Research On Artificial Intelligence Optimization Of Kriging Interpolation Based On Spatiotemporal Surface Attributes

With the development of artificial intelligence theory and technology,the integration of deep learning and other traditional models has become a trend in interdisciplinary application research.In the field of spatial interpolation,the Kriging interpolation method is widely used.Its core semivariogram function is based on the first law of geography and is a statistical method that depends only on the distance parameter.However,in the process of using it for interpolating water pollution diffusion in water bodies,physical laws cannot be represented,and the dependence on interpolated indicators and historical time series has not been considered,resulting in poor performance.Therefore,the application of Kriging interpolation in water pollution diffusion has been improved through deep learning models.Multiple influencing factors for water pollution transmission were considered,and the simulation output in extreme value situations was constrained by combining a physical model.The specific work is as follows:（1）Analyze the theory of spatial interpolation algorithms,design optimization methods,and construct application scenarios.Firstly,the Kriging interpolation algorithm was analyzed,and it was found that its core semivariogram function only determined the interpolation results based on distance.Then,the specific application-required influencing factors were considered in the semivariogram function,and the spatial interpolation theory was improved.Finally,by calculating the mean squared error sequence of the simulated value and the actual observation value of Kriging interpolation,combined with hydrological,meteorological,and other pollutant transmission data,a deep learning model was constructed,which could better simulate the physical process of pollutant transmission and replace the semivariogram function fitting method in Kriging interpolation.（2）Based on the inherent physical relationship between spatial feature attributes and application objects,VGG model was used to extract the surface features of spatial attributes to provide data and feature basis for the subsequent LSTM network model.Since the VGG model extracts the surface feature attributes of the research area,the output spatial feature dimension is high,and there is redundancy and noise.The principal component analysis method is used to reduce the dimension of high-dimensional features to obtain the optimal feature extraction result using the Euclidean distance method.The high-dimensional features are converted into multi-dimensional time series data as the image feature expression vector for the subsequent model.（3）A Kriging interpolation model based on spatiotemporal surface attributes was constructed.By combining VGG,LSTM,and Kriging models,the VGG-LSTM-Kriging combination model was formed,and its effectiveness was verified.The experimental results showed that the VGG-LSTM-Kriging model had the highest accuracy in predicting the water quality of dissolved oxygen（DO）,total nitrogen（TN）,total phosphorus（TP）,and permanganate(COD_Mn)in the Lijiang River basin,compared with other optimized methods based on Kriging interpolation:autoregressive integrated moving average model（ARIMA）,improved recurrent neural network（GRU）,and Transformer.The interpolation accuracy was better than the reference geographic statistical method optimization model,machine learning optimization model,and other deep learning optimization models.In summary,a spatiotemporal interpolation combination optimization model was constructed through CNN structure method to extract historical time series data and spatial surface information,and improve the semivariogram function of Kriging interpolation method.The model extracted spatial feature attributes,optimized the semivariogram function using a spatiotemporal deep learning model,and retained the Kriging interpolation structure completely.The experimental results showed that the proposed model performed well in the interpolation of spatiotemporal pollutant transmission simulation and can provide reference for the applicability research of Kriging in other fields.