Kriging with External Drift. #2 Data Analysis and Interpolation.

Kriging with External Drift. #2 Data Analysis and Interpolation.
To learn more about Geo RGB, visit us at:
giscourse.online/
Visit us in the Facebook page at:
/ georgbcommunity
Follow us in LinkedIn:
/ geor. .
Follow me in LinKedIn:
linkedin.com/in/marcel-cedrez
Contact us at:
[e-mail]
#geographicinformationsystem #QGIS #SpatialAnalysis #mapping #GoogleEarth #historicalimagery
In this tutorial, we will continue from the previous lesson where a full grid for performing Kriging with External Drift (KED) was created using only QGIS. The first step in this lesson is to analyze the variable of interest, in this case, the zinc concentration. We will perform a histogram and a box plot to observe the distribution of the dataset. Next, we will perform a data transformation to improve the distribution of the zinc concentration.
We will then conduct a bivariate analysis of the covariate, distance to the river, together with the variable of interest. The goal of this analysis is to adjust the parameters to obtain the best deterministic linear model using Ordinary Least Squares (OLS).
We will then analyze the spatial structure of the residuals to ensure that the deterministic model provides adequate residuals for performing KED. We will then fit an appropriate model on the semivariogram and use the script krige() of the package gstat to perform the interpolation. We will plot two maps, one for predictions and the other for variance (map of errors).
We will also execute the Ordinary Kriging (OK) using the same dataset for comparison.
Finally, we will export the final raster created in RStudio to QGIS and reverse the log10 transformation to the real scale of the parameter, zinc concentration in ppm. We will then compare the results obtained from KED and OK using the cross-validation leave-one-out technique to obtain the root mean square error (RMSE) of both models. The results will demonstrate that KED provides better results, but the improvement is minimal.