I have a task to make a researche about tillage detection possibilities. I have found and read many arcitcles so far but I would like to ask you about some ideas or studies you could know about. So my question is
How could I detect tillage (ploughing/plowing) occurence in agricultural parcels ?
I have found studies about usage of radar in agriculture and specifically about tillage but most of them if not all are about identification of tillage type (Chisel, Moldboard,…) and crop residue. There are too complicated, I do not want to identify what is the type of tillage or crop type. My only focus is to identify tillage occurence. Is there some simple method to identify tillage occurence based on the S1 data ? Mainly occurence of deep plughing with Moldboard so it will be easy to identify.
My question is if I need multitemporal data - compare imagery of the same area and detect changes or if it is possible to detect tillage base on one imagery using backscattering value. If tilled soil has bigger roughnes than untilled soil. Would it be enough to evaluate backscattering value if I know which areas are agricultural parcels ? And If so, what variable/parameter should I use (sigma0, something else)? Or could you write me some basic process worflow ?
I would like to tell one more time I need just some basic and simple method to identify tillage for testing purposes. Everything will help.
Start with coherence. A bare soil has high coherence if it is not disturbed in the period between a S1 pair. So, since you can combine S1A and S1B, that means not changed over a 6 day period. If a ploughing event is in that period, you will have no coherence. But in the next period, you should have high coherence (assuming the ploughed field is left uncultivated). You can then check the change in backscattering to see whether the surface became rougher (backscattering goes up, i.e. ploughed) or less rough (i.e. the ploughed field was turned into a seedbed. You need to control for the weather, because soil moisture change and/or snow will also have an influence.
I would like to ask if GRD or SLC product is better for my purpose. I want to detect tillage with 6 days or less difference between images.
Also I have one question about backscatter, to get backscatter value I am supposed to calibrate to sigma0 ?
The last thing is (I don’t want to ask in new question) does someone have tutorial from STEP websites about change detection ? (Error 404 - missing file when I want to download it -> http://step.esa.int/main/doc/tutorials/ ESA TRAINING COURSES -> Corregistration approaches and change detection)
GRD products only contain backscatter information. SLC are complex products (+ phase information). For the coherence solution proposed by @glemoine you will need SLC products.
About backscatter, yes you need to calibrate your GRD product and calculate sigma0 (in SNAP: Radar -> Radiometric -> Calibrate)
By the way, do you have training data. If you know when and where tillage happened , it may help you to see how this affect the SAR signal (at least visually). I am assuming you can match the S1 sensing time with your event.
Thanks for your response . I will have to just make some survey about fields near to me because there is no information about tillage. I know about this problem but there is no database about tillage occurence at some national level. I will have to ask farmers to get validation of results.
Do I have to use SLC for coherence and GRD for backscatter or I can use SLC for both so It will be easier ? Also I have read in this forum it is possible to calculate coherence for GRD.
I doubt there is a dedicated database about tillage, although it would be nice. Apart from asking farmers about tillage timing, you could use the crop calendar. If you know which crop is growing you could deduce when tillage took place. Of course it would be a rough estimation but then you could have a look to several S1 images during that time period an see if something changes.
About SLC/GRD products:
use SLC when you want to use the phase information (InSAR, DInSAR, etc.) or for polarimetry analysis.
If you want to use the backscatter information (normalized to sigma0 by radiometric calibration) then your choice should be the GRD products
My process for GRD images should then be
Apply Orbit File
Corregistrate (Radar-> Coregistration -> Coregistration)
and after I should calculate Coherence of two images and Sigma0 for both images ? The result should have been coherence image and two Sigma0 images ?
Is the order different or is something important missing before the Coregistration ? Because when I try to coregistrate two GRD images by Radar-> Coregistration -> Coregistration it produces master images (VV, VH) which were OK but the slave images were
just black. I have tried it more times and the slaves were just black always.
I will try both. But could you tell me if it is possible to get backscatter Sigma0 value from SLC product ? I know It was told SLC is fr complex information and GRD for backscatter but I would like to know if it is possible to get Sigma0 also from SLC product so I could calculate both coherence and Sigma0 just from one product.
Not sure if directly from SLC products but you could use your SLC to produce the GRD and then get your sigma0 values if you really want to skip the download of GRD products (although I doubt this is a recommended procedure)
Note that the SLC to GRD operator does not produce exactly the same product you can download from the Copernicus Data Access Hub
I have to calculate the coherence of change to monitor tillage from 2 images of Sentinel 1 _IW _SLC taken 12 days apart. It asks me to calculate the phase through interferometry, that is, InSAR. For that, should I process the 2 images independently to finally calculate the phase? My doubt comes because all the suggestions that I have seen here in this forum (and in other sources as well) process the 2 images but these always correspond to the same day and it is not my case.