Study Of Meteorological Data In Liaoning Province Based On Kriging And Bootstrap Methods

Geostatistics is a science which based on the theory of regional variables,with the variogram function as the main tool to study the spatial distribution of both random and structural(or spatial correlation and dependence)of the natural phenomenon.With the development and improvement of geostatistical techniques and the increasing interest in functional data,it has been possible to predict functional data without monitoring by using the FKED(functional kriging with external drift)model,which takes into account the effects of exogenous variables,including scalar and function variables.However,the uncertainty assessment of functional data is still a problem,we use a semi-parametric bootstrap method to predict the uncertainty of the functional data,to ensure that the spatial correlation structure remains in the self-help sample.This paper is divided into four chapters.The first chapter is introduction,introduction research background,research purpose,domestic and foreign research trends and so on.The second chapter is the theoretical knowledge,mainly introduces B-spline,generalized additive model,generalized additive mixed model and so on.The third chapter is case analysis,including the selection,processing and prediction of function data at unsupervised location,and gives the prediction interval of function data.The fourth chapter is the conclusion,summarize the shortcomings and future trends of this article.In the example analysis,we select the correlation analysis for the meteorological data of Liaoning Province,and give the prediction and forecast interval of the function data of unsupervised location.In predicting functional data,we mainly use the FKED model.In addition to the FKED model,we use the bootstrap method when predicting the confidence interval of functional data.This solves two problems.One is the order of the curves to obtain functional quantiles,and the other is the specification and estimation of the spatial dependency structure.And based on two methods,the confidence interval of functional data is given.At the same time,the results of the two methods were compared.Based on the results we can conclude that the method is computationally feasible and suitable for quantifying the uncertainty of predictive function data...