An Elastic LatticeKrig Generative Model for Non-ideal Spatial Datasets

When datasets have missing parts, we typically use data characteristics to select a suitable model to generate these missing parts. However, the poor quality of real data often affects the fitting accuracy of the ideal model, leading to lower-thanexpected results. In this paper, we improve the LatticeKrig model by using local basis function spanning. This enhancement aims to improve the model's fitting ability and robustness to poor quality data while avoiding potential overfitting. In our experiment, we demonstrated the model's fitting ability with ideal data and used it to fit the missing parts of garbage collection data from the Norwegian coast. The results show that our improved model achieves better fitting performance with non-ideal data.