ArticlePDF Available

Data Service Platform for Sentinel-2 Surface Reflectance and Value-Added Products: System Use and Examples

Authors:

Abstract and Figures

This technical note presents the first Sentinel-2 data service platform for obtaining atmospherically-corrected images and generating the corresponding value-added products for any land surface on Earth (http://s2.boku.eodc.eu/). Using the European Space Agency’s (ESA) Sen2Cor algorithm, the platform processes ESA’s Level-1C top-of-atmosphere reflectance to atmospherically-corrected bottom-of-atmosphere (BoA) reflectance (Level-2A). The processing runs on-demand, with a global coverage, on the Earth Observation Data Centre (EODC), which is a public-private collaborative IT infrastructure in Vienna (Austria) for archiving, processing, and distributing Earth observation (EO) data (http://www.eodc.eu). Using the data service platform, users can submit processing requests and access the results via a user-friendly web page or using a dedicated application programming interface (API). Building on the processed Level-2A data, the platform also creates value-added products with a particular focus on agricultural vegetation monitoring, such as leaf area index (LAI) and broadband hemispherical-directional reflectance factor (HDRF). An analysis of the performance of the data service platform, along with processing capacity, is presented. Some preliminary consistency checks of the algorithm implementation are included to demonstrate the expected product quality. In particular, Sentinel-2 data were compared to atmospherically-corrected Landsat-8 data for six test sites achieving a R2 = 0.90 and Root Mean Square Error (RMSE) = 0.031. LAI was validated for one test site using ground estimations. Results show a very good agreement (R2 = 0.83) and a RMSE of 0.32 m2/m2 (12% of mean value).
Content may be subject to copyright.
remote sensing
Technical Note
Data Service Platform for Sentinel-2 Surface
Reflectance and Value-Added Products:
System Use and Examples
Francesco Vuolo 1, *, Mateusz ˙
Zółtak 1, Claudia Pipitone 1,2, Luca Zappa 1, Hannah Wenng 1,
Markus Immitzer 1, Marie Weiss 3, Frederic Baret 3and Clement Atzberger 1
1Institute of Surveying, Remote Sensing & Land Information (IVFL), University of Natural Resources and
Life Sciences, Vienna (BOKU), Peter Jordan Str. 82, 1190 Vienna, Austria; mateusz.zoltak@boku.ac.at (M. ˙
Z.);
claudia.pipitone02@unipa.it (C.P.); luca.zappa2@studenti.unimi.it (L.Z.); hannah.wenng@gmail.com (H.W.);
markus.immitzer@boku.ac.at (M.I.); clement.atzberger@boku.ac.at (C.A.)
2Department of Civil, Environmental, Aerospace, Materials Engineering (DICAM), University of Palermo,
Viale Delle Scienze, Bld. 8, 90128 Palermo, Italy
3
Institut National de la Recherche Agronomique—Université d’Avignon et des Pays du Vaucluse (INRA-UAPV),
228 Route de l’Aérodrome, 84914 Avignon, France; marie.weiss@paca.inra.fr (M.W.);
baret@avignon.inra.fr (F.B.)
*Correspondence: francesco.vuolo@boku.ac.at; Tel.: +43-1-47654-85735
Academic Editors: Lenio Soares Galvao and Prasad S. Thenkabail
Received: 1 August 2016; Accepted: 6 November 2016; Published: 11 November 2016
Abstract:
This technical note presents the first Sentinel-2 data service platform for obtaining
atmospherically-corrected images and generating the corresponding value-added products for
any land surface on Earth (http://s2.boku.eodc.eu/). Using the European Space Agency’s
(ESA) Sen2Cor algorithm, the platform processes ESA’s Level-1C top-of-atmosphere reflectance
to atmospherically-corrected bottom-of-atmosphere (BoA) reflectance (Level-2A). The processing
runs on-demand, with a global coverage, on the Earth Observation Data Centre (EODC), which is
a public-private collaborative IT infrastructure in Vienna (Austria) for archiving, processing, and
distributing Earth observation (EO) data (http://www.eodc.eu). Using the data service platform,
users can submit processing requests and access the results via a user-friendly web page or using
a dedicated application programming interface (API). Building on the processed Level-2A data,
the platform also creates value-added products with a particular focus on agricultural vegetation
monitoring, such as leaf area index (LAI) and broadband hemispherical-directional reflectance
factor (HDRF). An analysis of the performance of the data service platform, along with processing
capacity, is presented. Some preliminary consistency checks of the algorithm implementation are
included to demonstrate the expected product quality. In particular, Sentinel-2 data were compared
to atmospherically-corrected Landsat-8 data for six test sites achieving a R
2
= 0.90 and Root Mean
Square Error (RMSE) = 0.031. LAI was validated for one test site using ground estimations. Results
show a very good agreement (R2= 0.83) and a RMSE of 0.32 m2/m2(12% of mean value).
Keywords: Sentinel-2; atmospheric correction; Sen2Cor; LAI; broadband HDRF
1. Introduction
Sentinel-2 is the newest generation Earth observation (EO) satellite of the European Space Agency
(ESA) for land and coastal applications [
1
]. The satellite was launched in June 2015 and is part of
Europe’s Copernicus program aiming at independent and continued global observation capacities [
2
].
Compared to Landsat satellites, Sentinel-2 offers an increased spectral and spatial resolution
with 13 spectral bands of 10 to 60 m spatial resolution. Together with its twin satellite (to be launched
Remote Sens. 2016,8, 938; doi:10.3390/rs8110938 www.mdpi.com/journal/remotesensing
Remote Sens. 2016,8, 938 2 of 16
beginning 2017), Sentinel-2 will cover the entire Earth every five days. The excellent performance
characteristics of Sentinel-2 were shown already for different applications, such as crop and forest
classification [2], sub-pixel landscape feature detection [3], mapping of built-up areas [4,5], as well as
monitoring of glacier and water bodies [6,7].
Currently, users can find data at the ESA’s Scientific Data Hub (SDH), or using alternative
platforms such as Amazon Web Service (AWS) or Google’s Earth Engine. Until now, however, none of
the three mentioned platforms provides products for arbitrary places on Earth beyond the standard
Level-1C level (top-of-atmosphere reflectance). To fill this gap, we developed a data service platform
within the Earth Observation Data Centre (EODC) that allows access to atmospherically-corrected
surface reflectance Sentinel-2 data (Level-2A) on-demand and globally. Level-2A data are required for
numerous applications dealing with multi-temporal analysis (e.g., land surface phenology, land cover
change) [
8
10
] as well as for accurate retrieval of land surface variables, such as leaf area index (LAI),
fraction of absorbed photosynthetically active radiation (FAPAR), or albedo [1113].
In this technical note we focus on products ready for delivery, including surface reflectance,
three-band image composites, LAI, and broadband hemispherical-directional reflectance factors
(HDRF), frequently deployed in our and other agricultural monitoring applications [
14
]. The paper
presents the system implementation and provides examples of products and instructions for access.
We also provide some first product examples and results of consistency checks. The technical note does
not provide a comprehensive validation nor an attempt to improve the integrated software packages
(e.g., related to the atmospheric correction).
In the following sections, we first introduce the algorithms used (Section 2.1) and describe options
to access data (Section 2.2). In Section 3we report preliminary results and data consistency checks.
The technical note concludes in Section 4with an overview of the data service platform performance,
and an outlook to future improvements, product availability and plans for exploitation.
2. Product Description and Access
The data service platform was implemented by the University of Natural Resources and Life
Science, BOKU [
15
] and provides access to individual Sentinel-2 granules (ortho-rectified image tiles
of 100
×
100 km
2
in UTM/WGS84 projection) processed at bottom-of-atmosphere (BoA) reflectance
(Level-2A). The service runs on the Earth Observation Data Centre (EODC), which is a collaborative IT
infrastructure for archiving, processing, and distributing Earth observation (EO) data [16].
Our data service platform processes the Sentinel-2 Level-1C images into Level-2A data using
the ESA’s Sen2Cor algorithm [
17
]. Sen2Cor is supported by the ESA as a third-party plugin for the
Sentinel-2 toolbox and it runs in the ESA Sentinel Application Platform (SNAP) or from the command
line. Additional layers produced by this algorithm are also available such as Aerosol optical thickness
(AOT), water vapor (WV), scene classification (SCL), and various quality indicators (QI).
To minimize atmospheric interference, all value-added products are calculated based on
atmospherically-corrected Level-2A data. For users interested in simple three-band composites,
the platform can create user-defined true and false color composites. All products are delivered in
JPEG 2000 or TIFF format, at three different spatial resolutions (60, 20, and 10 m).
Within the day of the satellite acquisition, Sentinel-2 Level-1C scenes are pulled from the national
mirror site and archived at EODC. As soon as the images are available, processing of Level-2A and
value-added products is performed based on standing orders of on-demand requests. Submission
of data queries and processing requests is made possible via a user-friendly web page or using an
application programming interface (API). The API also allows query on the image metadata and bulk
data access. An overview of the various products currently available on the Sentinel-2 data service
platform is presented in Figure 1and summarized in Table 1.
Remote Sens. 2016,8, 938 3 of 16
Remote Sens. 2016, 8, x FOR PEER REVIEW 3 of 15
Table 1. Products available on the Sentinel-2 data service platform (http://s2.boku.eodc.eu). Bold
crosses (x) indicate the original spatial resolution. Note that band B10 is not produced at Level-2A.
Product Name Center Wavelength (nm) Spatial Resolution (m)
10 20 60
BoA reflectance
B01 443
x
B02 490 xx x
B03 560 xx x
B04 665 xx x
B05 705 xx
B06 740 xx
B07 783 xx
B08 842 xx x
B8a 865
xx
B09 940
x
B10 1375
x
B11 1610
xx
B12 2190
xx
SCL n.a. x x
AOT n.a. x x x
WVP n.a. x x x
VIS n.a. x
LAI n.a. x
Broadband HDRF n.a. x
Figure 1. Examples of a Sentinel-2 100 × 100 km
2
images (tile 33UXP, covering the region between
Vienna and Bratislava, acquired on 6 May 2016) and value-added products available at the data
service platform. Note that clouds extracted from the Level-1C cloud mask are displayed (as hashed
symbol) in all other products. (a) RGB false color composite; (b) Scene classification; (c) Individual
band; (d) Broadband hemispherical-directional reflectance factor (HDRF); (e) Leaf Area Index.
2.1. Sentinel-2 Level2-A Data and Value-Added Products
Atmospherically-corrected bottom-of-atmosphere (BoA) Sentinel-2 data are produced using the
Sen2Cor processor (currently version 2.2.1), developed by ESA to perform atmospheric, terrain, and
Figure 1.
Examples of a Sentinel-2 100
×
100 km
2
images (tile 33UXP, covering the region between
Vienna and Bratislava, acquired on 6 May 2016) and value-added products available at the data service
platform. Note that clouds extracted from the Level-1C cloud mask are displayed (as hashed symbol)
in all other products. (
a
) RGB false color composite; (
b
) Scene classification; (
c
) Individual band;
(d) Broadband hemispherical-directional reflectance factor (HDRF); (e) Leaf Area Index.
Table 1.
Products available on the Sentinel-2 data service platform (http://s2.boku.eodc.eu).
Bold crosses (x) indicate the original spatial resolution. Note that band B10 is not produced at Level-2A.
Product Name Center Wavelength (nm) Spatial Resolution (m)
10 20 60
BoA reflectance
B01 443 x
B02 490 xx x
B03 560 xx x
B04 665 xx x
B05 705 xx
B06 740 xx
B07 783 xx
B08 842 xx x
B8a 865 xx
B09 940 x
B10 1375 x
B11 1610 xx
B12 2190 xx
SCL n.a. x x
AOT n.a. x x x
WVP n.a. x x x
VIS n.a. x
LAI n.a. x
Broadband HDRF n.a. x
2.1. Sentinel-2 Level2-A Data and Value-Added Products
Atmospherically-corrected bottom-of-atmosphere (BoA) Sentinel-2 data are produced using
the Sen2Cor processor (currently version 2.2.1), developed by ESA to perform atmospheric, terrain,
and cirrus correction of top-of-atmosphere Level-1C input data [
17
]. The processor is considered a
prototype and not validated for water and coastal regions. The correction is based on the application of
look-up-tables (LUTs), which were pre-calculated using the libRadtran radiative transfer routines.
The LUTs include two different types of aerosols (rural and maritime), two different types
Remote Sens. 2016,8, 938 4 of 16
of atmospheres (mid-latitude summer and winter), six different types of ozone concentrations,
and four different amounts of water vapor column [
18
]. As an example, Figure 1c shows the
atmospherically-corrected B8a BoA reflectance (at 865 nm). Sen2Cor includes a scene classification
module (example in Figure 1b) to map no data or defective pixels (pixel value = 0–1), four different
cloud clover class probabilities (7–10), and six different classes including shadows (2), cloud shadows
(3), vegetation (4), soils and deserts (5), water (6), and snow (11).
Other Sen2Cor outputs comprise (i) an estimation of the aerosol optical thickness (AOT) using
the dense dark vegetation (DDV) algorithm [
19
] and the (ii) retrieval of water vapor (WV) using the
pre-corrected differential absorption algorithm (APDA, [
20
]) analyzing Sentinel-2 bands B8a and B9.
Example outputs for AOT and WV are shown in Figure 2.
Remote Sens. 2016, 8, x FOR PEER REVIEW 4 of 15
cirrus correction of top-of-atmosphere Level-1C input data [17]. The processor is considered a
prototype and not validated for water and coastal regions. The correction is based on the application
of look-up-tables (LUTs), which were pre-calculated using the libRadtran radiative transfer routines.
The LUTs include two different types of aerosols (rural and maritime), two different types of
atmospheres (mid-latitude summer and winter), six different types of ozone concentrations, and
four different amounts of water vapor column [18]. As an example, Figure 1c shows the
atmospherically-corrected B8a BoA reflectance (at 865 nm). Sen2Cor includes a scene classification
module (example in Figure 1b) to map no data or defective pixels (pixel value = 0–1), four different
cloud clover class probabilities (7–10), and six different classes including shadows (2), cloud
shadows (3), vegetation (4), soils and deserts (5), water (6), and snow (11).
Other Sen2Cor outputs comprise (i) an estimation of the aerosol optical thickness (AOT) using
the dense dark vegetation (DDV) algorithm [19] and the (ii) retrieval of water vapor (WV) using the
pre-corrected differential absorption algorithm (APDA, [20]) analyzing Sentinel-2 bands B8a and B9.
Example outputs for AOT and WV are shown in Figure 2.
Figure 2. Examples of 60 m spatial resolution bands B1 (443 nm), B9 (940 nm), and B10 (1375 nm)
dedicated to atmospheric correction and cirrus cloud detection (Sentinel-2 image subset acquired on
22 June 2016) and retrieved water vapor (WV) and aerosol optical thickness (AOT). On the bottom
row, examples of true and false color RGB composites produced for the same acquisition are shown.
The legend for the scene classification is given in Figure 1.
The image file output and directory structures of all layers are similar to the Level-1C product
structure with lossless compressed images based on the JPEG 2000 format and produced at three
different resolutions, 60, 20, and 10 m.
Figure 2.
Examples of 60 m spatial resolution bands B1 (443 nm), B9 (940 nm), and B10 (1375 nm)
dedicated to atmospheric correction and cirrus cloud detection (Sentinel-2 image subset acquired
on 22 June 2016) and retrieved water vapor (WV) and aerosol optical thickness (AOT). On the bottom
row, examples of true and false color RGB composites produced for the same acquisition are shown.
The legend for the scene classification is given in Figure 1.
The image file output and directory structures of all layers are similar to the Level-1C product
structure with lossless compressed images based on the JPEG 2000 format and produced at three
different resolutions, 60, 20, and 10 m.
True and false color composites are produced on-demand by combining three different Sentinel-2
spectral bands based on user requests. Contrast stretching is offered to optimize the RGB display
(Figure 2bottom).
Remote Sens. 2016,8, 938 5 of 16
For mapping leaf area index (LAI), we employ a neural network (NNT) algorithm developed
by INRA [
21
]. The NNT algorithm was tailored for Sentinel-2 and trained using radiative transfer
simulations from PROSPECT and SAIL radiative transfer models [
22
,
23
]. The inputs (BoA surface
reflectance) and outputs (LAI) data are normalized and de-normalized, respectively, using given
coefficients. An example image is shown in Figure 1e.
The broadband (490–2160 nm) hemispherical-directional reflectance factor (HDRF) (Figure 1d),
obtained at the time of the satellite overpass was calculated as a weighted sum of the Level-2A
Sentinel-2 surface reflectance, with broadband weights representing the corresponding fraction of the
solar irradiance in each sensor band [
24
]. The broadband weights were adapted to take into account
the spectral configuration of Sentinel-2.
2.2. Data Discovery and Download via Web Interface
The web interface allows users to search for products (defined as the collection of elementary
granules within a single orbit), granules (the 100
×
100 km
2
image tiles) and images (image files of
an individual granule, such as spectral bands or value-added products). The search can be filtered
considering a range of acquisition dates, maximum cloud cover, and a set of coordinates (center point
or using a GeoJSON string to define a point, bounding box or polygon). The metadata catalogue is
regularly updated as new products are pulled from the national mirror sites and archived at EODC.
From the user perspective, major performance improvements in data search are related to the use
of (i) the actual geometry of valid data in the Sentinel-2 granules and (ii) an area-based cloud cover
statistic (using the Level 1C cloud mask) instead of the “bulk” cloud cover percentage provided with
the image metadata.
As an example, Figure 3shows the data service web page with the results of a query identifying
all atmospherically-corrected data available on the platform. Users can submit orders for any region of
interest for one-time processing or continuous tasking. In the latter case, the creation of the Level-2A
products is performed as soon as the Level-1C data are available on the server, generally within one
day from the satellite acquisition.
Remote Sens. 2016, 8, x FOR PEER REVIEW 5 of 15
True and false color composites are produced on-demand by combining three different
Sentinel-2 spectral bands based on user requests. Contrast stretching is offered to optimize the RGB
display (Figure 2 bottom).
For mapping leaf area index (LAI), we employ a neural network (NNT) algorithm developed by
INRA [21]. The NNT algorithm was tailored for Sentinel-2 and trained using radiative transfer
simulations from PROSPECT and SAIL radiative transfer models [22,23]. The inputs (BoA surface
reflectance) and outputs (LAI) data are normalized and de-normalized, respectively, using given
coefficients. An example image is shown in Figure 1e.
The broadband (490–2160 nm) hemispherical-directional reflectance factor (HDRF) (Figure 1d),
obtained at the time of the satellite overpass was calculated as a weighted sum of the Level-2A
Sentinel-2 surface reflectance, with broadband weights representing the corresponding fraction of
the solar irradiance in each sensor band [24]. The broadband weights were adapted to take into
account the spectral configuration of Sentinel-2.
2.2. Data Discovery and Download via Web Interface
The web interface allows users to search for products (defined as the collection of elementary
granules within a single orbit), granules (the 100 × 100 km2 image tiles) and images (image files of an
individual granule, such as spectral bands or value-added products). The search can be filtered
considering a range of acquisition dates, maximum cloud cover, and a set of coordinates (center
point or using a GeoJSON string to define a point, bounding box or polygon). The metadata
catalogue is regularly updated as new products are pulled from the national mirror sites and
archived at EODC. From the user perspective, major performance improvements in data search are
related to the use of (i) the actual geometry of valid data in the Sentinel-2 granules and (ii) an
area-based cloud cover statistic (using the Level 1C cloud mask) instead of the “bulk” cloud cover
percentage provided with the image metadata.
As an example, Figure 3 shows the data service web page with the results of a query identifying
all atmospherically-corrected data available on the platform. Users can submit orders for any region
of interest for one-time processing or continuous tasking. In the latter case, the creation of the
Level-2A products is performed as soon as the Level-1C data are available on the server, generally
within one day from the satellite acquisition.
Figure 3. Web interface to explore the data archive, access Sentinel-2 data and value-added products.
The map shows the footprint of the atmospherically-corrected granules processed over part of
Europe in July 2016.
Figure 3.
Web interface to explore the data archive, access Sentinel-2 data and value-added products.
The map shows the footprint of the atmospherically-corrected granules processed over part of Europe
in July 2016.
Remote Sens. 2016,8, 938 6 of 16
2.3. Data Exploration and Download via the Application Programming Interface (API)
The API offers a set of predefined connection points where HTTP requests can be submitted to
access metadata, granules, and individual image files or to activate processing services on the server
side. The user must provide the information needed for each request, specifying (i) the product level,
(ii) start and end of acquisition period; and (iii) coordinates of the region of interest. The request can
thereafter be submitted using an internet browser or using common programming languages such as
Python (Python Software Foundation), R (R Development Core Team, Vienna, Austria), or MATLAB
(The MathWorks, Inc., Natick, MA, USA). The set of parameters and HTTP requests are described in
the API documentation available on the web page. Some examples to query the catalogue of metadata,
to download images and to generate RGB false color composites are provided using R programming
environment in the Appendix A.
3. Example Products and Preliminary Validation
3.1. Data Processing and Performance
The data service platform runs on a computer cluster at EODC [
16
] that consists of 100 cores,
2.25 TB of RAM and 122 TB of disk storage, and it is directly connected to the Sentinel-2 data archive
(with 2 PB distributed hard disks and 1 PB tape storage). We currently deploy 4 cores with
12 GB
of RAM to serve the APIs and the web page, and 12 cores and 52 GB of RAM are assigned to
compute Level-2A data and value-added products. This hardware configuration is able to process
in near-real-time up to about 2,000,000 km
2
(equivalent to 200 Sentinel-2 granules) per day. In the
near future, it is planned to add additional resources (304 cores, 3.8 TB RAM and 1.1 PB disk storage).
The processing capacity should scale almost linearly with the amount of assigned system resources
(cores and RAM).
In the development phase, the data service platform has performed the atmospheric correction of
about 4000 Sentinel-2 granules and 2700 value-added products (i.e., LAI) for a group of core end-users.
On average, the data service platform required 37 min (to a max of 1 h) for the atmospheric correction
of one granule of 100
×
100 km
2
and 15 min (to a max of 24 min) for the production of value-added
products, such as LAI.
In the last month of operations (June 2016), we experienced a delivery time of final results from
three days (low priority) to one day (high priority) including the time span between image acquisition
and Level-1C product availability on the EODC archive.
3.2. Surface Reflectance
The bottom-of-atmosphere (surface) reflectance is a basic input to many EO applications ranging
from land surface phenology to land cover classification and change detection. To provide a preliminary
evaluation of the Sentinel-2 algorithm implementation, we conducted a pixel-based comparison
between Sentinel-2 and Landsat-8 surface reflectance data. The data were acquired on the same day
for different sites in Europe and the comparison was limited to the spectral bands in common to the
two satellites (bands 2, 3, 4, 5, 6, and 7 for Landsat-8, and bands 2, 3, 4, 8a, 11, and 12 for Sentinel-2).
With respect to Landsat-8, we used the atmospherically-corrected Landsat Surface Reflectance Climate
Data Record (CDR). Although considered provisional, the dataset reported a very good agreement
with other satellite data (e.g., MODIS) and AERONET measurements [
25
]. The Landsat CDR data also
compared favorably against manually fine-tuned atmospheric corrections [26].
Six test sites in Europe were chosen for the comparison between Sentinel-2 and Landsat-8 surface
reflectance (located in Greece, Turkey, Austria, Germany, Czech Republic, and France). Within each test
site, a number of randomly-selected points were chosen. Observations affected by clouds and cloud
shadows were identified using the Sentinel-2 scene classification (SCL) and the Landsat-8 fmask [
27
].
Observations affected by cloud and cloud-shadow were excluded from the analysis. To take into
account the differences in pixel size between Sentinel-2 and Landsat-8, we calculated the average
Remote Sens. 2016,8, 938 7 of 16
reflectance value in a buffer of 30 m for Sentinel-2 data. This buffered reflectance was compared to a
single pixel reflectance for Landsat-8 over a set of 4400 Landsat-8 pixels.
Results show an overall R
2
= 0.90 and RMSE = 0.031 when using all the six homologue bands,
with RMSE ranging from 0.023 for the green band to 0.043 for the near-infrared band. A detailed
overview of the comparison is provided for two contrasting (summer/winter) acquisition dates in
Figure 4. A spatial subset of the two acquisitions for the red and near-infrared bands is shown in the
scatterplots in Figure 4. Details regarding the image data are provided in Table 2.
Figure 4.
Scatterplots between BoA Landsat-8 and Sentinel-2 reflectance in red and near-infrared for
two test sites (top row: Greece, 8 August 2015 tile 34SEH; bottom row: Austria, 31 December 2015,
tile 33UUP). The reflectance from Sentinel-2 at 10 m spatial resolution was averaged within a buffer
of 30 m to match the pixel size of Landsat-8. BoA reflectance values are scaled by 10,000.
For Greece, considering the six homologue bands, results show a R
2
of 0.96 and RMSE
of 0.03 reflectance units. For Austria, results show a R2of 0.91 and RMSE of 0.027.
Figure 5shows some exemplary spectral profiles from atmospherically-corrected Landsat-8 and
Sentinel-2 pixels acquired on the same date. The spectral profiles represent different land cover
types including vegetation, soil, and water. They show a very good match over the six homologue
spectral bands.
Remote Sens. 2016,8, 938 8 of 16
Table 2. Contemporaneous Landsat-8 and Sentinel-2 acquisitions used for comparison.
Landsat-8 Sentinel-2
Greece
Tile 185/033 34SEH
Acquisition date 8 August 2015
Acquisition time 9.16 AM 9.25 AM
Sun Zenith Angle 3027
Austria
Tile 192/026 33UUP
Acquisition date 31 December 2015
Acquisition time 9.57 AM 10.22 AM
Sun Zenith Angle 7472
Remote Sens. 2016, 8, x FOR PEER REVIEW 8 of 15
For Greece, considering the six homologue bands, results show a R
2
of 0.96 and RMSE of 0.03
reflectance units. For Austria, results show a R
2
of 0.91 and RMSE of 0.027.
Figure 5 shows some exemplary spectral profiles from atmospherically-corrected Landsat-8 and
Sentinel-2 pixels acquired on the same date. The spectral profiles represent different land cover
types including vegetation, soil, and water. They show a very good match over the six homologue
spectral bands.
Figure 5. Examples of comparison between BoA Landsat-8 and Sentinel-2 reflectance for
randomly-selected points (including vegetation, soil and water). The images were acquired on the
same day. The reflectance from Sentinel-2 at 10 m spatial resolution was averaged within a buffer of
30 m to match the pixel size of Landsat-8. BoA reflectance values are scaled by 10,000.
Amongst the various randomly selected pixels, occasionally, we observed larger differences
between the atmospherically-corrected reflectances of the two satellites (Figure 6 left). Checking
those observations revealed that our comparison included pixels affected by undetected clouds
(scene classification or fmask). Some larger divergences were also observed over heterogonous pixel
locations and over complex surface terrains (Figure 7 left). The differences are probably a direct
result of the different spatial resolutions of the two satellite sensors.
Figure 5.
Examples of comparison between BoA Landsat-8 and Sentinel-2 reflectance for
randomly-selected points (including vegetation, soil and water). The images were acquired on the
same day. The reflectance from Sentinel-2 at 10 m spatial resolution was averaged within a buffer
of 30 m to match the pixel size of Landsat-8. BoA reflectance values are scaled by 10,000.
Amongst the various randomly selected pixels, occasionally, we observed larger differences
between the atmospherically-corrected reflectances of the two satellites (Figure 6left). Checking
those observations revealed that our comparison included pixels affected by undetected clouds (scene
classification or fmask). Some larger divergences were also observed over heterogonous pixel locations
and over complex surface terrains (Figure 7left). The differences are probably a direct result of the
different spatial resolutions of the two satellite sensors.
Remote Sens. 2016,8, 938 9 of 16
Remote Sens. 2016, 8, x FOR PEER REVIEW 9 of 15
Figure 6. Examples of spectral mismatch between Sentinel-2 and Landsat-8 BoA reflectances due to
undetected clouds or problems of atmospheric correction in regions adjacent to clouds. BoA
reflectance values are scaled by 10,000.
Figure 7. Spectral mismatch between Landsat-8 and Sentinel-2 observed over heterogeneous and
complex surface terrain. The maps on the right show a zoom in on point “210”. BoA reflectance
values are scaled by 10,000.
A field campaign was organized on 24 June 2016 in an agricultural area in Austria [28] to
measure surface reflectance with a field spectro-radiometer in coincidence (±2 days) of two
Sentinel-2 overpasses (tile 33UXP on 22 and 25 June, respectively). A second campaign was
organized on 31 August in the same area and a Sentinel-2 image was acquired on the same day. The
spectral reflectance was measured at ground over homogeneous targets using a Spectral Evolution
PSR-2500 radiometer operating in the range 350–2500 nm with a spectral resolution of 3.5 nm (in
visible, VIS, and near-infrared, NIR) and 22 nm (in the short-wave infrared, SWIR) [29]. The
Figure 6.
Examples of spectral mismatch between Sentinel-2 and Landsat-8 BoA reflectances due to
undetected clouds or problems of atmospheric correction in regions adjacent to clouds. BoA reflectance
values are scaled by 10,000.
Remote Sens. 2016, 8, x FOR PEER REVIEW 9 of 15
Figure 6. Examples of spectral mismatch between Sentinel-2 and Landsat-8 BoA reflectances due to
undetected clouds or problems of atmospheric correction in regions adjacent to clouds. BoA
reflectance values are scaled by 10,000.
Figure 7. Spectral mismatch between Landsat-8 and Sentinel-2 observed over heterogeneous and
complex surface terrain. The maps on the right show a zoom in on point “210”. BoA reflectance
values are scaled by 10,000.
A field campaign was organized on 24 June 2016 in an agricultural area in Austria [28] to
measure surface reflectance with a field spectro-radiometer in coincidence (±2 days) of two
Sentinel-2 overpasses (tile 33UXP on 22 and 25 June, respectively). A second campaign was
organized on 31 August in the same area and a Sentinel-2 image was acquired on the same day. The
spectral reflectance was measured at ground over homogeneous targets using a Spectral Evolution
PSR-2500 radiometer operating in the range 350–2500 nm with a spectral resolution of 3.5 nm (in
visible, VIS, and near-infrared, NIR) and 22 nm (in the short-wave infrared, SWIR) [29]. The
Figure 7.
Spectral mismatch between Landsat-8 and Sentinel-2 observed over heterogeneous and
complex surface terrain. The maps on the right show a zoom in on point “210”. BoA reflectance values
are scaled by 10,000.
A field campaign was organized on 24 June 2016 in an agricultural area in Austria [
28
]
to measure surface reflectance with a field spectro-radiometer in coincidence (
±
2 days) of
two Sentinel-2 overpasses (tile 33UXP on 22 and 25 June, respectively). A second campaign was
organized on 31 August in the same area and a Sentinel-2 image was acquired on the same day.
The spectral reflectance was measured at ground over homogeneous targets using a Spectral Evolution
PSR-2500 radiometer operating in the range 350–2500 nm with a spectral resolution of 3.5 nm (in visible,
VIS, and near-infrared, NIR) and 22 nm (in the short-wave infrared, SWIR) [
29
]. The manufacturer
reports a calibration accuracy of 5% (400 nm), 4% (700 nm), and 7% for (2200 nm). Spectral data were
Remote Sens. 2016,8, 938 10 of 16
collected between 10:30 and 12:30 (satellite overpass at 11:30) by averaging 10 scans for each target
over an area of approximately 10
×
10 m. The points were geo-located using GPS measurements.
A calibrated white reference panel (Spectralon) was used to measure solar irradiance at regular intervals
during the measurement period. The instrument was deployed with a 14
lens at a distance of 1 m
from the top of the surface (resulting in a field of view of 25 cm). In post-processing, the spectra were
smoothed using the Whittaker smoother [30] with a Lambda of 500.
Figures 8and 9shows some examples of the comparison for different targets for the two
campaigns. For most targets a very good agreement between Sentinel-2 signatures and the reflectance
measured with the PSR-2500 can be found. Figure 8also highlight the changes of the surface reflectance
within the two acquisition dates (22 and 25 June, respectively).
Remote Sens. 2016, 8, x FOR PEER REVIEW 10 of 15
manufacturer reports a calibration accuracy of 5% (400 nm), 4% (700 nm), and 7% for (2200 nm).
Spectral data were collected between 10:30 and 12:30 (satellite overpass at 11:30) by averaging 10
scans for each target over an area of approximately 10 × 10 m. The points were geo-located using
GPS measurements. A calibrated white reference panel (Spectralon) was used to measure solar
irradiance at regular intervals during the measurement period. The instrument was deployed with a
14° lens at a distance of 1 m from the top of the surface (resulting in a field of view of 25 cm). In
post-processing, the spectra were smoothed using the Whittaker smoother [30] with a Lambda of 500.
Figures 8 and 9 shows some examples of the comparison for different targets for the two
campaigns. For most targets a very good agreement between Sentinel-2 signatures and the
reflectance measured with the PSR-2500 can be found. Figure 8 also highlight the changes of the
surface reflectance within the two acquisition dates (22 and 25 June, respectively).
Figure 8. Comparison between field spectral measurements on 24 June and Sentinel-2 data acquired
on 22 June and 25 June 2016 over an agricultural region East of Vienna, Austria (Sentinel-2 tile
33UXP) from upper left to lower right: bare soil, winter wheat 1, potato, harvested field, winter
wheat 2, and soya. Reflectance values are scaled by 10,000.
Figure 9. Comparison between field spectral measurements on 31 August and Sentinel-2 data
acquired on the same day (Sentinel-2 tile 33UXP) from upper left to lower right: asphalt, bare soil,
crop residue, two soya fields at different phenological stages, and meadow. Reflectance values are
scaled by 10,000.
Figure 8.
Comparison between field spectral measurements on 24 June and Sentinel-2 data acquired
on 22 June and 25 June 2016 over an agricultural region East of Vienna, Austria (Sentinel-2 tile 33UXP)
from upper left to lower right: bare soil, winter wheat 1, potato, harvested field, winter wheat 2,
and soya. Reflectance values are scaled by 10,000.
Remote Sens. 2016, 8, x FOR PEER REVIEW 10 of 15
manufacturer reports a calibration accuracy of 5% (400 nm), 4% (700 nm), and 7% for (2200 nm).
Spectral data were collected between 10:30 and 12:30 (satellite overpass at 11:30) by averaging 10
scans for each target over an area of approximately 10 × 10 m. The points were geo-located using
GPS measurements. A calibrated white reference panel (Spectralon) was used to measure solar
irradiance at regular intervals during the measurement period. The instrument was deployed with a
14° lens at a distance of 1 m from the top of the surface (resulting in a field of view of 25 cm). In
post-processing, the spectra were smoothed using the Whittaker smoother [30] with a Lambda of 500.
Figures 8 and 9 shows some examples of the comparison for different targets for the two
campaigns. For most targets a very good agreement between Sentinel-2 signatures and the
reflectance measured with the PSR-2500 can be found. Figure 8 also highlight the changes of the
surface reflectance within the two acquisition dates (22 and 25 June, respectively).
Figure 8. Comparison between field spectral measurements on 24 June and Sentinel-2 data acquired
on 22 June and 25 June 2016 over an agricultural region East of Vienna, Austria (Sentinel-2 tile
33UXP) from upper left to lower right: bare soil, winter wheat 1, potato, harvested field, winter
wheat 2, and soya. Reflectance values are scaled by 10,000.
Figure 9. Comparison between field spectral measurements on 31 August and Sentinel-2 data
acquired on the same day (Sentinel-2 tile 33UXP) from upper left to lower right: asphalt, bare soil,
crop residue, two soya fields at different phenological stages, and meadow. Reflectance values are
scaled by 10,000.
Figure 9.
Comparison between field spectral measurements on 31 August and Sentinel-2 data acquired
on the same day (Sentinel-2 tile 33UXP) from upper left to lower right: asphalt, bare soil, crop residue,
two soya fields at different phenological stages, and meadow. Reflectance values are scaled by 10,000.
3.3. Leaf Area Index
For vegetation characterization and applications such as precision farming and irrigation
management, the crop’s leaf area index (LAI) is an important structural variable with direct links to
Remote Sens. 2016,8, 938 11 of 16
crop growth, water and energy balance. We assessed the quality of the Sentinel-2 derived LAI through
preliminary comparison with non-destructive (optical) field reference measurements.
LAI reference measurements were acquired with the Licor LAI-2200 Plant Canopy Analyzer [
31
]
from April to June 2016 over five different crops (Sugarbeet, Maize, Onion, Potatoes and Winter wheat)
for a total of 95 measurements. The field measurements were used to validate LAI retrievals from six
different Sentinel-2 acquisitions (tile 33UXP) over the study region of Marchfeld [
14
]. The maximum
time span between satellite acquisitions and ground measurements was 6 days (7 April). Results in
Figure 10 show a very good agreement (R2= 0.83) and a RMSE of 0.32 m2/m2(12% of mean value).
Remote Sens. 2016, 8, x FOR PEER REVIEW 11 of 15
3.3. Leaf Area Index
For vegetation characterization and applications such as precision farming and irrigation
management, the crop’s leaf area index (LAI) is an important structural variable with direct links to
crop growth, water and energy balance. We assessed the quality of the Sentinel-2 derived LAI
through preliminary comparison with non-destructive (optical) field reference measurements.
LAI reference measurements were acquired with the Licor LAI-2200 Plant Canopy Analyzer
[31] from April to June 2016 over five different crops (Sugarbeet, Maize, Onion, Potatoes and Winter
wheat) for a total of 95 measurements. The field measurements were used to validate LAI retrievals
from six different Sentinel-2 acquisitions (tile 33UXP) over the study region of Marchfeld [14]. The
maximum time span between satellite acquisitions and ground measurements was 6 days (7 April).
Results in Figure 10 show a very good agreement (R
2
= 0.83) and a RMSE of 0.32 m
2
/m
2
(12% of mean
value).
Figure 10. Scatterplot between ground and satellite-based LAI estimation for Marchfeld (n = 95). The
data were acquired from April to June 2016 and comprise six different Sentinel-2 scenes.
3.4. The Broadband Hemispherical-Directional Reflectance Factor (HDRF)
The broadband HDRF was compared to ground measurements of albedo obtained with a
Campbell CNR-1 net radiometer installed at 2 m height from top of canopy. Table 3 shows the
comparison for four different satellite acquisitions over two crop types. In general, we observe a
good agreement over a broad range of LAI values.
The Sentinel-2 broadband HDRF (obtained on 7 and 30 August 2015) was also compared with
two existing maps of the HDRF obtained from DEIMOS-1 [32] (acquired on 6 and 26 August 2015)
using the ATCOR-2 value-added product module [33]. Results show (Figure 11) a good agreement
between the two datasets, with a lower agreement for 30 August, probably due to the longer time
span (5 days) between the two acquisitions.
Table 3. Measured albedo (net radiometer) versus broadband HDRF from Sentinel-2 (weighted sum).
Crop Type Acquisition Date Leaf Area Index
(Sentinel-2)
Broadband HDRF
(Sentinel-2)
Measured Albedo
(CNR-1) at Noon
Soya 7 August 2015 3.2 0.24 0.22
Soya 30 August 2015 >6.0 0.28 0.26
Maize 25 June 2016 1.9 0.16 0.16
Maize 2 July 2016 2.9 0.17 0.17
Figure 10.
Scatterplot between ground and satellite-based LAI estimation for Marchfeld (n= 95).
The data were acquired from April to June 2016 and comprise six different Sentinel-2 scenes.
3.4. The Broadband Hemispherical-Directional Reflectance Factor (HDRF)
The broadband HDRF was compared to ground measurements of albedo obtained with a
Campbell CNR-1 net radiometer installed at 2 m height from top of canopy. Table 3shows the
comparison for four different satellite acquisitions over two crop types. In general, we observe a good
agreement over a broad range of LAI values.
Table 3.
Measured albedo (net radiometer) versus broadband HDRF from Sentinel-2 (weighted sum).
Crop Type Acquisition Date Leaf Area Index
(Sentinel-2)
Broadband HDRF
(Sentinel-2)
Measured Albedo
(CNR-1) at Noon
Soya 7 August 2015 3.2 0.24 0.22
Soya 30 August 2015 >6.0 0.28 0.26
Maize 25 June 2016 1.9 0.16 0.16
Maize 2 July 2016 2.9 0.17 0.17
The Sentinel-2 broadband HDRF (obtained on 7 and 30 August 2015) was also compared with
two existing maps of the HDRF obtained from DEIMOS-1 [
32
] (acquired on 6 and 26 August 2015)
using the ATCOR-2 value-added product module [
33
]. Results show (Figure 11) a good agreement
between the two datasets, with a lower agreement for 30 August, probably due to the longer time span
(5 days) between the two acquisitions.
As final check, we looked at time profiles of broadband HDRF, LAI and the scene classification
classes (Figure 12). The selected pixel is from cropland located in the Barrax region, Spain, from
November 2015 to end of June 2016. We observe a temporal evolution of LAI that is consistent with
the crop growing pattern for winter crops. Peaks in broadband HDRF values are consistent with the
scene classification.
Remote Sens. 2016,8, 938 12 of 16
Remote Sens. 2016, 8, x FOR PEER REVIEW 12 of 15
Figure 11. Scatterplots between Sentinel-2 and Deimos-1 broadband HDRF for two different dates.
As final check, we looked at time profiles of broadband HDRF, LAI and the scene classification
classes (Figure 12). The selected pixel is from cropland located in the Barrax region, Spain, from
November 2015 to end of June 2016. We observe a temporal evolution of LAI that is consistent with
the crop growing pattern for winter crops. Peaks in broadband HDRF values are consistent with the
scene classification.
Figure 12. Time series of LAI and broadband HDRF for an exemplary cropland pixel in Barrax, Spain
(588,916; 4,322,882, UTM/WGS84). The numbers on top of the chart indicate the scene classification
code (Bare soil = 5; Cloud cover with low to high probability: 7–9). All other points are classified as
vegetation (4, in green).
Figure 11. Scatterplots between Sentinel-2 and Deimos-1 broadband HDRF for two different dates.
Remote Sens. 2016, 8, x FOR PEER REVIEW 12 of 15
Figure 11. Scatterplots between Sentinel-2 and Deimos-1 broadband HDRF for two different dates.
As final check, we looked at time profiles of broadband HDRF, LAI and the scene classification
classes (Figure 12). The selected pixel is from cropland located in the Barrax region, Spain, from
November 2015 to end of June 2016. We observe a temporal evolution of LAI that is consistent with
the crop growing pattern for winter crops. Peaks in broadband HDRF values are consistent with the
scene classification.
Figure 12. Time series of LAI and broadband HDRF for an exemplary cropland pixel in Barrax, Spain
(588,916; 4,322,882, UTM/WGS84). The numbers on top of the chart indicate the scene classification
code (Bare soil = 5; Cloud cover with low to high probability: 7–9). All other points are classified as
vegetation (4, in green).
Figure 12.
Time series of LAI and broadband HDRF for an exemplary cropland pixel in Barrax, Spain
(588,916; 4,322,882, UTM/WGS84). The numbers on top of the chart indicate the scene classification
code (Bare soil = 5; Cloud cover with low to high probability: 7–9). All other points are classified as
vegetation (4, in green).
Remote Sens. 2016,8, 938 13 of 16
4. Conclusions and Outlook
This technical note presented the first data service platform for the provision of atmospherically
corrected Level-2A Sentinel-2 data and value-added products with a focus on vegetation monitoring.
A user-friendly web interface is available to submit processing requests and access individual
products. A dedicated application programming interface (API) supports bulk data processing and
it is accessible using common programming languages, such as R or Python. End-users can find the
full documentation and service features on the service web page and a dedicated R package at the
software repository https://github.com/IVFL-BOKU/sentinel2.
During the algorithm implementation phase, a number of consistency checks were performed
using existing satellite data, observations from Landsat-8 and dedicated ground measurements of LAI,
spectral reflectance, and broadband HDRF. The results obtained from this preliminary performance
analysis are very encouraging. They confirmed the spectral consistency with Landsat-8 and with
ground reflectance measurements and showed, for the first time, the high potential and quality of
Sentinel-2 data for the retrieval of LAI.
The data service platform operates on-demand and is ready to accept processing requests
from end-users. They can choose any region of interests, temporal window, and product(s) of
interest. After a fast registration process, end-users receive test account to generate and retrieve
atmospherically-corrected Sentinel-2 scenes and value-added products (one voucher for each
100
×
100 km
2
granule). For larger data volumes, end-users will need to pay a contribution to
cover costs for data archiving, processing, and maintenance of the algorithms.
In the future, we plan to extend the portfolio of value-added products, to offer the possibility to
plug-in different algorithms, and to integrate Sentinel-2 and Landsat-8 data processing, especially for
the production of smoothed and gap-filled time series ([
34
] under review). As a minimum service,
we plan to regularly process and store all Sentinel-2 scenes for Europe so that users can directly access
the new images as they become available.
Acknowledgments:
This work was partly financed by the project HQ-S2, FFG application ID 6277184.
The ground campaign and the integration of algorithms for value-added products has been performed within the
context of the H2020 FATIMA project (grant agreement No 633945). We also acknowledge the support from the
ERASMUS+ programme for Claudia Pipitone and Luca Zappa, visitors at BOKU from the University of Palermo
and from the University of Milano, respectively.
Author Contributions:
Francesco Vuolo conceived and designed the analysis and wrote the paper;
Mateusz ˙
Zółtak contributed to the data preparation and analysis; Claudia Pipitone performed the comparison
between Landsat-8 and Sentinel-2 surface reflectance; Hannah Wenng and Luca Zappa acquired and analyzed
the LAI data; Markus Immitzer performed the spectral measurements and data analysis; Marie Weiss and
Frederic Baret provided the LAI algorithm and contributed to the paper writing; Clement Atzberger supervised
the work and contributed to the paper writing.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A
Some examples to query the catalogue of metadata, to download images, and to generate RGB
false color composites are provided using R programming environment. For example, the query for
the Granules in the UTM zone 33U acquired for the year 2016 and having a cloud cover lower than
40% is:
library(jsonlite)
Url = ‘https://s2.boku.eodc.eu/granule?dateMin=2016-01-01&utm=33U&cloudCovMax=40’
granules = fromJSON(Url)
The query can be restricted to search and download atmospherically corrected images acquired
on 25 June for the granule in the UTM zone “33UXP”. The response also includes a list of
value-added products.
Remote Sens. 2016,8, 938 14 of 16
library(jsonlite)
login = URLencode(‘putYourLoginHere’, TRUE)
pswd = URLencode(‘putYourPasswordHere’, TRUE)
Url = sprint
(‘https://%s:%s@s2.boku.eodc.eu/image?dateMin=2016-06-25&dateMax=2016-06-25&utm=33UXP&atmCorr=1
’,
login, pswd)
images = fromJSON(Url)
for(i in seq_along(images$imageId)){
localFilename = paste0(images$utm[i], ‘_’, images$band[i], ‘_’, images$resolution[i], ‘.’, images$format[i])
download.file(images$url[i], localFileName, mode = ‘wb’, quiet = TRUE)}
HTTP requests can also be used to activate pre-defined processing services on the server side.
For instance, RGB true or false color composites can be created on-demand as follows:
library(curl)
library(jsonlite)
options(timeout = 600)
login = URLencode(‘putYourLoginHere’, TRUE)
pswd = URLencode(‘putYourPasswordHere’, TRUE)
Url =
sprintf(‘https://%s:%s@s2.boku.eodc.eu/image?dateMin=2016-06-22&dateMax=2016-06-22&utm=
33UXP&atmCorr=1&band=’, login, pswd)
rId = fromJSON(paste0(Url, ‘B04’))$imageId[1]
gId = fromJSON(paste0(Url, ‘B03’))$imageId[1]
bId = fromJSON(paste0(Url, ‘B02’))$imageId[1]
rgbUrl = sprintf(‘http://%s:%s@s2.boku.eodc.eu/RGB?r=%d&g=%d&b=%d’, rId, gId, bId)
curl_download (rgbUrl, ‘33UXP_2016-06-22_true_rgb.tiff’)
References
1.
Drusch, M.; del Bello, U.; Carlier, S.; Colin, O.; Fernandez, V.; Gascon, F.; Hoersch, B.; Isola, C.; Laberinti, P.;
Martimort, P.; et al. Sentinel-2: ESA’s Optical High-Resolution Mission for GMES Operational Services.
Remote Sens. Environ. 2012,120, 25–36. [CrossRef]
2.
Immitzer, M.; Vuolo, F.; Atzberger, C. First Experience with Sentinel-2 Data for Crop and Tree Species
Classifications in Central Europe. Remote Sens. 2016,8, 166. [CrossRef]
3.
Radoux, J.; Chomé, G.; Jacques, D.; Waldner, F.; Bellemans, N.; Matton, N.; Lamarche, C.; d’Andrimont, R.;
Defourny, P. Sentinel-2’s Potential for Sub-Pixel Landscape Feature Detection. Remote Sens. 2016,8, 488. [CrossRef]
4.
Pesaresi, M.; Corbane, C.; Julea, A.; Florczyk, A.; Syrris, V.; Soille, P. Assessment of the Added-Value of
Sentinel-2 for Detecting Built-up Areas. Remote Sens. 2016,8, 299. [CrossRef]
5.
Lefebvre, A.; Sannier, C.; Corpetti, T. Monitoring Urban Areas with Sentinel-2A Data: Application to
the Update of the Copernicus High Resolution Layer Imperviousness Degree. Remote Sens. 2016,8, 606. [CrossRef]
6.
Paul, F.; Winsvold, S.; Kääb, A.; Nagler, T.; Schwaizer, G. Glacier Remote Sensing Using Sentinel-2.
Part II: Mapping Glacier Extents and Surface Facies, and Comparison to Landsat 8. Remote Sens.
2016
,
8, 575. [CrossRef]
7.
Du, Y.; Zhang, Y.; Ling, F.; Wang, Q.; Li, W.; Li, X. Water Bodies’ Mapping from Sentinel-2 Imagery with
Modified Normalized Difference Water Index at 10-m Spatial Resolution Produced by Sharpening the SWIR
Band. Remote Sens. 2016,8, 354. [CrossRef]
8.
Bhandari, S.; Phinn, S.; Gill, T. Preparing Landsat Image Time Series (LITS) for Monitoring Changes in
Vegetation Phenology in Queensland, Australia. Remote Sens. 2012,4, 1856–1886. [CrossRef]
9.
Fisher, J.I.; Mustard, J.F. Cross-Scalar Satellite Phenology from Ground, Landsat, and MODIS Data.
Remote Sens. Environ. 2007,109, 261–273. [CrossRef]
10.
Liang, S.; Fang, H.; Morisette, J.T.; Chen, M.; Shuey, C.J.; Walthall, C.L.; Daughtry, C.S.T. Atmospheric
Correction of Landsat ETM + Validation and Applications: II. Validation and applications. IEEE Trans. Geosci.
Remote Sens. 2002,40, 2736–2746. [CrossRef]
Remote Sens. 2016,8, 938 15 of 16
11.
Atzberger, C.; Richter, K. Spatially Constrained Inversion of Radiative Transfer Models for Improved LAI
Mapping from Future Sentinel-2 Imagery. Remote Sens. Environ. 2012,120, 208–218. [CrossRef]
12.
Brest, C.L.; Goward, S.N. Deriving Surface Albedo Measurements from Narrow Band Satellite Data. Int. J.
Remote Sens. 1987,8, 351–367. [CrossRef]
13.
Qu, Y.; Liang, S.; Liu, Q.; He, T.; Liu, S.; Li, X. Mapping Surface Broadband Albedo from Satellite Observations:
A Review of Literatures on Algorithms and Products. Remote Sens. 2015,7, 990–1020. [CrossRef]
14.
Vuolo, F.; D’Urso, G.; de Michele, C.; Bianchi, B.; Cutting, M. Satellite-Based Irrigation Advisory Services:
A Common Tool for Different Experiences from Europe to Australia. Agric. Water Manag.
2015
,147, 82–95.
[CrossRef]
15.
Sentinel-2 Data Service Platform. Available online: https://s2.boku.eodc.eu/ (accessed on 9 November 2016).
16.
Wagner, W.; Wotawa, G.; Stowasser, R.; Staudinger, M.; Hoffmann, C.; Walli, A.; Federspiel, C.;
Aspetsberger, M.; Atzberger, C.; Briese, C.; et al. Addressing Grand Challenges in Earth Observation
Science: The Earth Observation Data Centre for Water Resources Monitoring. ISPRS Tech. Comm. VII
Mid-Term Symp. 2014,II-7, 81–88. [CrossRef]
17.
Müller-Wilm, U. Sentinel-2 MSI—Level-2A Prototype Processor Installation and User Manual. Available
online: http://step.esa.int/thirdparties/sen2cor/2.2.1/S2PAD-VEGA-SUM-0001-2.2.pdf (accessed on
6 October 2016).
18. European Space Agency. Sen2Cor 2.2.1—Software Release Note; European Space Agency: Paris, France, 2016.
19.
Kaufman, Y.J.; Wald, A.E.; Remer, L.A.; Gao, B.-C.; Li, R.-R.; Flynn, L. The MODIS 2.1-µm Channel-Correlation with
Visible Reflectance for Use in Remote Sensing of Aerosol. IEEE Trans. Geosci. Remote Sens.
1997
,35,
1286–1298. [CrossRef]
20.
Schläpfer, D.; Borel, C.C.; Keller, J.; Itten, K.I. Atmospheric Precorrected Differential Absorption Technique to
Retrieve Columnar Water Vapor. Remote Sens. Environ. 1998,65, 353–366. [CrossRef]
21.
Baret, F.; Weiss, M.; Bicheron, P.; Berthelot, B. Sentinel-2 MSI Products WP1152 Algorithm Theoretical Basis
Document for Product Group B; INRA-EMMAH: Avignon, France, 2010.
22.
Jacquemoud, S.; Verhoef, W.; Baret, F.; Bacour, C.; Zarco-Tejada, P.; Asner, G.P.; Francois, C.; Ustin, S.
PROSPECT+SAIL Models: A Review of Use for Vegetation Characterization. Remote Sens. Environ.
2009
,113,
S56–S66. [CrossRef]
23.
Verhoef, W. Light-Scattering by Leaf Layers with Application to Canopy Reflectance Modeling—The Sail
Model. Remote Sens. Environ. 1984,16, 125–141. [CrossRef]
24.
D’Urso, G.; Belmonte, A.C. Operative Approaches To Determine Crop Water Requirements From Earth
Observation Data: Methodologies And Applications. AIP Conf. Proc. 2006,852, 14–25.
25.
Vermote, E.; Justice, C.; Claverie, M.; Franch, B. Preliminary Analysis of the Performance of the Landsat
8/OLI Land Surface Reflectance Product. Remote Sens. Environ. 2015,185, 46–56. [CrossRef]
26.
Vuolo, F.; Mattiuzzi, M.; Atzberger, C. Comparison of the Landsat Surface Reflectance Climate Data Record
(CDR) and Manually atmospherically Corrected data in a Semi-Arid European Study Area. Int. J. Appl. Earth
Obs. Geoinf. 2015,42, 1–10. [CrossRef]
27.
Zhu, Z.; Wang, S.; Woodcock, C.E. Improvement and Expansion of the Fmask Algorithm: Cloud, Cloud
Shadow, and Snow Detection for Landsats 4–7, 8, and Sentinel 2 Images. Remote Sens. Environ.
2015
,159,
269–277. [CrossRef]
28.
Vuolo, F.; Neugebauer, N.; Bolognesi, S.; Atzberger, C.; D’Urso, G. Estimation of Leaf Area Index Using
DEIMOS-1 Data: Application and Transferability of a Semi-Empirical Relationship between two Agricultural
Areas. Remote Sens. 2013,5, 1274–1291. [CrossRef]
29.
Spectral Evolution, Lawrence, USA. Available online: http://www.spectralevolution.com/ (accessed on
9 November 2016).
30. Eilers, P.H.C. A Perfect Smoother. Anal. Chem. 2003,75, 3631–3636. [CrossRef] [PubMed]
Remote Sens. 2016,8, 938 16 of 16
31.
LI-COR Inc. LAI-2000 Plant Canopy Analyzer Instruction Manual Lincoln; LI-COR Inc.: Lincoln, NE, USA, 1992.
32.
Deimos-Imaging. Our Satellite: DEIMOS-1. Available online: http://www.deimos-space.com/ (accessed on
9 November 2016).
33.
GEOSYSTEMS. Atcor for Imagine 2013. Available online: http://www.geosystems.de/en/produkte/atcor-
for-imagine/download/ (accessed on 6 October 2016).
34.
Vuolo, F.; Ng, W.-T.; Atzberger, C. Innovative approach for smoothing and gap-filling of high resolution
multi-spectral time series: Example of Landsat data. IEEE Trans. Geosci. Remote Sens. 2016, under review.
©
2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
... The above Landsat 8 and Sentinel-2 optical data were obtained by passive remote sensing platforms [15]. Sentinel-2 is a multispectral high-resolution imaging satellite that carries a Multispectral Instrument (MSI) [16]. It has two satellites 2A and 2B, and each satellite conducts an Earth observation every 10 days under constant-observation conditions [16]. ...
... Sentinel-2 is a multispectral high-resolution imaging satellite that carries a Multispectral Instrument (MSI) [16]. It has two satellites 2A and 2B, and each satellite conducts an Earth observation every 10 days under constant-observation conditions [16]. The complementarity of the two satellites can achieve a temporal resolution of 5 days. ...
... The complementarity of the two satellites can achieve a temporal resolution of 5 days. The MSI has 13 bands that cover from 442 to 2202 nm, and the highest spatial resolution is 10 m [16]. With its three bands in the red edge, Sentinel-2 can provide rich information for crop detection and thereby greatly improve the estimation accuracy for chlorophyll content, the fractional cover of forest canopies, and leaf area index [17]. ...
Full-text available
Article
Although vegetation index time series from optical images are widely used for crop mapping, it remains difficult to obtain sufficient time-series data because of satellite revisit time and weather in some areas. To address this situation, this paper considered Wen County, Henan Province, Central China as the research area and fused multi-source features such as backscatter coefficient, vegetation index, and time series based on Sentinel-1 and -2 data to identify crops. Through comparative experiments, this paper studied the feasibility of identifying crops with multi-temporal data and fused data. The results showed that the accuracy of multi-temporal Sentinel-2 data increased by 9.2% compared with single-temporal Sentinel-2 data, and the accuracy of multi-temporal fusion data improved by 17.1% and 2.9%, respectively, compared with multi-temporal Sentinel-1 and Sentinel-2 data. Multi-temporal data well-characterizes the phenological stages of crop growth, thereby improving the classification accuracy. The fusion of Sentinel-1 synthetic aperture radar data and Sentinel-2 optical data provide sufficient time-series data for crop identification. This research can provide a reference for crop recognition in precision agriculture.
... While both models use an equivalent supervised classification method and produce analogous severity outputs, the comparison of outputs between sensors has not been comprehensively assessed. Although Sentinel 2 and Landsat 8 sensors cover similar spectral wavelengths, there are inherent differences that may impact remote sensing applications to various degrees [24][25][26]. For example, Sentinel 2 has a higher resolution (10 m pixel size) compared to Landsat 8 (30 m pixel size), while Landsat 8 has several bands with more narrow spectral coverage compared to Sentinel 2 (e.g., Red, NIR and SWIR1). ...
... For example, Sentinel 2 has a higher resolution (10 m pixel size) compared to Landsat 8 (30 m pixel size), while Landsat 8 has several bands with more narrow spectral coverage compared to Sentinel 2 (e.g., Red, NIR and SWIR1). Several studies have compared reflectance values for equivalent spectral bands between these two sensors [24][25][26], and small-scale studies have compared the effects of sensor type on fire extent and severity mapping [27,28]. These studies suggest that both sensors can reliably be used interchangeably in some applications. ...
... Our results align with previous studies which suggest that Sentinel 2 and Landsat 8 can be successfully used in combination for long-term monitoring applications, despite some inherent variation between the sensors [24][25][26][44][45][46][47]. Our study expands on the recent findings of smaller scale studies [27,28] which reported comparable outcomes using spectral indices derived from Sentinel 2 and Landsat 8 imagery to map fire extent and severity. ...
Full-text available
Article
Mapping of fire extent and severity across broad landscapes and timeframes using remote sensing approaches is valuable to inform ecological research, biodiversity conservation and fire management. Compiling imagery from various satellite sensors can assist in long-term fire history mapping; however, inherent sensor differences need to be considered. The New South Wales Fire Extent and Severity Mapping (FESM) program uses imagery from Sentinel and Landsat satellites, along with supervised classification algorithms, to produce state-wide fire maps over recent decades. In this study, we compared FESM outputs from Sentinel 2 and Landsat 8 sensors, which have different spatial and spectral resolutions. We undertook independent accuracy assessments of both Sentinel 2 and Landsat 8 sensor algorithms using high-resolution aerial imagery from eight training fires. We also compared the FESM outputs from both sensors across 27 case study fires. We compared the mapped areas of fire severity classes between outputs and assessed the classification agreement at random sampling points. Our independent accuracy assessment demonstrated very similar levels of accuracy for both sensor algorithms. We also found that there was substantial agreement between the outputs from the two sensors. Agreement on the extent of burnt versus unburnt areas was very high, and the severity classification of burnt areas was typically either in agreement between the sensors or in disagreement by only one severity class (e.g., low and moderate severity or high and extreme severity). Differences between outputs are likely partly due to differences in sensor resolution (10 m and 30 m pixel sizes for Sentinel 2 and Landsat 8, respectively) and may be influenced by landscape complexity, such as terrain roughness and foliage cover. Overall, this study supports the combined use of both sensors in remote sensing applications for fire extent and severity mapping.
... t/ha among all the estimation models, respectively. Moreover, since Landsat 8 and Landsat 9 are optical data, they can only obtain information on the forest canopy surface, which still suffers from saturation [27,75]. Combining Landsat 9 data with LiDAR or other data sources that can penetrate the canopy and obtain information on the vertical structure of the forest for forest parameter estimation is an effective way to relieve this problem [11,29]. ...
Full-text available
Article
Accurate estimation of forest above-ground biomass (AGB) is critical for assessing forest quality and carbon stocks, which can improve understanding of the vegetation growth processes and the global carbon cycle. Landsat 9, the latest launched Landsat satellite, is the successor and continuation of Landsat 8, providing a highly promising data resource for land cover change, forest surveys, and terrestrial ecosystem monitoring. Regression kriging was developed in the study to improve the AGB estimation and mapping using the Landsat 9 image in Wangyedian forest farm, northern China. Multiple linear regression (MLR), support vector machine (SVM), back propagation neural network (BPNN), and random forest (RF) were used as the original models to predict the AGB trends, and the optimal model was used to overlay the results of kriging interpolation based on the residuals to obtain the new AGB predictions. In addition, Landsat 8 images in Wangyedian were used for comparison and verification with Landsat 9. The results showed that all bands of Landsat 8 and Landsat 9 maintained a high degree of uniformity, with positive correlation coefficients ranging from 0.77 to 0.89 (p < 0.01). RF achieved the highest estimation accuracy among all the original models based on the two data sources. However, kriging regression can significantly reduce the estimation error, with the root mean square error (RMSE) decreasing by 55.4% and 51.1%, for Landsat 8 and Landsat 9, respectively, compared to the original RF. Further, the R2 and the lowest RMSE for Landsat 8 were 0.88 and 16.83 t/ha, while, for Landsat 9, they were 0.87 and 17.91 t/ha. The use of regression kriging combined with Landsat 9 imagery has great potential for achieving efficient and highly accurate forest AGB estimates, providing a new reference for long-term monitoring of forest resource dynamics.
... The satellite, launched on June 2015, is part of Europe's Copernicus program aiming at independent and continued global observation capacities. Compared to Landsat satellites, Sentinel-2 offers an increased spectral and spatial resolution with 13 spectral bands of 10 to 60 m spatial resolution (Vuolo et al., 2016). Moreover, in terms of temporal resolution, Sentinel-2 give an adequate update possibility with a combined constellation revisit frequency of 5 days. ...
Full-text available
Article
The concept of HNVF raises thanks to the integration of biodiversity theme into the CAP. The precise definition of the topic and the application of a correct identification procedure deeply affect the assessment needs of European RDPs. Nevertheless, the level of HNVF knowledge is rather limited due to a methodological variability and a structural lack of suitable data. The research aims at overtaking HNVF identification difficulties complying with specific Community requirements and the use of an efficient theoretical framework that allows accurate location and monitoring over time. The methodology results in characterization and accurate HNVF map for the Apulia Region that can effectively calibrate the implementation of the regional management policies.
... Les images satellites présentent souvent des erreurs géométriques (liées au système géographique d'enregistrement, dues à la rotation de la terre, à l'angle de balayage, aux mouvements de satellite, à la variation d'altitude…). Ces erreurs peuvent causer des erreurs de positionnement de l'image et par conséquent, des déformations des images ainsi que Plusieurs algorithmes ont été développés pour effectuer les CA sur les images S2, avec des études comparatives des différentes méthodes publiées (Dörnhöfer et al., 2016;Vuolo et al., 2016;Lantzanakis et al., 2017;Martins et al., 2017;Kukawska et al., 2017). Le SEN2COR reste le plus utilisé (Doxani et al., 2018;Sola et al., 2018). ...
Full-text available
Thesis
The presence of Mesozoic mafic rocks in southwestern Algeria has long been known. Recent studies indicate that this rock is mainly composed of tholeiitic dolerites emplaced at the end of Triassic, during the opening of the central Atlantic Ocean, 201 million years ago. These dolerites are part of a large magmatic province, called the Central Atlantic Magmatic Province (CAMP). Today, remnants of CAMP plumbing system include huge intrusive bodies (doleritic sills and dykes) and some lava flows preserved in Triassic basins. In the Tindouf Basin (Southwestern Algeria), CAMP rocks occur mainly as large sills and dykes especially in the northern flank of the basin. Doleritic pipes, that have never been studied so far, are spatially associated with these sills and dykes and thus supposed of CAMP origin. In this study, we present new field, petrographic, geochemical and geochronological data on the pipe, sills and dyke dolerites of the CAMP plumbing system of the northern flank of the Tindouf basin. The sills and dykes studied display homogeneous petrographic features, typical of CAMP tholeiites, whereas the doleritic pipes show more evolved characteristics with a hypovolcanic granophyric texture, containing alkali feldspars and quartz. Furthermore, we report for the first time the presence of abundant carbonatitic xenoliths in the pipe dolerites. Geochemically, all the investigated rocks are low-Ti tholeiites that can be subdivided into four groups, three of which correlate well with the intermediate, upper and recurrent units defined in the High Atlas of Morocco, whereas the fourth group, represented by the pipe dolerites, presents geochemical features that have never been reported before. However, their tholeiitic nature, their association in the field at the same geographic location with CAMP sills, and their trace element patterns, all indicate that pipe dolerites are of CAMP origin, representing the evolved products of Recurrent magmas. We performed 40Ar/39Ar biotite dating on four samples from the Tindouf basin (1 pipe, 1 sill of the recurrent unit and two hornfels in contact with doleritic sill). Plateau-ages of 193.7 ± 3.4 Ma, 197.6 ± 3.3 Ma, 202.5 ± 7.3 Ma and 199.8 ± 1.9 Ma was respectively obtained for the pipe, the recurrent sill and the two hornfels. Our study suggests that the small anticlines on the northern flank of the Tindouf Basin that host doleritic pipes in its cores can be explained as forced folds above igneous intrusions. Furthermore, we can explain the occurrence of carbonatitic xenoliths and sulphides in the pipe dolerites as the possible consequence of crustal assimilation of wall rocks by magma of the pipe conduits. Finally, the high CO2 and sulfur contents of the pipe dolerites, combined with the emplacement of large sills within Upper Devonian organic-rich shales of the Tindouf Basin, indicate a potentially significant emission of CO2 and sulfuric gases that may have contributed to climate change at the Triassic-Jurassic boundary.
... It runs in the ESA Sentinel Application Platform (SNAP) or from the command line. Additionally, topographic correction with a 90m digital elevation database from CGIAR-CSI (http://www.cgiar-csi.org) and cirrus corrections were applied [63]. Topographic correction here is purely radiometric and does not change the image geometry. ...
Full-text available
Thesis
Crop monitoring, especially in developing countries, can improve food production, address food security issues, and support sustainable development goals. Crop type mapping and yield estimation are the two major aspects of crop monitoring that remain challenging due to the problem of timely and adequate data availability. Existing approaches rely on ground-surveys and traditional means which are time-consuming and costly. In this context, we introduce the use of freely available Sentinel-2 (S2) imagery with high spatial, spectral and temporal resolution to classify crop and estimate its yield through a deep learning approach. In particular, this study uses patch-based 2D and 3D Convolutional Neural Network (CNN) algorithms to map rice crop and predict its yield in the Terai districts of Nepal. Firstly, the study reviews the existing state-of-art technologies in this field and selects suitable CNN architectures. Secondly, the selected architectures are implemented and trained using S2 imagery, groundtruth and auxiliary data in addition for yield estimation. We also introduce a variation in the chosen 3D CNN architecture to enhance its performance in estimating rice yield. The performance of the models is validated and then evaluated using performance metrics namely overall accuracy and F1-score for classification and Root Mean Squared Error (RMSE) for yield estimation. In consistency with the existing works, the results demonstrate recommendable performance of the models with remarkable accuracy, indicating the suitability of S2 data for crop mapping and yield estimation in developing countries
... The SNAP algorithm is a collection of sensor-specific ANNs trained with RTM simulations to retrieve GAI from multispectral canopy reflectance. Due to its generality and efficiency, the SNAP algorithm has been implemented as a core routine in the operational Sentinel-2 platform for vegetation biophysical variables estimation (Brown et al., 2021;Delloye et al., 2018;Li et al., 2015;Vuolo et al., 2016). ...
Article
The objective of this study is to evaluate the performances of a semi-empirical approach based on the Bayesian theory to retrieve Green Area Index (GAI) from multiple decametric satellites. It is designed to overcome some limitations in existing Radiative Transfer Model (RTM) inversion methods, including the high dimensionality of the inverse problem, the convergence problem due to possible equifinality, and the dependence of some RTM variables on the crop-specific architecture. The PROSAIL model is first inverted in a calibration step using the Hamiltonian Monte Carlo (HMC) algorithm over a global dataset of ground GAI measurements (for maize, wheat, and rice) and the corresponding reflectance observations from Landsat-8, Sentinel-2, and Quickbird to derive crop-specific distributions of PROSAIL input variables. These distributions were then used as prior information to predict GAI over an independent set of reflectance observations. Results show that the full Bayesian approach provides close estimates of GAI to ground truth, with respective Root Mean Square Error (RMSE) of 1.01, 1.33, and 0.97 for maize, wheat, and rice (R²=0.67, 0.76 and 0.63, respectively). The performances are better than those approaches generally reported using radiative transfer models that are non-crop-specific, like the SNAP algorithm for Sentinel-2, but are slightly behind the purely empirical models based on machine learning. However, the proposed approach provides an explicit insight of the joint distribution of PROSAIL variables that are valid for any satellite platform. This constitutes a major advantage against purely empirical models, as it enables to fully exploit large observational datasets from multiple sensors and generalize to other platforms.
Article
One of the crucial components in improving the quality of forests is post-fire vegetation regrowth monitoring. This is done by analyzing the time series of satellite data and studying the severity of forest fires to improve the ability to monitor the dynamic changes of the forest. This study investigates the regeneration process of existing vegetation types in different severities of The Rim fire in Sierra Nevada, California, using the time series of vegetation indices obtained from the MODIS sensor. The Vegetation Return Period (VRP) and the Recovery Rate (RR) after the fire were evaluated to monitor the regrowth of vegetation types. According to the results, the VRP values of the species for low, moderate, and high severity were estimated to be between 22 to 33 months, 33 to 47 months, and about 5 years, in this area. The 8-year changes in the time series of vegetation indices confirm that some vegetation types in this region have not fully recovered. In addition, spatio-temporal variations of the burned regions were examined with Landsat images at 2-year post-fire intervals until 2021. The results showed that in three 2-year periods after the fire, 16,074 hectares, 48,722 hectares, and 27,391 hectares of land were, respectively, converted into unburned areas, and until 2019, about 60% of the burned areas were recovered. Researchers and land managers can use the results of such studies to identify areas that need more attention after a fire.
Full-text available
Article
Canopy nitrogen content (CNC, kg/ha) provides crucial information for site-specific crop fertilization and the usability of Sentinel-2 (S2) satellite data for CNC monitoring at high fertilization levels in managed agricultural fields is still underexplored. Winter wheat samples were collected in France and Belgium in 2017 (n = 126) and 2018 (n = 18), analysed for CNC and S2-spectra were extracted at the sample locations. A comparison of three established remote sensing methods to retrieve CNC was carried out: (1) look-up-table (LUT) inversion of the canopy reflectance model PROSAIL, (2) Partial Least Square Regression (PLSR) and (3) nitrogen-sensitive vegetation indices (VI). The spatial and temporal model transferability to new data was rigorously assessed. The PROSAIL-LUT approach predicted CNC with a root mean squared error of 33.9 kg/ha on the 2017 dataset and a slightly larger value of 36.8 kg/ha on the 2018 dataset. Contrary, PLSR showed an error of 27.9 kg N/ha (R2 = 0.52) in the calibration dataset (2017) but a substantially larger error of 38.4 kg N/ha on the independent dataset (2018). VIs revealed calibration errors were slightly larger than the PLSR results but showed much higher validation errors for the independent dataset (> 50 kg/ha). The PROSAIL inversion was more stable and robust than the PLSR and VI methods when applied to new data. The obtained CNC maps may support farmers in adapting their fertilization management according to the actual crop nitrogen status.
Full-text available
Article
Crop production and productivity monitoring play a crucial role for food security and livelihoods, international trade and sustainable agricultural practices. Earth Observation (EO) data provides high spectral, spatial and temporal data for various agricultural applications. However, mapping and monitoring small crop fields and complex landscapes are still challenging, in particular when attempting to trace the historical evolution of land use changes. To address this issue, a study was set up in the Udon Thani Province of Thailand, with small agricultural parcels and highly fragmented landscapes, covering an area of approximately 11,000 km². Three decades of crop type dynamics were monitored and assessed using different combinations of multi-temporal Sentinel-1, Sentinel-2 and Landsat data and the random forest (RF) classifier. The combined multi-temporal EO datasets proved the most efficient for mapping crop types. Classification results achieved overall accuracy (OA) of 87.9%, 88.1%, 84.8% and 92.6% for the four base-years 1989, 1999, 2009 and 2019, respectively. Thanks to the availability of high-quality reference labels, the crop type map of 2019 showed the highest overall and class-specific accuracies. The 2019 classification model separated many crop classes well, especially sugarcane, cassava, rice and para rubber. On the contrary, for 1989, 1999 and 2009 drops in accuracy had to be accepted, as direct field reference observations were unavailable and reference information had to be sourced through photo-interpretation or trimming approaches. Overall, however, the RF method together with multi-temporal EO satellite data from multiple platforms showed high potential and excellent efficiency in crop type classification in complex landscapes. The most dominant classes of crop types for the four base-years were rice, sugarcane, and cassava, respectively. Land cover changes indicated that transitions of 1,529 km² (14%) occurred between 1989 and 1999, mainly as increase in sugarcane and rice areas. From 1999 to 2009, significant land changes were observed covering 2,340 km² (21%), primarily as increased cassava and para rubber cultivation. During the most recent period 2009 to 2019, an additional 3,414 km² (31%) were transformed, mainly through the expansion of para rubber and sugarcane plantations. The main drivers for the observed land use changes in the three decades were commodity prices and agricultural policies. The cost-efficiently derived results provide valuable information to inform land use management decisions of policymakers and other stakeholders, including the consideration of environmental aspects.
Full-text available
Article
Monitoring with high resolution land cover and especially of urban areas is a key task that is more and more required in a number of applications (urban planning, health monitoring, ecology, etc.). At the moment, some operational products, such as the “Copernicus High Resolution Imperviousness Layer”, are available to assess this information, but the frequency of updates is still limited despite the fact that more and more very high resolution data are acquired. In particular, the recent launch of the Sentinel-2A satellite in June 2015 makes available data with a minimum spatial resolution of 10 m, 13 spectral bands, wide acquisition coverage and short time revisits, which opens a large scale of new applications. In this work, we propose to exploit the benefit of Sentinel-2 images to monitor urban areas and to update Copernicus Land services, in particular the High Resolution Layer imperviousness. The approach relies on independent image classification (using already available Landsat images and new Sentinel-2 images) that are fused using the Dempster–Shafer theory. Experiments are performed on two urban areas: a large European city, Prague, in the Czech Republic, and a mid-sized one, Rennes, in France. Results, validated with a Kappa index over 0.9, illustrate the great interest of Sentinel-2 in operational projects, such as Copernicus products, and since such an approach can be conducted on very large areas, such as the European or global scale. Though classification and data fusion are not new, our process is original in the way it optimally combines uncertainties issued from classifications to generate more confident and precise imperviousness maps. The choice of imperviousness comes from the fact that it is a typical application where research meets the needs of an operational production. Moreover, the methodology presented in this paper can be used in any other land cover classification task using regular acquisitions issued, for example, from Sentinel-2.
Full-text available
Article
Mapping of glacier extents from automated classification of optical satellite images has become a major application of the freely available images from Landsat. A widely applied method is based on segmented ratio images from a red and shortwave infrared band. With the now available data from Sentinel-2 (S2) and Landsat 8 (L8) there is high potential to further extend the existing time series (starting with Landsat 4/5 in 1982) and to considerably improve over previous capabilities, thanks to increased spatial resolution and dynamic range, a wider swath width and more frequent coverage. Here, we test and compare a variety of previously used methods to map glacier extents from S2 and L8, and investigate the mapping of snow facies with S2 using top of atmosphere reflectance. Our results confirm that the band ratio method works well with S2 and L8. The 15 m panchromatic band of L8 can be used instead of the red band, resulting in glacier extents similar to S2 (0.7% larger for 155 glaciers). On the other hand, extents derived from the 30 m bands are 4%–5% larger, indicating a more generous interpretation of mixed pixels. Mapping of snow cover with S2 provided accurate results, but the required topographic correction would benefit from a better orthorectification with a more precise DEM than currently used.
Full-text available
Article
Land cover and land use maps derived from satellite remote sensing imagery are critical to support biodiversity and conservation, especially over large areas. With its 10 m to 20 m spatial resolution, Sentinel-2 is a promising sensor for the detection of a variety of landscape features of ecological relevance. However, many components of the ecological network are still smaller than the 10 m pixel, i.e., they are sub-pixel targets that stretch the sensor’s resolution to its limit. This paper proposes a framework to empirically estimate the minimum object size for an accurate detection of a set of structuring landscape foreground/background pairs. The developed method combines a spectral separability analysis and an empirical point spread function estimation for Sentinel-2. The same approach was also applied to Landsat-8 and SPOT-5 (Take 5), which can be considered as similar in terms of spectral definition and spatial resolution, respectively. Results show that Sentinel-2 performs consistently on both aspects. A large number of indices have been tested along with the individual spectral bands and target discrimination was possible in all but one case. Overall, results for Sentinel-2 highlight the critical importance of a good compromise between the spatial and spectral resolution. For instance, the Sentinel-2 roads detection limit was of 3 m and small water bodies are separable with a diameter larger than 11 m. In addition, the analysis of spectral mixtures draws attention to the uneven sensitivity of a variety of spectral indices. The proposed framework could be implemented to assess the fitness for purpose of future sensors within a large range of applications.
Full-text available
Article
Monitoring open water bodies accurately is an important and basic application in remote sensing. Various water body mapping approaches have been developed to extract water bodies from multispectral images. The method based on the spectral water index, especially the Modified Normalized Difference Water Index (MDNWI) calculated from the green and Shortwave-Infrared (SWIR) bands, is one of the most popular methods. The recently launched Sentinel-2 satellite can provide fine spatial resolution multispectral images. This new dataset is potentially of important significance for regional water bodies' mapping, due to its free access and frequent revisit capabilities. It is noted that the green and SWIR bands of Sentinel-2 have different spatial resolutions of 10 m and 20 m, respectively. Straightforwardly, MNDWI can be produced from Sentinel-2 at the spatial resolution of 20 m, by upscaling the 10-m green band to 20 m correspondingly. This scheme, however, wastes the detailed information available at the 10-m resolution. In this paper, to take full advantage of the 10-m information provided by Sentinel-2 images, a novel 10-m spatial resolution MNDWI is produced from Sentinel-2 images by downscaling the 20-m resolution SWIR band to 10 m based on pan-sharpening. Four popular pan-sharpening algorithms, including Principle Component Analysis (PCA), Intensity Hue Saturation (IHS), High Pass Filter (HPF) and à Trous Wavelet Transform (ATWT), were applied in this study. The performance of the proposed method was assessed experimentally using a Sentinel-2 image located at the Venice coastland. In the experiment, six water indexes, including 10-m NDWI, 20-m MNDWI and 10-m MNDWI, produced by four pan-sharpening algorithms, were compared. Three levels of results, including the sharpened images, the produced MNDWI images and the finally mapped water bodies, were analysed quantitatively. The results showed that MNDWI can enhance water bodies and suppressbuilt-up features more efficiently than NDWI. Moreover, 10-m MNDWIs produced by all four pan-sharpening algorithms can represent more detailed spatial information of water bodies than 20-m MNDWI produced by the original image. Thus, MNDWIs at the 10-m resolution can extract more accurate water body maps than 10-m NDWI and 20-m MNDWI. In addition, although HPF can produce more accurate sharpened images and MNDWI images than the other three benchmark pan-sharpening algorithms, the ATWT algorithm leads to the best 10-m water bodies mapping results. This is no necessary positive connection between the accuracy of the sharpened MNDWI image and the map-level accuracy of the resultant water body maps.
Full-text available
Article
Monitoring of the human-induced changes and the availability of reliable andmethodologically consistent urban area maps are essential to support sustainable urban developmenton a global scale. The Global Human Settlement Layer (GHSL) is a project funded by the EuropeanCommission, Joint Research Centre, which aims at providing scientific methods and systems forreliable and automatic mapping of built-up areas from remote sensing data. In the frame of the GHSL,the opportunities offered by the recent availability of Sentinel-2 data are being explored using anovel image classification method, called Symbolic Machine Learning (SML), for detailed urban landcover mapping. In this paper, a preliminary test was implemented with the purpose of: (i) assessingthe applicability of the SML classifier on Sentinel-2 imagery; (ii) evaluating the complementarityof Sentinel-1 and Sentinel-2; and (iii) understanding the added-value of Sentinel-2 with respect toLandsat for improving global high-resolution human settlement mapping. The overall objective is toexplore areas of improvement, including the possibility of synergistic use of the different sensors. Theresults showed that noticeable improvement of the quality of the classification could be gained fromthe increased spatial detail and from the thematic contents of Sentinel-2 compared to the Landsatderived product as well as from the complementarity between Sentinel-1 and Sentinel-2 images Assessment of the Added-Value of Sentinel-2 for Detecting Built-up Areas. Available from: https://www.researchgate.net/publication/299572090_Assessment_of_the_Added-Value_of_Sentinel-2_for_Detecting_Built-up_Areas [accessed May 10, 2016].
Full-text available
Article
The study presents the preliminary results of two classification exercises assessing the capabilities of pre-operational (August 2015) Sentinel-2 (S2) data for mapping crop types and tree species. In the first case study, an S2 image was used to map six summer crop species in Lower Austria as well as winter crops/bare soil. Crop type maps are needed to account for crop-specific water use and for agricultural statistics. Crop type information is also useful to parametrize crop growth models for yield estimation, as well as for the retrieval of vegetation biophysical variables using radiative transfer models. The second case study aimed to map seven different deciduous and coniferous tree species in Germany. Detailed information about tree species distribution is important for forest management and to assess potential impacts of climate change. In our S2 data assessment, crop and tree species maps were produced at 10 m spatial resolution by combining the ten S2 spectral channels with 10 and 20 m pixel size. A supervised Random Forest classifier (RF) was deployed and trained with appropriate ground truth. In both case studies, S2 data confirmed its expected capabilities to produce reliable land cover maps. Cross-validated overall accuracies ranged between 65% (tree species) and 76% (crop types). The study confirmed the high value of the red-edge and shortwave infrared (SWIR) bands for vegetation mapping. Also, the blue band was important in both study sites. The S2-bands in the near infrared were amongst the least important channels. The object based image analysis (OBIA) and the classical pixel-based classification achieved comparable results, mainly for the cropland. As only single date acquisitions were available for this study, the full potential of S2 data could not be assessed. In the future, the two twin S2 satellites will offer global coverage every five days and therefore permit to concurrently exploit unprecedented spectral and temporal information with high spatial resolution.
Full-text available
Article
This study contributes to the quality assessment of atmospherically corrected Landsat surface reflectance data that are routinely generated by the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS). This dataset, named Landsat Surface Reflectance Climate Data Record (Landsat CDR), is available at global scale and offers unprecedented opportunities to land monitoring and management services that require atmospherically corrected Earth observation (EO) data. Our assessment is based on the comparison of the Landsat CDR data against a set of Landsat and DEIMOS-1 images processed to a high degree of accuracy using an industry-standard atmospheric correction algorithm (ATCOR-2). The software package has been used for many years and its correction procedures can be considered consolidated and well-established. The dataset of Landsat and DEIMOS-1 images was acquired over a semi-arid agricultural area located in Lower Austria and was independently corrected by using a manual fine-tuning of ATCOR-2 parameters to reach the highest possible accuracy. Results show a very good correspondence of the surface reflectance in each of the six reflective spectral channels as well as for the NDVI (Normalized Difference Vegetation Index). An additional comparison against a NDVI time series from MODIS revealed also a good correspondence. Coefficients of determination (R2) between the two multi-year and multi-seasonal Landsat/DEIMOS datasets range between 0.91 (blue band) and 0.98 (nIR, SWIR-1 and SWIR-2). The results obtained for our semi-arid test site in Austria confirm previous findings and suggest that automatic atmospheric procedures, such as the one implemented by LEDAPS are accurate enough to be used in land monitoring services that require consistent multi-temporal surface reflectance data.
Article
This paper introduces a novel methodology for generating 15-day, smoothed and gap-filled time series of high spatial resolution data. The approach is based on templates from high quality observations to fill data gaps that are subsequently filtered. We tested our method for one large contiguous area (Bavaria, Germany) and for nine smaller test sites in different ecoregions of Europe using Landsat data. Overall, our results match the validation dataset to a high degree of accuracy with a mean absolute error (MAE) of 0.01 for visible bands, 0.03 for near-infrared and 0.02 for short-wave-infrared. Occasionally, the reconstructed time series are affected by artefacts due to undetected clouds. Less frequently, larger uncertainties occur as a result of extended periods of missing data. Reliable cloud masks are highly warranted for making full use of time series.
Article
The surface reflectance, i.e., satellite derived top of atmosphere (TOA) reflectance corrected for the temporally, spatially and spectrally varying scattering and absorbing effects of atmospheric gases and aerosols, is needed to monitor the land surface reliably. For this reason, the surface reflectance, and not TOA reflectance, is used to generate the greater majority of global land products, for example, from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) sensors. Even if atmospheric effects are minimized by sensor design, atmospheric effects are still challenging to correct. In particular, the strong impact of aerosols in the visible and near infrared spectral range can be difficult to correct, because they can be highly discrete in space and time (e.g., smoke plumes) and because of the complex scattering and absorbing properties of aerosols that vary spectrally and with aerosol size, shape, chemistry and density.