proposed method, the Landsat TM/ETM+, MERIS and MODIS
time series are considered as calibration references because
they are from reference sensors. However, the rescaling of
the input reflectance to a reference sensor is not systematic;
it is done in some cases, as shown in Table 3.
From our point of view, besides the stability of the mission itself, the most important aspect is to quickly evaluate the agreement between the input TOA reflectance data
and the reference time series discussed above. A statistical
approach is carried out. In case of a large deviation from
the reference, more investigations are initiated. The investigations include procedures similar to the one explained in
[23]. The accuracy of this method is estimated to be within
5 %: a 5 % difference between input data and calibration reference data means that it is within the uncertainty budget.
By using Landsat data, analysis showed that the TOA
reflectance of the Libya-4 site is extremely stable from a
temporal point of view. The Table 3 lists statistics of the
temporal stability of the radiometric calibration. They are
grouped by Landsat mission (Landsat 5 TM, Landsat 7
ETM+ and Landsat 8 OLI). They are given depending on
the band number and are deduced from TOA measurement time series. The differences between results are mostly due to relative spectral response differences. Over a typical year, the temporal stability of the time series is below
1% in case of visible and near infrared bands and below
1.5% for short wave infrared bands. These results form a
good baseline for comparing with radiometric stability of
Copernicus missions
june 2015

ieee Geoscience and remote sensing magazine

The region of interest (ROI) covers area of a square degree and is centered on the point that defines the geographical coordinate of the site. All input images are automatically
geometrically registered to a reference image, the ROI is extracted, Digital counts converted into TOA reflectance based
on the extraterrestrial solar irradiance defined in [30] and
statistics are computed. It was expected that the data would
have little trend and only exhibit a random scatter depending on the band number. Sometimes the number of available products is not sufficient to estimate a trend and therefore the comparison with reference time series provides only
a preliminary assessment.
The shortcoming of this method is that the influences
of atmosphere and Bi-directional Reflectance Distribution
Function (BRDF) are not accounted for. To measure the
impact of the first parameter,
By using Landsat data,
the sample data are always
observed during the winanaLysis showed that
ter (end of December) and
the toa refLectance of
during the summer (end of
the LiBya-4 site is
June). To control the impact
extremeLy staBLe from a
of the second parameter and
temporaL point of view.
consequently allow a direct
these resuLts form a
comparison of different misgood BaseLine for
sions, data from nadir obcomparing with
servation without body/inradiometric staBiLity of
strument pointing applied is
requested. As reported by S.
copernicus missions.
Saunier [21], when discarding off-nadir observations in
case of ALOS/AVNIR-2, the
stability of radiometric calibration is estimated to be within
1%, considering all bands during a period of 2 years.
b) Geometry accuracy
For the geometric calibration, the 'La Crau' test site is used;
located in the southern part of France, near the Mediterranean Sea at coordinates 43.513°N, 4.875°E, with a terrain
elevation ranging from 50 m to 240 m above the WGS84
ellipsoid. The 'La Crau' test field has been equipped for
the purposes of the ALOS Commissioning phase as ESA
contribution for the JAXA Cal/Val [21]. Consequently, a
set of 40 Ground Control Points (GCPs), recorded with
Differential Global Positioning System (DGPS) have been
collected, their accuracy is within 10 cm. The GCP set is
distributed over an area of 75 km (East-West) by 35 km
(North-South).
The GCP set is considered as a reference source. In some
cases, the points are directly used to validate the geometry
of Very High Resolution (VHR) optical data, in other cases
through the geometric refinement process, the points allow a
raster reference to be produced, subsequently used to validate
High Resolution (HR) and Medium Resolution (MR) data.
The production of the raster reference, starting from
the GCP set, is a streamlined procedure that may include
31

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2015