Abstract: Spatio-temporal interpolation is a fundamental task of spatio-temporal data analysis. Modeling of spatio-temporal dependencies in geospatial data plays a key role in spatio-temporal interpolation. When geospatial data is non-stationary and sparsely distributed, modeling of spatio-temporal dependencies is still challenging. On that account, this study developed a spatio-temporal interpolation method based on Yang Chizhong filtering. This method combined statistical and geometric methods to model spatio-temporal dependencies in geospatial data. Specifically, Yang Chizhong filtering and spatio-temporal product-sum model were first employed to construct the spatio-temporal fundamental variation function that quantitatively describes spatio-temporal dependencies in geospatial data. Then, an optimal linear unbiased estimation model for spatio-temporal data interpolation was built using the spatio-temporal fundamental variation function. We utilized simulated dataset, annual average temperature dataset in mainland China from 2000 to 2009 and daily average PM_2.5 concentration dataset in Beijing from May 2014 to April 2015 for experimental verification. Experimental results on both simulated and real-world datasets showed that the proposed method outperforms the three state-of-the-art methods, e.g., spatio-temporal Kriging, point estimation model of biased hospitals-based area disease estimation, and lightweight ensemble methods. The proposed method does not require the assumption of spatio-temporal stationarity, and can better adapt to sparsely distributed geospatial data.

Key words: spatio-temporal interpolation, Yang Chizhong filtering, spatio-temporal product-sum model, spatio-temporal dependency, spatio-temporal non-stationary