Hello,
I’ve formed some Sentinel-1 TopSAR coherence images of the San Francisco area and the bursts look different intensities - see the screen grab of an example below. Does anyone know what I’ve missed out/am doing wrong?
The two source images I used are S1A_IW_SLC__1SDV_20180812T140745_20180812T140812_023214_0285D5_C268 and S1A_IW_SLC__1SDV_20190220T140744_20190220T140811_026014_02E63F_FE85
The graph of the processing I’ve used is shown below:
(this took 11 hours to run!)
I’ve also run the same graph, but without the calibration modules, it took about an hour and the coherence image looked pretty much identical to the first with the same burst intensity issue.
(…on a seperate note - does anyone have any top tips for quickening up this processing too?)
this is not possible, coherence requires complex products. But in the screenshot it is shown that he used complex calibration. I would have done the same, but I don’t have an explanation for the strong variations.
I’ve done it with every 12 days for over the time period of a year (without the calibration modules). The effect is definitely more noticeable with some pairs of images than others. Even with some of the closer times, the effect is present, but usually not as noticeable.
Has any noticeable change occurred in this area during this specific time?
Basically the aforementioned knowledge about the study area should be collected before any of processing, to easier reason any of the phenomena could be seen,
But one thing comes to my mind, I suggest reprocess the only this IW, separately by applying ESD after backgeocoding ,
Please take a look at the following post, to find out the effect of ESD,
Nothing to worry about. You can show that coherence is altered among other by the incidence angle. Here’s another example
Also, if you have to use big graph, don’t forget to modify the amount of RAM that the JVM is allowed to use. Go in Tools >> Options >> Performance. In the Cache Size box, set the amount equal to 3/4 of your total amount of RAM of your computer. In the following figure, taht corresponds to 48032 Mo in my specific case.
Thanks qglaude, that’s encouraging and helpful about the RAM.
(What was taking an hour, now only takes 2.5 mins!!)
However, at the cross-over points between each burst the incidence angle is the same, so I would have thought that the coherence image shouldn’t have a noticeable line between the two bursts?
Yes and no. You’re right in the sense that the angle from the satellite to the pixel is almost the same, however the pulses sent by the sensor are from very different incidence angle (not sure it is called this way).
I’m not sure if you are familiar with SAR processing, but you should not believe that the sensor sends an electromagnetic wave to each pixels and than captures its return. The incidence angle of the emitted pulse is different from the incidence angle of the considered pixel.
The range-resolution varies between sub-swaths even if the incidence-angle is identical. The true coherence of open water is zero and the effective number of looks affects the bias (spurious coherence) in the coherence estimation. The bias only affects low-coherence areas, over land-ice the coherence seems to be essentially the same over the sub-swaths.
Yes and no. You’re right in the sense that the angle from the satellite to the pixel is almost the same, however the pulses sent by the sensor are from very different incidence angle (not sure it is called this way).
qglaude, I understand how a SAR image is formed, but I would have thought that the data for each burst was collected over the same time interval with the pulses of each burst being interleaved. I would have thought that the phase-centre for the synthetic aperture of each burst is virtually in the same position, so that the incidence and squint angle for each burst would be the same at the points where two bursts meet.
Hi mengdahl,
Can you confirm: is the slant range resolution different for each burst? I’ve had a look at your link but I can’t find where it says it is.
if it is, that makes sense. (However if the slant range resolution is the same for different bursts, even though the ground range will change with incidence angle, it will be the same at the points where two bursts meet.)
Dave
@mengdahl - thanks you’re right - the slant-range res is different for each band. I’ve found the specs. I’d imagine that makes quite a difference at low coherence levels.
