Terrain Correction, Terrain Flattening or TOPS Coregistration produces blank images with SRTM 3sec DEM
With SNAP versions prior to 6.0.5 the URL to the SRTM 3sec data is broken therefore it cannot download the dem from that location. Please update SNAP to the latest version. In case you develop operational routinges and require a specific version of SNAP, you can update the .snap/etc/snap.auxdata.properties to fix the URL.
For best results in Terrain Flattening, SRTM 1Sec (AutoDownload) is recommended.
S1 TOPS Coregistration takes much time or even fails
The "S1 TOPS Coregistration" module is a chain of multiple separate operators (Orbit File, Split, BackGeocoding, ESD). Each output of one operator is the input for the next operator. As none of these outputs is physically written before the final output product, the "S1 TOPS Coregistration" consumes much of the computer's RAM. Even if your computer has 16 GB of RAM or more, this is often not sufficient to execute all steps in sequence, especially when you select many bursts in the Split operator. If the "S1 TOPS Coregistration" keeps to fail, it is recommended to execute all steps separately, especially reduce the number of bursts as much as possible and make sure that you only use VV polarization. This is usually much faster than executing the coregistration as a graph.
In what order should I apply calibration in my processing?
Calibration usually requires the application of look up tables (LUT) provided in the product. These LUTs correspond to the original pixels in the product. Therefore, calibration should be applied prior to any filtering or resampling of the data.
In what order should I apply terrain correction in my processing?
Terrain correction applies a map projection as well as interpolation to correct for layover, foreshortening and shadow. It's generally advised to perform terrain correction late in your processing chain such that filtering and classification is performed on the image in SAR geometry.
What are the general processing steps for polarimetric processing?
Polarimetric processing requires dual pol, compact pol, or quad pol SLC data. Calibrate the SLC data enabling the parameter to output complex data. Apply polarimetric matrix generation, polarimetric speckle filtering, polarimetric decomposition or classification and finally terrain correction. Please find detailed instructions in this tutorial: SAR Polarimetry and Analyses
I created a graph, but it takes very long to finish
The graph builder helps to combine multiple operators to a longer processing chain. It allows to define processing steps and save the workflow as a XML file. This is especially helpful if you want to apply the same processing to a series of input products (Batch Processing). However, graphs files are not very efficient regarding RAM and data in the cache. There are some options to increase the RAM which can be used by SNAP (see below under 'Related questions'), but it is limited by the RAM which is installed in the computer.
Graph files are not recommended for the analysis of single products, because of the following reasons:
Executing a graph can take much longer than executing the involved steps individually. This is especially the case for very large graphs.
Graphs do not show the intermediate products. It is hard to keep track of the processing chain in case of errors. Again, executing steps one after another and checking their outcomes gives you more control on the analysis.
Accordingly, working with graphs is most efficient on machines with large computing capacity and for the automation of workflows (e.g. to apply the same processing on multiple images).
SAR images are displayed and processed in the SAR acquisition geometry. Depending on if the satellite is in an ascending or descending orbital pass and right or left facing, the image may appear as flipped when compared to the typical North is up convention. After the map projection step in terrain correction, images should appear in the correct orientation for the desired projection.
How can I derive a DEM from SAR data?
Radar interferometry (InSAR) allows to retrieve surface heights from a pair of complex SAR images. There are many things to be considered, for example, the selection of images with suitable acquisition dates and geometries. If you are new to InSAR, please have a look at the documents compiled by ESA: InSAR Principles -Guidelines for SAR Interferometry Processing and Interpretation (ESA TM-19):
Part A is for readers with a good knowledge of optical and microwave remote sensing, to acquaint them with interferometric SAR image processing and interpretation.
Part B provides a practical approach and the technical background for beginners with InSAR processing.
Part C contains a more mathematical approach, for a deeper understanding of the interferometric process. It includes themes such as super-resolution and ERS/Envisat interferometry.
One crucial point to understand is that radar interferometry is very sensitive to various parameters which make the results unusable. Movement of objects at the ground and water vapour in the atmosphere lead to noise in your data. Especially vegetation on the ground leads to temporal decorrelation of the phase. It is therefore important to check the quality of the products after each step (coregistration, interferogram formation, phase filtering, unwrapping, conversion to height), because errors which occurred at early stages are propagated throughout the entire processing.
Because of the sensitivity of the phase towards the movement of canopies, is not possible to extract the heights of vegetation from a single image pair with repeat-pass interferometry (images acquired at different dates). Accordingly, any DEM will have errors over vegetated areas. The severity of these errors depends on the acquisition geometry (a perpendicular baseline of 150 - 300 meters is recommended) and the time which has passed between both image acquisitions. Currently, the 6-day revisit cycle of Sentinel-1 does not allow to create reliable DEMs of most of the earth's surface. Please also have a look at this comment.
Please have a look at the following tutorials on radar interferometry:
I have performed an InSAR analysis, but the result is not correct
SAR interferometry can deliver exact measurements of the earth's surface and its deformation, but it is prone to a number of error sources which can affect the result:
Temporal phase decorrelation (time between master and slave image is too long)
Atmospheric disturbance and ionospheric delay (atmospheric artifacts in the phase, e.g. caused by rainfall during image acquisition)
Unsuitable perpendicular baseline (distance between the position of the satellites at the time of image acquisitions is either too large or too small)
Inaccurate coregistration (master and slave image are not overlayed precisely)
Unwrapping errors (low coherence areas lead to random unwrapping results)
If the conditions for InSAR (little to no vegetation, suitable baseline, no rainfall during image acquisitions,...) are not given, your interferogram will contain noise and artifacts and not produce accurate results. You should keep an eye on the coherence as an indicator for the quality which can be expected from your product. Low coherence will mostly lead to unusable results.
Although radar interferometry allows measuring surface changes, it is not a suitable technique for rapid or abrupt processes such as landslides, erosion, rockfalls etc. because of the following reasons:
These processes are non-coherent changes, which means that they alter the surface characteristics of the investigated surfaces. This leads to changes in scattering mechanisms and finally to decorrelation of the phase. DInSAR does not work for areas which are covered by vegetation (volume decorrelation) or areas which have undergone significant changes.
The maximum displacement occurring between two images is determined by the wavelength of the sensor. Even coherent processes which exceed the length of the wave (e.g. 5.6 cm for Sentinel-1) cannot be measured reliably.
Atmospheric disturbances can furthermore introduce fringe patterns which are not related to surface displacement, so it is important to have images with suitable (dry) atmospheric conditions.
The amount of measured displacement is also dependent on the looking direction and incidence angle because the displacement is measured along the line of sight (LOS). It is worth spending some time searching for a suitable image pair which captures the movement relative to the sensor best possible.
An alternative to DInSAR for measuring mass movements is the offset tracking module in SNAP.
I calculated displacement with DInSAR, but the resulting range of values is unrealistic
This can have two reasons:
Processing errors. Radar interferometry is very sensitive to many potential error sources. Especially if the interferogram quality is not good, unwrapping will produce false results (both regarding their spatial patterns and the value ranges of the displacement) or induce unwanted ramps. Please carefully have a look at this question: I have performed an InSAR analysis, but the result is not correct
Even in case of good data quality (high coherence and clear interferogram patterns), it is important to understand that only relativedisplacement between two images is calculated. The resulting raster can still contain a general offset which has to be corrected manually. To do this, you identify an area where you do not expect surface displacement with high coherence, read its displacement value (e.g. 15 mm per year) and add this offset to the original displacement raster in the Band Maths. The result will be a displacement raster corrected for this offset. An example is given in this tutorial (chapter "Check and correct for offsets").
To determine the offset in your displacement product, it is not recommended to look at extreme values. Minimum and maximum values are often only statistical outliers and produced in cases of unwrapping errors. Instead, digitize a line across both stable and changing areas in the raster (as explained in this tutorial, Figure 26) to check the pixel value in areas which are expected stable. The offset in these areas should be zero, so the value they show should be added or subtracted to the result. Please also have a look at these discussions in the forum
Although backscatter intensity is correlated with the dielecticity of a surface or volume (Figure 8), there is no constant relationship between the measurement and soil moisture, because backscatter intensity is also affected by further parameters, such as surface roughness or vegetation cover. The retrieval of soil moisture from radar data, therefore, requires a study design which includes the collection of soil moisture measurements in the field (at the time of image acquisition), the application of multiple input images, or the integration of polarimetric information. Many approaches have been published on this topic, but there is no standardized way of soil moisture retrieval at high resolution. Global products exist at the spatial resolution of 500 meters or 1 kilometre (SMOS, ASCAT).
Please also have a look at these discussions in the forum: