IEEE Geoscience and Remote Sensing Magazine - June 2015 - 26

Usually, the services are expecting data observed at a
given spectral band (Visible, Near Infrared). However, together with the spectral bandwidth, the relative spectral
responses (RSR) strongly depend on the sensor. As the RSR
information is not always shared by the data provider sensor interoperability may not be correctly assessed from the
user point of view, even if the
methods for the comparison
of RSRs are available [8]. AlTHE CQC TEAM PERFORMS
though there is a direct relaINDEPENDENT QUALITY
tionship between RSR and
CHECkING AND vALIDATION
the physical measurements
OF THE LEvEL 1 DATA
made, the underlying uncertainties of the atmospheric
PRODUCTS. BESIDES
corrections and bidirectional
THIS, THE CQC TEAM IS
reflectance corrections canRESPONSIBLE FOR
not be correctly assessed
COLLECTING, ANALYzING,
without RSR information.
HARMONIzING AND
As for the spectral bandExPOSING TO THE
width definition, the Spatial
COPERNICUS SERvICES
Resolution (SR) is the key criTHE QUALITY INFORMATION
teria involved in the selection
of remote sensing data, more
PROvIDED BY THE
important than the pixel spacNUMEROUS DATA
ing. The SR allows assessing
PROvIDERS.
the level of detail in the image [9] and is fundamental for
comparing the capability of
two different EO sensors. The main issue is that the SR definition is understood differently among actors. The CQC team
analyzed the various approaches adopted for the computing
of modulation transfer function, and also recommends the
way to report this key parameter to the community.
The geometric accuracy of an image is a critical point
when merging multi-source data and strongly depends on
the SR but is not solely restricted to this element.
In the current literature, various reports describe tedious
work for co-registering data [10], obtaining an accuracy that
is not under control. The topic is becoming more important because of new requirements for time series production and sub-pixel accuracy. Despite this, it is not common practice for the data providers to share information
on the stability of the geolocation accuracy and even less
on the coherence of the internal geometry of the orthorectified products. Those parameters play an important role
and should be monitored for all applications involving the
long term tracking of various phenomena. Previous studies [11], [12] showed that even a small error in registration
might have a large impact on the accuracy of global change
measurements: a registration accuracy of 1/5 of a pixel is
required to achieve a change detection error of less than
10 %. Conversely, previous validation studies on optical High Resolution images (Landsat, SPOT) have shown
that the geometric accuracy was varying depending on the
geographical location and might, in some cases, not be in
agreement with the nominal specifications.
26

The radiometric calibration parameters have also raised
questions on traceability. Some services apply atmospheric
corrections in order to obtain the bottom of atmosphere
measurements from the top of atmosphere measurements.
The transformation from radiance to TOA reflectance is
standard but the parameters involved may differ significantly due to the solar extra-terrestrial irradiance reference
choice for instance. In addition, the comparison of data
from different sensors involves the traceability of the dependency between the physical measurement and spectral response of the sensor. This raises the question if the service is
able to use the calibration parameter and spectral response
in a transparent way.
The last but not least critical issue is the product format.
The assimilation of multi-source Copernicus data is quite
tedious, as each data provider has his own understanding
of the level of information that should be made available to
the user and also adopts his own format: from a poor ASCII
file format to complex binary format. Furthermore, it happens sometimes that the format depends on the processing
level. Consequently, the usage of products is suffering from
inappropriate or inaccessible information. In addition, the
user is in some cases forced to design and apply his own
metadata transformation to obtain a common representation of the data.
It is now evident that the work required to draw up a status
on the main issues and subsequently initiate harmonization
activities is complex. Over the course of the last three years,
the CQC team has invested a significant amount of time on
all of these topics, some result of which are reported herein.
This paper is organized into 4 sections. The introduction
describes the context and purposes of the CQC service. The
second section describes the methods and the system used
for managing and processing of the quality information.
The third section details quality control results, validation
results and harmonization of the findings. A summary section concludes the paper.
II. QUALITY INFORMATION PROCESSING:
METHODS AND SYSTEM
ESA offers a centralized data archive dedicated to Copernicus services. The archive includes several datasets, consisting of a collection of multi-source remote sensing data
depending on the mission group broken into the following
categories:
◗ Sentinel-1 for radar imaging data;
◗ Sentinel-2 for multispectral high-resolution optical data;
◗ Sentinel-3 for multispectral medium-resolution optical,
infrared and radar altimeter data;
◗ Sentinel-4 for atmospheric monitoring from geostationary orbit;
◗ Sentinel-5 for atmospheric monitoring from polar orbit;
◗ Sentinel-6 reserved to Sentinel-3 follow on.
The data is more likely to be fully exploited when quality information is included. Therefore, the baseline concept has been to have an independent system in charge of
ieee Geoscience and remote sensing magazine

june 2015

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