The image below is the study (urban) area (central Florida) which features with buildings. My question is about choosing reference point. I studied the online material which says integrated residual height and velocity are supposed to be peak at 0 value. However, there might be that majority of PS are on the roof/structure. In this case, does the rule to choose the reference point (velocity peak at 0 and residual height peak at 0) still works?
I also attached the histograms according to the reference point I have chosen.
first of all,
velocity and residual height are two different parameters belonging to two different physical characteristics, very different from each other and usually not correlated.
then, even if the majority of radar reflections would be higher than the ground level, still, reflections would be at different height values. Unless you are in an are where all constructions have the same shape and orientation and all of them are equipped with the same reflecting features at the same height. Pretty unlikely.
So, I’d say that in 99% of cases, if there is a clear peak in the height distribution, that peak is at ground level. in your case the peak is visible and it is at about -5m, showing that the reference point is on a roof, at about 5m.
About the reference point choosing, i still have a question. When i read the SARPROZ tutorial ‘Demo: APS Estimation Multitemporal analysis with Sarproz’, it says that best reference point should reflects that both velocity and residual height distribution peak at 0. However, in the Site Processing – APS module, i can see 25 auto reference candidate points. I checked every single one of them, and not any of them can do both – 0 peak at velocity ad residual height.
From the manual, it is very clear that i should choose a point on the ground and stable, which corresponding to residual height and velocity peak at 0. My questions is that when there is no such a point, which rule is preferrable. If choose a stable point, it could shows a bias in residual height distribution, like what i showed in the previous images.
By looking at your results again, I notice that the deviation of the estimated height is particularly high: is this Sentinel? If so, the estimated residual height is not very accurate, because the normal baselines are very small. Do not get crazy with the ref point selection.
The suggested ref points are chosen also considering their phase stability. So, they should be fairly equivalent.
If you want, at the end of your analysis you can still correct for height biases.
It is sentinel-1A data. One interesting thing is that in Site Processing – APS estimation, i generated the graph, and it gives me 25 auto point, and i chose a ‘best’ point with good distribution. Then, i moved to Site Processing -Sparse point processing. When i checked the distribution of the parameters, there is a obvious bias in the parameters distribution (esp. for residual height and velocity). I think the distribution calculated here is from all the sparse points, instead of all the pixels like APS does. Do you have any advice on this?
So from your suggestion, i guess that velocity is kind of more important than residual height when choosing the ref point.