Unexplained differences between different orbits

Microwave Toolbox
#1

Hello, I monitor the evolution of vegetation on an agricultural plot and I use images of 3 different orbits that cover the same area. I calculate sigma calibration (VH/VV ) according to time. I notice significant differences between the different orbits. Can you help me figure out where those gaps might be coming from? thank you in advance.
Graphique_Resultat-David.ods (33.6 KB)

#2

There are many reasons for different backscatter of agricultural crops

  • Different orbits cause different angles between the look direction of the satellite and the row orientation of the crops. Changes of a few degrees can already cause a difference of 5-10 dB (source). I would try to avoid to combine images of different orbits for agricultural studies.
  • Especially row-like crops cause directional scattering which is very sensitive towards angles (source)
  • Cross-polarization (VH or HV) is sensitive to ear bending of wheat crops (source). This can change quite quickly within several days.
  • Stalk or root lodging of cereals, e.g. caused by high density of grains, leads to an increase of backscatter (source)
  • Rainfalls lead to higher moisture content and larger contribution of surface scattering from the ground, especially over grasslands and spiked grains.
2 Likes
What is the suitable polarization and the orbital pass for crop classification using sentinel-1?
S1 Coherence - Agricultural Zone
Sigma nought values change working with Sentinel-1 SLC
Sigma nought values change working with Sentinel-1 SLC
#3

Hello, many thanks for your answer. In fact, it is a plot of hay that was mowed at the end of April and for which I try to check this date with the evolution of sigma(VH/VV). I am surprised by the variation between the different orbits, especially between ascending and descending orbits.
Thank you

#4

Could you post a clearer plot of what you are seeing?

#5

Hello,
The plot is located in Soumans (France) POINT(2.3002961323439663 46.30479816608185). GRD Images are processed according to the following process: Subset, Apply Precise Orbit, Calibration, dB conversion, Terrain correction, write as GeoTif. With QGIS, I extract the parcel and calculate the Sigma(VH/VV).
Here are two pictures of the plot:
Orbit 59 (Ascending) the 4th of May, mean value is -3.8

Orb 59
Orb 59682×576 3.33 KB

and Orbit 132 (Ascending) the 3rd of May,mean value is -6,00
Orbite 132
Orbite 132684×584 4.73 KB

Thank you for your attention.

#6

Here is the geographic coverage of the two images

image
image913×552 274 KB

The overlapping part correspond to portions observed with different incidence angles.
Did you compare with images acquired with same relative orbits ?

#7

hello,
I did the calculations for the 3 orbits 110, 59 and 132 and for each image between the beginning of April and the 20th of May, that is between 8 and 9 images per orbit.
For my part, I improved the process to include ‘Thermal Noise Removal’, ‘Remove Border Noise’ and ‘Terrain Correction Flattening’ operations to reduce the influence of incident angle. The results change, but the answers still don’t converge.
Sigma (VH/VV) differs from Sigma (VH) and Sigma (VV) indicators, each fairly close to one orbit to another. So I continue to look at how to explain these phenomena.