FLATSIM Product Manual

Products delivered by the FLATSIM service

We present here the products generated by the FLATSIM project as of January 2021. FLATSIM stands for ForM@Ter LArge-scale multi-Temporal Sentinel-1 InterferoMetry. This project aims at processing Sentinel 1 data on large areas over the whole available time range. Processed areas are selected on a scientific basis.

Some processing information can be found here: Thollard, F.; Clesse, D.; Doin, M.-P.; Donadieu, J.; Durand, P.; Grandin, R.; Lasserre, C.; Laurent, C.; Deschamps-Ostanciaux, E.; Pathier, E.; Pointal, E.; Proy, C.; Specht, B. FLATSIM: The ForM@Ter LArge-Scale Multi-Temporal Sentinel-1 InterferoMetry Service. Remote Sens. 2021, 13, 3734. https://doi.org/10.3390/rs13183734

Once the computation is done, the user is given access to the results. The results are bundled through three types of packages: interferograms, auxiliary data and time series. Each package is provided as a zip file.

There are as many interferogram packages as interferograms but there is only one auxiliary package and only one timeseries package.

Most of the filenames follows the same naming convention:

CNES_{TYPE}_{GEOMETRY}_{DATE1}_{DATE2}_{LOOKS}rlks.{EXT}

where

TYPE documents the file content. See hereafter for a more detailed description.

GEOMETRY is set either to radar for radar geometry and geo for ground geometry (in lat/lon coordinates).

DATE1 (resp. DATE2). These optional fields corresponds to the date at which the first (resp. second) image has been taken. They are provided as eight digits in the form YYYYMMDD, as for example 20040529.

LOOKS is the number of looks used to downsample the image in radar geometry. When a grid is transformed into ground geometry, the lat/lon spacing has been chosen close to the file sample spacing in radar geometry.
EXT is the extension of the file that defines its type:
- tiff contains the data themselves;
- png serves as a preview of the above file. It comes with a legend_XXX file that provides the legend information, typically a colorbar legend with the range information. Only one band is displayed in the png;
- meta a file that contains meta data that are specific to our purpose and cannot be filled in the original tiff file format. It contains for example informations about the parameters of the processing chain or information on the image itself as its DOI, and sensor or acquisition information, as the orbit direction, the relative orbit number, ….

Except for the DEM which is provided only in radar geometry, all the products are generated in both radar and ground geometry. A preview in png is also provided.

In each package, both graphs and small text files that contains statistics of the processing are provided that allows quality assessment of the processing that generated the images. A short description for FLATSIM products is given just below, whereas a more detailed product definition, common for all NSBAS generated products is appended in the next items.

The interferogram packages

Each interferogram is embedded in an interferogram package. These packages contain Atmospheric Phase screen, wrapped and unwrapped unfiltered differential interferograms, and wrapped filtered differential interferograms, and spatial coherence.

APS file (Interferogram Atmospheric Phase Screen from Global Atmospheric Model)

The APS is the total phase delay induced by atmospheric pressure and temperature as predicted by the atmospheric model, defined as the integral of the delay between satellite and the ground elevation for each point in the radar scene (See Doin et al., 2009). The currently available model comes from ECMWF : (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5).

Wrapped differential interferograms (InW)

InW is the complex wrapped interferogram. Wrapped interferograms have a resolution in radar geometry defined by the number of looks taken in azimuth and range (2x8 and either 8x32 or 16x64 looks). The _sd keyword means the interferogram is corrected using the enhanced spectral diversity method. It is followed by the _era keyword meaning that the interferogram has been corrected from the above calculated atmospheric phase screen.

Wrapped differential filtered interferogram (InWF)

As above in 8 looks or 16 looks, but filtered by a triangular sliding window average of footprint 13x13 pixels performed on wrapped (complex) interferograms in radar geometry (as done in Doin et al., 2015).

Unwrapped differential interferograms (InU)

As above, but InU is the filtered unwrapped interferogram. The unwrapping method used is an unwrapping algorithm that starts from a seed in a very coherent area and propagates progressively away from it using preferential paths along the most coherent areas (Grandin et al., 2012). In ground geometry, the pixel spacing is set to be slightly less than the mean pixel size in radar geometry.

Spatial coherence (Coh)

Files Coh file provide the spatial coherence of the wrapped interferogram. It can thus be considered as a proxy of the quality/reliability of the interferogram. Typically, the spatial coherence will be very low at points located on water – e.g. sea, lakes, rivers, snow – and high ( up to 1) at location where there is no vegetation – e.g. deserts, rocky mountains without snow, roads, ….

The timeseries package

The time series data cube (DTs)

This data cube product contains phase delay images at each time step of the time series. It is cumulated through time. When, for a given pixel, no information is available on a given date (because of snow, for example), the value is set NaN.

Differential displacement between two dates can be recovered by substracting the value for the first date to the value for the second date. The time series for each pixel is known plus or minus a constant value. Two strategies are used to define this constant: In radar geometry, the first defined value in the time series is set to zero. In ground geometry, the reference (Phi_ref, t_initial) is chosen to be the origin of a linear fit for each pixel time series. This representation avoids for the APS of the first image to appear in negative on all other dates. However, it is not ideal in case of strongly non linear deformation. The user can choose a different strategy to define the constant as those given here.

The geotiff file has nb_image bands (nb_image is the number of time steps). The unit is radian along the LOS, positive away from satellite.

Band name provides the date of the time step in the format YYYYMMDD. The associated preview file contains only one band that corresponds to the first time step of the cube.

Mean LOS velocity (MV-LOS)

The mean los velocity product has 2 bands :

band 1 is the mean velocity in rad/yr, positive away from satellite

band 2 is a shaded view of the topography (simulated amplitude from DEM simulation)

As in the interferogram package, a set of files are provided in order to asses the quality of the processing, which can be seen as a proxy of the reliability of the data. We now present them.

The time series quality product (Net)

This product shows various indicators of the quality of the time series inversion, in radar or ground geometry.

The geotiff file has 5 bands:

band 1 shows the rms misclosure of the interferogram network for each pixel, in radian

band 2 shows the number of interferograms used in the time series inversion for each pixel

band 3 shows the number of images used in the time series inversion for each pixel

band 4 is a proxy for the temporal coherence, computed from triplets (between 0 and 1)

band 5 is a proxy for a possible bias in the timeseries, computed from triplets (in rad)

The Stk-In_list products

These products come as either a meta data file (namely CNES_Stk-In_list_InW.meta) or a png file. The metadata file provides information on the timeseries, as for example the list of dates used as input to the processing, the list of the interferogram used, the referenced image that was selected, …

Auxiliary data package

This set of products provides the user with auxiliary information like informations on the processing parameters, some logs of the processing, …. The main products of this package are:

CosENU: LOS unit vector expressed in local reference frame (East, North, Up components in bands 1,2,3, respectively).

DEM: Elevation in meter in radar geometry. Comes in two resolutions.

LuT: Lookup Tables used to do the mapping between radar and ground geometry. In radar geometry, the table gives the latitude and longitudes of each pixel. In ground geometry, the table gives the pixel number in range and azimuth.

TCoh: Proxy for temporal Coherence and averaged amplitude and dispersion index of the stack of all interferograms or acquisitions.

Other plots and txt files shown in the package are used by scientists validating the FLATSIM products to make sure that some critical steps have been successfully performed: the burst selection, the enhanced spectral diversity step, the unwrapping step and the ramp removal step.

Contributors