Table 2

Summary of cross-validation by leaving one borehole out. The number of point observations in the study area was 35. Cross-validations were undertaken by ordinary kriging (), cokriging (), bedrock kriging (), and Poisson equation () with an optimal load parameter estimated by leaving one observation out. Kriging on residuals between observations and Poisson equation () were done for the Poisson solution where the optimal parameter was estimated based on all observations. The superscripts indicate either normal score transform (*) or lognormal transform (′) of observations. In addition to the evaluation criteria , , , , and (see references below), the correlation coefficient R and the slope coefficient from linear regression is included.

6.51 6.47 7.46 7.26 11.30 2.15
0.58 0.61 0.48 0.58 0.49 0.86
1.73 4.51 1.41 1.99 0.92 3.97
mean (0.48 0.47 0.51 0.49 0.66 0.57
std (0.25 0.28 0.30 0.29 0.22 0.17
mean (12.10 12.15 5.67 5.95 8.64 1.98
std (16.01 15.99 10.98 9.55 12.47 1.12
0.55 0.59 0.48 0.51 0.40 0.99
0.31 0.37 0.23 0.25 0.19 0.89
6.51 6.47 7.46 7.26 11.30 2.15
0.58 0.61 0.48 0.58 0.49 0.86
1.73 4.51 1.41 1.99 0.92 3.97
mean (0.48 0.47 0.51 0.49 0.66 0.57
std (0.25 0.28 0.30 0.29 0.22 0.17
mean (12.10 12.15 5.67 5.95 8.64 1.98
std (16.01 15.99 10.98 9.55 12.47 1.12
0.55 0.59 0.48 0.51 0.40 0.99
0.31 0.37 0.23 0.25 0.19 0.89

, Mean absolute error (Equation 9); , Accuracy (Equation 10); , Precision (Equation 11); , Probability score (Equation 12); , Continuous ranked probability score (Gneiting & Raftery 2004); , Correlation coefficient between observations and estimates; , Slope coefficient estimated by linear regression.

Close Modal