This forum is shutting down! Please post new discussions at

SDM model - output values


I am running the SDR model in a region in Brazil, but the output raster have values apparently too high. I would like to know if it is expected or if I need to fix something.

For example (output raster):

Usle raster
min 0
max 2,970,429.5

sed_retention index raster
min 0
max 1,443,183.25

Cell size of the output raster (as define by the DEM raster that I  used): 999.7, 999.7 m

Some information about input raster:

Soil erodibility (K)
min 0.0132
max 0.0534
ton h MJ-1 mm-1 (as defined in the source that I got the values)

Rainfall erosivity index (R)
min 4,184
max 11,806
MJ mm (ha h)−1 year−1 (as defined in the source that I got the values)

Obs: I have already filled hydrological sinks in DEM raster using Arc Hydro Tools.

Thank you!



  • RichRich Administrator, NatCap Staff
    Hi Fernando, sorry i don't have an intuitive sense of how much sediment should be exported, but looking at only the max can be a little skewing.  Also 1km square cells are pretty large...

    Do you have a sense of why you think that number is too big?  Also if it were me, I might try to see if those large export sizes were common in my outputs, or outliers.  If outliers, it might be worth tracing through the raster stack output to see what might have triggered it.  In the past I've seen some large USLE numbers because of a DEM that had a very steep slope in one point due to noise in the DEM itself.

    I hope something is helpful there.  I'd be happy to help further if you'd like to post more.
  • Hi Rich,

    thank you for your fast answer. I've evaluated the sed_retention_index raster in the ArcGis and also made a histogram in R... I could see that the large numbers are outliers and they are associated to areas with very steep slope.

    Also, when I converted to ton/ha/year, the values do not seems too high. In the sed_retention_index raster, values ranged from 0 to 14,431.83 ton/ha/year; mean 13.10 and median 3.00 ton/ha/year. So, I feel more confortable with these values...

    Thank you again!

  • RichRich Administrator, NatCap Staff
    Hi Fernando, I'm glad you posted again!  We just patched a bug we found in SDR on Friday that we were searching for in part because of your message (I was going to dig out your thread and post on it, but you saved me the time).  If you'd like to get at it now, you can download this development version of InVEST and try out SDR to see if it fixes those outliers:

    And if you're in no rush, we should get out InVEST 3.3.3 soon (maybe today) if you feel like using an official release, although there won't be any code differences between the two.
  • fmrfmr Member
    edited February 2017
    Hi Rich,

    thanks to let me know that a new version is available! I want to use this one... 

    I changed a detail in my input raster (R, K and LULC raster) and I ran in Invest 3.3.3. The sed_retention index raster of the both versions (v.3.3.2 and 3.3.3) are very similar and indeed the v 3.3.3 reduced the value of the outliers. The max values reduced from 14,431.83 to 11,905.3 ton/ha/year and the mean from 13.10 to 12.70. The median continued similar (3.00).

    Curiously, using my new input raster (which are basically the same as the previous one, but with  more cells) I had some warnings in the log file... I do not know if it is serious, could you please check that? I could also send you my files by e-mail...

    Thank you.
    Post edited by fmr on
  • RichRich Administrator, NatCap Staff
    Hi Fernando, those warnings are okay.  They're saying the floating point representation of some numbers will get truncated when copied into the output shapefile fields.  Seeing as how the decimals are so long, you won't lose any significant digits.

    Another user reported this warning too, but I'm unable to recreate it on my end.  Would you be willing to dropbox your data to me at  If so, I could try to fix this and patch it for the next release.

    Otherwise nothing to worry about on your end.
  • Hi Rich! 
    It is good to know that those warnings are OK! Nice!
    I have just sent to you a link with my data.
    Thank you!
  • RichRich Administrator, NatCap Staff
    Thanks Fernando, I got it and can recreate on my end, though I'm not quite sure what to do about it yet.  On one hand it's easy to add additional digits of precision to the floating point field widths.  On the other it might not make sense to report that the model retains exactly 208477516.68314272 tons of sediment either.  

    But anyway, it's okay to disregard those warnings on your end.

  • Hi Rich,
    For me it is OK, I think that it is not necessary to report so many decimals!
    Good to know that the warnings were just that... Thank you a lot for your help!
  • fmrfmr Member
    Hello Rick,

    I have used the "sed_retention_index" raster to map the service of sediment retention. But I would like to confirm if it is better to use  the "sed_retention_index" raster than the "sed_export" raster.

    I think that the  "sed_retention_index" is a better option because it takes into account the SDR ratio (and then the flow path and connectivity index). It seems to me better than use only the "sed_retention". Is it correct?

    Thank you.
  • RichRich Administrator, NatCap Staff
    Hi Fernando, sediment export *IS* calculated using SDR.  The retention index is a rough measure of how well the landcover is retaining sediment at that pixel.  Here's the user's guide description if it's helpful:

    sed_retention_index.tif (tons/pixel): Index of sediment retention, used to identified areas contributing more to retention with reference to a bare watershed. This is NOT the sediment retained on each pixel (see Section on the index in “Evaluating Sediment Retention Services” above).

Sign In or Register to comment.