10

This is my first post here.

I have some points from a soil depth survey and I need to interpolate the depth for a whole area. What is the most appropriate technique for this and is there anything I need to take into account when using the technique? A pointer to a tutorial would be brilliant!

Thanks

Improve this question
4
  • 1
    You mean you want to create a new surface? As opposed to a simple mean of the depth. Do you want to infer variation in sparsely sample parts based on the variation in more closely sample patches?
    – Willy
    May 1, 2012 at 12:46
  • Yes, I want to create a new surface. I want to infer the depth over an area based on a number of point depth samples.
    – jpugh
    May 1, 2012 at 12:53
  • Do you have any preference in technique ? Software (Open Source or Arc/proprietary), script etc?
    – James Milner
    May 1, 2012 at 15:53
  • I want to use Open Source if possible and I've done a bit of Python (I'm learning it so I can use GDAL/OGR and QGIS).
    – jpugh
    May 2, 2012 at 8:37

5 Answers 5

Reset to default
10

This is not a pure GIS problem. When interpolating soil depths, you need to apply the GIS techniques on the basis of a geomorphological hypothesis. In other words, what is the geological history of the area you are mapping? What are the soil formation process in operation? Are there landscape features that limit the processes (eg. barriers such as rivers, etc)? What are the imlications of catena and other soil and landscape processes? Generally I begin such studies by defining geolomorphological and landscape regions (upper slope, mid slope, lower slope, alluvial, colluvial, etc) areas, and then do my modelling. So, you might have a model for one geological region and a different model for another geological region. Or models based on landscape (upper, mid, lower slope). THEN you apply the GIS technique.

Improve this answer
3
  • 3
    Yikes! I just KNEW this was going to be more complicated than just slapping a function on the data! I'm not a soil scientist but can you point me in the direction of any simple literature so that I can at least know how to speak to a soil scientist or know what caveats to put on my results? Thanks.
    – jpugh
    May 2, 2012 at 8:33
  • 4
    A big +1 for emphasizing the science and modeling aspects of this problem.
    – whuber
    May 2, 2012 at 13:54
  • I think you're best to try & work with a geomorphologist. They have a skillset that allows them to interpret landscape processes as they relate to soil formation. Bore-holing folk and soil engineers are rarely able to help because they tend to be interested in site level stuff rather than landscape scale stuff. You often need to look well beyond the boundaries of what you think is your site in order to understand this. Otherwise, try and find some archival soil surveys. Even if they're tens of years old they will still be useful. Good luck.
    – Ian Allan
    May 4, 2012 at 12:22
2

You will want to interpolate your soil data using Kriging or Euclidean allocation processes. Here are some helpful links:

http://resources.esri.com/help/9.3/arcgisdesktop/com/gp_toolref/geoprocessing/surface_creation_and_analysis.htm

http://forums.esri.com/Thread.asp?c=93&f=997&t=161054

http://forums.esri.com/Thread.asp?c=93&f=995&t=237635

You can run kriging both in ArcGIS and Grass/R (open source).

Improve this answer
5
  • 3
    Euclidean allocation, although appropriate in some circumstances, is probably unsuitable for soil depth data. One of the first lessons one learns when studying interpolation methods is that data characteristics and analytical objectives determine the applicable methods; no single method is universally appropriate for all datasets. (That's one reason there are so many of them...)
    – whuber
    May 1, 2012 at 19:30
  • @artwork21. Thanks for the links, these look really helpful.
    – jpugh
    May 2, 2012 at 8:27
  • @whuber. Thanks for the correction wrt Euclidean allocation. Do you know any good sites with simple explanations to help me understand interpolation?
    – jpugh
    May 2, 2012 at 8:29
  • Is kriging the way to go here?
    – jpugh
    May 2, 2012 at 8:40
  • 3
    If you're not already expert at interpolation and statistical analysis, then kriging is not the way to go IMHO: it's a highly technical arcane field with specialized terminology and practices. The semi-automated versions found in many GISes merely use the underlying computational machinery of kriging to interpolate data: with luck it may work ok, but it certainly has none of the attributes of kriging (such as being a "best unbiased [linear] predictor"). Worse, kriging is computationally intensive. One should look at simpler procedures first, including splines, natural neighbors, etc.
    – whuber
    May 2, 2012 at 13:44
2

Ian is quite right in his comments, but supposing the soil model consists of two simple layers, then the problem becomes one of interpolation method.

The interpolation method then needs to be suitable for the intended use of the final product. What resolution /scale are management activities going to be carried out on? What scale was the data gathered on? What is known scale of variation of the features of interest?

Soil depth is seemingly benign term. But it can the depth of a layer which is only defined by pH or soil colour. These features may have variation on a very fine resolution, always moreso than the scale of sampling. But the management action might be only applied on a whole patch, encompassing a great number of sampling points.

So explore these issues a little and the apporpriate methodology and tools will reveal themselves.

Improve this answer
1

QGIS now has a Kriging plugin (see here).

Ian Allan has a good point and well made. As a GIS guy who has worked alongside soil scientists I know that soil can really muddy the GIS waters (sorry I couldn't resist!). Seriously though, if you can find a soil scientist who is willing to spend a bit of time with you, they will be worth their weight in gold. You could start by reading up on the subject (e.g. here, here and here - forgive the slight bias towards peat, but these are really just a starter! A bone fide soil scientist is your friend here).

Improve this answer
0

Regarding soil depth, you should definitively use geostatistical methods, which give you more strength to the results. These methods prform best in conjunction with environmental covariates, cause you probably have a sparse network of soil depth locations, and the soil is not topography. Please take a look at these resources:

ISRIC-WISE soil database interpolation

Regression kriging example, exactly for soil depth

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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