Kriging interpolation

[tomaszaba]

Use {automap}

[tomaszaba]

Returned surface map (raster) looks odd 🤨:

Where could this be stemming from ...?🧐

[tomaszaba]

  locality  survey interp  bias min_interp max_interp
  <chr>      <dbl>  <dbl> <dbl>      <dbl>      <dbl>
1 El Fasher   19.7   24.3  4.60       1.54       41.3
2 Tawila      26.6   21.7 -4.90       1.64       41.9 

The high bias between original survey results (after removing rows with missing XY coordinates) and interpolated results might be stemming from the fact that sampling points were clumped in a specific region, leaving great part of the locality polygon without observed sampling points. The interpolation in its turn predicted the values for the areas in question by using values derived from previous interpolations...

How to fix this?

One way to address this could be to tell/set the interpolation model to consider a maximum distance; in this way, the model will ignore the area outside the given distance buffer. With {gstat} this can be achieved by using the argument maxdist - "only observations within a distance of maxdist from the prediction location are used for prediction or simulation"

💡The use of of the argument maxdist in gstat::krige() or automap::autoKrige() could be a great strategy in surveys that exclude some areas, such as due to conflict, etc.