Frequently Asked Questions
How is APSIS© different to other InSAR methods?
All common InSAR techniques (e.g. PSInSAR, SBAS) assume that, to make a precise measurement, a target (point or distributed) must be a consistently good reflector in ALL observations. This could mean consistency in any number of parameters such as amplitude, phase dispersion and coherence.
This works very well over hard targets, like buildings and rocky outcrops, which is why most conventional InSAR applications relate to surveys of hard targets and structures.
Vegetated and natural surfaces do not meet this criterion as their reflectivity and scattering properties, almost by definition, change from season-to-season. Thus, conventional InSAR methods fail to work well in such areas.
What APSIS© does is recognise that, even during periods of low reflectivity, vegetated and natural surfaces can still form good, high coherence, pairs with other observations, even if those observations are far apart in time.
Thus, by considering the whole range of possible high coherence pairs for a pixel, we can reconstruct a new set of measurements for which it is possible to determine motion. If we apply this technique across the landscape, to all types of terrain, we find full coverage over all land cover types is possible.
Below compares a conventional InSAR method (SBAS) with an APSIS© survey over a coal mining area in the East Midlands of the UK using Sentinel-1 data.
Conventional InSAR Coverage
Map of the area
Can radar penetrate a vegetation canopy?
Yes, of course, but the amount of signal lost through scattering depends on things like the leaf state, the measurement geometry and the vegetation canopy. Also, shorter wavelengths (e.g. C-band, X-band) are affected more than longer wavelengths (e.g. L-band).
Does two-pass InSAR work over vegetation canopies?
Yes, but not everywhere and not at the same time.
Two-pass InSAR is where two SAR images taken at different dates are combined to form an interferogram. The quality of the resulting interferogram will be affected by the amount of signal penetration but also any changes that occurred between the two dates and this can be quite subtle (wind, moisture etc.). Any signal that does not penetrate the canopy is likely to be randomly bounced around the canopy before being reflected back to the sensor. This is called diffuse scattering and any InSAR observation of diffuse scattering is likely to be of very poor quality, with very low coherence. Because of all of these influences, it is a very common misconception that InSAR never works over vegetation. But there are some notable exceptions….
Zhong Lu et al. (2005) found a ‘surprising’ capability for C-band InSAR data to be able to maintain good strong fringes over flooded forests in Louisiana, despite the common view that the signal would be scattered by the canopy. That work initiated a capability to use InSAR to routinely measure the surface water level over the swamp forests of the southern USA that continues to this day.
Through our work, we also know that, occasionally, strong fringes can be seen over tropical forests, and below is an example from the North Selangor Peat Swamp Forest, which is an area we have comprehensively surveyed on behalf of academics from the University of Nottingham and John Moores University.
Images generated from C-band Sentinel-1 data of the North Selangor Peat Swamp Forest, Malaysia.
A Google Earth image
A two-pass differential interferogram
We see this everywhere. If we have two-pass interferograms that are taken using pairs that are very close in time we can often see fringes over many different vegetation types. Below is another example, this time from the Black Forest in Germany.
Images generated from C-band Sentinel-1 data of the Black Forest, Germany
A Google Earth image
A two-pass differential interferogram
What causes those InSAR fringes over thick vegetation?
Any fringes seen over a dense forest canopy cannot originate from the diffuse-reflecting canopy because the coherence would be too low. In this case, we speculate that high quality fringes, of high coherence, must originate from specular reflection, most likely from the ground, either directly or, as they presuppose from the Louisiana/Florida swamp forests, from a double bounce from surface to trunk. This agrees with Polarimetric SAR analyses of scattering over such areas where the assumption of the dominance of diffuse reflection has had to be revised to account for fringes in co- and cross-polarised interferograms.
Below is a figure that illustrates this, taken from an FAO report on tropical peat monitoring.
Can we measure ground motion through a forest canopy?
Yes. Below is a map of subsidence around the former Clipstone coal mine in the East Midlands, UK which lies next to Sherwood Forest. A conventional InSAR survey which can only make measurements of hard targets covering the town itself and some of the infrastructure but is only able to identify some elements of the motion. If we filter the coherence using the APSIS© method and utilise high coherent measurements through the canopy, we can see the full extent of the subsidence, which is oval-shaped extending under agriculture and thick forest canopies.
Images of subsidence in Clipstone, Notts using the same stack of Sentinel-1 data.
Subsidence using conventional InSAR
Subsidence using the APSIS© method
How can I make precise and dense sets of InSAR measurements through vegetation?
Terra Motion Limited have a solution, called APSIS©. It is an enhancement of the Patent Pending Intermittent SBAS technique invented by the University of Nottingham in 2012. It has been used successfully by the research community since then for monitoring a whole range of phenomena in urban and rural terrains, through forests, agriculture and natural surfaces. Most notable is the work in the development of peat condition assessment criteria and the monitoring of earthwork dams.
Do I need corner reflectors?
No, APSIS© does not need corner reflectors to be in place to perform a survey. Corner reflectors are artificial point targets that give a bright, coherent reflection in a SAR image and are often used as a benchmark for InSAR surveys or for densifying the poor coverage of a conventional InSAR (PSInSAR, SBAS) survey.
Corner reflectors may be used as a benchmark but the difficulty is choosing a location that represents the motion in the pixel. For example, a building piled into the rock will not rise and fall as the soil does with the seasons. An APSIS© survey may not be able to resolve the two if the soil signature dominates the return so attaching a corner reflector to the building serves little use as a reference for the measurements.
Soils and vegetation are not hard, uniform surfaces like buildings so, if a reflector is placed in the field there can be no guarantee that that point will move as the rest of the soil moves. Furthermore, our experiences with corner reflectors over peatland surfaces indicate that they may compress and damage the peat and also tend to sink slowly over long periods. Given that the seasonal motion of the peat surface may be measured in millimetres, corner reflectors are a poor choice to densify an already poor network.
In addition, a survey that relies on corner reflectors is not able to exploit the vast archive of SAR data that currently exists. For example, most land areas currently have a Sentinel-1 archive going back 4-5 years - it is simply not possible to retrospectively install a benchmark or densify a PS network. Corner reflectors can also be expensive to install and maintain, especially in remote and dangerous terrain.
Does APSIS© work in areas of snow and ice?
Yes, but it depends on the particular characteristic of the snow or ice cover, its liquid water content, the surface roughness and its grain size. For seasonal snow, it is often the case that the radar scattering is dominated by the ground that it sits upon. For re-frozen snow, we see more volume scattering. Certainly wet snow gives a very low response, like standing water. In summary, snow is like vegetation in that good InSAR measurements can be made, but these are intermittent in quality.
In that case, it comes as no surprise that the APSIS© method appears to work very well over seasonal snow areas. Any temporal variations in coherence due to intermittent cover can be overcome by an examination of the permutations and combinations of different interferometric pairs to complete the network of observations, usually allowing us to form a complete time series for the majority of pixels in an area.
Below is an example average velocity, over a four-year period, of a permafrost area in the North West Territory in Canada mapped with C-band Sentinel-1 data using the APSIS© method. These maps can be used to monitor the collapse of melting permafrost areas during the summer periods, its effect on infrastructure and contribution to greenhouse gas release.
Under certain circumstances, the motion of ice can also be monitored using InSAR. However, the velocities found over ice sheets and glaciers are often too fast for InSAR to be of any use. For example, any displacement greater than 1.4cm between C-band interferograms may be difficult to measure with any confidence. Most glacial motion far exceeds this limit and pixel offset tracking is a preferred alternative to InSAR in such cases.
If InSAR can be used to monitor glacial motion, a more simple approach is often used, considering only consecutive pairs with the shortest temporal baseline and applying a simple stacking approach to determine motion characteristics. In the example below, we have applied a stacking technique to the measurement of glaciers on Ellesmere Island, Canada.
In summary, we have been fully able to apply APSIS© to areas of seasonal snow cover without any problem. However, we would hesitate to use the technique over areas of persistent cover, such as ice caps and glaciers.