-2

I am not sure if this is the right place to ask the question, but I don't know where to ask. So please excuse me if not so.

I am considering two sampling strategies, the first is locating samples on the three black lines as the first picture shows, the second is random sampling across the whole area as the second picture shows. Both ways will use the same number of samples, the difference is the pattern. samples by the line: enter image description here random:
enter image description here

My questions are: 1. Will they generate different variograms and produce different kriged maps?? 2. Which one is better? 3. Can I use the two sampling strategies in ONE kriging? I mean if I have two areas, can I use random way in one area and 'line sampling' in another area, and use both data together to generate a variogram and mapping?

Vince
  • 20,017
  • 15
  • 45
  • 64
Summer
  • 123
  • 2
  • 9
  • I think this would be better asked in Cross Validated Stack Exchange site, but you would have to describe more precisely the analysis (what it is being studied, what is expected from such analysis, the study area, etc.). If you go that venue, use the experimental-design tag. If you want to keep the question here, it is also on-topic, but you still need to add more details as suggested. – Andre Silva Mar 10 '19 at 00:28

1 Answers1

3

There's massive literature for sampling schemes for spatial interpolation. Here's some thoughts:

On 1: yes they will generate different empirical variograms and different kriged maps. But if the underlying data is independent of your sampling scheme then the outputs should be in agreement within standard errors. But if your data are somehow related to your sampling scheme (maybe you have geological features running along that transect such that you sample the same thing lots of times) then all bets are pretty much off.

On 2: "better" depends on a multitude of things. Firstly, what are your criteria for "better"? Prediction error at some set of points? Integrated prediction error over regions? Inference on possible covariates? Better predictive probabilities of some threshold exceedence?

On 3: Yes, you can use both sampling strategies, combine the points, and do a single interpolation. More data always helps.

General opinion on sampling schemes is: structure your scheme to reflect what you think the underlying spatial correlation range is likely to be; have a mix of close points (to get a handle on short-range correlation) and far points (for long-range correlation); random sampling is generally better on most measures than sampling along transects except things like "time taken to do survey" or "fuel cost to survey sites". Transect sampling is often a convenience measure because a researcher can walk in a straight line taking samples rather than randomly wander (and get lost...).

R makes it easy to investigate this - generate an underlying raster (like the one in your image) and try various sampling schemes, generate variograms, compare, generate predictions, compare within standard errors. Play.

Spacedman
  • 63,755
  • 5
  • 81
  • 115
  • Thanks for your answer @Spacedman. I have searched for some literature but they all compared different sampling designs and didn't use two at once. – Summer Mar 09 '19 at 10:56
  • I have to sample crop yield data in two adjacent fields, one is random sampling and another is transect sampling. I have a feeling that data from random samplings should be interpolated separately from transect sampling (all other conditions are the same), but I cannot explain the 'why'. One reason, to my knowledge', is that the transect sampling might generate less accurate map that random sampling. Another thing that there is no sample close to the border/edge of the transect sampling field (as the image shows) also concerns me and thus I think the mapping would be better off without it. – Summer Mar 09 '19 at 10:57
  • Could you help with the above and or kindly show me where I can find relevant literature? Thanks. – Summer Mar 09 '19 at 11:03
  • 1
    Sorry @Summer for butting in. I am a bit worried if you are trying to apply your transect sampling scheme on the area you had in the other post because it poses unstationarity issue. Certainly the sampling along the transect will add bias on your results. If you are not in a hurry I would recommend to calculate representativeness which is a similar measure like variogram - range. (It is a half of the radius of area which is similar to your center cell). – Kazuhito Mar 09 '19 at 12:34
  • @Summer You may think Spacedman's first part On 1 is reasonably generic. I think you would better to state why/how your transect pattern (NE-SE) is planned that way. – Kazuhito Mar 09 '19 at 12:42
  • What does the background colour show, is it just an illustration or does it represent something about the space you will sample? – AnserGIS Mar 09 '19 at 16:13