Median polish kriging model for circular-spatial data
H. Surjotedjo, Yekti Widyaningsih, Siti Nurrohmah · 2019
In this paper, a new model for kriging or predicting the spatial value of a circular random field at a non-observed location is introduced. This model employs the median polish algorithm that has been adapted for circular data. The median polish algorithm for circular data is used to remove spatial trends, often present in circular-spatial data. Removing spatial trend in the circular-spatial data is required to meet the stationarity assumption, prior to predicting a circular-spatial value using the circular ordinary kriging method. The new model is evaluated on simulated and real-world datasets. Leave-one-out cross-validation is used to evaluate performance of the new model. The Mean Absolute Cosine Error (MACE) and Mean Cosine Difference Error (MCDE) are two metrics used to measure the performance of the new model. The experiments on both types of dataset (i.e. simulated and real-world) show that the new model can be considered as a promising tool for predicting the realization value of a circular random field at non-observed points.