sarproz-group

analysis of APS results

This topic contains 2 replies, has 2 voices, and was last updated by  cp02 5 months ago.

  • Author
    Posts
  • #2668

    cp02
    Participant

    Dear Prof. Perissin,
    I’d like to ask you for any suggestions and opinions about the analysis to better understand the parameters involved in the processing.

    I used Sarproz to analyse 21 Sentinel 1A images over one year. In the preliminary processing I analysed an area of 8 km (radius) for the coregistration of all images. In the final part I used the module “Small Area processing” to analyse only my study area, which includes a dam.
    In the first part, after the preliminary geocoding and the InSAR processing, for the APS estimation module I used the STAR-graph and other parameters included in the attached figure (weights–> coher; smart=5).
    To estimate the non-linear movements and the correlation with water levels during this period for my study area I set up the option smart=5, and after importing the data as “aux data” in the “Select Dataset” module (normalizing the data between -0.1 and 0.1) I checked the estimate for the “azimuth pos” and “seasonal trend”, reading the estimated “Height” in the PS processing within the “Small area processing” module.

    Using the APS estimation I found different results using the Inv. Par and Inv. Res. Could you please suggest to me which is the best choice and the difference using both? And also, I’d like to ask you if there is any possibility of using external data for atmospheric correction.

    Please, find attached results from this analysis.
    Thank you and congrats for the software.

    Attachments:
    You must be logged in to view attached files.
  • #2675

    periz
    Keymaster

    1. 20 images is a relatively low number. so, you have to be careful in choosing things properly
    2. using the coherence as a weight means using the spatial coherence to keep coherent data and discard not coherent one. if you use the STAR graph, you should not expect high coherence, because of STAR graph long baselines. it means discarding data. it makes no sense. use the coherence as weight only when you have short baselines redundant graphs. don’t do it here.
    3. the smart option (temporal filter) is a powerful non-parametric estimation. however, it has more degrees of freedom and I do not suggest to use it here with only 20 images. go with a linear trend.
    4. I also suggest you to keep limited ranges of height and velocity to estimate at the beginning
    5. do not use the inverted parameters APS. it forces a solution even if wrong.
    6. if you want to estimate the correlation with water levels, do not use the smart filter. also, you need more images or you will overfit your data. the same if you estimate seasonal trends.
    best

  • #2689

    cp02
    Participant

    Thanks for your help and suggestions.
    Best regards

You must be logged in to reply to this topic.