Sentinel For Applications in Agriculture

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W6, 2019
ISPRS-GEOGLAM-ISRS Joint Int. Workshop on “Earth Observations for Agricultural Monitoring”, 18–20 February 2019, New Delhi, India
SENTINEL FOR APPLICATIONS IN AGRICULTURE
Mario A. Gomarasca1,*, Antonella Tornato2, Daniele Spizzichino2, Emiliana Valentini2, Andrea Taramelli2,§, Giuseppe Satalino3,
Massimo Vincini4, Mirco Boschetti5, Roberto Colombo6, Livio Rossi7, Enrico Borgogno Mondino8, Luigi Perotti9, Walter Alberto9,
Fabio Villa9
1,*
Summer School Director, Italian Remote Sensing Association (AIT) & CNR-IREA Milano, [email protected]
2
ISPRA, Rome, Italy,§IUSS, Pavia, Italy,3CNR IREA Bari, Italy,4CRAST-UNIPC, Piacenza, Italy,5CNR IREA Milano, Italy,
6
DISAT-UNIMIB, Milano, Italy,7e-GEOS SpA, Rome, Italy,8DISAFA-UNITO, Torino, Italy,9IMAGEO srl Spin-off, Torino, Italy
Commission III, WG III/10
KEY WORDS: Copernicus, Sentinel, Agriculture, Earth Observation, Geo-information, Geomatics
ABSTRACT:
The European Union and the European Space Agency (EU/ESA) have promoted since 1998 (Baveno Manifesto*) the GMES
Programme (Global Monitoring for Environment and Security), nowadays called Copernicus (www.copernicus.eu). In the
agriculture domain, the use of Copernicus Sentinel imagery and its services are providing several new opportunities. The knowledge
of fundamentals of Earth Observation/Geographic Information EO/GI, namely Geomatics, for the development of innovative
strategies for professional skills adequacy and capacity building, supporting Copernicus user uptake, becomes mandatory
(Gomarasca, 2009). The target is to help bridging gaps between supply and demand of education and training for geospatial
sector (www.eo4geo.eu). The innovative and strategical novelties are the complete free access to Sentinel time series imagery
and digital image processing software “Sentinel toolboxes” such as SNAP (Sentinel Application Platform) for different
environments (Windows, Mac, Unix). The paper introduce topics as crop mapping and monitoring, biophysical parameters,
phenology and yield estimations, through several concluded or ongoing international projects such as: ERMES -FP7
(http://www.ermes-fp7space.eu/it/homepage/, Busetto et al. 2017) and SATURNO (https://www.progettosaturno.it/, Nutini et al.,
2018) devoted to the regional agricultural monitoring. As conclusion, SNAP software for image processing of Sentinel data was
demonstrated and tested together with Earth Engine software for specific vertical agriculture applications. The topics reported in
this paper have been part of the Summer School ‘Sentinel for Applications in Agriculture’ supported by the Copernicus
programme, several scientific associations (AIT, ASITA, EARSeL - European Association of Remote Sensing Laboratories),
the European Erasmus+ project EO4GEO, University Departments and Geo-Information Companies.
1. INTRODUCTION Programme manager owing the infrastructure and data rights on

behalf of the Union. The Space segment and services are based
The Copernicus Programme, the EU Earth Observation and on information from a dedicated constellation of satellites
monitoring programme, was established by Regulation (EU) No (Sentinels), as well as of the third-party satellites known as
377/2014 of the European Parliament and of the Council, on “contributing space missions”, complemented by “in situ”
April 3rd, 2014 (http://www.copernicus.eu/). (meaning local or on site) measurement data and the added
value products. By making the vast majority of its data,
The programme was designed to provide a European response analysis, forecast and maps available and accessible,
to Global needs such as the environment management, climate Copernicus contribute towards the development of new
change effects mitigation and to ensure civil and citizen innovative applications and services, tailored to the need of
security. More in detail the programme is a cornerstone of the specific groups of user, which touch on a variety of economic
European Union’s efforts to monitor the Earth and it’s many and cultural or recreational activities, from Urban Planning,
ecosystem, whilst ensuring that its citizens are prepared and sailing and insurance to archeology. From the economic point
protected in the faces of crises and natural and man-made of view the programme is supported by an adequate and
disasters. Copernicus is also a useful tool for economic substantial budget. The Overall programme is coordinate by
development and a driver for digital economy. The programme European Commission while the space component is
entered its operational phase with the launch of Sentinel-1A in coordinated by ESA and public and private partners. Future
2014 and its governance is based on the Copernicus Regulation development concerning the space component will be a joint
adopted the same year witch established the Commission as the EC-ESA divided into two different sets of activities: Sentinel
This contribution has been peer-reviewed.

https://doi.org/10.5194/isprs-archives-XLII-3-W6-91-2019 | © Authors 2019. CC BY 4.0 License. 91
future Expansion and Next Generation Sentinels. The other product. This product needs to be pre-processed before its use,
fundamental pillar of the programme are the six cross cut following a typical chain of pre-processing that includes:
thematic services: Land monitoring, marine environment, calibration, co-registration, multi-looking and geocoding. This
Atmosphere, Climate Change, Emergency Management and pre-processing can be performed by using the Sentinel-1
Security. The Copernicus Programme is designed and operates Toolbox (S1TBX), which consists of a collection of processing
to make available to several public and private user tools, data product readers and writers, embedded into the
communities trusted and quality multi-source products and SNAP common architecture
services. The engagement of Member States takes place in the (https://sentinel.esa.int/web/sentinel/toolboxes/sentinel-1). The
frame of the National User Forum, and is characterized by some crop classification by using Sentinel-1 data is a specific
common denominators: coping with the European and national problem of data classification. A wide literature is available
obligations, facilitating and sustaining the scientific and concerning classification methods in general (e.g. supervised or
technologic innovation to realize downstream services and unsupervised) and crop classification methods from SAR data
societal benefits through the National Space Economy and in particular (Oliver and Quegan, 1998). What have to be taken
Policy. Finally, since the development of operational services into account is that crop classification accuracy depends not
has to be user-driven oriented, the criteria for Sentinel evolution only on the specific method used but also on various factors
are: such as:
• consolidated user needs: Copernicus Core users • level of data pre-processing (speckle filtering, multi-
require additional observational capacity, temporal filtering, …);
• gap analysis: new mission taking into account • SAR features used (backscatter coefficients, single-
synergies with other missions and the pool of dual polarization, multi-temporal data, …);
current/novel products. • type of crops (winter-summer crops, small stem
• technological readiness: the mission could be realized crops, broad leaf plants, …);
within the timeframe 2025-2030. Detailed criteria for • number of crops.
application/technology/in-situ. A comprehensive overview of the results achievable by means
• affordability: mission can be realized with realistic of supervised approaches as applied to multi-polarized and/or
assumptions on available budgets to meet multi-temporal SAR data can be found in Skriver et al., 2011,
requirements effectively or efficiently. where the importance of the polarizations (e.g. VH and VV)
and of the use of multi-temporal SAR data for improving the
1. Description of the activities and discussion crop classification accuracy are illustrated. Their main
drawback is that the classification methodology is data driven
1.1 Sentinel-1 for agriculture (i.e., training fields sampled on the investigated site are
required); therefore, the classifier performance is normally
The potentiality of Sentinel-1 for crop mapping comes from the space and time dependent. As an example, a winter crop map
sensitivity of Synthetic Aperture Radar (SAR) data to crop obtained from multi-temporal Sentinel-1 images acquired in
structure and vegetation water content. Although optical 2015 over the Apulian Tavoliere (Southern Italy) is given in
sensors (e.g. Sentinel-2) are well suited for agricultural Figure 1. The map identifies seven soil/crop classes and it has
applications, the use of Sentinel-1 can offer an additional and been obtained by a classification procedure of seven multi-
complementary source of information for crop mapping, as temporal VV and VH Sentinel-1 pre-processed images (at
Sentinel-1 SAR images: i) are not affected by cloud cover, ii) approximately 100m pixel resolution) acquired during spring
have a global coverage, and iii) have a frequent temporal 2015 and by using the Maximum Likelihood classifier. The
revisit. The Sentinel-1 constellation has been designed to overall accuracy of the map, evaluated over a testing data set,
acquire high resolution dual-polarized C-band SAR data with a achieves 85%. As can be observed, the majority of the
temporal gap of 6 days. This is achieved by using two satellites agricultural area is covered by cereals, i.e. wheat, barley and oat
(i.e. S-1A and S-1B) sharing the same orbit plane with a 180° crops.
orbital phasing difference. The C-SAR instrument on-board can
operate in four exclusive acquisition modes (i.e. i) Strip map Sentinel-1 data have a good potential to identify agricultural
(SM), ii) Interferometric Wide swath (IW), iii) Extra-Wide crops due to its sensitivity to crop structure and vegetation
swath (EW), and iv) Wave mode (WV)), with different water content. This potential can be exploited to get crop maps
resolution (down to 5 m) and coverage (up to 400 km). The with a good classification accuracy (e.g. 85%) by using multi-
obtained acquisitions are made available as SAR products with temporal and dual polarization (VV and VH) Sentinel-1 data.
different level of processing and resolution (e.g. level-1 SLC,
level-1 GRD) 1.2 Sentinel-2 for agriculture
(https://sentinel.esa.int/web/sentinel/missions/sentinel-
Sentinel-2 constellation is built to analytically acquire high-
1/overview). SAR products can be searched and downloaded by
resolution, multi-spectral images with a high revisit time
means of the Copernicus Open Access Hub
globally. The constellation is composed by two (Sentinel-A and
(https://scihub.copernicus.eu/). Among them, the most used
B) polar orbiting satellites space out at 180° visible in Figure 2.
product for agricultural applications is the IW Level-1 GRDH
The Multi-Spectral Instrument (MSI) is the sensor carried as

using the Sentinel-2 Toolbox (application part of the

Sentinel Application Platform (SNAP));
Figure 3. Comparison of Sentinel-2 with Landsat 7 and 8

(source: https://landsat.gsfc.nasa.gov/sentinel-2a-launches-our-
compliments-our-complements/ )
1.3 Farm level monitoring: phenology assessment

Figure 1. Winter crop map over the Apulian Tavoliere site Timely and accurate information on crop growth and seasonal
(Italy) derived from multi-temporal Sentinel-1 images in 2015. dynamics are increasingly needed to develop monitoring
systems aimed to detect seasonal anomalies, support site
specific management and estimate crop yield at the end of the
payload and has a push-broom configuration. The MSI sensor
season. In particular, frequent decametric information
acquires in 13 different spectral band (Visible, NIR and SWIR)
nowadays being provided exploiting the new generation of
with a ground sample resolution (GSD) of 10 m for (B2, B3, B4
Earth Observation (EO) platforms are fundamental for farm
and B8) and a swat of 290 km.
level monitoring. The current availability of operational data
from decametric multispectral systems, such as ESA Sentinel 2
and NASA/USGS Landsat-OLI, represents the fundamental
precondition to develop solutions able to retrieve information at
the agricultural management unity (i.e., the field). In this
framework, during the summer school it was presented a study
aimed at assessing the usefulness of dense time series of
decametric Leaf Area Index (LAI) maps for phenological
monitoring on rice-cultivated areas. More detail about this
application can be found in (Boschetti et al. 2018). Decametric
multi-source LAI time series were produced by the Department
of Earth Physics and Thermodynamics (Universitat de
València) by inverting the PROSAIL radiative transfer model
with Gaussian process regression techniques (Campos-Taberner
et al., 2016) using available Sentinel 2 and Landsat-7/8
Figure 2. Sentinel-2 overview (source:
https://sentinel.esa.int/web/sentinel/missions/sentinel- mission’s optical data acquired from early May up to the end of
2/overview) September for the year 2016. LAI maps at 10-30 m resolution,
One of Sentinel-2 objectives is to create a synergy with already according to sensor source, were registered and resampled to
existing land monitoring missions (e.g. USGS Landsat Sentinel 2 images providing a 26 bands multi-temporal dataset
Thematic Mapper (TM) and Operational Land Imager (OLI) for subsequent analysis. Phenological estimates were performed
and the SPOT series). As visible in Figure 3. the bands of with a modified version of the PhenoRice algorithm (Boschetti
Sentine-2 MSI are comparable to the ones of the Landsat et al., 2017). The time series of images were analyzed as
satellites. follows: first, the LAI time series of each pixel was smoothed
using a Savitzky-Golay two-iteration method able to assign
The data products delivered to the public in a free and open weights to each LAI value according to EO product cloud mask
access policy have 2 levels: information and LAI retrieval estimated accuracy.
• Level-1A: Top-Of-Atmosphere reflectance distributed Subsequently, the smoothed signal, characterized by a 7 days
online; regular LAI time series, was analyzed using PhenoRice to
• Level-2C: Bottom-Of-Atmosphere reflectance reconstruct estimate of the dates of occurrence of several rice
distributed online but can also be generated user side crop phenological stages (crop establishment corresponding to

DVS = 0, emergence DVS = 1, tillering DVS = 1.3, flowering the SNAP VI processor, and used to apply crop-specific LAI-
DVS = 2, and maturity DVS = 4) and to compute seasonal VIs empirical relationships in band math. LAI, CCD and LCD
metrics (e.g., length of the vegetative and reproductive phases). estimates can be also obtained by using the SNAP Biophysical
processor (Figure 4), employing a neural network (NNT)
algorithm developed by INRA tailored for Sentinel-2 and
trained using radiative transfer simulations from PROSPECT
and SAIL RT models. Differences between empirical and RTM
inversion-based estimates as implemented within SNAP for the
two crops can be evaluated and critically discussed based on the
capability to integrate prior crop-specific information, (e.g.
average leaf inclination, taxonomy) critical for the accuracy of
the estimation. While such capabilities are limited in the current
release of the SNAP Biophysical processor, crop-specific
empirical relationships of Biophysical parameters vs. VIs can
be easily used in SNAP for empirical estimates of LAI, CCD
and LCD.
Figure 4. Examples of Phenological estimates at field level for a Beside the temporal estimation of biophysical parameters, high-
case study rice farm in Lomellina (North of Italy, Pavia temporal resolution Sentinel 2 VIs Time-Series (TS) can be
province, latitude 45°15’ N ± 1.5’, longitude 8° 34’ E ± 2’, 112 used for yield assessment and to extract key phenological
m ± 15 m a.s.l.). Cultivated varieties (a) with different sowing metrics at the field level.
dates (d) and phenological estimates @ 10 m resolution:
Phenological metrics that can be obtained include the onset of
tillering (b), flowering (c), maturity (e) and season length (f).
greenness (start of season), time of peak VI, maximum VI, rate
of green-up, length of season, rate of senescence and time-
In figure 4. the variability of the crop condition in few integrated area under the curve, a metric that can be used for
kilometers according to cultivated varieties and agro-practices, yield forecast. Similarly, to the inversion of RT models current
in European context, is shown. It is therefore evident how SNAP release incorporates basic tools for simple TS analysis.
parcel level information is needed to characterized and monitor
such a complex condition. Estimates performances were
assessed both exploiting field observation and modelling results
(see Boschetti et al. 2018 for details). Differences between
reference and EO estimates at flowering phenological stage
(DVS 2) range between – 15 to 10 days, in the majority of the
cases differences are comprised within 5 days and the RMSE is
about 6.5 days. The preliminary test conducted in this study
highlights how time series of decametric data can contribute to
parcel-scale crop monitoring providing new insights on crop
management practices, including information on cultivated
varieties. This information can push towards crop modelling Figure 5. LAI estimate by using the SNAP Biophysical
application for yield estimation at field scale and first test have processor
been conducted to assess contribution in assimilation
framework (Gilardelli et al., 2019). Three multi-temporal Sentinel 2 part of images, corresponding
to Piacenza province portions, geographically starting from
1.4 Crop monitoring: biophysical parameters, phenology ‘Tadini Centre Locality Gariga di Podenzano Italy, towards
and yield south up to Emilian Apennines mountains are shown in Figure
The analysis of high-spatial and temporal resolutions Sentinel 2 5. The images, all acquired in 2018 (February, April, July),
data is a powerful tool for crops monitoring. At the field level have been already processed and have been using for the
the estimation of crops’ key biophysical parameters (Leaf Area Common Agricultural Policy Controls (CAP) in Italy; the band
Index - LAI, Leaf and Canopy Chlorophyll Density – combination 8-11-4 (near infrared, short infrared, red) is the
respectively LCD and CCD) can be addressed by Empirical best, by literature and in operation, for agricultural monitoring
relationships with Vegetation Indexes (VIs) or by the inversion and crop groups detection. Due to the different resolutions,
of Leaf and canopy radiative transfer (RT) models short IR has 20 m pixel vs 10 of NIR, the CAP procedure
(PROSPECT+SAIL) suggests to do not apply the more correct pan sharpening
SNAP provides the opportunity to apply and compare both method, which offers a better geometry (smoothing) but a
methods (i.e. empirical VIs’ relationships and inversion of RT worse spectral signature maintenance. This data set has been
models) for the estimation of biophysical parameters from used for train the capability of the multi-temporal information
Sentinel 2 data. Many VIs sensitive to LAI, CCD and LCD can of Sentinel 2 (72 possible yearly acquisitions in south Europe)
be obtained at 10 or 20 m resolution using SNAP band math or to identify the crop phenological phases as additional

information for the detection of ‘cultivation groups’ by spectral concerns the quality of a measure (or of an intervention) and
signatures. not exactly the way the problem is faced. It can be said that,
ordinarily, people retain that technological approaches are more
1.5 European CAP precise (and intelligent) than traditional one, generating this
great misunderstanding.
The Common Agriculture Policy (CAP) of European Union
To make “technological forestry/agriculture” a “precise” one, a
needs a verified money distribution to 8 million of EU farmers.
great work has still to be done concerning the entire expected
Remote sensing geomatics applications allow to manage and
workflow: improvement of users’ consciousness of both data
control the single distributed subsidies overall the EU28 (27
and tools is the main challenge to ensure that PFF could be
next years) agricultural holdings.
adopted at farm/forest company level. In particular,
Remote Sensing and the geomatics science have been multispectral remote sensing from imaging systems is expected
accompanying and explaining the main purposes of this policy, to heavily support this process. Recently, three facts have
the regulation evolution along the years, in parallel with the determined an exponential and sudden acceleration of remote
technology improvement, both in terms of resolution/spectral sensing technology transfer to PFF: a) the availability of open
bands and satellite acquisition capacity. Particular focus is now access data archives of native and pre-processed global datasets
assigned to Sentinel data new contribution, not only for (e.g NASA Landsat and MODIS, ESA Copernicus Sentinel
improving the efficiency of the existing approach and data, etc.); b) the introduction into the market of low cost
methodology but, as often happens, for providing the base and systems for image acquisition (i.e. RPAS, Remotely Piloted
the technical support of the new EU regulations: passing from Aerial Systems) equipped with multispectral and thermal
“controls” at sample level to “monitoring” for the entire sensors; c) software development, that greatly has improved the
continental agro-environmental territory which will be applied degree of automation in data processing. Commercial players
in Europe after 2020. This different method, due to the Sentinel have been the firsts that approached the technology transfer,
constellation availability, open to a continuous monitoring of forecasting great income opportunities, especially in the
each agronomic parcel, along the different phenological seasons agricultural context; this, generally, occurred with no care about
in Europe. Obviously, this new systematic approach can be the appropriate scientific consciousness; the high level of
suggested and applied for other international automation, given by software improvement, encouraged this
applications/replications, even at regional level. trend. Surprisingly, traditional remote sensing experts initially
did not participate to technology transfer. Consequently, this
RPAS light devices (more diffused, especially for their major business-based approach is now showing some limitations
feasibility in authorization) are continuously growing in mainly related to operators’ unconsciousness about some
applications and in popular awareness, even for agriculture technical aspects that are needed to make results and deductions
management and precision farming. However, at the important reliable (Mondino, 2017). This is not a negligible issue,
and satisfactory results (geometrical and thematic), must be especially when quantitative measures are required to drive
always associated the corresponding limitations in terms of agronomic/forest interventions. The current moment is crucial:
working time and costs (management and SW). Several official scientists have still not answered all the questions concerning
experiences have given an overview of the capability, through properness and effectiveness of remote sensing to the
operational tests performed. For instance, the Italian Ministry of operational compart (Borgogno-Mondino, 2017). Moreover,
Agriculture asked and paid specific application trials aimed at consistency of RS costs with those required by these at-low-
evaluating the pasture graze ability, the soil erosion, the crop income sectors is not still demonstrated, nor the actual benefits
species distinguishing, the vineyard NDVI evolution, etc., with that this technology can provide at farm/forest company level.
the goal to demonstrate when a benefit/cost balance can justify Geomatics scientific community is therefore called to support
the light RPAS use. Generally, the above ratio appears positive this transitory phase by:
when specific or complicate detections are requested, while for a) testing sensors and tools providing information concerning
simple crop or land use analysis or updating the relative costs proper ranges of their usability;
(devices, software and working time) seem too high for b) transferring to users and sensors/sw producer’s knowledge
justifying the methodology against the traditional rapid field and skills about metric and spectral issues related to RS;
visits. c) proposing and formalizing operational standards for both
technology and data processing.
1.6 Agriculture and Precision Farming
Among a great variety of unanswered questions, the following
Precision Farming and Forestry (PFF) are, nowadays, well ones seem to require particular attention in this technology
sounding keywords for everyone involved in forest and transfer phase:
agronomic practices. Nevertheless, at the state of the art, the a) how can automate data (free) processing to generate reliable
same concept would be better addressed to as “Technological spectral indices time series? Exploitation of all the available
Forestry/Agriculture”, being technology the most relevant issue auxiliary information is expected to filter out bad observations
that strongly characterizes this type of approach, making it at pixel level;
different and, probably, “advanced”, in respect of more
traditional practices. Precision is something different: it

b) which type of models can be adopt to translate spectral

information into phenological/crop parameters? Can model be
assumed as general or a continuous process of calibration has to
be taken into account?
c) which is the minimum mapping unit to be considered in
agriculture while monitoring crops from satellite? Have we to
operate at pixel, cadastral parcel or segment level when reading
the behavior of a cultivated field? The feeling is that the
properness of spectral aggregation achieved at cadastral parcel
level has to be overcome to get a true description of the ongoing
phenomena.
d) How RPAS can effectively support agriculture and which is
expected to be their relationship with satellite acquisitions?
Figure 6. FLEX and Sentinel-3 joining forces (image credit:

ESA)
1.7 FLEX7Sentinel 3Tandem Mission
Sentinel-3 is primarily an ocean mission, however, the mission 1.8 Drone survey and data elaboration
is also able to provide atmospheric and land applications. The
mission provides data continuity for the ERS, Envisat and This section is dedicated to illustrate the potentiality of drone
SPOT satellites. Sentinel-3 makes use of multiple sensing survey as support tool for agriculture and precision farming.
instruments to accomplish its objectives; SLSTR (Sea and Land Nowadays UAV-RPAS-Drone instrument is a powerful a tool
Surface Temperature Radiometer), OLCI (Ocean and Land to produce high resolution orthophotos and high resolution
Colour Instrument), SRAL (SAR Altimeter), DORIS, and Digital Elevation/Surface Models (DTMs and DSMs). An
MWR (Microwave Radiometer). An interesting future unmanned aerial vehicle (UAV), commonly known as drone or
development of space observation is the combination between RPAS, is an aircraft without a human pilot aboard. UAVs are a
Sentinel 3 and FLEX in a tandem mission (Figure 6). The component of an unmanned aircraft system (UAS); which
ESA’s FLEX mission aims to provide global maps of include a UAV, a ground-based controller, and a system of
vegetation fluorescence, which can be converted into an communications between the two parts. Topographic modeling
indicator of photosynthetic activity. These data would improve data processing is becoming more accessible and more widely
our understanding of how much carbon is stored in plants and used due to recent advances in digital photogrammetry methods
their role in the carbon and water cycles. FLEX will be the first and UAV-Drone instruments. Nowadays classic aerial
space mission designed to observe fluorescence by using a photogrammetry survey is often replaced by Structure from
novel technique measuring the main part of the chlorophyll Motion (SfM) techniques, where the non-metric camera is
fluorescence spectrum that originates from the core of the mounted on a UAV, and many photographs taken vertically or
photosynthetic machinery. Sun induced fluorescence will be slantwise towards the ground. The SfM workflow generates,
measured with a high-resolution imaging spectrometer first of all, 3D models in arbitrary object coordinates and the
acquiring data in the 500–780 nm spectral range, with a precision of the model depends on image quality/quantity and
sampling of 0.1 nm in the oxygen bands (759–769 nm and 686– on the 3D reconstruction algorithm. Moreover, the cartographic
697 nm), while red edge, chlorophyll absorption and PRI absolute accuracy of the final orthophoto and DSMs depends on
(Photochemical Reflectance Index) will be computed at 0.5–2.0 the accuracy of the coordinates used in the georeferencing
nm. FLEX mission will orbit in tandem with Sentinel-3 of the process. Commonly a complete drone survey includes a first
Copernicus constellation. Taking advantage of Sentinel-3's step with area selection and drone system selection and setup, a
optical and thermal sensors will lead to an integrated package of second step that includes the flight plan definition and target
measurements to assess plant health. With the Sentinel-2 positioning; the third step is the outside Drone flight including
satellites also in orbit, there is a unique opportunity of using this the GNSS target survey and final step is dedicated to data
data synergistically from all three missions for vegetation export and elaboration through Structure for Motion
studies. Sentinel-3 instruments OLCI and SLSTR provide the photogrammetric software (Agisoft Photoscan
needed atmospheric information, as well as contribute to www.agisoft.com) in order to obtain Digital Surface Model and
retrieve vegetation information (LAI, chlorophyll) useful for the high resolution orthophoto of the surveyed area.
interpretation of the fluorescence signal.

1.9 Sentinel Application Platform (SNAP) and Google

Earth Engine
A common architecture for all Sentinel Toolboxes has been

developed and called the Sentinel Application Platform (SNAP)
(Brockmann Consult, Array Systems Computing and C-S).
The SNAP architecture is ideal for Earth Observation

processing and analysis due to several technological
innovations: extensibility, portability, Modular Rich Client
Platform, generic Earth Observation (EO) data abstraction,
Tiled Memory Management, and a Graph Processing
Framework. The image processing of Sentinel-2 data for
agricultural applications is one of the key features available in
SNAP. SNAP permits to build processing chains using its main
features and to apply them to specific agricultural applications.
In SNAP it is possible to open a Sentinel 2 image and to exploit
Figure 7. Elements of Drone Survey. From top left clockwise – the basic tools developed for image processing and analysis
Drone Phantom 4 DJI, flight plan with parameters, DSM (e.g. raster tools, optical tools, band math and supervised
hillshade and orthophoto 2cm GSD. classification). Basic computation of synthetic indexes (NDVI,
EVI) is already available as default but can be modified in order
to meet each user needs. Furthermore, it is possible to carry out
A synthetic process how to carry out the UAV-Drone survey is supervised and unsupervised classifications. For example,
following: random forest supervised classifications can be applied both on
Level-1A and on Level-2C satellite images. SNAP is also
• Define take-off, landing and flight sites, in particular for suitable for land cover monitoring. For example, by combining
the survey with a UAV, always ensuring flight safety. 9 different Level-2C (one per month). it is possible to create a
• Define the right flight planning together with drone pilot, time-series stack. The result can be later on queried using the
ensuring adequate transverse and longitudinal overlaps to SNAP time series tool. This result is obtained by computing an
obtain good 3D intersections. NDVI for one image. Later, a chain graph is create using the
• Define position of well-distributed GCPs (Ground Control simple NDVI. Afterwards, the generated graph it has been
Points) by means of targets placed on the ground inside applied to all the 9 images using the batch processor. This
the survey area (and visible during the flights). permits to create the evolution of the NDVI for each pixel of
• Carry out differential, rapid static GNSS survey of all the subset, as shown in the snapshot of the result (Figure 8).
defined GCPs in order to extract 3D coordinates of the The obtained result can be later integrated or validated using
topographic network useful to georeferencing and scaling higher resolution imagery e.g. aerial or RPAS.
the 3D model and the derived cartographic products.
Once field surveys were carried out and all the data were
collected, the main SfM data processing steps are:
• Matching photo survey with EXIF data;
• Building the point cloud in a local reference system;
• Texturing 3D model and building an orthomosaic;
• Georeferencing model in the cartographic reference
system;
• Exporting the georeferenced orthomosaic and DSM;
During the survey setup and in the case of multispectral data
comparison (i.e. with Sentinel images) the UAV-Drone can be
equipped with multispectral and/or thermal sensors. The SfM
techniques and software permits to use also multispectral or Figure 8. NDVI evolution using the Time series tool.
thermal images to produce final high resolution orthophoto. In
this case we can use Drone orthophoto to analyze areas Google Earth Engine (EE) combines a multi-petabyte catalog of
identified by Sentinel data in order to obtain a complete satellite imagery and geospatial datasets with planetary-scale
synoptic vision with infrared or thermal data. analysis capabilities and makes it available for scientists,
researchers, and developers to detect changes, map trends, and
quantify differences on the Earth's surface.
(https://earthengine.google.com/). The goal of the platform is to
provide to its users the tools to perform analysis using earth

observed data. Moreover, the platform is free to use, and every Environment Volume 187, 15 December 2016, Pages 102–118
user owns the personal scripts and can decide whether to http://dx.doi.org/10.1016/j.rse.2016.10.009
publish or keep them private. Fundamental is the
Gilardelli C. , Stella T., Confalonieri R., Ranghetti L., Campos-
comprehension of the main features, capabilities, and benefits
Taberner M., Garcia-Haro F.J, Mirco Boschetti M., (2019).
for agricultural applications. These aspects are essential to give Downscaling rice yield simulation at sub-field scale using
an overview of EE and in a second moment explore possible remotely sensed LAI data. European Journal of Agronomy 103
applications of the existing algorithms of the standard library of (2019) 108–116. https://doi.org/10.1016/j.eja.2018.12.003
EE to remote sensed imagery.
Gomarasca M.A., 2009. Basics of Geomatics. Springer, London
Multi-temporal analysis can be implemented with ease in EE. New York, 656 pp ISBN: 9781402090134, DOI 10.1007/978-1-
For example, it is possible to perform a quick NDVI trend using 4020-9014-1
the Sentinel-2 Level-1C imagery over a point of interest with
Nutini, F., Confalonieri, R., Crema, A., Movedi, E., Paleari, L.,
only a few lines of code. In Figure 2 it is visible the interface of Stavrakoudis, D., Tabacchi, M., Cerioli, S., Tesio, F., Boschetti,
EE, in the top-center the code used to filter one year the M., 2018. Application of a satellite based approach to monitor
Sentinel-2 imagery over a specific point of interest and compute rice nitrogen status and to support precision agriculture
the NDVI, on the top-left the image the NDVI value of the techniques, in: Società Italiana di Agronomia (Ed.), Atti Del
point of interest (cyan) at the center of the image and the date of XLVII Convegno Della Società Italiana Di Agronomia, Marsala
the value. This permits to have a quick overview of the situation (TP), 12-14 September 2018, pp. 17–19.
of a certain crop. The main downside of EE is the unavailability
Mondino, E. B. 2017, June. Remote Sensing from RPAS in
of atmospherically corrected Sentinel-2 Level-2A data. Agriculture: An Overview of Expectations and Unanswered
Questions. In International Conference on Robotics in Alpe-
Adria Danube Region (pp. 483-492). Springer, Cham.
2.CONCLUSIONS
Oliver, C. J., Quegan, S. 1998. Understanding synthetic
In this paper an overview of the Copernicus programme has aperture radar images. Artech House, Inc., Norwood, MA.
been presented as a result of an organic collection and
Rossi L. Drone imagery use: pilot test in Italy – 22 MARS JRC
presentation of aspect related to the utilization and application
Conference Lisbon (PT) 24-25 November 2017.
of Sentinel data in agriculture applications. The field of
applications is very wide and a basic approach has been Rossi L. 2018. Controls with Remote Sensing vs Monitoring:
introduced considering the fundamental aspect what could be the role of Sentinel2 - The actual use of imagery session-
taken in account when approaching with specific context. Workshop on checks and management of agricultural land in
Considering the relatively new possible approaches and the IACS – Vilnius Lithuania 28-30 May 2018.
enormous set of automated data that are available and can be
Skriver, H., Mattia, F., Satalino, G., Balenzano, A., Pauwels, V.
used, several unanswered questions are proposed stimulating
R. N., Verhoest, N. E. C., & Davidson, M. 2011. Crop
the scientific community in approaching this frontier, opening
classification using short-revisit multi-temporal SAR data.
strong operative possibilities to be studied and applied.
IEEE Journal of Selected Topics in Applied EO and RS, 4(2),
423–431, DOI: 10.1109/JSTARS.2011.2106198
REFERENCES
Boschetti M., Busetto L., Ranghetti L., Garcia-Haro F.J.,

Campos-Taberner M. and Confalonieri R., 2018, "Testing
Multi-Sensors Time Series of Lai Estimates to Monitor Rice
Phenology: Preliminary Results," IGARSS 2018 - 2018 IEEE
Int. Geoscience and Remote Sensing Symposium, Valencia,
2018, pp. 8221-8224. doi: 10.1109/IGARSS.2018.8518494
Busetto L., et al., 2017. Downstream services for rice crop

monitoring in Europe: From regional to local scale. IEEE
Journal of Selected Topics in Applied Earth Observations and
Remote Sensing Volume: Issue 99. pp.1–19.
10.1109/JSTARS.2017.2679159
Borgogno Mondino, E., & Gajetti, M. 2017. Preliminary

considerations about costs and potential market of remote
sensing from UAV in the Italian viticulture context. European
Journal of Remote Sensing,50(1), 310-319.
Campos-Taberner M., García-Haro F.J., Camps-Valls G., Grau-

Muedra G., Nutini F., Crema A., Boschetti M., 2016. Multi-
temporal and multi resolution leaf area index retrieval for
operational local rice crop monitoring. Remote Sensing of


Sentinel For Applications in Agriculture

Uploaded by

Copyright:

Available Formats

Sentinel For Applications in Agriculture

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sentinel For Applications in Agriculture

Uploaded by

Copyright:

Available Formats

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W6, 2019

SENTINEL FOR APPLICATIONS IN AGRICULTURE

Commission III, WG III/10

KEY WORDS: Copernicus, Sentinel, Agriculture, Earth Observation, Geo-information, Geomatics

1. INTRODUCTION Programme manager owing the infrastructure and data rights on

This contribution has been peer-reviewed.

This contribution has been peer-reviewed.

using the Sentinel-2 Toolbox (application part of the

Figure 3. Comparison of Sentinel-2 with Landsat 7 and 8

1.3 Farm level monitoring: phenology assessment

This contribution has been peer-reviewed.

This contribution has been peer-reviewed.

This contribution has been peer-reviewed.

b) which type of models can be adopt to translate spectral

Figure 6. FLEX and Sentinel-3 joining forces (image credit:

This contribution has been peer-reviewed.

1.9 Sentinel Application Platform (SNAP) and Google

A common architecture for all Sentinel Toolboxes has been

The SNAP architecture is ideal for Earth Observation

This contribution has been peer-reviewed.

Boschetti M., Busetto L., Ranghetti L., Garcia-Haro F.J.,

Busetto L., et al., 2017. Downstream services for rice crop

Borgogno Mondino, E., & Gajetti, M. 2017. Preliminary

Campos-Taberner M., García-Haro F.J., Camps-Valls G., Grau-

This contribution has been peer-reviewed.

You might also like