I think kriging is the best solution, especially local kriging to interpolate for small areas the using spline or any other method these local interpolations may be smoothly merged together. All these operations may be carried out in ArcGIS or Surfer.

Hello Syed - I assume that each of your data points also has spatial data associated with with, such as lat/lon or UTM and elevation. Rather than trying to plot the soil description as text (e.g., clay, clay-sand, sand, etc), you could give each of those descriptors a numerical value. ArcMap, Isatis, Petrel, and other software packages might struggle to plot non-numeric strings. My one recommendation for free software is GoogleEarth. You could plot the non-numeric data or numerica data as icons. The website earthpoint.us is a great resource for converting Microsoft Excel spreadsheets to GoogleEarth's KML format (http://www.earthpoint.us/ExcelToKml.aspx). There is a $5/month charge to access that feature, but it is well worth it. Good luck, Ben

Hello Syed,

the first thing I suggest you is to recode your point data assigning to each soil type a number. I also suppose that for each point you have the following information elements:

- ID

- Soil Type

- X coordinate

- Y coordinate

After converting "Soil Type" from a "name" to a "number" then you can apply some interpolation (also called gridding by some software).

The steps that most software require are the following ones:

1) Convert your points into a text file, with a header line and then each line will contain the 4 information above separated by a tab (or other character). This is easily done for example using MS Excel or OpenOffice.

2) Execute the interpolation tool, first importing the points from the text file and then chosing the resulting raster projection, extension and pixel size. Also the interpolation method has to be indicated.

However there is an important issue related to the values that the interpolator can use. Let say you have:

a) Clay class =1

b) Clay-sand class=2

c) Sand class =3

d) Water class=4

When interpolating the areas between two classes the interpolator could assign intermediate values (e.g. by applying linear interpolation) or simply assign the value of the nearest point (e.g. by applying nearest neighbour interpolation). In some cases it can cause errors.

For example when interpolating an area between two points classified as a) and c), the intermediate value of b) can be assigned. This in principle can be correct. But when we consider two points classified as b) and d), the intermediate area can be classified as c), making a mistake.

You can mitigate this problem by chosing the valued with care, but I believe that the nearest neighbour approach should be used in this case.

Some software with a gridding tool is ERMapper (Gridding wizard) or ERDAS Imagine (Surfacing tool) or ENVI (Rasterize point data).

Hope this helps.

Giulio

The conversion of point data into polygon data is achieved using Geostatistics. This was the work of Kriging. There are many software free and propriety which can achieve this. There are many models and implementation, however selection of the most suitable required experience.

The geostatistical interpolation methods tend to fail in the determination of the boundaries between the soil classes. I recommend using aerial photographs or satellite imagery and elevation data (DEM) for a final adjustment.

Risso answer is right!

In case we have only point data, Geostatistics is used for that.

Of course land surveying, satellite images or any other geospatial source will enhance the output.

Can u get hold of a soil atlas of that area? If yes then digitise that and try overlaying ur points over that map. That will give you a better picture

I think one of the key factors that help you to select the mapping procedure is the distribution of the points in the area and the distance between points. if your points are closely spaced and cover the whole area, the interpolation may be quite favourable. it not, I recommend to overlay your data above an enhanced satellite image in the GIS and delineate the various types of soil based on visual interpretation of the image. Good luck

Thank you all

I have similar problem.

I have a set of 65 X Y coordinated points with % of slit, soil and sand. I plotted the points on USDA triangle and found the classes. So I have a separate column of texture for these points data in excel.

Now comes the question of interpolating textures.

1. If I assign value of 1 to say clay class, 2 to say silty clay and so on... after interpolation the raster created will has value of say 1.2 by IDW; what is meaning of this value. OR shall I go for density map with colours assigned to each circle to show just variation of classes.

2. If I interpolate from 3 rasters created from sand, clay, silt percentages, how can I write rule in GIS to classify and create a new raster layer defining texture classes,

I am using ARCGIS 9.3.

Any help appreciated.

I think kriging is the best solution, especially local kriging to interpolate for small areas the using spline or any other method these local interpolations may be smoothly merged together. All these operations may be carried out in ArcGIS or Surfer.

If your data is quantitative and you have at each location clay, silt, sand contents, than I agree with Mr. Liming Ye and you should interpolate each fraction separately, using methods such as kriging or regresion, or regression-kriging. If your data is categorical, that is if you know only the textural class and not the proportion of the different fractions, than you should use methods such as the logistic regression and derive spatial probabilities of textural classes. Afterwards, you can combine the spatial probabilities in order to derive a texture map, using, for instance, SoLIM software

Sounds like you are interested in geostatstical methods for compositional data. I would start here:

http://books.google.com/books?id=3tj2f2_IOSkC&printsec=frontcover&source=gbs_ge_summary_r&cad=0#v=onepage&q&f=false

This book provides methods to properly handle the spatial dependence in your point locations while also handling the dependence among variables due to the fact that they sum to 100%. Interpolating independently will lead to inconsistent solutions and also fails to take advantage of the relationships among variables.

Here is how I would approach the problem :

First normalize the sand silt and clay contents to contain within 100. Assign a quantitative number to those points, based on your normlized data. Then go for a normal probability plot for the distribution to define the classes(class intervals). Based on USDA classification you can also give nomenclature to those soil classes. Now define a mask for the points. After that create a interpolation using nearest neighbor method(Kriging is more suitable for irregular / discrete distribution - like minerals deposits). Once that interpolation is ready - corroborate with elevation, geomorphology and underlain geology(including depth of weathering) to fine tweak those boundaries to a more reasonable context. The final result may resemble a good fit of soil layer. But then again you may have to decide from published dta and documents - how good the result is.

I share with you this link to Hengl's paper http://dx.doi.org/10.1016/j.geoderma.2007.04.022 as an ancillary resource to the book of Pawlowsky-Glahn & Olea's book linked by John Kern. I hope you'll find them useful!

Here is another approach for totally qualitative data - where quantification is not possible.

Plot the data points - Use ArcGIS or QGIS. Create theissen polygons for the location points. Now based on those points - assign soil attribute to each individual polygons as it appears for that location within the polygon. By these you will be able to generate a crude soil type dsitribution map. Now use all other background information available like elevation, relief, slope, drainage, depth of weathering, geomorphology, geology etc to finally derive at a pertinent result. Fine tune the same with the available data and records for that area. Hope that helps

Statistica (Statsoft USA) also has the facility to create surface plots that could help generate soil maps from point data. And its quite easy to use.

There are many methods for interpolating a surface from a point cloud. These vary from the simple IDW to the Kriging method. Most of the GIS solution around (both closed and open source) support these methods and so the best for each case will depend on your requirements e.g accuracy of the representation, availability of a validation dataset and so on. One way of evaluating could be designating a part of your initial point cloud data as validation dataset and using the remainder in the interpolation, after which you compute statistics of correlation between interpolated values against the actual values from the validation dataset.

@ David Kuria

We can get interpolated data at validation point on surface map in Arcgis, But to compare that value with respect to actaul, Is it done by using visual interpretation or some utlility to export in say excel or dbf format. So that multiple validation data can be compared side by side.

Can any body describe?

Dear Sainath, how a text field (clay, silt, and sand) can be interpolated, which is not possible, and if we give a specific numeric value to each soil type than the result is not good.

Pay attention for some soil variables that are compositional (particle size fractions). Take a look here (https://www.researchgate.net/post/Interpolating_compositional_data_ie_soil_particle_size_fraction) and here (http://rbras.org.br/lib/exe/fetch.php/pessoais:abtmartins:vsi.pdf)

I am trying the same, but interpolation can't be used. the only solution I find is to make a classified post map by replacing soil types by values. I don't like it but I am still looking for another solution also, @syed Adnan : I see that this post is very old, so if u found a solution please let me know :)

Dear Rasha,

I got soil map from soil survey of Pakistan, digitized and converted it to raster which was required for my study. So its the good option to get it from you soil survey dept. ;-)

A short answer to your question is the following. First you need to interpolate your data to a regular grid. Then you can use a GIS software such as ArcGIS to generate your map.

YOU CAN DO THE FOLLOWING: 

CRATE A THIESSEN POLGON FEATURE CLASS FROM POINT FEATURE CLASS

THEN CONVERT CREATED THIESSEN POLYGON FEATURE CLASS TO RASTER