Key Technology And Its Application For Spatio-Temporal Kriging

With the human exploration of natural process, the source of information acquisition extends from local to global, then expand to various shperes, even to universe space, at the same time, the format of data is also transformed from spatial data to spatio-temporal data. The so-called spatio-temporal data is the data which both have temporal dimension and spatial dimension. In recent years, domestic scholars and foreign scholars have made a numerous research on the analysis and application of the spatial-temporal data, including Empirical Orthogonal Function, Bayesian-Markov Chain, Grey Forecasting model, Spatio-temporal Geostatistical et al. Among them, the Spatio-temporal Geostatistical is being became the research hotspots because of it can use spatio-temporal prediction and stochastic simulation to calculate and show the spatial-temporal variation and dynamic process of spatial variation for the variable. Therefore, basic on the framework of the Spatio-temporal kriging theory, the author analyzes and summarizes the spatio-temporal data from spatial-temporal theory variogram model, spatio–temporal prediction and spatial–temporal analysis. Moreover, the author develop the Spatio-Temporal Kriging Software in Matlab environment in order to achieve the spatio-temporal kriging technology and functional.In this study, the author predits the PM_2.5 concentration of Shandong Province by using Spatio–Temporal Ordinary Kriging（STOK）, Spatio–Temporal Trends Kriging（STTK） and Ordinary Kriging（OK）. Moreover, in STOK prediction, the author takes six spatial-temporal variogram model into consideration, and explores the impact of different spatial-temporal variogram model on spatial-temporal prediction accuracy. For STTK prediction, the author sets the higest corresponding heterogeneity orders of spatial nonhomogeneity and temporal non-stationarity both are second order for the part of trend, and uses STOK to deal with the part of residual. As for OK, selects the data which have the same time property, and predicts it. Base on the three prediction results, the author set the root mean square error（RMSE）, mean absolute error（MAE）, the maximum error（MAE） and the minimum error（MINE） to the index of accuracy verification, then, calculates and evaluates the four accuracy index in different time periods and comprehensive periods. Finally, takes the best prediction results as data source, the author analyzes the spatial–temporal distribution and pollution property of PM_2.5 in Shandong Province. The achievements of this study can bu summarized as follows:（1）. For STOK, the sort for relative fitting accuracy of six spatial–temporal variogram model is: MM > DM > BM > GM > CH1 > CH2, and the prediction accuracy of RMSE is: STOK_DM（12.911） > STOK_MM（13.0124） > STOK_GM（14.2160） > STOK_BM（14.2626） > STOK_CH1（15.4129） > STOK_CH2（15.9724）, the prediction accuracy of MAE are: STOK_MM（9.853） > STOK_DM（10.0664） > STOK_GM（10.4671） > STOK_BM（10.0664） > STOK_CH2（11.6616） > STOK_CH1（12.0023）. As far as the research object is concerned, the separation model not only better than non-separation model on model fitting, but also better than non-separation model on spatio-temporal prediction.（2）. In the contrast of spatio–temporal prediction accuracy, the prediction accuracy of STOK is changing with its spatial–temporal variogram model, and the prediction accuracy of four trend models are quite similar. Overall, the sort of spatio–temporal prediction accuracy is: STOK_DM ≈ STOK_MM > STTK > STOK_GM > STOK_BM > STOK_CH1 > STOK_CH2. In the contrast of spatio–temporal prediction and spatial prediction, the STOK and STTK both are better than OK in different time periods and comprehensive periods. Simultaneously, compared with OK, STOK and STTK prediction accuracy are respectively increased more than 41.01% and 31.32% in comprehensive periods, which illustrate that spatio–temporal prediction is obviously better than spatial prediction.（3） The overall pollution situation of PM_2.5 in Shandong province is more serious. In space, the yearly average concentrations of PM_2.5 are more than 100μg/m3 in Central and West of Shandong Province, and the number of daily average concentrations higher than World Health Organization（WHO） ”Air quality guidelines” transition goal III is greater than 290. For North Central and South Central in Shandong Province, the yearly average concentrations of PM_2.5 are between 75μg/m3 to 150μg/m3, the number of daily average concentrations higher than transition goal III is around 220. For the East, the yearly average concentrations of PM_2.5 are below 50μg/m³, and the number of daily average concentrations lower than transition goal I is more than 146.In time, the most serious pollution period are January, February, November and December, June, July and August are almost no pollution. The contaminative levels sort of each season is: winter > autumn > spring > summer. Meanwhile, according to the distribution of Gini coefficient in Shandong Provinve, the value in West Central are lower than 0.3, the value in East are between 0.3 to 0.4, its show that the West Central of Shandong Provinvce exist persistent high risk of contamination for PM_2.5, and the East of Shandong Provinvce exist interim slight risk of contamination for PM_2.5.Furthermore, from West to East, the contaminative levels and duration are both reduce.