You are on page 1of 37

Copernicus Global Land Operations – Lot 1

Date Issued: 29.11.2018


Issue: I1.20

Copernicus Global Land Operations


“Vegetation and Energy”
”CGLOPS-1”
Framework Service Contract N° 199494 (JRC)

ALGORITHM THEORETICAL BASIS DOCUMENT

SURFACE SOIL MOISTURE


COLLECTION 1KM
VERSION 1.0

Issue 1.20

Organization name of lead contractor for this deliverable:

Book Captain: Bernhard Bauer-Marschallinger


Contributing Authors: Christoph Paulik
Stefan Schaufler
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Dissemination Level
PU Public X
PP Restricted to other programme participants (including the Commission Services)
RE Restricted to a group specified by the consortium (including the Commission Services)
CO Confidential, only for members of the consortium (including the Commission Services)

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 2 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Document Release Sheet

Book captain: Bernhard Bauer-Marschallinger Sign Date: 29.11.2018

Approval: Roselyne Lacaze Sign Date: 03.12.2018

Endorsement: Michael Cherlet Sign Date:

Distribution: Public

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 3 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Change Record

Issue/Rev Date Page(s) Description of Change Release

20.12.2016 All Initial version – description of algorithm I1.00

I1.00 27.07.2017 All Update after external review I1.10

I1.10 29.11.2018 All Update related to products public release I1.20

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 4 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

TABLE OF CONTENTS

Executive Summary .................................................................................................................... 9


1 Background of the document ............................................................................................. 10
1.1 Scope and Objectives............................................................................................................. 10
1.2 Content of the document....................................................................................................... 10

1.3 Related documents ............................................................................................................... 10


1.3.1 Applicable documents ................................................................................................................................ 10
1.3.2 Input............................................................................................................................................................ 10
1.3.3 Output ......................................................................................................................................................... 11
1.3.4 External documents .................................................................................................................................... 11

2 Review of Users Requirements ........................................................................................... 12


3 Methodology Description .................................................................................................. 14
3.1 The Sentinel-1 Acquisitions .................................................................................................... 14
3.2 The retrieval Algorithm ......................................................................................................... 16
3.2.1 Outline ........................................................................................................................................................ 16
3.2.2 Basic underlying assumptions ..................................................................................................................... 17
3.2.3 Alternative methodologies currently in use ............................................................................................... 18
3.2.4 Input data.................................................................................................................................................... 18
3.2.5 Output product ........................................................................................................................................... 20
3.2.6 Methodology............................................................................................................................................... 20
3.2.7 Limitations .................................................................................................................................................. 29

3.3 Quality Assessment ............................................................................................................... 32


3.3.1 Evaluation Experiments over Italy 2015-2017 ............................................................................................ 32
3.3.2 Masking before re-sampling ....................................................................................................................... 32
3.3.3 Soil Moisture Noise Estimation ................................................................................................................... 33
3.3.4 Slope Estimation ......................................................................................................................................... 33

3.4 Risk of failure and Mitigation measures ................................................................................. 35


3.4.1 Sensor failure .............................................................................................................................................. 35
3.4.2 Quality degradation .................................................................................................................................... 35

4 References ........................................................................................................................ 36

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 5 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

List of Figures

Figure 1: S-1 Product Modes (from ED1) ...................................................................................... 15


Figure 2: Sentinel-1 Revisit and Coverage Frequency (as in 02/2018 from ESA) .......................... 15
Figure 3: Sentinel-1 Mode - Polarisation - Observation Geometry Map (as in 02/2018 from ESA) 16
Figure 4: Example of dry and wet backscatter reference curves for winter conditions with minimum
vegetation cover and summer with maximum vegetation cover and their relationship with the
incidence angle. From (Pathe et al., 2009). ............................................................................ 18
Figure 5: The workflow of the SSM1km production ....................................................................... 21
Figure 6: Number of S-1 observations over Europe between October 2014 and October 2016..... 23
Figure 7: Slope parameter before (left) and after (right) the multiple regression step, for an area
over Belgium. ......................................................................................................................... 24
Figure 8: SSM1km Water Mask over Europe (with country borders for orientation). Pixels in dark
blue masked in the SSM1km data.......................................................................................... 27
Figure 9: SSM1km Sensitivity Mask over Europe (with country borders for orientation). Pixels in
orange masked in the SSM1km data. .................................................................................... 28
Figure 10: SSM1km Elevation Slope Mask over Europe (with country borders for orientation).
Pixels in orange masked in the SSM1km data. ...................................................................... 29
Figure 11: Sentinel-1 SSM image over southern Sweden and eastern Denmark on 2016-07-12.
Scalloping causes jumps in the soil moisture images, along the edges of the backscatter
image’s sub-swaths. .............................................................................................................. 31
Figure 12: Re-sampled images to 500m pixel size, without masking of high and low backscatter
values on the left and with masking on the right. .................................................................... 32
Figure 13: From upper left to lower right: Number of observations per pixel from S-1A; areas where
initial S-1A slope estimate is valid; S-1A slope from regression; Enivsat ASAR WS slope;
difference map of initial ASAR and S-1A regression slope. .................................................... 34

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 6 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

List of Tables

Table 1: Target requirements of GCOS for soil moisture (up to 5cm soil depth) as Essential Climate
Variable (GCOS-154, 2011) ................................................................................................... 12
Table 2: Characteristics of S-1 IW mode (from ED1) ..................................................................... 19

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 7 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

List of Acronyms
ASAR Advanced Synthetic Aperture Radar (Envisat satellite)
ASAR GM ASAR Global Mode (observation mode of the ASAR sensor)
ASAR WS ASAR Wide Swath (observation mode of the ASAR sensor)
ASCAT Advanced Scatterometer (MetOp satellites)
CGLS Copernicus Global Land Services
CGIAR-CSI Consultative Group on International Agricultural Research - Consortium for
Spatial Information
CDF Cumulative Distribution Function
CSAR C-band Synthetic Aperture Radar (sensors onboard Sentinel-1 satellites)
DEM Digital Elevation Model
ECMWF European Centre for Medium Range Weather Forecasting
ENVISAT Environmental Satellite
ERS European Remote Sensing Satellites
EUMETSAT European Organization for the Exploitation of Meteorological Satellites
EW Extra Wide swath
H-SAF Hydrology Satellite Application Facility
IW Interferometric Wide swath
LIA Local Incidence Angle of a SAR observation
S-1 Sentinel 1 A and B
SAR Synthetic Aperture Radar
SCATSAR-SWI Scatterometer-Synthetic Aperture Radar Soil Water Index
SLC Single Look Complex (Sentinel-1 product type)
SMAP Soil Moisture Active Passive (US soil moisture satellite)
SNAP Sentinel Application Platform (processing toolbox from ESA)
SMOS Soil Moisture and Ocean Salinity (ESA satellite)
SRTM Shuttle Radar Topography Mission
SSM Surface Soil Moisture
SWI Soil Water Index
TOPSAR Terrain Observation with Progressive Scanning SAR
TU Wien Technische Universität Wien
WMO World Meteorological Organization

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 8 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

EXECUTIVE SUMMARY

The Copernicus Global Land Service (CGLS) is earmarked as a component of the Land service to
operate “a multi-purpose service component” that provides a series of bio-geophysical products on
the status and evolution of land surface at global scale. Production and delivery of the parameters
take place in a timely manner and are complemented by the constitution of long-term time series.
From 1st January 2013, the Copernicus Global Land Service is providing Essential Climate
Variables like the Leaf Area Index (LAI), the Fraction of Absorbed Photosynthetically Active
Radiation absorbed by the vegetation (FAPAR), the surface albedo, the Land Surface
Temperature, the soil moisture, the burnt areas, the areas of water bodies, and additional
vegetation indices, are generated every hour, every day or every 10 days on a reliable and
automatic basis from Earth Observation satellite data.
Satellite radar remote sensing has shown its capability to retrieve soil moisture at spatial scales in
the tens of kilometres operationally (Dorigo and de Jeu, 2016). Examples for such products are the
Copernicus Soil Water Index (SWI) product and ASCAT Surface Soil Moisture (SSM) distributed
through H-SAF (Hasenauer et al., 2006) as well as products from L-Band missions like SMOS
(Kerr et al., 2010) or SMAP (Entekhabi et al., 2010). Higher resolution products up to 1km
resolution have been distributed as science products in the past.
This document describes the algorithm used in the operational processing chain for the production
of Collection 1km SSM from the Sentinel-1 (S-1) constellation. The algorithm builds upon the
experience gained with similar data acquired with Envisat ASAR (Pathe et al., 2009) but is adapted
to the specific attributes of the S-1 sensor and orbit configuration (Hornáček et al., 2012; Bauer-
Marschallinger et al., 2018).

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 9 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

1 BACKGROUND OF THE DOCUMENT


1.1 SCOPE AND OBJECTIVES

The current CGLS Soil Water Index (SWI) is based on observations from the ASCAT sensors on
board the MetOp-A and MetOp-B satellites that leads to near daily revisit times globally. This revisit
cycle enables the product to describe temporal soil moisture dynamics well, but comes at the cost
of a spatial sampling of 0.1°, which is much coarser than the 1/112° spatial sampling of most other
products distributed through the Global Land Service. Because of that, the current SWI product
does not facilitate analysis of hydrological patterns on a local level that many users require. It is
also more difficult to use it in combination with other products from the CGLS. High-resolution
Synthetic Aperture Radar (SAR) data from the recently started European Sentinel-1 (S-1) program
is used for an evolution towards a 1km SSM product. This product (named SSM1km) aims to
provide coverage every few days, depending on location. It can also serve as an input product for
the SCATSAR-SWI that was developed in a previous evolution of the service, which will be
provided under the name SWI1km.

1.2 CONTENT OF THE DOCUMENT

This document is structured as follows:


 Chapter 2 recalls the users requirements, and the expected performance
 Chapter 3 gives an overview of the sensor, describes the retrieval methodology and
outlines limitations and mitigation strategies.
1.3 RELATED DOCUMENTS

1.3.1 Applicable documents

AD1: Annex I – Technical Specifications JRC/IPR/2015/H.5/0026/OC to Contract Notice 2015/S


151-277962 of 7th August 2015
AD2: Appendix 1 – Copernicus Global land Component Product and Service Detailed Technical
requirements to Technical Annex to Contract Notice 2015/S 151-277962 of 7th August 2015
AD3: GIO Copernicus Global Land – Technical User Group – Service Specification and Product
Requirements Proposal – SPB-GIO-3017-TUG-SS-004 – Issue I1.0 – 26 May 2015.

1.3.2 Input

Document ID Descriptor
CGLOPS1_SSD Service Specifications of the Global Component of
the Copernicus Land Service.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 10 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

1.3.3 Output

Document ID Descriptor
CGLOPS1_PUM_SSM1km-V1 Product User Manual summarizing all
information about the Collection 1km SSM
Version 1 product
CGLOPS1_QAR_SSM1km-V1 Report describing the results of the
scientific quality assessment of the
Collection 1km SSM Version 1 product

1.3.4 External documents

Document ID Descriptor
ED1 Sentinel-1 User Guide,
https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-sar

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 11 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

2 REVIEW OF USERS REQUIREMENTS


According to the applicable documents [AD2] and [AD3], the user’s requirements relevant for S1-
SSM are:
 Definition:
o Moisture content in the top 5 cm of the soil
 Geometric properties:
o Pixel size of output data shall be defined on a per-product basis so as to facilitate
the multi-parameter analysis and exploitation.
o The baseline datasets pixel size shall be provided, depending on the final product,
at resolutions of 100m and/or 300m and/or 1km.
o The target baseline location accuracy shall be 1/3 of the at-nadir instantaneous field
of view.
o Pixel co-ordinates shall be given for centre of pixel.
 Geographical coverage:
o geographic projection: lat long
o geodetical datum: WGS84
o pixel size: 1/112° - accuracy: min 10 digits
o coordinate position: pixel centre
o global window coordinates:
 Upper Left: 180°W-75°N
 Bottom Right: 180°E, 56°S
 Accuracy requirements:
o Baseline: wherever applicable the bio-geophysical parameters should meet the
internationally agreed accuracy standards laid down in document “Systematic
Observation Requirements for Satellite-Based Products for Climate”. Supplemental
details to the satellite based component of the “Implementation Plan for the Global
Observing System for Climate in Support of the UNFCCC. GCOS-#154, 2011”.
(Table 1)
o Target: considering data usage by that part of the user community focused on
operational monitoring at (sub-) national scale, accuracy standards may apply not
on averages at global scale, but at a finer geographic resolution and in any event at
least at biome level.

Table 1: Target requirements of GCOS for soil moisture (up to 5cm soil depth) as Essential
Climate Variable (GCOS-154, 2011)

Variable Horizontal Temporal Accuracy Stability


Resolution resolution
3 3 3 3
Volumetric soil moisture 50 km Daily 0.04 m /m 0.01 m /m /year

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 12 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

GCOS notes that the targets above “are set as an accuracy of about 10 per cent of saturated
content and stability of about 2 per cent of saturated content. These values are judged adequate
for regional impact and adaptation studies and verification and development of climate models. It is
considered premature to consider global scale changes.” It adds that “stating a general accuracy
requirement is difficult for this type of observation, as this depends not only on soil type but also on
soil moisture content itself. The stated numbers thus should be viewed with some caution”.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 13 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

3 METHODOLOGY DESCRIPTION
This section describes the methodology used to retrieve SSM from S-1 observations.

3.1 THE SENTINEL-1 ACQUISITIONS

The Sentinel-1 constellation consists of two identical satellites (Sentinel-1A launched in 2014,
Sentinel-1B launched in 2016). Both carry a C-Band Synthetic Aperture Radar (CSAR) instrument
for making observations in one of four modes (shortened from ED1, see also Figure 1):
 Stripmap (SM) - A standard SAR stripmap imaging mode.
 Interferometric Wide swath (IW) - Data is acquired in three swaths using the Terrain
Observation with Progressive Scanning SAR (TOPSAR) imaging technique. This is the
primary observation mode over land and also the mode of interest for soil moisture
retrieval.
 Extra Wide swath (EW) - Data is acquired in five swaths using the TOPSAR imaging
technique. EW mode provides very large swath coverage at the expense of spatial
resolution.
 Wave (WV) - WV is SENTINEL-1's operational mode over open ocean.
The mode of primary interest for soil moisture applications is IW mode. IW mode data is provided
as Level 1 Ground Range Detected (L1-IWGRD) products, which are the input data for the
Sentinel-1 SSM1km processing chain.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 14 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 1: S-1 Product Modes (from ED1)

Figure 2: Sentinel-1 Revisit and Coverage Frequency (as in 02/2018 from ESA)

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 15 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 3: Sentinel-1 Mode - Polarisation - Observation Geometry Map (as in 02/2018 from ESA)

The observation scenario of the S-1 constellation governs the spatio-temporal coverage of the
IWGRD data. The maps in Figure 2 and Figure 3 show the revisit times for different acquisition
modes and polarizations. IW observations in VV mode are acquired over most land regions. The
VV polarization is the one most useful for SSM applications due to better sensitivity to soil moisture
than the cross-polarized observations in VH.
When operating with two satellite spacecrafts (as from 2016 ongoing), the revisit time of the
Sentinel-1 mission over Europe is good with a coverage every 1.5-4 days. Over the rest of the
Earth, the coverage frequency can drop to 6 days. Further, the lower observation frequency makes
a longer observation record necessary to build consistent backscatter time series, delaying the
generation of reliable retrieval parameter database. Therefore, a global coverage of the SSM1km
is envisaged in a later product release.

3.2 THE RETRIEVAL ALGORITHM

3.2.1 Outline

The S-1 SSM retrieval algorithm is an adaption of the TU Wien change detection method (Wagner,
1998) for low resolution ERS and ASCAT data to higher resolution SAR data. The algorithm was
adapted to Envisat ASAR global mode data by Pathe et al., 2009 and validated over Australia,
Africa and large parts of South America (Dostálová et al., 2014; Mladenova et al., 2010). The
method normalizes the observed backscatter to a reference incidence angle to make them
comparable and to minimize the influence of vegetation on the backscattered signal. Scaling the

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 16 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

normalized backscatter between the historically driest (dry reference) and wettest (wet reference)
observed value leads to a relative soil moisture value. An error model is used to calculate soil
moisture error estimates through error propagation. A recent peer-reviewed publication (Bauer-
Marschallinger et al., 2018) explains in detail the SSM1km retrieval algorithm, underpinned by its
scientific background.

3.2.2 Basic underlying assumptions

Radar measurements show a dependence on the local incidence angle (LIA) under which they are
observed. This dependence is different for different land cover and terrain types. For bare or lightly
vegetated regions with smooth terrain, the observed backscatter decreases with the incidence
angle. This relationship becomes weaker for regions with denser vegetation and higher surface
roughness (Pathe et al., 2009).
Figure 4 illustrates the general behavior of C-band radar backscatter over an area with a seasonal
cycle. It shows an example of the incidence angle relationship of backscatter in dry and wet
conditions for summer (high vegetation) and winter (low vegetation). The figure also shows the
existence of cross over angles for which there is no difference in backscatter between high and low
vegetation conditions. These crossover angles differ between wet and dry conditions. In the case
of the ASCAT and ERS sensors, these seasonal vegetation effects can be corrected because
these sensors can measure backscatter at different incidence angles simultaneously.
Unfortunately, this is not possible for Sentinel-1. Then, a constant vegetation correction is applied,
based on the normalization of the backscatter to a 40° incidence angle (see section 3.2.6.2). This
constant vegetation correction can lead to biases during dry periods but, as can be seen from
Figure 4, the change in backscatter due to soil moisture changes is generally higher than the
change due to vegetation effects.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 17 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 4: Example of dry and wet backscatter reference curves for winter conditions with minimum
vegetation cover and summer with maximum vegetation cover and their relationship with the
incidence angle. From (Pathe et al., 2009).

3.2.3 Alternative methodologies currently in use

Hornáček et al., 2012 suggested an algorithm that should mitigate the slope definition issues
outlined in section 3.2.6.2.2. The therein algorithm does not use a slope for normalization of the
backscatter to a reference incidence angle but rather calculates dry and wet references for every
set of characteristic incidence angles of an S-1 orbit. This is possible because of the S-1
acquisition strategy, which measures any point on the Earth surface from up to only three
incidence angles. This method can potentially produce more robust model parameters, but since it
can only use data from the same relative repetitive orbits, it needs more data than the implemented
method.

3.2.4 Input data

3.2.4.1 Sentinel-1 IW GRDH

The main input data to the SSM1km retrieval algorithm is Sentinel-1 backscatter from the IW mode
in VV polarization (see also section 3.1). Table 2 shows the main characteristics of the IW
acquisition mode. The most important characteristics for the SSM1km retrieval are the incidence
angle range of 29.1 to 46°, and the radiometric accuracy of 1dB (3σ), as well as the spatial
sampling rate of 5 by 20 meters.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 18 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Table 2: Characteristics of S-1 IW mode (from ED1)

Characteristic Value
Swath width 250 km
Spatial Sampling 5 x 20 m
Incidence angle range 29.1° - 46.0°
Sub-swaths 3
Azimuth steering angle ± 0.6°
Azimuth and range looks Single
Polarisation options Dual HH+HV, VV+VH
Single HH, VV
Maximum Noise Equivalent Sigma Zero (NESZ) -22 dB
Radiometric stability 0.5 dB (3σ)
Radiometric accuracy 1 dB (3σ)
Phase error 5°

The S-1 Level 1 data is available in two formats, Single Look Complex (SLC) and Ground Range
Detected (GRD). The SSM1km processor uses GRD data since it is already focused SAR data that
has been detected, multi-looked and projected to ground range.

3.2.4.2 SRTM Digital Elevation Model

For terrain correction of the dataset, a digital elevation model (DEM) is necessary. ESA’s Sentinel-
1 toolbox within the Sentinel Application Platform (SNAP) automatically downloads and uses for
this purpose the CGIAR-CSI SRTM 90m v4.1 (Jarvis et al., 2008), which is gap-filled and with a
90m sampling in the WGS84 datum. The vertical error of the DEM’s is reported to be less than
16m.

3.2.4.3 Precise orbit information

The orbit state vectors included in the SAR product metadata are not accurate and can be refined
by information from precise- or restituted- orbit files, which are provided by ESA as auxiliary
datasets. Whereas the first are most precise but available only within days or weeks, the second
are generally available within hours after sensing time and precise enough for SSM applications.
For NRT operations, the restituted orbit files are downloaded and forwarded as input data to the
Sentinel-1 toolbox.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 19 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

3.2.5 Output product

The output product is a netCDF4 file with the two layers SSM and SSM_NOISE according to the
format specifications outlined in the CGLOPS1_SSD. The output product has a daily resolution,
meaning that all S-1 scenes observed between 00:00:00 and 23:59:59 UTC will be considered, run
through the SSM retrieval algorithm and included in the output product.
A water mask, a sensitivity mask (showing areas of low soil moisture sensitivity like cities), and a
slope mask are applied to SSM1km data, in order to mask non-meaningful values or values of low
quality (see section 3.2.6.5).
For more details about the output product specifications please see CGLOPS1_PUM_SSM1km-
V1.

3.2.6 Methodology

Figure 5 displays the flowchart of the SSM1km retrieval algorithm. The main steps are described in
more details below.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 20 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 5: The workflow of the SSM1km production

3.2.6.1 Pre-processing

The Sentinel-1 toolbox implements Range Doppler Terrain Correction (Small and Schubert, 2008)
which is used to reduce the geometric distortions present in SAR data due to the observation
geometry and the topography. The output of the terrain correction is a SAR image with a 10m
sampling. This image is then resampled to a 500m sampling, whereas averaging of the values is
done in the linear domain in order to correctly add backscatter intensities.
As a first step during the resampling process, very high (equivalent to -5dB) and very low
(equivalent to -20dB) backscatter values are masked out. Very high backscatter values originate
from cities and other man-made objects that act as corner reflectors whereas very low backscatter
values originate from standing water surfaces. These backscatter measurements are highly
unlikely to carry any soil moisture signal and thus are discarded. Because the value range in linear
Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium
Issue: I1.20 Date:29.11.2018 Page: 21 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

domain spreads over many powers of ten, the masking is of particular importance, as those 10m-
points are kept from spreading into the 500m-point during the resampling, polluting the output
image with non-representative values.
Secondly, a uniform aggregation of remaining 10m-pixels to 500m-pixels by simple averaging is
done.
As a last step, a spatial filtering of the aggregated image in order to remove aliasing effects is done
(and preserving signals larger than 1km), applying a Gaussian kernel with a half-energy-width of
1km.
This procedure was chosen, since common resampling engines do not comply with sampling laws
and produce aliasing effects. Furthermore, with this setup, the resampling procedure was speeded
up substantially.

3.2.6.2 Model parameter generation

The SSM scaling requires model parameters for the estimation of relative soil moisture content
(see 3.2.6.3). These parameters are inferred from the Sentinel-1 backscatter archive, comprising
three years of observations at the time of writing this document (2015 to 2017). For each grid point,
the backscatter time series are built, statistically examined and parameter layers are saved to disk,
holding values on backscatter distribution, observation frequency and incidence angle
dependency.

3.2.6.2.1 Dry- and Wet-Reference parameters

To determine the relative soil moisture content, one must know the local backscatter at dry and wet
conditions. These two values are estimated through the determination of the 10%- and 90%-
percentiles of the local backscatter time series, assuming that, by this choice, the estimation is not
violated by outliers. During soil moisture retrieval, the dry and wet references are then interpreted
as 10% and 90% SSM, and extrapolated linearly to 0% and 100% SSM, respectively.
Currently, changes in land use/cover are not accounted for in the model. For the kilometric scale, it
is presumed that anthropogenic/hazardous land transformations have no effect.

3.2.6.2.2 Slope parameter

The slope parameter describes the dependency of the backscatter to the local incidence angle ( )
for each grid point (Figure 4). This is modelled by an indirect linear relationship between and
backscatter. However, a sufficient number of different observation angles that spread over a large
range of incidence angles is necessary for a reliable estimation of the slope parameter .

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 22 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

(1)

Improved slope estimation


Due to the Sentinel-1 acquisition strategy, the backscatter measurements are not equally
distributed. For example, Figure 6 shows Europe with the sum of acquired measurements for each
ground point during October 2014 and October 2016. Here, the pattern of observation frequency is
clearly visible.

Figure 6: Number of S-1 observations over Europe between October 2014 and October 2016

Consequently, the range of incidence angles varies around the Earth’s surface. Furthermore, the
quality of the slope estimation depends on this incidence angle range. In some areas, the slope
estimation from equation (1) is sufficient, but in some areas it is not. To overcome this problem, a
multiple linear regression is performed to improve the slope estimation in those poorly estimated
areas. The established approach for the multiple regression in equation (2) is reasoned by the fact
that the correlation of the slope with the sensitivity and with the total mean is high. The sensitivity
S0 is the range of backscatter values at one location over time, here defined as the difference
between the 5%- and 95%-percentile. The regression for the slope is then:

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 23 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

(2)

The correlation weights a, b, c were determined in areas with a high number of observations where
a good slope estimation is possible. These areas make up ca. 80% of total area under observation,
giving confidence that all land cover classes are considered well. Afterwards, the correlation
weights were used to estimate the slope in areas with poor coverage, as the relationship
(expressed with a, b, c) is assumed to be stable and thus transferable in the space domain.
Figure 7 shows the comparison of the initial slope parameter (from Equation 1) and the slope
estimation using the regression weights (Equation 2).

Figure 7: Slope parameter before (left) and after (right) the multiple regression step, for an area over
Belgium.

3.2.6.3 Soil moisture retrieval

The resampled backscatter values ( )) where is the incidence angle and is the time is
normalized to 40° by using the slope model parameter ( ) and equation (3). Equation (3) describes
the linear relationship between incidence angle and backscatter as described in section 3.2.2
(Figure 4).

(3)

Retrieval of SSM is performed by application of equation (4). The normalized backscatter value
is scaled between the driest ( ) and the wettest ( ) historical value of the grid point in
question.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 24 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

(4)

This linear scaling in the SSM1km algorithm depends on statistically derived reference values for
dry and wet soil at each pixel. In most cases, the radar observations lie between these two
extreme values and the observations can be converted meaningfully to soil moisture values from
0% to 100% relative soil moisture.
However, in cases of one of the following, the SSM retrieval is ill posed:
• extreme dry conditions,
• frozen soil,
• snow covered soil,
• floodings.
Therefore, all SSM retrievals exceeding the local pixel’s dry- and wet-reference are replaced by
flag values.

3.2.6.4 Soil Moisture Noise Estimation

The SSM’s noise is estimated by propagating the error from the S-1 backscatter observations
through the retrieval model. S-1 backscatter uncertainty is given by the radiometric accuracy of the
observations with a standard deviation of 0.33 dB for a 10x10 m pixel (see Table 2). Resampling to
1km resolution reduces this noise level. The noise level of S-1 IW mode at 1km was estimated to
be between 0.05 and 0.07 dB by Hornáček et al., 2012. We revised this estimate to 0.2 dB after
inspection of the data and assuming higher correlations between the errors at the 10m scale.
Equation (5) shows the error propagation formula, with as the noise of the S-1 IW data at 1km,
as the noise of the slope parameter, the error of the dry reference and as
the error of the wet reference.

(5)

With the normalized sensitivity S40, defined as:

(6)

The errors are assumed to be independent of each other. The uncertainties of the model
parameters are estimated based on assumptions about the potential error sources. Errors can
occur in the model parameters due to incomplete sampling over the seasons, which was a problem
Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium
Issue: I1.20 Date:29.11.2018 Page: 25 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

for Envisat ASAR data but is not so problematic for S-1 data due to the higher and more regular
revisit rate. They can also occur due to inherent uncertainties in the statistical methods used to
estimate the model parameters. Another source of errors is the static vegetation correction, which
leads to seasonally varying errors in the model parameters. Following the reasoning in Hornáček et
al. (2012) and Pathe et al. (2009) and considering these error sources, we have chosen the error
of the slope to be 10% of the slope parameter meaning . The errors for the dry and wet
reference are estimated to be 10% of the sensitivity .

The maximum retrieval error occurs either in very dry (SSM=0) or very wet (SSM=1) conditions
and when the incidence angle of an observation is farthest away from the reference of 40 degrees
at 29.1 degrees. Inserting the assumptions about the errors into (5) allows an estimate of the
maximum SSM error:

(7)

3.2.6.5 Soil Moisture Masking

The Sentinel-1 SSM algorithm does not apply, or rather does not make sense, for all surfaces. It is
straightforward to mask out water surfaces to remove misleading values over the sea, lakes and
rivers. This is facilitated with the SAR backscatter parameters. Here, the water mask is created
from the 5%-percentile parameter, where each pixel with a value lower than -17dB is classified as
water. In the SSM1km images, those pixels are set automatically to a flag value (=251), since SSM
cannot be retrieved over water. Figure 8 illustrates the water mask over Europe.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 26 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 8: SSM1km Water Mask over Europe (with country borders for orientation). Pixels in dark blue
masked in the SSM1km data.

The sensitivity S40 is an important measure for the accuracy and reliability of the S-1 SSM
algorithm. Therefore, a sensitivity mask is generated, masking out pixels with low sensitivity. A
threshold of 1.2dB is used. This yields a mask that detects urban areas. In the SSM1km images,
those pixels are set automatically to a flag value (=252). Figure 9 illustrates the sensitivity mask
over Europe.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 27 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 9: SSM1km Sensitivity Mask over Europe (with country borders for orientation). Pixels in
orange masked in the SSM1km data.

The terrain correction in the pre-processing step does not completely remove errors stemming
from topographic effects (see Section 3.2.7). To identify and mask out locations with high
topographic complexity, an analysis of the CGIAR-CSI SRTM 90m v4.1 (Section 3.2.4.2) is carried
out and the elevation slope is computed with GDAL’s gdaldem tools. From this elevation slope, a
(binary) slope mask at 500m is produced, masking pixels with an elevation slope higher than 30%
(equivalent to a gradient of 17°). For the resampling from 90m to 500m, gdalwarp cubic resampling
method has been applied. In the SSM1km images, those pixels are set automatically to a flag
value (=253). Figure 10 illustrates the slope mask over Europe.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 28 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 10: SSM1km Elevation Slope Mask over Europe (with country borders for orientation). Pixels
in orange masked in the SSM1km data.

3.2.7 Limitations

Retrieval Algorithm limitations


Soil moisture cannot be retrieved over deserts, high vegetation areas like tropical forests and
during frozen or snow covered conditions.

Freeze/Thaw Masking
Radar remote sensing of soil moisture is error-prone in the presence of frozen or thawing
conditions, as well as in presence of snow cover. An implementation of a dynamic 1km freeze/thaw
masking should be considered in the future, for removing erroneous values during cold periods.

Vegetation Correction
The S-1 SSM1km algorithm as-is shows limitation towards a proper modelling of vegetation
influences on the SAR backscatter measurements, owing to the static vegetation correction of the
used change-detection approach. This can lead to biases during the vegetation period, especially
over areas with high vegetation density (e.g. forests). Future developments of the Sentinel-1
SSM1km algorithm e.g. towards to the ASCAT SSM algorithm have potential to account for
vegetation signals in the SAR observations.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 29 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Terrain Corrections
The Sentinel-1 Toolbox terrain correction does not completely remove the influence of topography,
in part due to complexity of terrain effects and in large part due to the insufficient resolution of the
used DEM. The limitation appears mainly over high mountain ranges into effect. Once a
comprehensive high-resolution DEM is available, this can be improved substantially. In this
implementation, the slope mask can be used to mask out points images that suffer from
topographic effects.

Sentinel-1 Intra-Image-Calibration
Sentinel-1 observations in IW mode consist of three parallel sub-swaths that need to be calibrated,
which is done by ESA’s processing facilities. Without this calibration, or when not accurately
calibrated, images show linear artefacts, biasing the level of backscatter per sub-swath. This is
commonly referred to as scalloping. Although the calibration was improved during the initial
observation period in 2014, for a number of orbits scalloping is still present. The SSM1km retrieval,
being a change detection algorithm, is especially sensitive to scalloping, which can be seen in the
example in Figure 11.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 30 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Figure 11: Sentinel-1 SSM image over southern Sweden and eastern Denmark on 2016-07-12.
Scalloping causes jumps in the soil moisture images, along the edges of the backscatter image’s
sub-swaths.

An assessment of the error evoked by these artefacts is under investigation at present. The orbit-
based approach as mentioned in Section 3.2.3 would mitigate the erroneous pattern, as building
the scaling parameters from data only from one single orbit, affected by the same bias.

Radio Frequency Interference (RFI)


In some small amount of Sentinel-1 images, artefacts from RFI were detected, appearing as block-
shaped irritations. These stem from ground C-band transmitting systems and cannot be removed
by ESA, or by the SSM algorithm. This issue was raised in the Sentinel online forums
(http://forum.step.esa.int/t/artifacts-in-sentinel-1-grd-data-how-to-detect-filter/3499/3). It should be
mitigated by automatically flagging and removing affected scenes, preferably during the Level-1
processing.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 31 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

3.3 QUALITY ASSESSMENT

3.3.1 Evaluation Experiments over Italy 2015-2017

As an update to this ATBD document: A recent peer-reviewed publication (Bauer-Marschallinger et


al., 2018) explains first in detail the SSM1km retrieval algorithm and then investigates its
performance over Italy and its neighbours. Over that area, which comprises challenging terrains
and heterogeneous land cover, a consistent set of model parameters and product masks was
generated, unperturbed by coverage discontinuities. An evaluation of therefrom generated
SSM1km data, involving in-situ measurements and a 1km Soil Water Balance Model (SWBM) over
Umbria, yielded high agreement over plains and agricultural areas, with low agreement over
forests and strong topography. Furthermore, positive biases during the growing season were
detected. However, excellent capability to capture small-scale soil moisture changes as from
rainfall or irrigation is evident from this evaluation.

3.3.2 Masking before re-sampling

The masking of outliers before resampling is necessary since high scatterers on the 10m scale
pollute the 500m pixels when resampling in the linear domain. Figure 12 illustrates the impact of
masking on the re-sampled images. On the left, we see the effect of very high backscatter (white)
in cities and villages that are very distinct and very low backscatter values (black) over rivers and
lakes, which were not masked out. On the right, the river was masked out in areas where it covers
the whole 500m pixel and the city impact on surroundings is attenuated after masking the highest
values.

Figure 12: Re-sampled images to 500m pixel size, without masking of high and low backscatter
values on the left and with masking on the right.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 32 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

3.3.3 Soil Moisture Noise Estimation

The soil moisture noise model applied to Envisat ASAR data was evaluated by Doubková et al.,
(2012) with encouraging results. They predicted the RMSE and correlation coefficient R between
the Envisat ASAR SSM product and a hydrological model using the error estimates and variances
of the respective datasets. The RMSE was predicted within an accuracy of 4 % of saturated SSM
over 89 % of Australia.

3.3.4 Slope Estimation

The quality assessment of the improved slope estimates from multiple regression of parameters,
as described in Section 3.2.6.2.2 is done in two ways. First, a comparison with the initial slope
estimation (from direct regression of backscatter from incidence angle) is done, checking for
consistency between slope estimates from initial and improved method and how they impact on
backscatter normalisation. Second, a comparison of the improved slope with slope estimates from
the Envisat ASAR Wide Swath (WS) data heritage is performed.
3.3.4.1 Comparison to initial estimate

It is of interest how the improved slope estimations influence the backscatter normalisation in
comparison to the initial slope estimations. To assess this, Sentinel-1 observations from October
2014 to October 2016 are normalized (Equation (3)) by using 1) the initial slope and 2) the
improved slope. Subsequently, the average RMSD between the two normalized backscatter image
stacks is calculated. For a 1.2 Mio km² land area in central Europe, an average RMSD value for
normalised backscatter of 0.08 dB is measured. Considering the sensor accuracy at the scale of
1km, estimated to 0.2 dB, the RMSD is acceptably low. Thus, in respect to the accuracy level, the
estimates from the initial method are replicated successfully with the slope from the multiple
regression method.
3.3.4.2 Comparison to Envisat ASAR WS

Envisat ASAR provided comprehensive SAR observations at 150m resolution over Europe for the
years 2004 to 2012. A comparison of slope estimates from ASAR Wide Swath (WS) and the
improved Sentinel-1 slope estimates over land pixels in the Belgium area is shown in Figure 13.
Although observations stem from different sensors, observing at different incidence angles and a
different period, the slope estimate patterns correspond sufficiently and strengthen our confidence
in the slope estimated via multiple regression. The difference image is dominated by physical
patterns, and not by coverage patterns from Sentinel-1. The satellite coverage pattern governs the
selection of areas where the initial slope estimate is possible (Figure 13: green; not possible in
grey areas). In these two classes, the average root-mean-square-difference (RMSD) between
slopes from ASAR and S-1A is calculated. For the green areas, this is 0.032 dB/° and for the grey
areas 0.033 dB/°. Extending the same analysis to complete central Europe yields 0.032 dB/° and
0.034 dB/°, respectively. Thus, the relation of the regression-model-slope of Sentinel-1 to the
reference-slope of ASAR appears to be stable and consistent for both area classes. The
Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium
Issue: I1.20 Date:29.11.2018 Page: 33 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

regression model can be considered as suitable to estimate slope values even in areas where no
initial slope estimation is possible. For the slope difference map, ASAR slope values have been
scaled linearly to the Sentinel-1 value domain because we focus on patterns, and not on absolute
slope levels (which are different for ASAR and S-1) in this comparison.

Figure 13: From upper left to lower right: Number of observations per pixel from S-1A; areas where
initial S-1A slope estimate is valid; S-1A slope from regression; Enivsat ASAR WS slope; difference
map of initial ASAR and S-1A regression slope.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 34 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

3.4 RISK OF FAILURE AND MITIGATION MEASURES

3.4.1 Sensor failure

While sensor failure is always a risk, the probability for complete sensor failure is low since
there are two S-1 satellites in orbit. Failure of one sensor will affect the spatio-temporal
coverage of the product, potentially reducing the range of applications to SSM1km can be
used for. A mitigation strategy for this kind of failure could be a reduction of the temporal
resolution of the output product to e.g. a decadal product, to get better data coverage.
Mitigation strategies for a failure of both satellites do not exist since S-1 like data is not
produced by other satellites at this time.

3.4.2 Quality degradation

If the quality of the retrievals is degraded, the mitigation options depend on two main factors.
1. Can the quality-degraded data be detected automatically?
2. Is it limited to a certain region?
If one of these questions can be answered positively then procedures can be implemented to reject
the affected data or raise the error level appropriately. If automatic detection is not possible, then
raising the overall error level of the input data can be an option. In the worst case all data of a
sensor would have to be rejected.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 35 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

4 REFERENCES

Bauer-Marschallinger, B., Freeman, V., Cao, S., Paulik, C., Schaufler, S., Stachl, T., Modanesi, S.,
Massari, C., Ciabatta, L., Brocca, L. and Wagner, W., 2018. Toward Global Soil Moisture
Monitoring With Sentinel-1: Harnessing Assets and Overcoming Obstacles. IEEE
Transactions on Geoscience and Remote Sensing, (99), pp.1-20.

Dorigo, Wouter, and Richard de Jeu. 2016. “Satellite Soil Moisture for Advancing Our
Understanding of Earth System Processes and Climate Change.” International Journal of
Applied Earth Observation and Geoinformation 48: 1–4.

Dostálová, A., Doubková, M., Sabel, D., Bauer-Marschallinger, B., Wagner, W., 2014. Seven years
of Advanced Synthetic Aperture Radar (ASAR) Global Monitoring (GM) of surface soil
moisture over Africa. Remote Sens. 6, 7683–7707.

Doubková, M., Van Dijk, A.I., Sabel, D., Wagner, W., Blöschl, G., 2012. Evaluation of the predicted
error of the soil moisture retrieval from C-band SAR by comparison against modelled soil
moisture estimates over Australia. Remote Sens. Environ. 120, 188–196.

Entekhabi, Dara, Eni G Njoku, Peggy E O’Neill, Kent H Kellogg, Wade T Crow, Wendy N
Edelstein, Jared K Entin, et al. 2010. “The Soil Moisture Active Passive (SMAP) Mission.”
Proceedings of the IEEE 98 (5). IEEE: 704–16.

Hasenauer, S, W Wagner, K Scipal, V Naeimi, and Z Bartalis. 2006. “Implementation of near Real
Time Soil Moisture Products in the SAF Network Based on METOP ASCAT Data.” In
Proceedings EUMETSAT Meteorological Satellite Conference, 12–16.

Hornáček, M., Wagner, W., Sabel, D., Truong, H.-L., Snoeij, P., Hahmann, T., Diedrich, E.,
Doubková, M., 2012. Potential for high resolution systematic global surface soil moisture
retrieval via change detection using Sentinel-1. Sel. Top. Appl. Earth Obs. Remote Sens.
IEEE J. Of 5, 1303–1311.

Jarvis, A., Reuter, H.I., Nelson, A., Guevara, E., others, 2008. Hole-filled SRTM for the globe
Version 4. Available CGIAR-CSI SRTM 90m Database Httpsrtm Csi Cgiar Org.

Kerr, Yann H, Philippe Waldteufel, J-P Wigneron, Steven Delwart, François Cabot, Jacqueline
Boutin, M-J Escorihuela, et al. 2010. “The SMOS Mission: New Tool for Monitoring Key
Elements Ofthe Global Water Cycle.” Proceedings of the IEEE 98 (5). IEEE: 666–87.

Mladenova, I., Lakshmi, V., Walker, J.P., Panciera, R., Wagner, W., Doubkova, M., 2010.
Validation of the ASAR global monitoring mode soil moisture product using the NAFE’05
data set. Geosci. Remote Sens. IEEE Trans. On 48, 2498–2508.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 36 of 37
Copernicus Global Land Operations – Lot 1
Date Issued: 29.11.2018
Issue: I1.20

Pathe, C., Wagner, W., Sabel, D., Doubkova, M., Basara, J.B., 2009. Using ENVISAT ASAR global
mode data for surface soil moisture retrieval over Oklahoma, USA. Geosci. Remote Sens.
IEEE Trans. On 47, 468–480.

Small, D., Schubert, A., 2008. Guide to ASAR geocoding. Issue 1, 2008.

Wagner, W., 1998. Soil moisture retrieval from ERS scatterometer data. European Commission,
Joint Research Centre, Space Applications Institute.

Document-No. CGLOPS1_ATBD_SSM1km-V1 © C-GLOPS1 consortium


Issue: I1.20 Date:29.11.2018 Page: 37 of 37

You might also like