Get to know the different data layers available in a GDS analysis
Results from InSAR measurements like the Global Deformation System (GDS) analysis are most commonly viewed in displacement values, like millimeters of change. In addition to that key total deformation value, there are a number of other data layers provided with each analysis that help to contextualize the data. This article will walk through the different data layer types that can be displayed in Iris and explains how each can be used in your interpretation. It will also explain
Line of Site versus East-West-Vertical
Line of Site (LOS)
Line of site data refers to results are using the data from only one traverse direction of the satellite source, either ascending or descending, but not both. The measurement provided is the change in the distance between the satellite source and the point of measurement in millimeters. SAR satellites, including Sentinel-1, travel in polar orbits, meaning the total displacement reported in LOS data represents not just vertical (up/down) change, but also horizontal change in the east-west direction. To get precise east-west and vertical measurements, monitoring sites need coverage with both ascending and descending orbits of the satellite.
East-West-Vertical (EWZ) Data
When a site is covered by both ascending and descending passes of a satellite, then the right-looking and left-looking LOS analyses can be collated and the true east-west and vertical components of deformation can be calculated through trigonometry. Because the satellites travel in a polar orbit, north-south components of movement can never be derived.
When viewing EWZ data in Iris you will have the option to select between the East-West and Up/Down components for each of the data layers and derivatives available including uncertainty, velocity, and acceleration. As a convention, western movement is given a negative value and eastern movement is given a positive value.
Spatial filtering
Deformation time series results from InSAR are relative in two ways, time and space. For time, line of site data is referenced to the the first date in the analysis, epoch 0, with values reported as change since epoch 0 (which will always have an initial value of 0). To reference the points in space, a reference system needs to be chosen. Typically, a good spatial reference point would be one that is known to be stationary, however this can be hard to achieve over a large site like a mining operation. To deal with this issue, Descartes Labs handles spatial referencing in two different ways and provides both options to clients with each analysis.
LOS
The default data layer available, LOS, has had a spatial filtering algorithm applied to the points to remove larger patterns in the data in order to highlight more local deformation signals. To apply the spatial filtering, a long wavelength trend is fitted to the data and then subtracted from the raw data for each epoch. In most small-moderate scale sites this trend is roughly 1 km, while with larger areas such as cities or regional features, the long wavelength trend can be 3-5 km.
Left: unfiltered data. Center: the estimated long wavelength background trend. Right: Spatially filtered results with background trend removed.
LOS (Unfiltered)
When the LOS (unfiltered) data layer is selected, a new orange point marker with an anchor symbol will appear in the map window. This point becomes the spatial reference point for the dataset, meaning all values displayed are in change relative to the orange marker. In this mode, users can place the orange icon on a location where they have high confidence has remained stable over the analysis window, often a building on flat ground. Now, the green icon can be move through the scene and all measurements are in reference to this stable position.
LOS Uncertainty
Uncertainty in Line of sight measurements can be useful when investigating a deformation signal and determining whether or not it is based on highly confident data or if the movement is within the noise of the data. Assessing uncertainty is a complex task and the details of Descartes Labs novel and groundbreaking methodology can be read about in further detail here.
Velocity and Acceleration
Another interesting way to visualize GDS results to understand movement at a site is through velocity and acceleration. These two values are delivered as single values based on a trend curve that is fit to the past 180 days of data for each point. Because these are static data layers, the time-series deformation plot will be inactive when either velocity or acceleration are selected.
Velocity values are provided in mm/day, and acceleration values are provided in mm/day2.
Coherence
Coherence is a property of InSAR data that reflects the noise in the radar signal. Hard, solid materials like rock, bare soil, concrete, or buildings will have high coherence while soft, wet, or moving objects like vegetation, water or wet ground, and earth moving activities will produce very noisy signals and lower coherence. Coherence is calculated as a value for 0 to 1, with 1 being highly coherence and high confidence data.
Descartes Labs applies a coherence threshold to its GDS results to ensure only high quality points are returned with each analysis. Typically this value is set at 0.6, but may be individually modified during initial setup to insure sufficient coverage over a site. If you notice gaps in data over your area of interest, it is likely because these points did not have high enough coherence to be validated in the result.
Viewing the coherence layer over you site and observing how it has changed over time can give you insight into how events like seasonal vegetation or rainfall can affect the quality of your InSAR signal throughout the analysis window.