I have prepared these two maps of Electrical Conductivity (EC). I used both IDW and Kriging to prepare them. Which one should I choose? The sampling sites are also pointed in the map.
You can see that the Kirging map is quite strange!
I am in a fix. Please help!
Kriging is better than IDW but it looks like there is a problem with how the autocorrelation structure is being defined, so you're just getting the mean. It's hard to say what is happening without further details, you need to check the variogram model being used. It's very possible that spatial interpolation is not a good idea with this data.
Dear Baig,
These two methods have some advantages over each other. For instance, IDW works better in which there is hard breaks or changes within data.
Stephen Blair Stewart Sir, I tried with the other variogram models as well (exponential, circular, .......), but it shows the same result.
Not to put too fine a point on it, but eyeballing it, both are relatively meaningless. Interpolation is not magic, there needs to be a certain density and distribution of samples. Algorithms will blindly give you a result no matter what you put into them. There needs to be enough points for there to be a 'neighborhood'.
If a sizable portion of your points are on the convex/concave hull, that is a problem ( make a Delaunay triangulation https://mathworld.wolfram.com/DelaunayTriangulation.html). If you are using ArcGIS, this all described fairly well with rules of thumb in the Help documentation.
Kriging perform better than IDW but need more tune to be used. Your map shows values uniformly set to the average. May be you have to study the spatial correlation of your data and set the variogram parameter according to your data. May be you have to use a larger range.
The other question I would have is the 'representation' of the data, i.e. the assumption that any values between the measured points are members of a more or less continuous field. Another way to think of this is sampling rate. How much variability is there over what sort of distances. This is dependent on the physical properties of the phenomena being addressed. It is sometimes useful to triangulate with another 'proxy' dataset which ideally might have some relationship to the interpolated values one is generating. ( I'm not a 'soils' person ) Maybe Ph is related to conductivity, for instance. Another way to check is by decimating the original data, does the interpolated surface change greatly when sample points rare removed - if removing one or a few random points generates and entirely different result, on my look, it is under sampled.
hi dear
please have a look the following:
Article Comparison of spatial interpolation techniques for mapping s...
https://research.wur.nl/en/publications/soil-salinity-mapping-by-different-interpolation-methods-in-mirzaand also some of my publications about soil salinity and geostatistics
Kriging, no doubt. However, to get a correct kriging map, you must use correct and reliable variogram model that defines the spatial correlation between your data points. hope it helps.
Dear Baig,
Kriging mostly depicts good results,espevially when you are using some model. But if, there is enough number of variations in the data, then you can go for IDW. Depending upon the prime aim of your work, you can choose it accordingly.
It depends on the pattern, variations and availability of data. If there is more variation in data and you can't fit well the variogram model with the data, go with the IDW. If otherwise, Variogram fits will with the data, I recommend the Kriging. Just play with the alternatives and decide accordingly.
Dear Baig Abdullah,
It is better to judge according to difference measures
look at the following article
DOI: 10.3923/tasr.2012.467.475
Best Wishes,
Saeed
Hello dear Shoumic. I would like recomendation to you, this paper and you can downloading. But if you let me, I wana talk about this article to you and if you attention section Kriging in the this paper, one thing is very important, and that is uncertainty. I don't know, you how many information about this word (uncertainty). Therefor, I would like one book to introducing you, and you can using it about uncertainty. If you read 5 page on this book, that providing more information about uncertainty to you and then you could better juge about using method Kriging or Idw. If you read page 5 on this book, that's providing more information about uncertainty to you, and then after study that, you could better, one select two methods.
link paper: https://www.intechopen.com/chapters/52704
Book address: STATISTICAL METHODS IN THE ATMOSPHERIC SCIENCE, AUTHOR: DANIEL S. WILKS, FOURTH EDITION, 2019. PROFESSOR EMERITUS DEPARTMENT OF EARTH & ATMOSPHERIC SCIENCE CORNELL UNIVERSITY, ITHACA NY, USA.
Thanks a lot.
Arman Jahedi Thank you for that paperChapter Comparison of Spatial Interpolation Techniques Using Visuali...
Should be required reading. Excellent illustrations.
Kriging is better but consumes more computer resources. The Kriging result showed in the question is a failure, I guess the reason was too small radius search for interpolation. Try increasing the radius search and you will get a better result.
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Fixatives