Two satellite programs have been used in InSAR Norway: Sentinel-1 and Radarsat-2. They have different acquisition modes that result different data properties that can change the end-product properties.
In addition, the radar images are processed with two different InSAR techniques that can change the end-product properties.
Spatial resolution (the distance between two objects that can be separated on the ground) and temporal resolution (the time interval between images taken in the same area) are summarised in the table below:
| Satelite | InSAR Norway Coverage | Processing Technique | Temporal Resolution | Spatial Resolution |
| Sentinel-1 | Nationwide | Persistent Scatterer Interferometry (PSI) | 12 days until the autumn of 2016. 6 days afterwards. | About 5 x 20 meters (5 meters in east-west and 20 meters in north-south direction) |
| Radarsat-2 | Regional (Northern and Western Norway) | Small Baseline Subset (SBAS) | 24 days | About 30 x 30 meters (30 meters in east-west and 30 meters in north-south direction) |
| Radarsat-2 | Regional (the urban areas of Oslo and Trondheim) | Persistent Scatterer Interferometry (PSI) | 24 days | About 15 x 5 meters (15 meters in east-west and 5 meters in north-south direction) |
The Sentinel-1 program consists of two satellites, Sentinel-1A and Sentinel-1B.
The first satellite was launched in 2014 and the second in 2016. Each satellite orbits the Earth for 12 days until it is the same position again. With one satellite in orbit, the temporal resolution was thus 12 days. The temporal resolution improved to 6 days with the second satellite became operational, in October 2016.
Since mid-December 2021, Sentinel-1B has been out of commission, reducing the temporal resolution to 12 days again.
Why are InSAR measurements relative?
InSAR measures the change in distance between the satellite and the Earth's surface. These movements are one-dimensional and spatially and temporally relative:
One-dimensionality:
InSAR measurements depend on the measurement geometry of the radar looking down to the right in the satellite's Line-of-Sight (LoS). InSAR only measures one-dimensional movements along the radar's LoS.
If the actual movement on the ground deviates from the satellite’s LoS, the extent of movement will be underestimated or not detected. More information about how geometry is measured.
Spatial relativity:
InSAR documents the relative movement between two points on the ground. In the case of InSAR Norway, the movement is relative to a local reference point.
InSAR is a relative technique, and our data is not yet put into a reference system. Currently all data are processed in approximately 5 x 5 km tiles, each with a local reference. Before putting the tiles together for the national map, the tiles are adjusted so that the peak of the velocity histogram is 0 mm/yr.
This means that the map so far only shows relative and local deformation. We plan to integrate the InSAR data with data from the GNSS stations of the national mapping authority to get everything in a common, absolute reference frame.
Temporal relativity:
The motion is relative to a reference time that varies from dataset to dataset and from area to area. For the Sentinel-1 2019 delivery called "Sentinel-1 Deformation (old)" in map service, the reference was in August 2019 (see information about datasets).
In the latest datasets, «Sentinel-1 Deformation», the reference time is set for the first radar image recording.
Why are measurements not everywhere and all the time?
InSAR Norway data covers a large part of the country, but there are still gaps in data coverage. Three main factors directly affect data coverage in spatial and temporal extent:
Geometric effects:
The coverage of InSAR results is affected by the terrain. The radar has a fixed aiming direction down to the ground. Therefore, there may be some areas without data due to shadow effects in steep mountain slopes facing away from the satellite.
On the other side of the mountain, the one facing the satellite, a layover effect can affect the coverage. When the mountaintop is closer to the satellite than the foot of the mountain, the reflected signals from several different places become impossible to distinguish, leading to an ambiguous return signal. These areas are filtered out.
Dense vegetation, water, snow and ice:
The nationwide Sentinel-1 InSAR map is based on a method that utilises small and clear points on the ground, such as rooftops and rocky outcrops, which show a stable interferometric phase behaviour over time. This method works well in urban areas and barren rock slopes.
However, this technique does not perform well in areas with dense vegetation, agricultural land, bogs, snow or glaciers. If the surface reflectance changes too much between successive radar images, for example due to changes in vegetation of snow cover, the result is low coherence and/or false estimates of the extent of movement.
Fast-moving ground deformation:
InSAR is not a valid technique for detecting sudden ground movement. The maximum measurable displacement corresponds to half a satellite sensor's wavelength in the time interval used to compare radar images.
That is to say, the maximum length is about 2.8 cm (for the satellites used in InSAR Norway) in 6 days (as is the case of two consecutive Sentinel-1 images). Many image pairs are used with different time intervals. Due to the winter data gaps, InSAR cannot monitor areas moving more than 100 mm/year or at very non-linear velocities due to seasonal movements.
In areas where ground displacement is very fast, a few dark-coloured permanent scattering points (PS) are often visible, but large sections may have no measurement points. In this case, the points are filtered out due to poor signal quality. If an area is missing data, we cannot assume there is no movement: it means that InSAR cannot obtain information.
The Sentinel-1 satellites have acquired images every 12 days since 2014 and every six days since 2016. This is a very long time series of InSAR measurements. Nonetheless, in northern latitudes, continuous snow cover five to seven months annually results in significant gaps in data.
Snowfall and melting snow result in a loss of signal or a misestimation of the extent of movement. Only data from snow-free months are used, and the time series display large data gaps every winter. However, we can minimise the gaps by adapting the processing to local conditions and including as many snow-free images as possible.
In coastal areas in southern and central Norway and urban areas, the measurement series have been extended by several months. Northern Norway and the mountains of southern Norway have an extended snow season, so the data sets continue to run from June to October.
It is important to remember that all ground motion measurement values shown between the snow-free seasons are estimates. As estimates, the data must be interpreted with caution to avoid over-, or underestimating, the extent of displacement.
What are the primary sources of error in InSAR?
InSAR ground motion is measured from phase differences (a fraction of the wavelength) between two signals transmitted from a radar satellite to the Earth's surface. Phase differences are affected by many more factors than just movements on the ground. These factors include atmospheric variations, topography and changes in surface properties (see more info).
To isolate the movement, the other phase shift elements must be removed, and sources of error must be reduced in processing. Although there are advanced solutions to do this, there will always be some errors in the end products. Here we describe the main problems found in InSAR Norway data and how to identify them:
Noise/Deviation:
Some points have wrong values due to noise in the input data. Even if a point has passed a series of filtering criteria based on quality measurements, deviating results are inevitable. Without detailed information about ground conditions, it is not possible to know whether these measuring points are correct or not.
A general rule is always to use caution when interpreting individual points that show deviating values compared to neighbouring areas.
Humidity and seasonal effects:
In wetlands, such as bogs, seasonal variations are often due to changes in water content. Because only data from summer months are used, such variations can create problems for processing.
The problems appear either as areas with large point-by-point variations (a mixture of red and blue points) or as areas with evenly positive values (blue points) or evenly negative values (red points). This typically appears in marsh areas.
Snow cover:
The quality of the measurements is affected by changes in surface properties, for example, due to snow. InSAR Norway uses data from approximately June to October to avoid most snow-cover problems.
The actual period of data used is chosen based upon the local snow statistics. However, there may still be snow in some images during this period, especially at high altitudes, which leads to processing challenges and possible errors. These can result in errors.
Atmospheric effect:
Change in the atmosphere leads to phase changes that can be misinterpreted as a movement. Assuming these variations are not time-dependent, one can filter them out using long data series.
Nevertheless, in some cases, the filtration may be incomplete in deep valleys, where the air achieves a stable stratification. Occasionally, deep valleys can display false-positive values (light blue colours).
Data calibration:
The nationwide data set has been composed of several smaller areas (approx. 5 x 5 km), processed separately. If data from different times are used to process nearby areas, it can create a shift in the values along a straight line.
Each area is processed independantly and then calibrated to a local reference point (see section "Why are InSAR measurements relative?").
Before the areas are combined into the nationwide datasets, each area is adjusted so that most of the points in an area have a movement of 0 mm/year. When there has been excessive movement, the assumption is invalid; hence all the values increase or decrease.
Sudden jumps:
Sudden jumps in time series can indicate that results are affected by the phase ambiguity (see more info about this).
InSAR measures the phase difference between two waves, corresponding to a fraction of the wavelength. The results are then converted from cyclic phase differences to continuous distance differences. The challenge is a displacement value resulting from whole number of cycles of the wavelength. Any number of whole numbers cycles factored with the wavelength will give identical phase values.
This phase ambiguity is one of the biggest challenges of processing interferometric data. The phase ambiguity can lead to a change in value in the results, especially when there are long data gaps in the time series.
How accurate/precise is the data?
The exact horizontal position of the reflectors on the ground is unknown. A point in the map service can originate from a reflector with a strong signal or from several reflectors located in a cell. The Sentinel-1 satellites have a resolution of (20 m x 5 m). Therefore, points can deviate up to 10 m from the location indicated in the map service.
InSAR can measure movement with millimetre level accuracy but the accuracy varies from point to point due to variable sources of error.
Points that have passed the quality control are available in the map service and have time series with mean square deviation from a deformation model (a third-order polynomial and a periodic seasonal component) of less than 5 mm. The uncertainty in relation to the model is given in the results of the mean square error parameter (RMSE).
- To assess the quality of the motion values, it is essential to look not only at the average velocity of simple points but also:
- To evaluate results in connection with neighbouring points: If specific points display divergent values from adjacent points, they must be interpreted with caution. We also recommend viewing other ascending/descending datasets. If a particular motion is seen in more than one data set from the same geometry, it strongly indicates that the movement is real and not just noise.
- To analysing the entire time series (displayed in the time series window): The average velocity used to colour points on map service is estimated based on the whole time series. However, seasonal trends, extreme velocities, or seasonal ground surface (vegetation, snow) affect the results, particularly when long data gaps occur in the wintertime series. Therefore, it is essential to look carefully at time series before continuing to use data.
- To check quality measurements (available in the data viewer): Root mean square error parameter (RMSE) provides information about deviations from the deformation model. Two other documented parameters are indicators of stability of measuring points over time: temporal coherence (values are between 0 and 1. Higher values are better) and amplitude quality (values: between 0 and 1. Lower values are better).
More information about map services, the time series window and the data viewer.
More information about documented parameters (including quality goals).