<![CDATA[This study compares the performance of Ordinary Kriging (OK) and Cokriging (CK) methods in spatial estimation based on simulated data. Twelve scenarios are arranged based on a combination of sample size (50, 250, 500) and correlation levels between variables (ρ=0.1, 0.6, 0.9), with each scenario repeated 30 times. Spatial data are generated randomly within the geographical boundaries of Indonesia, variables are generated based on spherical variograms with nugget or sill or dan range or ,, and model evaluation is carried out using Leave-One-Out Cross Validation (LOOCV) with RMSE and metrics. The results show that Cokriging consistently produces more accurate estimates than Ordinary Kriging in all scenarios. In the best configuration (CK, n=500), RMSE = 1.04 and = 0.945 were obtained, while the best performance of OK only reached RMSE = 1.06 and = 0.873. All levels of correlation in Cokriging showed good performance, especially when the amount of data is sufficient. Therefore, Cokriging is recommended as a superior spatial interpolation method in the context of multivariate and spatial data, especially when relevant secondary information is available.]]>
Copyrights © 2025
