Professional Documents
Culture Documents
Next ESA SAR Toolbox (NEST) : A Cookbook
Next ESA SAR Toolbox (NEST) : A Cookbook
Cover illustration adapted from a SAR interferogram of the July 2007 earthquake in Tanzania, courtesy of Anneleen Oyen, TU Delft
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
TABLE OF CONTENTS
1
GETTING STARTED .............................................................................................................................. 3
4 ORTHORECTIFICATION .................................................................................................................... 18
REFERENCES .............................................................................................................................................. 31
ACKNOWLEDGEMENTS ............................................................................................................................ 31
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
The Next ESA SAR Toolbox (NEST) is a user-friendly open-source Java application used for reading,
post-processing, analysing and visualising Synthetic Aperture Radar (SAR) data from ESA missions.
These include ERS 1 and 2, Envisat, as well as third-party SAR data from JERS, ALOS, TerraSAR-X,
Radarsat 1 and 2, and Cosmo-Skymed. NEST helps the remote sensing community by handling ESA
SAR products and complementing existing commercial packages.
This cookbook is an attempt at providing familiarity with using NEST for basic processing of SAR
images.
Installation steps:
• Create account at the NEST website (http://earth.esa.int/nest). You will receive a password by
email. Use this to login to the website.
• After logging in, go to the Downloads link and click on the latest software version available for
download. This cookbook uses Pre-Release 4B-0.5. Download the installer file compatible with
your computer, e.g. NEST-4B-windows-installer.exe on a Windows 32-bit computer.
• Double-click on the downloaded installer file and follow the setup steps. Choose an appropriate
installation path (e.g. C:\Program Files\). This should take a few minutes to complete.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Copy the NestData folder from the USB stick to your local computer, for example to C:\.
• Data paths must be specified via Edit → Settings. Set the AuxDataPath to C:\NestData, or
to wherever you copied the NestData folder. Setting this should automatically set the paths to
DEMs and orbit files correctly.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Our first example is the Great Manmade River (GMR) project in Libya. Libya is a desert country, and
finding freshwater has always been a problem. Industrialisation put even more strain on water supplies,
even as coastal aquifers became contaminated with seawater.
The GMR is a network of pipes that supplies water to the Sahara Desert in Libya, from the Nubian
Sandstone Aquifer System fossil aquifer beneath Libya’s southern desert. It is the largest underground
network of pipes and aqueducts in the world, and consists of more than 1,300 wells, most more than 500
m deep. It supplies ~6,500,000 m³ of fresh water per day to coastal cities such as Tripoli, Benghazi and
Sirt.
Left: A map of the Great Manmade River project in Libya. Middle: Excavation work in the desert. Right: An underground pipe.
We shall use Envisat ASAR images from 2004, 2008 and 2009 to study this area. Take a look at the
Google Earth image of this area, below. A network of pipes is clearly visible.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• File → Product Readers gives a variety of products to choose from. For our example, click on
Open ENVISAT ASAR Product.
• In the resulting dialogue-box, locate C:\NestData\Libya\ and look at the Envisat Level 1
product in that folder (ASA_IMS_1PNUPA20040605_202846_000000162027_00272_11849_4776.N1).
Take a careful look at the file name.
Question: Study the Metadata associated with the image, and answer the following questions:
a) In approximately which direction is the satellite flying?
b) What are the latitude and longitude coordinates of the 4 corners of the image?
c) What is the approximate area covered on the ground?
d) How much time did it take to acquire this image?
e) What is the polarization? How can this be interpreted?
f) What is the wavelength of the radar signal in millimetres?
g) What is the time between subsequent radar pulses?
h) In which direction does the image have better (higher) spatial resolution: azimuth or range?
• The various bands of the SAR image can be visualised by double-clicking on the corresponding
band. Observe the Intensity band.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Question: Does it look like the image from Google Earth? If not, why?
• After you have opened an Image View you can modify the colours of the image using the
Colour Manipulation window.
• Further information (including geocoding) about the SAR image can be viewed by clicking on
Analysis → Information.
1.2.3 Multilooking
Generally, a raw SAR image appears speckled with inherent speckle noise. To reduce this inherent
speckled appearance, NEST can perform space-domain multilooking of the SAR image, by averaging a
single look image with a small sliding window.
When you compared the SAR image with the Google Earth image, you probably noticed that the SAR
image looks compressed along the horizontal direction. Multilook processing can also be used to produce
an application product with nominal image pixel size.
NOTE: Multilooking is fairly computationally-intensive, so skip the rest of this subsection 1.2.3 if you
are using a slow computer.
• Observe that a new product is visible in the Products view. Visualise the Intensity_VV band of
this new product.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Question: Does this image now resemble the Google Earth image?
• Open Raster Product can also be used to open Level 1 SAR data, or any other raster data. Here,
we will use it to open multilooked data from the previous subsection.
• If you did not perform multilooking in the previous subsection, go to File → Open Raster
Product, and locate C:\NestData\Libya\ in the resulting dialogue-box. Click on the
multilooked data file ENVISAT-ASA_IMS_1PNUPA20040605_202846_000000162027_00272_11849_4776.N1_ML.dim.
This is the output file created in the multilooking step (Subsection 1.2.3).
• In the Products view, visualise the Intensity_VV band of the multilooked data.
Creating a project will help organise your data by storing all related work in one folder.
• Click on File → New Project. A dialogue-box will prompt you for a project folder location and
project file name. Call your project Project1.xml and place it in C:\NestData.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Sometimes, one may not be interested in the entire image of a product. Using the subset operator, you
may specify an area of the product to be loaded. The area can be selected by entering the top left corner
and the width and height. A sub-sampling in the X or Y directions can also be specified.
• In the Project view, right-click on Processed Products and click on Create Folder. Call the
new folder Subsets.
• Click on Utilities → Spatial Subset from View, and set the X and Y values as shown below.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Select the new product, Subset_0, in the Products view and save it in BEAM DIMAP format
into the Subsets folder.
• Similarly, create subsets of the other three multilooked data files and save them in BEAM
DIMAP format into the Subsets folder.
Question: Four new subsets are visible in the Products view. Visualise these subsets. Do they all
look the same?
• We will continue further processing only on these subsets, so right-click individually on all the
other ML.dim files and Close them.
The orbit state vectors provided in the metadata of a SAR product are generally not very accurate and can
be refined with the precise orbit files which are available days to weeks after the generation of the
product. The orbit file provides accurate satellite position and velocity information. Based on this
information, the orbit state vectors in the abstract metadata of the product are updated.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
For Envisat ASAR products such as ours, DORIS precise orbit files generated by the Centre de
Traitement Doris Poseidon (CTDP) and Delft University of Technology can be applied.
• In the Project view, create a folder called ORBIT within the Subsets folder.
• Select the first subset in the Products view and click on SAR Tools → Apply Orbit File. Set
the target directory to the ORBIT folder.
• Similarly, apply orbit file to the other three subsets. Four new AppOrb products are now visible
in the Products view; retain these and close all the others.
The objective of SAR calibration is to provide imagery in which the pixel values can be directly related to
the radar backscatter of the scene. Though uncalibrated SAR imagery is sufficient for qualitative use,
calibrated SAR images are essential for quantitative use of SAR data.
Typical SAR raw data processing, which produces Level 1 images, does not include radiometric
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Select the first product in the Products view and click on SAR Tools → Radiometric
Correction → Calibrate. Set the target directory to the Calibrated Products folder.
• Similarly, apply the calibration operator to the other three subsets. Four new Calib products are
now visible in the Products view; retain these and close all the others.
2.4 Coregistration
Images obtained from different measurements may be coregistered such that the data is transformed into
one coordinate system. NEST accurately coregisters two or more input images by spatial registration and
resampling to correct translational, rotational and scale differences. One image is selected as the master
image while the others are slave images. Correlation metrics are used to automatically relate the slave
images to the master image.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Click on SAR Tools → Automatic Coregistration. Coregistration processing may take a few
minutes.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• In the Products view, retain the coregistered_stack and close all the other products.
• In the Products view, right-click on the coregistered_stack and click on Open RGB Image
View.
• Set the ‘clean’ SAR images of June 2004, July 2004 and May 2009 as the red, green and blue
channels respectively. Observe the resulting RGB image.
• Right-click on the first band of the coregistered_stack in the Products view and click on
Linear to/from dB. Repeat this for the other three bands. Four new db bands are thus created.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Open RGB image view and set the db SAR images of June 2004, July 2004 and May 2009 as
the red, green and blue channels respectively.
Question: Look the resulting RGB image and compare this with the linear-scale RGB image. What
do you observe?
3 SPECKLE FILTERING
SAR images have inherent salt-and-pepper like texturing called speckle, which degrades the quality of the
image and makes interpretation of features more difficult. Speckle is caused by random constructive and
destructive interference of the de-phased but coherent return waves scattered by elementary scatterers
within each resolution cell. Speckle noise reduction can be applied either by spatial filtering or multilook
processing.
• Create a new folder called Single_speckle_filter under Coregistered Products in the Project
view.
• Go to SAR Tools → Speckle Filtering → Single Product. Choose the linear-scale images
from June 2004, July 2004 and May 2009 as the source bands.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
The band arithmetic tool is used to create new image sample values derived from existing bands, tie-point
grids and flags. The source data can originate from all currently open and spatially compatible input
products. The source data is combined by an arithmetic expression to generate the target data. By default,
a new image view is automatically opened for the new sample values.
We will use the band arithmetic tool to average the three bands within our stack of coregistered single-
speckle-filtered images.
• Go to Utilities → Create Band by Band Maths and perform averaging as shown below.
Question: Visualise the new AVERAGE band that is created. Also, convert the AVERAGE band
into the dB scale and visualise the output. Compare AVERAGE_db with the db images from the
coregistered_stack.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Use the band arithmetic tool to average the three bands within the stack of coregistered
multitemporal speckle-filtered images.
Question: Visualise the new AVERAGE band that is created, and compare with the AVERAGE
band created using single product speckle filtering. Which do you think is better? Also, convert the
AVERAGE band into the dB scale and visualise the output.
4 ORTHORECTIFICATION
Due to topographical variations of a scene and the tilt of the satellite sensor, distances can get distorted in
SAR images. Image data not directly at the sensor’s nadir location will have some distortion, as depicted
below. Terrain corrections are intended to compensate for these distortions so that the geometric
representation of the image will be as close as possible to the real world.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
NEST uses the available orbit state vector information, the radar timing annotations, the slant to ground
range conversion parameters together with the reference DEM data to derive the precise geolocation
information.
In order to demonstrate orthorectification, we will use a different study area that contains much more
topography than the GMR in Libya. We choose our second study area as the Swiss-French Alps in
Europe, around the region of Lake Geneva. The Google Earth image of this area is shown below, along
with a view of Lake Geneva and the Chablais Alps from Caux.
Graph Builder allows the user to assemble graphs from a list of available operators and connect operator
nodes to their sources. Right-clicking is used to add an operator. As operators are added, their
corresponding OperatorUIs are created and added as tabs to a property sheet. The OperatorUIs accept
user input for the operator’s parameters.
Operator graph nodes are connected by moving the mouse over the left edge of an operator node until a
circle appears. The mouse is dragged over the operator node one wishes to designate as the source. Every
node except for a Reader node will require a source.
In this example, we will first perform precise orbit correction, multilooking, absolute calibration and
speckle filtering on three images from around Lake Geneva. We will build a graph for this and perform
batch processing.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Open the three Envisat images within C:\NestData\Alps and take a look at them.
Question: Observe how the mountains look, and try to locate Lake Geneva. Based on how the
mountains appear, would you say that the satellite track is ascending or descending? Why?
• Apply orbit file by right-clicking in the dialogue box as shown below, and set the orbit type to
DORIS Precise (ENVISAT).
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Link all the operation boxes with arrows (by clicking and dragging with the mouse). Ensure that
the target directory is Calibrated Products of the current project, as shown above.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
The processing chain has been defined for one product, and has been saved as Graph_Alps.xml. We will
now run this chain over the three Envisat images of the dataset via batch processing.
• Go to Graphs → Batch Processing, and click on Add All Open to add all the open Envisat
images of the Alps to the batch processing list.
Question: Click on Run, and compare the output images with the original Envisat images. Do you
notice the differences?
4.3 Coregistration
As a next step, we will perform coregistration of the images, range-Doppler terrain correction, and
then visualise them.
• Perform automatic coregistration as shown below. Add the Calibrated Products as inputs,
selecting the image of 12 Jan 2008 as master. Select the target directory to be
C:\NestData\Project2\Processed Products\Coregistered Products, and click on Process.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Question: Visualise the coregistered stack of images in RGB bands, both in linear and dB scales.
We will now perform range-Doppler terrain correction on the images, using the SRTM (Shuttle Radar
Topographic Mission) DEM of the region. The Range Doppler Terrain Correction Operator
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Under Processed Products in the Project View, create a new folder called TerrainCorrected.
• Click on Geometry → Terrain Correction → Range-Doppler Terrain Correction, and Run
as shown below.
Question: Visualise the RD terrain-corrected stack of images in RGB bands, both in linear and dB scales.
Also, using File → Product Writers → Export View as Google Earth KMZ, export an RD terrain
corrected image into .kmz format and visualise it on Google Earth.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Coregistration: The simulated SAR image (master) and the original SAR image (slave) are coregistered and
a WARP function is produced. The WARP function maps each pixel in the simulated SAR image to its
corresponding position in the original SAR image.
Terrain correction: Traverses the DEM grid that covers the imaging area. For each cell in the DEM grid,
its corresponding pixel position in the simulated SAR image is computed using SAR model. Then its
corresponding pixel position in the original SAR image can be found with the help of the WARP
function. Finally the pixel value for the orthorectified image can be obtained from the original SAR image
using interpolation.
• Click on Geometry → Terrain Correction → SAR-Simulation Terrain Correction, and Run
it on a single coregistered image (not on the stack of coregistered images).
Question: Compare the SAR-simulation terrain correction result with that of RD terrain correction. What
differences do you observe? Also, using File → Product Writers → Export View as Google Earth
KMZ, export the SAR-simulation terrain corrected image into .kmz format and visualise it on Google
Earth.
5 SAR POLARIMETRY
Radar polarimetry deals with the full vector nature of polarized electromagnetic waves. It is the science of
acquiring, processing and analysing the polarization state of a radar signal. The polarization information
contained in the radar waves backscattered from a given medium is highly related to its geometrical
structure, reflectivity, shape and orientation, its geophysical properties such as humidity, roughness, and
so on.
A simple way of utilizing this property is by performing simple arithmetic operations on the four
polarization channels (HH, HV, VH and VV) of quad-pol radar data. These operations result in the
following physical interpretation:
|HH+VV| |HH-VV| |HV|
In our third example, we will use quad-pol ALOS PALSAR data over the city of San Francisco in the
USA to demonstrate that the above property can be used for simple land use classification.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Question: Observe the different polarization bands in this product. Do they look different? Can you
locate the city of San Francisco clearly in all the bands?
• Edit the expression on the red channel to abs(Intensity_HH - Intensity_VV) using the Edit
Expression operator symbol beside the red channel dialogue box, as shown below.
• Similarly, edit the expression on the green and blue channels to abs(Intensity_HV) and
abs(Intensity_HH + Intensity_VV) respectively. The final RGB channel configuration is
shown below. Click on OK.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
Question: Observe the radar image above. Can you see where there is water? Can you locate urban and
rural areas in the image? Can you locate the city of San Francisco? Zoom into it (shown below). Can you
differentiate between buildings, roads and vegetation?
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
The interferogram formation operator computes (complex) interferogram, with or without subtraction
of the flat-earth (reference) phase. The source bands are the set of (usually coregistered) bands of the
complex product. This set is resampled prior to interferogram formation.
Our fourth and final example is the Bam earthquake in Iran on 26 Dec 2003. It was a 6.6 Mw earthquake
that was particularly destructive, with the death toll amounting to 26,271 people and an additional 30,000
injured. About 60% of the buildings in Bam were destroyed. The old quarter and a 2,000-year-old citadel
(severely damaged by the earthquake, shown below) were built primarily of mud brick.
A pair of Envisat SAR images were acquired before (3 Dec 2003) and after (11 Feb 2004) the earthquake,
and their interferogram therefore showed the ground motion in that period.
• Create a folder called Interferograms within the Calibrated Products directory of the current
project.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
• Select the Phase_ifg1 band of cplx_ifg and perform multilooking. Save the output cplx_ifg_ML
to the Interferograms folder.
Question: Using the colour manipulation feature, try to fit a colour scheme to the fringes, such that
each fringe cycles through a set of colours.
Question: There appear to be two main fringe loops in the image. How many fringes can you count in
each loop? How many other fringes are there? How many millimetres of displacement do these fringes
correspond to? In which direction?
Optional question: Create a coherence map between the two SAR images in the coregistered stack.
Tiger Capacity Building Facility II
Training Course on Active and Passive Microwave Remote Sensing
REFERENCES
‘NEST User Training Course’ by Luis Veci, Andrea Minchella, Marcus Engdahl and Petar Marinkovic,
25-27 November 2009, ESA-ESRIN, Frascati, Italy.
‘Polarimetric SAR data analysis’ by Eric Pottier, University of Rennes 1, France (tutorial given on 2 May
2011 at the Ecole de Physique des Houches, Chamonix, France).
ACKNOWLEDGEMENTS
The authors would like to thank ESA for providing the SAR data used in this cookbook, and Petar
Marinkovic of PPO Labs for his technical help and support.