Too high LS factor values SDR model

Hey, NatCap folks!
Hope you are doing good and had a great symposium, as always! 

I`m having some very high values for the USLE`s LS factor in recent SDR runs (as they are presented in the temporary folder). The original DEM is the 30m pixel official DEM for Sao Paulo State (http://www.ambiente.sp.gov.br/cpla/modelo-digital-de-elevacao-mde-do-estado-de-sao-paulo ) from where was clipped the elevation values from a watershed in the Ribeira valley. The extract was then processed in the RouteDEM and the pit filled DEM was used as elevation input in SDR model. 

The LS values reached 2900 in some scattered plots and the average value was almost 22, while another LS map calculated to SP state shows maximum value of 17 and 10 in average, for the same watershed (have in mind this other map was derived from SRTM). 

Here goes Dropbox links for: (I haven`t succeed to upload it using the "Attach a file" button)
1) the extracted DEM file (before to run in RouteDEM) https://www.dropbox.com/s/hhu4y4khxy6p51a/DEM_extracted_SPmap.rar?dl=0 

I `d appreciate if you get to take a look and perhaps identify what could cause such big LS values. Is there some pre-processing technique to avoid those outlier values? 

many thanks,
João

Comments

  • PerrinePerrine Moderator, NatCap Staff
    Hi Joao, 

    Thanks for your feedback! 
    We've been aware of this issue for a while, and are working on practical ways to address it --if you have suggestions, they are always welcome!

    To understand what is happening, I'd recommend reading these two threads, where users report a similar issue:

    In brief, I'd say that the outliers themselves may not be a big issue (since they will affect only a few pixels, and the model is typically used at larger spatial scales). 
    The mean value of the LS factor is in my view a bigger concern; I'd suggest to use the calibration factor to tweak these values if you have reasons to believe LS was over/under-estimated.

    Cheers

  • Thanks, Perrine, for your quick response!

    Yes, I m pretty sure the LS values, even the mean values, are over-estimated. 
    Which one of the SDR calibration factors should I use to try this adjustment? Should I reduce or increase its value (in order to reduce LS values)?

    many thanks,
    Joao
  • Dear, Perrine,
    let me reformulate my question: the calibration factor you mentioned for SDR model is the k (Borselli) parameter, right?

    thanks, cheers
    João
  • PerrinePerrine Moderator, NatCap Staff
    Hi Joao, 

    Correct, it's the k_b factor. 
    A decrease in this factor should result in a decrease in sed_export.

    Best,

    Perrine
  • Many thanks, Perrine!

    cheers

    Joao
Sign In or Register to comment.