InSAR TECHNICAL BACKGROUND

InSAR  stands for Interferometric Synthetic Aperture Radar. Although it is a high-tech method to measure deformation, its principles are fairly simple. This guide explains how we measure millimeter-scale movements and what you can do with the data.

Thousands of satellites are currently orbiting the earth. The images obtained from these satellites are able to provide an understanding of many parameters about the Earth’s surface. Passive satellite systems measure the radiation from the sun that is reflected to the satellite off the Earth to record information about the Earth, whereas active satellite systems, such as radar, sends a signal to the Earth’s surface from the satellite and measures the signal that is reflected.

Radar images acquired from these active satellite systems provide information used for measuring deformation. Radar is able to transmit its signals through clouds, and is independent the radiation from the sun, giving it large advantages over passive satellite systems.

1. InSAR

1.1 What is InSAR?

InSAR (Interferometric Synthetic Aperture Radar) is a technique that maps millimeter-scale deformations of the earth’s surface with radar satellite measurements. Given the continuous change of the Earth’s surface, the ability to yield measurements at night and throughout any weather condition makes this technique extremely valuable.

1.2 Measuring deformation

The satellite measures both the amplitude and the phase of the radar signal reflected from the Earth’s surface. The amplitude is the strength of the signal that is recorded, and the phase is the fraction of a complete waves cycle that reaches the sensor. The phase measurement is extremely important for measuring deformation. A difference in phase between two sequential measurements means that something has changed. The figure below shows how we derive deformation from a difference in phase. If the Earth’s surface subsides, the emitted radar signal has to travel a further distance to reach the surface. This results in an extra fraction of the radar wave being reflected and recorded, known as the phase difference (shown in red). The length of a complete wave cycle is in the order of centimeters and differs depending on the satellite. Because this length is known very precisely, we can determine the surface deformation with millimeter precision.

Fig 1: InSAR measurements before and after subsidence occurs.

1.3 Relative measurements

Reference point
The deformation measurements obtained are very precise because of the accuracy of the phase measured by InSAR. However, the measurement of the absolute distance in which the surface deformed is less accurate. To account for this uncertainty on the absolute distance, all measurements are taken with respect to a reference point. We can then make optimal use of the high precision measurements of InSAR and provide deformation values with millimeter precision. The reference point is generally placed at a location that is stable in terms of deformation. This makes interpreting the deformation map easier. However, since the earth is constantly changing, even the most stable locations may experience deformation. To overcome this problem, measurements should always be interpreted relative to each other. Any deformation of the reference point will affect the deformation values of the other measurement points. For example, if the reference point is subsiding, all stable points will appear to ascend. On the other hand, if such an effect is not visible in the deformation map, the reference point was probably well chosen.

Relative in time
In our deformation map we provide time series for each data point. The starting point of the time series has a deformation value of 0. All of the following deformations are given with respect to this first measurement.

1.4 Persistent Scatterers and Distributed Scatters

Scatterers
The resolution of the data determines the amount of measurements that can be performed. This can be compared to a photo camera; within each pixel, only one value can be obtained. However, multiple objects reflecting the signal may be present inside the pixel, so called scatterers. Depending on the reflectance properties of the objects within each pixel, different methods are used to measure the deformation. In an ideal world, each pixel only contains one scatterer, which means the measured signal will contain little disturbances. In reality, however, pixels hold multiple scatterers. The two main types of scatterers are Persistent Scatterers (PS) and Distributed Scatterers (DS).

Persistent Scatterers (PS)
Persistent Scatterers are pixels that contain objects that have high reflectance properties and are time-coherent, meaning the reflectance remains constant throughout time. These objects reflect the majority of the signal received by the satellite from a certain pixel. Persistent Scatterers are commonly found in urban areas, such as on buildings or other man-made structures. When a pixel contains one scatterer that reflects much stronger than the others, the weaker scatterers are almost negligible for the measured values. In other words: the signal to noise ratio is high. This is depicted in the image below, where the red signal is stronger than the rest of the signals.

Fig 2: Persistent Scatterer; the pixel contains one dominant reflection.


Distributed Scatterers (DS)

Distributed Scatterers are pixels where multiple objects show a weaker reflection, at a relatively similar strength. If a pixel is classified as a Distributed Scatterer, our algorithm will search for neighboring pixels with similar reflection characteristics. If multiple neighbors show comparable reflections, the relatively ‘weak’ signal is seen as trustworthy because the signal to noise ratio again becomes larger. The derived deformation, however, is not from a single reflector, but from a larger, homogeneous area. Examples of this are a piece of road or the prepared substructure of a train track. The image below shows an example of a distributed scatterer pixel.

Fig 3: Distributed Scatterer; the pixel contains multiple similar reflections. 

1.5 Horizontal / vertical deformations

Oblique satellite viewing angle
The satellite is observing the earth under an angle with respect to the vertical, generally between 20 and 40 degrees. We measure deformations on the Earth’s surface in the viewing direction of the satellite, while the actual deformation may have both a vertical and a horizontal component. Depending on the type of deformation problem we can often make an assumption on the direction of the actual deformation. For example, it is reasonable to assume that the deformation of a street is predominantly vertical. In that case, we can recalculate the measured deformations as if they were purely vertical. These and other assumptions are always made in close consultation with the customer.

Decomposition by using multiple satellites

Because of the orbit of the satellite, it passes a certain location both from an ascending orbit (which is more or less northbound) as well as from a descending orbit (which is more or less southbound). This is explained in further detail in the satellites section. Because of its fixed viewing direction (generally to the right) this means images are acquired from both the west by south and the east by south. For larger objects that contain multiple data points in both the ‘ascending’ and ‘descending’ orbits, and in combination with at least one independent assumption, a decomposition of the measured deformation can be done in the vertical and horizontal direction. Note, however, that due to the flying direction of the satellite, the measurements are much less sensitive to north-south movements.

Fig 4: Horizontal/vertical decomposition of the measurement signal for line infrastructure, such as a levee. (Note we assume that there is no significant deformation in the longitudinal direction)

The left figure shows the deformation signal when using one satellite orbit. The measured deformation in the satellite’s line-of-sight (black arrow) can be caused by all ‘real’ deformation vectors along the green line (assuming there is no deformation in the longitudinal direction of the levee). As an example, the vertical and horizontal-transversal deformations are shown in red (arrows).

 

The right figure shows the vector decomposition when using two satellite orbits (‘ascending’ and ‘descending’). Again, the measured deformations by the satellite are depicted in black. However, this time, under the same assumptions, there is an intersection point between the two green lines along which the ‘real’ deformation vector is located. The deformation that occurs in this example is not only vertical, but has a slight horizontal-transversal component which is depicted by the yellow arrow.

1.6 Interferograms

Interferograms are the result of subtracting the phases of two SAR images, a process called Differential-InSAR (D-InSAR). An example of an interferogram is shown in the figure below.


Fig 5: Interferogram.

The phase difference is depicted by the color of the pixels in an interferogram. The interferogram above shows repeating color cycles, which are clearly visible in the northern part. Because the same phase value is repeated in every cycle of the radar wave, the colors repeat. The following figure zooms in on the area.

Fig 6: Wave cycles in an interferogram.

Deriving Deformation
It can be seen that the same color cycle (from red to blue to red) repeats three times, in the direction of increasing phase. This means that the northern part has subsided three times a full wavelength with respect to the southern part between the two image acquisitions that were used to produce the interferogram. Because we know the exact length of a full wave cycle, the deformation can be derived. This process is called unwrapping.

However, if there was a strong atmospheric activity during one or both of the image acquisitions used for producing the interferogram, the deformation signal can be polluted with atmospheric noise. In order to resolve deformation from atmosphere, a time series of interferograms can be used to model the atmospheric signals. This is further explained in the article about atmosphere.

2. Measurement Locations

How suitable a certain location is for performing our measurements depends on how well the objects reflect the radar signal, whether the reflection is consistent enough in time to produce a measurement time series and the measurement resolution.

2.1 Reflecting objects

Not all objects on the Earth’s surface reflect a proper signal back to the satellite, as illustrated in the figure below. The shape, orientation and structure of a surface influence the way it reflects the signal. The best reflections come from angular objects such as buildings and other solid infrastructure. However, when the radar signal bounces of a smooth surface, such as water during windless conditions, it acts as a mirror and almost no signal is received by the satellite. By using satellites that emit a short wavelength, we are able to obtain a high measurement point density on roads, which are also relative smooth surfaces.

Fig 7: Radar signal reflecting on the earth’s surface.

2.2 Consistent reflections

By combining a series of satellite acquisitions we can track the deformation of a certain location over time. This can be done under the condition that the reflection is roughly the same in every satellite image, so the location remains recognizable. This means that changing objects, such as vegetation, are not suitable for our measurements. Buildings, infrastructure and other hard surfaces often do reflect consistently throughout time. These objects are called coherent reflectors and form the basis for our precise deformation measurements. Depending on the satellite and area, it is possible to obtain thousands to tens of thousands of measurement points per square kilometer.

Measuring the same object over time

Every time the satellite revisits a location, it measures exactly the same object as the previous times. This is because of the consistent reflections of suitable objects, and the fact that we can relate satellite images on a sub-pixel level. If our algorithm is uncertain about the consistency of a reflection, the entire measurement point is removed from the deformation map. This is the reason that not every object contains measurement points.

2.3 Measurement resolution

Resolution can be defined as the size of the area over which the satellite gathers one measurement value. A higher InSAR measurement resolution results in a large increase in the number of measurement points. The figure below compares a typical SkyGeo standard resolution product (left) with a typical SkyGeo high resolution product (right).

Fig 8: Left: standard resolution, right: high resolution.

This increase in the number of points can be explained by the fact that multiple reflecting objects that were first measured together in one resolution ‘cell’ (left), now have their own separate ‘cell’. Because the satellite can only measure one value per cell, we acquire more reflections with a higher resolution. This is illustrated in the figure below.

Fig 9: Capturing more reflections with a higher resolution.

Using higher resolution data has three advantages:

  • There is less room for noise in smaller cells so the measured value in the cell is less noisy and it is therefore more likely that a reflector shows a consistent reflection throughout time (see also Consistent reflections). This leads to even more measurement points.
  • The measured deformation values are more precise.
  • The absolute positions of the measurement points are more precise (cf. Location precision).

2.4 Point location (X,Y,Z)

The satellite measures the deformation that occurs within a certain resolution cell (see also Measurement resolution). However, the major part of the signal comes from the well reflecting object(s) within that cell. Our algorithm estimates the X,Y location and the height (Z) of this dominant reflection. This reflector is visualized as a measurement point with a certain height in the SkyGeo deformation map. The precision of the measurement point location depends on the satellite resolution and is shown in the table below.

Table: Point location precision
X,Y Z
Standard resolution 2-3 m 2 – 2.5 m
High resolution 1-2 m 1-1.5 m

3. Measurement Quality

Quality is inherent to measuring; how precise and how reliable is the data? With InSAR we measure the deformation at certain locations, but there is an uncertainty in the determined deformation as well as in the location that this deformation occurs.

3.1 Measurement precision

Individual measurements

Because the InSAR technique uses the phase of the radar signal (see Measurement principles), the deformation value at one location can be measured with high precision. This precision depends on the weather during the time of measurement and the resolution of the satellite. An indication of the precision of our individual measurements in regards to resolution are shown in the table below.

Table: Individual measurement precision
Individual measurement precision
Standard resolution 6-8 mm
High resolution 2-3 mm

 

Estimated linear trend

The colors in the SkyGeo Interactive Map generally represent the deformation velocity of the point, estimated by fitting a linear trendline through the individual measurements over time. The precision of this deformation velocity, assuming the deformation is purely linear, depends on the satellite and the length of the time series. If there are more measurements, the estimated velocity is more precise. An indication of the precision of our deformation velocity estimates are shown in the table below.

Table: Deformation velocity precision
Deformation velocity precision
Standard resolution 1-2 mm/yr
High resolution <1 mm/yr

 

3.2 Location precision

Although we can measure deformation very precisely, the absolute X,Y,Z location of the measurement points is less accurate. An indication of the precision of our measurement point locations are shown in the table below.

Table: Point location precision
X,Y Z
Standard resolution 2-3 m 2-2.5 m
High resolution 1-2 m 1-1.5 m

 

3.3 Data reliability – unwrapping

As well as understanding how precise the data is, it is important to know how reliable it is. In order to conduct our deformation measurements, we use the phase recorded by the satellite (see Measurement principles). This is the fraction of a complete wave cycle that is in addition to the number of full wavelengths the signal has traveled.

Phase ambiguity
Phase measurements can be thought of like the time on a clock. There is a finite number of times that can be displayed on a clock and every twelve hours those times repeat. Similarly, there is a finite number of phase measurements that can be recorded within one wave cycle (0 to 2pi). Once a wave cycle has completed, the phase measurements then repeat. This is visualized in the images below, where the phase measurements are shown in a circle, like a clock.

Fig 10: Both clock time and radar phase are ambiguous.

However, the values have a new meaning once a new revolution has started. For a clock, a new revolution means a new 12 hour cycle has started. To fully understand what time it is, one needs to know the date and/or which part of the day it currently is. Similarly, a new revolution on the phase cycle means a full wave cycle has been completed. To fully understand the meaning of the phase measurements recorded, one needs to know the number of full wave cycles associated with the phase measurements. Resolving this to determine the actual deformation is called ‘unwrapping’.

Unwrapping
In order to derive the actual deformation from consecutive phase measurements in time, correlations in time and/or space are used to ‘unwrap’ the measurements. For example, we can assume that the occurred deformation is linear during (a part of) the measurement time span or that the deformation rate change is smooth in space. If a wrong assumption is made, this may lead to so called ‘unwrapping errors’.

In our algorithm we do a strict control on this process to ensure a high degree of data reliability. In practice, this means that for the measurement points in the SkyGeo deformation map the above process is executed correctly at least 99.0% of the time.

3.4 Point Quality

Point Quality is a relative measure of the quality of the respective measurement point. The value is displayed in the metadata of the data point and ranges from 0 to 1. Generally, a lower quality value indicates a higher amount of noise in the time series.

To ensure the quality of the datasets, a minimum quality cut-off is applied before delivery. You can further filter the data points on their quality by using the SkyGeo Maps filter sliders.

Computation
To acquire a quality value, a model reflecting the expected deformation behavior is fit to the measurements within a single data point. This model can for example contain linear, quadratic and seasonal components and is adjusted according to the investigated area. After that, each measurement is compared to the value of the model at that date in time. The difference between the measurement value and the model value determines the Point Quality. A higher value indicates a better match between the model and the measurements.

Interpretation
Point Quality is generally a good indication of the quality of the measurements in the time series. If the model for expected deformation reflects the actual movement behavior of the measurement point, it is a measure of the amount of noise in the time series. However, if you compare a seasonal deformation signal with a linear model, Point Quality is lower than it would be when a seasonal model was used for the comparison, meaning the type of model used affects the outcome. For some points, this effect can lead to a less correct measure of quality.

4. Geostatistics

To provide you with the best information for maintaining your assets, we can apply geostatistics to select and aggregate data points on user-specific assets.

4.1 High/Low Separation

When a building is constructed, generally a foundation is laid to stabilize the building. Because of this foundation, buildings are typically more stable than natural features such as vegetation. In order to distinguish between different deformation regimes, it is often important to separate measurement points located on a building or other infrastructure and measurement points located on the ground. These are referred to as high and low data points, respectively.

Fig 11: Difference between high and low data points

In order to separate the ‘high’ and ‘low’ points, the height obtained from our algorithms for each measurement point are compared to a Digital Terrain Model (DTM) with relatively accurate continuous elevation values. If the height value obtained for a measurement point is more than a set threshold of X m higher than the DTM value at the same location, the point is classified as a high point. If the height value obtained for a measurement point is below a second set threshold of Y m than the DTM value, the point is considered not trustworthy and is discarded. If the difference between the obtained height value and the DTM value falls within these two thresholds, the point is classified as a low point. This is visualized in the diagram below. The values for X and Y mainly depend on the resolution and quality of the satellite data. Common values are X = 2 m for high resolution, X = 3 m for standard resolution and Y = 10 m for both types of resolution.

Fig 12: The thresholds for high/low classification.

4.2 Weighted Averaging

Weighted averaging is a technique to aggregate data points on pipelines by using the relative distance of the points to the underground structure. The first step in the aggregation process is high/low separation. In this step, the data points from the ground surface and from higher infrastructure are separated. This is a way to ensure that the points that are taken into account are from the ground surface. Secondly, virtual points are sampled every X meter along the pipeline course.  After that, the data points in a radius of Y meter are selected and aggregated to the virtual point, scaled by the square of the inverse distance. Common settings are X = 10m and Y = 30m.

An example of weighted averaging on pipeline shapes is shown in the figure below.

Fig 13: Weighted averaging on pipelines

Statistics
A uniformly subsiding pipeline does not need to be a problem for pipeline stability; it is the differential settlements that should be worried about. Therefore, we generally provide pipeline asset managers with the following statistics:

  • Weighted average subsidence
  • Largest subsidence difference within a certain along-track area
  • Largest subsidence within a certain along-track area
  • Percentile values of subsidence

4.3 Shape Aggregation

In this article you will learn about the techniques we use to aggregate the SkyGeo data points on asset shapes like roads, buildings and zip codes, and what assumptions we make in the process. Furthermore, you will learn about the statistics that are derived from the aggregation.

An example of shape aggregation on building shapes is shown in the figure below.

Fig 14: Shape aggregation on buildings.

Aggregation process
The first step in the aggregation process is high/low separation. In this step, the data points from the ground surface and from higher infrastructure are separated. This is a way to ensure that the points that are taken into account are from the height level of the surface of interest. After that, the data points within a distance of 3 m with respect to the asset shape are selected in order to account for the x/y position precision.

Statistics
After the points are aggregated, the statistics are computed per asset shape. The following statistics are available:

  • No data. Only displays the shapes of the data product, without data or colors.
  • Average deformation. Average linear deformation velocity (in mm/yr).
  • Standard deviation. The standard deviation of the linear deformation velocity of the data points that were aggregated in every shape (in mm/yr). The dropdown menu parameters with values behind ‘standard deviation’ only show the shapes with a standard deviation in the specified range.
  • Number of points. The number of data points that were aggregated per shape. The values behind ‘number of points’ specify the range of the color scale that is visualized.
  • Percentile. The percentile values of the linear deformation velocity in the shape (in mm/yr). The X’th percentile indicates the X percent of the data points in the shape, which have a lower deformation (i.e. higher subsidence, since subsidence is negative) value.
  • Default. The default setting for this deformation map. This is generally is the average deformation parameter.

5. Satellites

5.1 Satellite acquisitions

Radar satellites orbit the earth at an altitude between 500-800 km with a velocity of approximately 7.5 km/s. They scan the whole earth in strips and because of their orbits and the rotation of the earth they will repeat the exact same cycle after a certain amount of days, depending on the satellite.


Fig 15: Satellites orbiting the earth. Source: J.C. Szidloski

Because of the orbit of the satellite, it passes a certain location both from south to north (the ascending orbit) as well as from north to south (the descending orbit). Because the satellite looks at the earth under an angle and the viewing direction is generally fixed (mostly right-looking), the earth is observed from two different directions. These two measurements can be combined to resolve horizontal and vertical deformations.

Fig 16: Ascending and descending orbit of the satellite.

Furthermore, because we are dependent on the natural reflections of objects on the earth’s surface, the orbit direction (ascending or descending) affects which objects reflect signal back to the satellite and, thereby, can be measured. Note that due to the different viewing directions and reflections, ascending and descending images cannot be combined into a single InSAR time series.

5.2 Radar satellite characteristics

The characteristics of the radar satellites we most commonly use are visualized in the table below. Note that the actual repeat cycle depends on how the satellite is tasked. The resolution can depend on the satellite settings; the values shown here are what we generally use.

Satellite Time period Repeat cycle (days) Resolution
Sentinel-1 2014-now 12 Standard
TerraSAR-X 2007-now 11 High
COSMO-SkyMed 2007-now 16 High
RadarSat-2 2007-now 24 Standard
ALOS 2006-2011 46 Medium
Envisat 2003-2010 35 Standard
ERS 1992-2001 35 Standard

6. Quality Assessment

We make sure you get data with the highest possible quality. In order to achieve this we perform a thorough internal quality assessment and provide you with a summary report of the steps we took for producing your data. This is explained in more detail in the sections Internal Quality Control and Factual Report.

6.1 Internal Quality Control

This section explains how our engineers make sure you get the best data. It elaborates on both our internal quality control procedure as well as the components that have been checked.

Procedure
Once the responsible engineer(s) is/are satisfied with the product they produced, the internal quality control process commences. This procedure is carried out according to the following steps:

  1. The responsible engineer(s) deliver(s) the product internally.
  2. The product is investigated by a peer engineer and checked against our quality control checklist.
  3. The responsible engineer(s) improve(s) the product based on the comments that were received.
  4. The peer engineer checks if this iteration fulfills our quality standards; if not, step 3 is re-iterated.
  5. The product is delivered to the account manager, who checks whether the product is good enough to send out to the customer; if not, step 3 is re-iterated.
  6. The account manager delivers the product to the customer.

Quality checklist
The data is thoroughly checked on different aspects:

  1. The product is consistent with the offer letter.
  2. The customer username and password to access SkyGeo Maps work.
  3. The quality of the geodetic network is sufficient.
  4. The point quality is sufficient. We check for unlikely values and/or locations and, if applicable, compare the values to those of external datasets.
  5. The noise in the dataset is acceptable. We check and address possible causes of noise, like atmosphere.
  6. The point density is sufficient. We check if the areas of interest are covered properly.
  7. The reference point choice reflects the deformation of the area of interest in the way the customer wants.
  8. The data points are displayed on the right locations on the map (X/Y/Z).
  9. The geostatistical operations are performed correctly.

6.2 Factual Report

In this article you will learn about the Factual Report; a standardized report on the processing steps we took to provide you with your products.

The Factual Report is meant to provide an overview of the processing steps we took; both for interested customers that want to know more about the approach and to ensure future reproducibility of the results.

Downloading the Factual Report
The Factual Report can be downloaded by clicking on the download button in the upper right corner of the Interactive Map.  However, instead of selecting ‘data’, one should select ‘docs’.

Contents
The Factual Report has a default structure, which is shown in the figure below.

The Document Status describes the status of the document. Here, it states for which customer and which viewer the document was produced and at what date. Furthermore, it describes to which InSAR datasets the report applies.

The Area of Interest depicts the area of interest in the form of an outline on a map and an estimation of the size of the area.

The Observation Period chapter describes the satellites that were used to produce the datasets and their main characteristics. Furthermore, the timespan over which the observations were acquired is depicted, together with the median of the time interval between two subsequent acquisitions. The distribution of dates at which a measurement was done, is shown the ‘Temporal baseline’ figure. An example of such a figure is provided below.

Fig 17: Temporal baseline figure.

Data Delivery and File Conventions explains the format of the downloadable files. The names of the layers in the sidebar can be different from the names of the CSV file that can be downloaded; this chapter links the two names. Furthermore, it describes the meaning of the different columns that are delivered in the CSV file.

The Quality Assurance chapter states the points the datasets have been checked on before delivery.

In the Appendices several statistics and properties of the different datasets are summarized:

  • The number of data points
  • The direction in which the delivered values are measured
  • Data averages
  • Data extremes (percentiles)
  • Used deformation models for determining the quality and linear deformation velocity
  • Reference point
  • Type of atmosphere modeling

Atmosphere
Atmosphere is a term we use for disturbances in the measured deformation signal, caused by atmospheric disturbances the signal encounters when traveling from the satellite to the Earth’s surface and back. These disturbances are caused by a combination humidity, pressure and temperature in the atmosphere, delaying the signal. Some areas generally have a more turbulent atmosphere than others. Around the equator, for example, the air contains more water vapor than around the poles, resulting in a more boisterous atmosphere. Next to that, the time of the day also has an influence on the likeliness that disturbances are occurring. The sun heats the Earth’s surface and thereby produces heat and water vapor, increasing atmospheric turbulence.

Recognizing atmosphere
Challenges arise because the atmosphere is different from one satellite acquisition to another. This alters the influence it has on the deformation signal, making it harder to distinguish deformation from atmospheric disturbances. However, precisely because atmosphere is not correlated in time, but deformation is, we have a way to separate the two. Furthermore, atmosphere is correlated in space over certain distances, providing a second possible differentiator from the signal of interest.

 

Fig 18: The decorrelation of atmospheric disturbances in time.

 

Filtering atmosphere
Because atmosphere is spatially correlated over certain distances, it does not form a large obstacle for measuring deformation over small areas. For larger areas, however, it might influence the signal greatly. Therefore, we apply atmosphere estimation in order to resolve the signal of interest. The figure below compares the time series of a certain data point before and after removing the atmosphere from the data.

We perform the atmosphere estimation by choosing high quality data points with a regular deformation pattern (calibration points), either stable or deforming perfectly linear. After that, we assume that all time series measurements deviating from this regular pattern are due to atmospheric noise. We use this noise in a statistical approach to produce an interpolated atmosphere map over the area during all satellite acquisitions. These maps are used to separate the actual deformation signal from the atmospheric disturbances.

Difficulties
The statistical approach improves with the availability of data; when there is more information at hand, it is easier to model the disturbances. Therefore, when there are not a lot of image acquisitions in time, atmosphere estimation can be a challenge. Furthermore, the amount and distribution of calibration points influences the quality of the interpolated atmosphere maps. Large deformation bowls, for example, can sometimes be an obstacle for estimating the disturbances. Because the deformation is far from regular in these areas, calibration points are sparse. Therefore, the atmospheric interpolation has to be propagated over a large distance, potentially containing atmospheric variations and thereby increasing the likeliness of making errors in the estimations.

7. Validation

In the past 50 years, the use of satellite technology to obtain useful information about the Earth’s surface has improved both in understanding and practice. With the advancement in technology and the ability to measure many different parts of the electromagnetic spectrum, the information recorded by satellites can be adapted for a broad range of applications, including land deformation, natural hazard detection/mapping, change detection, land use mapping, urban growth and much more.

Due to the rapid advancements in satellite technology and the relatively recent commercialization of the use of the resulting data, satellite products are fairly new to the public. Our goal is to help you understand why using satellites, and specifically InSAR for determining surface movements, is a viable and accurate procedure.

We have summarized a small batch of the hundreds of scientific journal articles that validate the InSAR technique. Take a look at the following journal articles to see how InSAR is justified as a reasonable measurement technique for deformation.

Here are a few quotes taken directly from the above articles:

“The results show that the measurement techniques, i.e. leveling and InSAR, are in the agreement within the error bounds.”

Marinkovic, P., Ketelaar, G., van Leijen, F., & Hanssen, R. (2007, November). InSAR quality control: Analysis of five years of corner reflector time series. In Proceedings of Fringe 2007 Workshop (ESA SP-649), Frascati, Italy (pp. 26-30).

“The calculated subsidence rates for monuments MAGF, MANI, OSDO, and the subsidence magnitude at COTD calculated using both GPS data and InSAR data compare favorably, thereby improving confidence in results derived from both methods.”

Sneed, M., & Brandt, J. T. (2007). Detection and Measurement of Land Subsidence Using Global Positioning System Surveying and Interferometric Synthetic Aperture Radar, Coachella Valley, California, 1996-2005. US Department of the Interior, US Geological Survey.

“The radar observations of land subsidence are in good agreement with recent levelling and GPS observations. However, the radar data provides a more detailed mapping of both the amplitude and spatial extent of land subsidence.”

Motagh, M., Djamour, Y., Walter, T. R., Wetzel, H. U., Zschau, J., & Arabi, S. (2007). Land subsidence in Mashhad Valley, northeast Iran: results from InSAR, levelling and GPS. Geophysical Journal International168(2), 518-526.

“InSAR and time-continuous GPS deformation data overlap, allowing a direct comparison of the data. The comparisons possible for the Los Angeles basin demonstrate that the two methods produce quantitatively similar results.”

Helz, R. L. (2005). Monitoring ground deformation from space. US Department of the Interior, US Geological Survey.

 

7.1 Summary

To summarize, InSAR has many benefits for monitoring surface deformation. The main benefits to using InSAR are listed:

Reliability
With the high precision phase data that the satellite records, we are able to generate very accurate deformation data.

Simplicity
With the use of satellites, data collection can be done remotely, eliminating in-field measurements, which are more complicated and time consuming.

Low Cost
Eliminating field measurements lowers costs by decreasing the amount of time required to collect data as well as the high priced tools used for surveying.

Ability to measure large areas
By using satellites, large areas are able to be accurately covered in a little amount of time.

Weather independent
Since radar penetrates through clouds, measurements can be taken during all weather conditions, creating a continuous dataset.

Ability to obtain measurements at night
Radar satellites emit their own signal and do not use the emitted radiation from the sun, thereby allowing measurements to be taken at any time of the day.

InSAR TECHNICAL BACKGROUND

InSAR  stands for Interferometric Synthetic Aperture Radar. Although it is a high-tech method to measure deformation, its principles are fairly simple. This guide explains how we measure millimeter-scale movements and what you can do with the data.

Thousands of satellites are currently orbiting the earth. The images obtained from these satellites are able to provide an understanding of many parameters about the Earth’s surface. Passive satellite systems measure the radiation from the sun that is reflected to the satellite off the Earth to record information about the Earth, whereas active satellite systems, such as radar, sends a signal to the Earth’s surface from the satellite and measures the signal that is reflected.

Radar images acquired from these active satellite systems provide information used for measuring deformation. Radar is able to transmit its signals through clouds, and is independent the radiation from the sun, giving it large advantages over passive satellite systems.

1. InSAR

1.1 What is InSAR?

InSAR (Interferometric Synthetic Aperture Radar) is a technique that maps millimeter-scale deformations of the earth’s surface with radar satellite measurements. Given the continuous change of the Earth’s surface, the ability to yield measurements at night and throughout any weather condition makes this technique extremely valuable.

1.2 Measuring deformation

The satellite measures both the amplitude and the phase of the radar signal reflected from the Earth’s surface. The amplitude is the strength of the signal that is recorded, and the phase is the fraction of a complete waves cycle that reaches the sensor. The phase measurement is extremely important for measuring deformation. A difference in phase between two sequential measurements means that something has changed. The figure below shows how we derive deformation from a difference in phase. If the Earth’s surface subsides, the emitted radar signal has to travel a further distance to reach the surface. This results in an extra fraction of the radar wave being reflected and recorded, known as the phase difference (shown in red). The length of a complete wave cycle is in the order of centimeters and differs depending on the satellite. Because this length is known very precisely, we can determine the surface deformation with millimeter precision.

Fig 1: InSAR measurements before and after subsidence occurs.

1.3 Relative measurements

Reference point
The deformation measurements obtained are very precise because of the accuracy of the phase measured by InSAR. However, the measurement of the absolute distance in which the surface deformed is less accurate. To account for this uncertainty on the absolute distance, all measurements are taken with respect to a reference point. We can then make optimal use of the high precision measurements of InSAR and provide deformation values with millimeter precision. The reference point is generally placed at a location that is stable in terms of deformation. This makes interpreting the deformation map easier. However, since the earth is constantly changing, even the most stable locations may experience deformation. To overcome this problem, measurements should always be interpreted relative to each other. Any deformation of the reference point will affect the deformation values of the other measurement points. For example, if the reference point is subsiding, all stable points will appear to ascend. On the other hand, if such an effect is not visible in the deformation map, the reference point was probably well chosen.

Relative in time
In our deformation map we provide time series for each data point. The starting point of the time series has a deformation value of 0. All of the following deformations are given with respect to this first measurement.

1.4 Persistent Scatterers and Distributed Scatters

Scatterers
The resolution of the data determines the amount of measurements that can be performed. This can be compared to a photo camera; within each pixel, only one value can be obtained. However, multiple objects reflecting the signal may be present inside the pixel, so called scatterers. Depending on the reflectance properties of the objects within each pixel, different methods are used to measure the deformation. In an ideal world, each pixel only contains one scatterer, which means the measured signal will contain little disturbances. In reality, however, pixels hold multiple scatterers. The two main types of scatterers are Persistent Scatterers (PS) and Distributed Scatterers (DS).

Persistent Scatterers (PS)
Persistent Scatterers are pixels that contain objects that have high reflectance properties and are time-coherent, meaning the reflectance remains constant throughout time. These objects reflect the majority of the signal received by the satellite from a certain pixel. Persistent Scatterers are commonly found in urban areas, such as on buildings or other man-made structures. When a pixel contains one scatterer that reflects much stronger than the others, the weaker scatterers are almost negligible for the measured values. In other words: the signal to noise ratio is high. This is depicted in the image below, where the red signal is stronger than the rest of the signals.

Fig 2: Persistent Scatterer; the pixel contains one dominant reflection.


Distributed Scatterers (DS)

Distributed Scatterers are pixels where multiple objects show a weaker reflection, at a relatively similar strength. If a pixel is classified as a Distributed Scatterer, our algorithm will search for neighboring pixels with similar reflection characteristics. If multiple neighbors show comparable reflections, the relatively ‘weak’ signal is seen as trustworthy because the signal to noise ratio again becomes larger. The derived deformation, however, is not from a single reflector, but from a larger, homogeneous area. Examples of this are a piece of road or the prepared substructure of a train track. The image below shows an example of a distributed scatterer pixel.

Fig 3: Distributed Scatterer; the pixel contains multiple similar reflections. 

1.5 Horizontal / vertical deformations

Oblique satellite viewing angle
The satellite is observing the earth under an angle with respect to the vertical, generally between 20 and 40 degrees. We measure deformations on the Earth’s surface in the viewing direction of the satellite, so the deformation measurement has both a vertical and a horizontal component. Depending on the type of deformation we can often make an assumption on the orientation of the deformation. For example, it is reasonable to assume that the deformation of a street is predominantly vertical. In that case, we can recalculate deformations as if they were purely vertical. If it is not safe to make such an assumption, it is possible to decompose the signal by using multiple satellites with a different viewing angle.

Decomposition by using multiple satellites

Because of the orbit of the satellite, it passes a certain location both from south to north (the ascending orbit) as well as from north to south (the descending orbit). This is explained in further detail in the satellites section. Because of its fixed viewing direction (generally to the right) this means images are acquired from both the west and the east. For larger objects that contain multiple data points in both the ‘ascending’ and ‘descending’ orbits, a decomposition of the measured deformation can be done in the vertical and horizontal (east-west) direction. However, due to the flying direction of the satellite, the measurements are relatively insensitive to north-south movements.

Fig 4: Horizontal/vertical decomposition of the measurement signal.

The left figure shows the deformation signal when using one satellite orbit. The measured deformation in the satellite’s line-of-sight (black arrow) can be caused by all ‘real’ deformation vectors along the green line. As an example, the purely vertical and purely horizontal deformations are shown in red (arrows).

The right figure shows the vector decomposition when using two satellite orbits (‘ascending’ and ‘descending’). Again, the deformations measured by the satellite are depicted in black. However, this time, there is an intersection point between the two green lines along which the ‘real’ deformation vectors are located. The deformation that occurs in this example is not only vertical, but has a slight horizontal component in the east-west direction and is depicted by the yellow arrow.

1.6 Interferograms

Interferograms are the result of subtracting the phases of two SAR images, a process called Differential-InSAR (D-InSAR). An example of an interferogram is shown in the figure below.

Fig 5: Interferogram.

The phase difference is depicted by the color of the pixels in an interferogram. The interferogram above shows repeating color cycles, which are clearly visible in the northern part. Because the same phase value is repeated in every cycle of the radar wave, the colors repeat. The following figure zooms in on the area.

Fig 6: Wave cycles in an interferogram.

Deriving Deformation
It can be seen that the same color cycle (from red to blue to red) repeats three times, in the direction of increasing phase. This means that the northern part has subsided three times a full wavelength with respect to the southern part between the two image acquisitions that were used to produce the interferogram. Because we know the exact length of a full wave cycle, the deformation can be derived. This process is called unwrapping.

However, if there was a strong atmospheric activity during one or both of the image acquisitions used for producing the interferogram, the deformation signal can be polluted with atmospheric noise. In order to resolve deformation from atmosphere, a time series of interferograms can be used to model the atmospheric signals. This is further explained in the article about atmosphere.

2. Measurement Locations

How suitable a certain location is for performing our measurements depends on how well the objects reflect the radar signal, whether the reflection is consistent enough in time to produce a measurement time series and the measurement resolution.

2.1 Reflecting objects

Not all objects on the Earth’s surface reflect a proper signal back to the satellite, as illustrated in the figure below. The shape, orientation and structure of a surface influence the way it reflects the signal. The best reflections come from angular objects such as buildings and other solid infrastructure. However, when the radar signal bounces of a smooth surface, such as water during windless conditions, it acts as a mirror and almost no signal is received by the satellite. By using satellites that emit a short wavelength, we are able to obtain a high measurement point density on roads, which are also relative smooth surfaces.

Fig 7: Radar signal reflecting on the earth’s surface.

2.2 Consistent reflections

By combining a series of satellite acquisitions we can track the deformation of a certain location over time. This can be done under the condition that the reflection is roughly the same in every satellite image, so the location remains recognizable. This means that changing objects, such as vegetation, are not suitable for our measurements. Buildings, infrastructure and other hard surfaces often do reflect consistently throughout time. These objects are called coherent reflectors and form the basis for our precise deformation measurements. Depending on the satellite and area, it is possible to obtain thousands to tens of thousands of measurement points per square kilometer.

Measuring the same object over time

Every time the satellite revisits a location, it measures exactly the same object as the previous times. This is because of the consistent reflections of suitable objects, and the fact that we can relate satellite images on a sub-pixel level. If our algorithm is uncertain about the consistency of a reflection, the entire measurement point is removed from the deformation map. This is the reason that not every object contains measurement points.

2.3 Measurement resolution

Resolution can be defined as the size of the area over which the satellite gathers one measurement value. A higher InSAR measurement resolution results in a large increase in the number of measurement points. The figure below compares a typical SkyGeo standard resolution product (left) with a typical SkyGeo high resolution product (right).

Fig 8: Left: standard resolution, right: high resolution.

This increase in the number of points can be explained by the fact that multiple reflecting objects that were first measured together in one resolution ‘cell’ (left), now have their own separate ‘cell’. Because the satellite can only measure one value per cell, we acquire more reflections with a higher resolution. This is illustrated in the figure below.

Fig 9: Capturing more reflections with a higher resolution.

Using higher resolution data has three advantages:

  • There is less room for noise in smaller cells so the measured value in the cell is less noisy and it is therefore more likely that a reflector shows a consistent reflection throughout time (see also Consistent reflections). This leads to even more measurement points.
  • The measured deformation values are more precise.
  • The absolute positions of the measurement points are more precise (cf. Location precision).

2.4 Point location (X,Y,Z)

The satellite measures the deformation that occurs within a certain resolution cell (see also Measurement resolution). However, the major part of the signal comes from the well reflecting object(s) within that cell. Our algorithm estimates the X,Y location and the height (Z) of this dominant reflection. This reflector is visualized as a measurement point with a certain height in the SkyGeo deformation map. The precision of the measurement point location depends on the satellite resolution and is shown in the table below.

Table: Point location precision
X,Y Z
Standard resolution 2-3 m 2 – 2.5 m
High resolution 1-2 m 1-1.5 m

3. Measurement Quality

Quality is inherent to measuring; how precise and how reliable is the data? With InSAR we measure the deformation at certain locations, but there is an uncertainty in the determined deformation as well as in the location that this deformation occurs.

3.1 Measurement precision

Individual measurements

Because the InSAR technique uses the phase of the radar signal (see Measurement principles), the deformation value at one location can be measured with high precision. This precision depends on the weather during the time of measurement and the resolution of the satellite. An indication of the precision of our individual measurements in regards to resolution are shown in the table below.

Table: Individual measurement precision
Individual measurement precision
Standard resolution 6-8 mm
High resolution 2-3 mm

 

Estimated linear trend

The colors in the SkyGeo Interactive Map generally represent the deformation velocity of the point, estimated by fitting a linear trendline through the individual measurements over time. The precision of this deformation velocity, assuming the deformation is purely linear, depends on the satellite and the length of the time series. If there are more measurements, the estimated velocity is more precise. An indication of the precision of our deformation velocity estimates are shown in the table below.

Table: Deformation velocity precision
Deformation velocity precision
Standard resolution 1-2 mm/yr
High resolution <1 mm/yr

 

3.2 Location precision

Although we can measure deformation very precisely, the absolute X,Y,Z location of the measurement points is less accurate. An indication of the precision of our measurement point locations are shown in the table below.

Table: Point location precision
X,Y Z
Standard resolution 2-3 m 2-2.5 m
High resolution 1-2 m 1-1.5 m

 

3.3 Data reliability – unwrapping

As well as understanding how precise the data is, it is important to know how reliable it is. In order to conduct our deformation measurements, we use the phase recorded by the satellite (see Measurement principles). This is the fraction of a complete wave cycle that is in addition to the number of full wavelengths the signal has traveled.

Phase ambiguity
Phase measurements can be thought of like the time on a clock. There is a finite number of times that can be displayed on a clock and every twelve hours those times repeat. Similarly, there is a finite number of phase measurements that can be recorded within one wave cycle (0 to 2pi). Once a wave cycle has completed, the phase measurements then repeat. This is visualized in the images below, where the phase measurements are shown in a circle, like a clock.

Fig 10: Both clock time and radar phase are ambiguous.

However, the values have a new meaning once a new revolution has started. For a clock, a new revolution means a new 12 hour cycle has started. To fully understand what time it is, one needs to know the date and/or which part of the day it currently is. Similarly, a new revolution on the phase cycle means a full wave cycle has been completed. To fully understand the meaning of the phase measurements recorded, one needs to know the number of full wave cycles associated with the phase measurements. Resolving this to determine the actual deformation is called ‘unwrapping’.

Unwrapping
In order to derive the actual deformation from consecutive phase measurements in time, correlations in time and/or space are used to ‘unwrap’ the measurements. For example, we can assume that the occurred deformation is linear during (a part of) the measurement time span or that the deformation rate change is smooth in space. If a wrong assumption is made, this may lead to so called ‘unwrapping errors’.

In our algorithm we do a strict control on this process to ensure a high degree of data reliability. In practice, this means that for the measurement points in the SkyGeo deformation map the above process is executed correctly at least 99.0% of the time.

3.4 Point Quality

Point Quality is a relative measure of the quality of the respective measurement point. The value is displayed in the metadata of the data point and ranges from 0 to 1. Generally, a lower quality value indicates a higher amount of noise in the time series.

To ensure the quality of the datasets, a minimum quality cut-off is applied before delivery. You can further filter the data points on their quality by using the SkyGeo Maps filter sliders.

Computation
To acquire a quality value, a model reflecting the expected deformation behavior is fit to the measurements within a single data point. This model can for example contain linear, quadratic and seasonal components and is adjusted according to the investigated area. After that, each measurement is compared to the value of the model at that date in time. The difference between the measurement value and the model value determines the Point Quality. A higher value indicates a better match between the model and the measurements.

Interpretation
Point Quality is generally a good indication of the quality of the measurements in the time series. If the model for expected deformation reflects the actual movement behavior of the measurement point, it is a measure of the amount of noise in the time series. However, if you compare a seasonal deformation signal with a linear model, Point Quality is lower than it would be when a seasonal model was used for the comparison, meaning the type of model used affects the outcome. For some points, this effect can lead to a less correct measure of quality.

4. Geostatistics

To provide you with the best information for maintaining your assets, we can apply geostatistics to select and aggregate data points on user-specific assets.

4.1 High/Low Separation

When a building is constructed, generally a foundation is laid to stabilize the building. Because of this foundation, buildings are typically more stable than natural features such as vegetation. In order to distinguish between different deformation regimes, it is often important to separate measurement points located on a building or other infrastructure and measurement points located on the ground. These are referred to as high and low data points, respectively.

Fig 11: Difference between high and low data points

In order to separate the ‘high’ and ‘low’ points, the height obtained from our algorithms for each measurement point are compared to a Digital Terrain Model (DTM) with relatively accurate continuous elevation values. If the height value obtained for a measurement point is more than a set threshold of X m higher than the DTM value at the same location, the point is classified as a high point. If the height value obtained for a measurement point is below a second set threshold of Y m than the DTM value, the point is considered not trustworthy and is discarded. If the difference between the obtained height value and the DTM value falls within these two thresholds, the point is classified as a low point. This is visualized in the diagram below. The values for X and Y mainly depend on the resolution and quality of the satellite data. Common values are X = 2 m for high resolution, X = 3 m for standard resolution and Y = 10 m for both types of resolution.

Fig 12: The thresholds for high/low classification.

4.2 Weighted Averaging

Weighted averaging is a technique to aggregate data points on pipelines by using the relative distance of the points to the underground structure. The first step in the aggregation process is high/low separation. In this step, the data points from the ground surface and from higher infrastructure are separated. This is a way to ensure that the points that are taken into account are from the ground surface. Secondly, virtual points are sampled every X meter along the pipeline course.  After that, the data points in a radius of Y meter are selected and aggregated to the virtual point, scaled by the square of the inverse distance. Common settings are X = 10m and Y = 30m.

An example of weighted averaging on pipeline shapes is shown in the figure below.

Fig 13: Weighted averaging on pipelines

Statistics
A uniformly subsiding pipeline does not need to be a problem for pipeline stability; it is the differential settlements that should be worried about. Therefore, we generally provide pipeline asset managers with the following statistics:

  • Weighted average subsidence
  • Largest subsidence difference within a certain along-track area
  • Largest subsidence within a certain along-track area
  • Percentile values of subsidence

4.3 Shape Aggregation

In this article you will learn about the techniques we use to aggregate the SkyGeo data points on asset shapes like roads, buildings and zip codes, and what assumptions we make in the process. Furthermore, you will learn about the statistics that are derived from the aggregation.

An example of shape aggregation on building shapes is shown in the figure below.

Fig 14: Shape aggregation on buildings.

Aggregation process
The first step in the aggregation process is high/low separation. In this step, the data points from the ground surface and from higher infrastructure are separated. This is a way to ensure that the points that are taken into account are from the height level of the surface of interest. After that, the data points within a distance of 3 m with respect to the asset shape are selected in order to account for the x/y position precision.

Statistics
After the points are aggregated, the statistics are computed per asset shape. The following statistics are available:

  • No data. Only displays the shapes of the data product, without data or colors.
  • Average deformation. Average linear deformation velocity (in mm/yr).
  • Standard deviation. The standard deviation of the linear deformation velocity of the data points that were aggregated in every shape (in mm/yr). The dropdown menu parameters with values behind ‘standard deviation’ only show the shapes with a standard deviation in the specified range.
  • Number of points. The number of data points that were aggregated per shape. The values behind ‘number of points’ specify the range of the color scale that is visualized.
  • Percentile. The percentile values of the linear deformation velocity in the shape (in mm/yr). The X’th percentile indicates the X percent of the data points in the shape, which have a lower deformation (i.e. higher subsidence, since subsidence is negative) value.
  • Default. The default setting for this deformation map. This is generally is the average deformation parameter.

5. Satellites

5.1 Satellite acquisitions

Radar satellites orbit the earth at an altitude between 500-800 km with a velocity of approximately 7.5 km/s. They scan the whole earth in strips and because of their orbits and the rotation of the earth they will repeat the exact same cycle after a certain amount of days, depending on the satellite.


Fig 15: Satellites orbiting the earth. Source: J.C. Szidloski

Because of the orbit of the satellite, it passes a certain location both from south to north (the ascending orbit) as well as from north to south (the descending orbit). Because the satellite looks at the earth under an angle and the viewing direction is generally fixed (mostly right-looking), the earth is observed from two different directions. These two measurements can be combined to resolve horizontal and vertical deformations.

Fig 16: Ascending and descending orbit of the satellite.

Furthermore, because we are dependent on the natural reflections of objects on the earth’s surface, the orbit direction (ascending or descending) affects which objects reflect signal back to the satellite and, thereby, can be measured. Note that due to the different viewing directions and reflections, ascending and descending images cannot be combined into a single InSAR time series.

5.2 Radar satellite characteristics

The characteristics of the radar satellites we most commonly use are visualized in the table below. Note that the actual repeat cycle depends on how the satellite is tasked. The resolution can depend on the satellite settings; the values shown here are what we generally use.

Satellite Time period Repeat cycle (days) Resolution
Sentinel-1 2014-now 12 Standard
TerraSAR-X 2007-now 11 High
COSMO-SkyMed 2007-now 16 High
RadarSat-2 2007-now 24 Standard
ALOS 2006-2011 46 Medium
Envisat 2003-2010 35 Standard
ERS 1992-2001 35 Standard

6. Quality Assessment

We make sure you get data with the highest possible quality. In order to achieve this we perform a thorough internal quality assessment and provide you with a summary report of the steps we took for producing your data. This is explained in more detail in the sections Internal Quality Control and Factual Report.

6.1 Internal Quality Control

This section explains how our engineers make sure you get the best data. It elaborates on both our internal quality control procedure as well as the components that have been checked.

Procedure
Once the responsible engineer(s) is/are satisfied with the product they produced, the internal quality control process commences. This procedure is carried out according to the following steps:

  1. The responsible engineer(s) deliver(s) the product internally.
  2. The product is investigated by a peer engineer and checked against our quality control checklist.
  3. The responsible engineer(s) improve(s) the product based on the comments that were received.
  4. The peer engineer checks if this iteration fulfills our quality standards; if not, step 3 is re-iterated.
  5. The product is delivered to the account manager, who checks whether the product is good enough to send out to the customer; if not, step 3 is re-iterated.
  6. The account manager delivers the product to the customer.

Quality checklist
The data is thoroughly checked on different aspects:

  1. The product is consistent with the offer letter.
  2. The customer username and password to access SkyGeo Maps work.
  3. The quality of the geodetic network is sufficient.
  4. The point quality is sufficient. We check for unlikely values and/or locations and, if applicable, compare the values to those of external datasets.
  5. The noise in the dataset is acceptable. We check and address possible causes of noise, like atmosphere.
  6. The point density is sufficient. We check if the areas of interest are covered properly.
  7. The reference point choice reflects the deformation of the area of interest in the way the customer wants.
  8. The data points are displayed on the right locations on the map (X/Y/Z).
  9. The geostatistical operations are performed correctly.

6.2 Factual Report

In this article you will learn about the Factual Report; a standardized report on the processing steps we took to provide you with your products.

The Factual Report is meant to provide an overview of the processing steps we took; both for interested customers that want to know more about the approach and to ensure future reproducibility of the results.

Downloading the Factual Report
The Factual Report can be downloaded by clicking on the download button in the upper right corner of the Interactive Map.  However, instead of selecting ‘data’, one should select ‘docs’.

Contents
The Factual Report has a default structure, which is shown in the figure below.

The Document Status describes the status of the document. Here, it states for which customer and which viewer the document was produced and at what date. Furthermore, it describes to which InSAR datasets the report applies.

The Area of Interest depicts the area of interest in the form of an outline on a map and an estimation of the size of the area.

The Observation Period chapter describes the satellites that were used to produce the datasets and their main characteristics. Furthermore, the timespan over which the observations were acquired is depicted, together with the median of the time interval between two subsequent acquisitions. The distribution of dates at which a measurement was done, is shown the ‘Temporal baseline’ figure. An example of such a figure is provided below.

Fig 17: Temporal baseline figure.

Data Delivery and File Conventions explains the format of the downloadable files. The names of the layers in the sidebar can be different from the names of the CSV file that can be downloaded; this chapter links the two names. Furthermore, it describes the meaning of the different columns that are delivered in the CSV file.

The Quality Assurance chapter states the points the datasets have been checked on before delivery.

In the Appendices several statistics and properties of the different datasets are summarized:

  • The number of data points
  • The direction in which the delivered values are measured
  • Data averages
  • Data extremes (percentiles)
  • Used deformation models for determining the quality and linear deformation velocity
  • Reference point
  • Type of atmosphere modeling

Atmosphere
Atmosphere is a term we use for disturbances in the measured deformation signal, caused by atmospheric disturbances the signal encounters when traveling from the satellite to the Earth’s surface and back. These disturbances are caused by a combination humidity, pressure and temperature in the atmosphere, delaying the signal. Some areas generally have a more turbulent atmosphere than others. Around the equator, for example, the air contains more water vapor than around the poles, resulting in a more boisterous atmosphere. Next to that, the time of the day also has an influence on the likeliness that disturbances are occurring. The sun heats the Earth’s surface and thereby produces heat and water vapor, increasing atmospheric turbulence.

Recognizing atmosphere
Challenges arise because the atmosphere is different from one satellite acquisition to another. This alters the influence it has on the deformation signal, making it harder to distinguish deformation from atmospheric disturbances. However, precisely because atmosphere is not correlated in time, but deformation is, we have a way to separate the two. Furthermore, atmosphere is correlated in space over certain distances, providing a second possible differentiator from the signal of interest.

 

Fig 18: The decorrelation of atmospheric disturbances in time.

 

Filtering atmosphere
Because atmosphere is spatially correlated over certain distances, it does not form a large obstacle for measuring deformation over small areas. For larger areas, however, it might influence the signal greatly. Therefore, we apply atmosphere estimation in order to resolve the signal of interest. The figure below compares the time series of a certain data point before and after removing the atmosphere from the data.

We perform the atmosphere estimation by choosing high quality data points with a regular deformation pattern (calibration points), either stable or deforming perfectly linear. After that, we assume that all time series measurements deviating from this regular pattern are due to atmospheric noise. We use this noise in a statistical approach to produce an interpolated atmosphere map over the area during all satellite acquisitions. These maps are used to separate the actual deformation signal from the atmospheric disturbances.

Difficulties
The statistical approach improves with the availability of data; when there is more information at hand, it is easier to model the disturbances. Therefore, when there are not a lot of image acquisitions in time, atmosphere estimation can be a challenge. Furthermore, the amount and distribution of calibration points influences the quality of the interpolated atmosphere maps. Large deformation bowls, for example, can sometimes be an obstacle for estimating the disturbances. Because the deformation is far from regular in these areas, calibration points are sparse. Therefore, the atmospheric interpolation has to be propagated over a large distance, potentially containing atmospheric variations and thereby increasing the likeliness of making errors in the estimations.

7. Validation

In the past 50 years, the use of satellite technology to obtain useful information about the Earth’s surface has improved both in understanding and practice. With the advancement in technology and the ability to measure many different parts of the electromagnetic spectrum, the information recorded by satellites can be adapted for a broad range of applications, including land deformation, natural hazard detection/mapping, change detection, land use mapping, urban growth and much more.

Due to the rapid advancements in satellite technology and the relatively recent commercialization of the use of the resulting data, satellite products are fairly new to the public. Our goal is to help you understand why using satellites, and specifically InSAR for determining surface movements, is a viable and accurate procedure.

We have summarized a small batch of the hundreds of scientific journal articles that validate the InSAR technique. Take a look at the following journal articles to see how InSAR is justified as a reasonable measurement technique for deformation.

Here are a few quotes taken directly from the above articles:

“The results show that the measurement techniques, i.e. leveling and InSAR, are in the agreement within the error bounds.”

Marinkovic, P., Ketelaar, G., van Leijen, F., & Hanssen, R. (2007, November). InSAR quality control: Analysis of five years of corner reflector time series. In Proceedings of Fringe 2007 Workshop (ESA SP-649), Frascati, Italy (pp. 26-30).

“The calculated subsidence rates for monuments MAGF, MANI, OSDO, and the subsidence magnitude at COTD calculated using both GPS data and InSAR data compare favorably, thereby improving confidence in results derived from both methods.”

Sneed, M., & Brandt, J. T. (2007). Detection and Measurement of Land Subsidence Using Global Positioning System Surveying and Interferometric Synthetic Aperture Radar, Coachella Valley, California, 1996-2005. US Department of the Interior, US Geological Survey.

“The radar observations of land subsidence are in good agreement with recent levelling and GPS observations. However, the radar data provides a more detailed mapping of both the amplitude and spatial extent of land subsidence.”

Motagh, M., Djamour, Y., Walter, T. R., Wetzel, H. U., Zschau, J., & Arabi, S. (2007). Land subsidence in Mashhad Valley, northeast Iran: results from InSAR, levelling and GPS. Geophysical Journal International168(2), 518-526.

“InSAR and time-continuous GPS deformation data overlap, allowing a direct comparison of the data. The comparisons possible for the Los Angeles basin demonstrate that the two methods produce quantitatively similar results.”

Helz, R. L. (2005). Monitoring ground deformation from space. US Department of the Interior, US Geological Survey.

 

7.1 Summary

To summarize, InSAR has many benefits for monitoring surface deformation. The main benefits to using InSAR are listed:

Reliability
With the high precision phase data that the satellite records, we are able to generate very accurate deformation data.

Simplicity
With the use of satellites, data collection can be done remotely, eliminating in-field measurements, which are more complicated and time consuming.

Low Cost
Eliminating field measurements lowers costs by decreasing the amount of time required to collect data as well as the high priced tools used for surveying.

Ability to measure large areas
By using satellites, large areas are able to be accurately covered in a little amount of time.

Weather independent
Since radar penetrates through clouds, measurements can be taken during all weather conditions, creating a continuous dataset.

Ability to obtain measurements at night
Radar satellites emit their own signal and do not use the emitted radiation from the sun, thereby allowing measurements to be taken at any time of the day.

Request an InSAR sample

With measurements of your area of interest.

INSAR SAMPLE