I run into issues, when trying to calculate the topographic position index (TPI). I get values ranging from -13 to 17. And the resulting layer looks a bit odd, too:

enter image description here

If I understand it correctly, this index should range from 1-10 (Poster of TPI-creator)

My DEM as a resolution of 30m and I used the default values fo the Tpi-Tool. I'm running QGIS 3.16.3 in Windows 10 and I tried to use "SAGA > Terrain Analysis - morphometry > Topographic position index (tpi)" from the toolbox. (I also tried "GDAL > Raster analysis > Topographic Position Index (TPI)", resulting in values ranging from -80-70ish). I made sure I have the right projection assigned to my DEM. I also made sure there are no holes or anything in the DEM.

It gives me the error message:

The following layers were not correctly generated.
• C:/Users/xxxx/AppData/Local/Temp/processing_flYZSV/1abd3e1d82594e88b04d28cd346343e3/TPI.sdat
You can check the 'Log Messages Panel' in QGIS main window to find more information about the execution of the algorithm.

However, in the log-panel I can't find something helpful:

021-04-16T17:31:35     INFO    SAGA execution commands
   ta_morphometry "Topographic Position Index (TPI)" -DEM "C:/Users/xxxxx/AppData/Local/Temp/processing_flYZSV/10afe001939746859ef23972335147d5/DEM30mSavaSofia.sgrd" -STANDARD true -RADIUS_MIN 0.0 -RADIUS_MAX 100.0 -DW_WEIGHTING 0 -DW_IDW_POWER 1 -DW_IDW_OFFSET true -DW_BANDWIDTH 75.0 -TPI "C:/Users/xxxx/AppData/Local/Temp/processing_flYZSV/642542ca03774ce59275236d376a8fc0/TPI.sdat"
2021-04-16T17:32:00     INFO    SAGA execution console output

   D:\xxxx>set SAGA=C:/OSGEO4~1/apps\saga-ltr
   D:\xxxx>set SAGA_MLB=C:/OSGEO4~1/apps\saga-ltr\modules
   D:\xxxx>call saga_cmd ta_morphometry "Topographic Position Index (TPI)" -DEM "C:/Users/xxxx/AppData/Local/Temp/processing_flYZSV/10afe001939746859ef23972335147d5/DEM30mSavaSofia.sgrd" -STANDARD true -RADIUS_MIN 0.0 -RADIUS_MAX 100.0 -DW_WEIGHTING 0 -DW_IDW_POWER 1 -DW_IDW_OFFSET true -DW_BANDWIDTH 75.0 -TPI "C:/Users/xxxx/AppData/Local/Temp/processing_flYZSV/642542ca03774ce59275236d376a8fc0/TPI.sdat"

   ##### ## ##### ##
   ### ### ## ###
   ### # ## ## #### # ##
   ### ##### ## # #####
   ##### # ## ##### # ##
   SAGA Version: 2.3.2 (64 bit)
   library path: C:\OSGEO4~1\apps\saga-ltr\modules\
   library name: ta_morphometry
   library : Morphometry
   tool : Topographic Position Index (TPI)
   author : O.Conrad (c) 2011
   processors : 8 [8]
   Load grid: C:/Users/xxxx/AppData/Local/Temp/processing_flYZSV/10afe001939746859ef23972335147d5/DEM30mSavaSofia.sgrd...
   Grid system: 30.284487; 5718x 5048y; 231561.120511x 8322967.608079y
   Elevation: DEM30mSavaSofia
   Topographic Position Index: Topographic Position Index
   Standardize: yes
   Radius: 0.000000; 100.000000
   Weighting Function: no distance weighting

So I wonder, what could be the cause? And what else could I try?

PS: I'm sorry I don't know how to attach sample data, otherwise I would do.

  • Why do you think it should be from 1 to 10? That PDF shows maps with "Positive (ridge)" and "Negative (valley)" in the legend. That (I think) is the TPI, in the bottom left panel there is a "Landform" classification which has 10 values, that's derived from the TPI indices calculated previously, it seems.
    – Spacedman
    Apr 16 at 16:17
  • @Spacedman thank's for your hint. I was not aware of this intermediate step. How then would I get there from my current output?
    – Vincé
    Apr 16 at 16:23
  • 1
    Understand your statistics! The Topographic Position Index is calculated using mean deviations, each cell minus its NxN focal value. You will get negative values. The threshold values for a classified surface are not universal because topographic variation is not uniform. Give the source references (De Reu et al., 2014; Gallant & Wilson 2000; Giles 1998) a read. Table 2 in De Reu et al., (2014) gives an idea on how to classify landforms from TPI. BTW, the TPI, with some optional modifications, is available in spatialEco::tpi so you do not have to step out to SAGA or GRASS. Apr 16 at 19:47
  • 1
    And, if you read the poster you link to, you may notice that, under the Landforms section, Andy provides the index scaling and ifelse statements that he used in creating landforms. Apr 16 at 19:53
  • Dear @JeffreyEvans then I'll use your extensive R-Package! Thanks for pointing it out. For me this question would is solved with this. If you'd post it as an answer I would upvote it - just in case anyone is making my mistake as well.
    – Vincé
    Apr 17 at 6:48

As @Spacedman and @Jeffrey Evans pointed out, it is indeed correct to get netative values for the TPI.

The result I however expected from the poster was a TPI based land classification as can be seen here (a screenshot from said poster). enter image description here

There's a good post describing how the different TPI and TRI based land-classifications look here.

And this post describes well how the TPI is calculated.

In the end, it turned out, I had just used the wrong tool for what I expected: Instead of using:

Processing Toolbox > SAGA > Terrain analysis - Morphometry > Topographic position index (tpi)

I should have used:

Processing Toolbox > SAGA > Terrain analysis - Morphometry > Tpi based landrom classification

One has to search for "tpi" not for "topographic position index" in the toolbox's search field in order to finde the later one, which I was not aware of.

As @Jeffrey Evans pointed out there's a classification and comparison between 2 TPI-layers at different scales necessary as an intermediate step.

This results in the following raster: enter image description here

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.