Research On Climate Model Data Organization And Compression Method Based On Hierarchical Tensor

Climate model data is the basic data of research,with characteristics such as massiveness,high dimensionality and structural sensitivity.With the continuous improvement of modern global geographical change technology and the continuous strengthening and updating of global systems,model data is rapidly accumulating,and management and compression of Climate Model Data in long term become one of the key difficulties in model data analysis and citation.At present,the growth rate of schema data has far exceeded the growth rate of storage scale,which makes the common methods of schema data compression tend to the compression limit.The research on loss-compression of climate model data has increasingly become the frontier and hotspot of basic computing and analysis framework of geographic model.However,how to construct a feature-preserved loss-compression method for Climate Model data,which can support data updating continuously and data computing efficiently,is still a difficult problem in this field.Tensors,which maintain the original structure of data,have the characteristics of unified expression of multi-variables and multi-dimensions.Tensors often used to express high-dimensional data.Hierarchical tensor decomposition has a tree-like hierarchical structure,which has advantages in computer data storage and retrieval.It can achieve continuous addition and compression of pattern data,and provides a superior mathematical tool for dynamic compression of pattern data.However,the existing hierarchical tensor decomposition model is difficult to maintain the consistency of the compression error during the decomposition process,resulting in the error volatility of the data compression in the long time series mode,thus affecting the calculation and analysis based on the mode data(such as trend calculation,etc.)precision.In order to meet the application requirements,it is necessary to construct a hierarchical tensor compression method with stable errors.Based on the temporal and spatial characteristics of climate model data and the theory of hierarchical tensor decomposition,this paper studies the data organization method of pattern data feature detection and block strategy based on information volume.The research idea of loss-compression based on hierarchical tensor decomposition under multi-constraint conditions is proposed.A data compression strategy with constant errors for climate model data with characteristic sensitivity is constructed,and the data error is evaluated and analyzed.On this basis,the global Climate Model Data T(temperature)/U(warp direction wind)variables as an example,the model data compression test and error distribution analysis.By comparing and analyzing this method with the current mainstream loss-compression algorithm,the results prove the advantages of this method in data error stability.Finally,a compression management system of climate model data based on hierarchical tensor formed to realize the compression and storage of climate model data.The research work of this paper mainly includes the following aspects:(1)Climate model data block organization based on feature detection.According to the spatial and temporal characteristics of climate model,the information entropy of each dimension is calculated.Data feature management carried out by feature statistics of similarity,balance and maximum information.Judging the condition of data reorganization,forming a data partitioning strategy considering information entropy and computer storage capacity.A schema data organization framework based on multi-dimensional features of data realized.(2)Adaptive compression of climate model data under multiple constraints.Based on the hierarchical tensor decomposition theory,this paper forms a data model of time constraints and compression parameters to select suitable compression parameters.Constructing the relationship between compression parameters and data error.This paper establishes a time-parameter-error correlation function for climate model data,thus realizing the control and storage of loss-compression data with controllable error.(3)Climate Model Data error evaluation.This paper studies the error assessment and analysis theory of climate model data.Based on different data blocking strategies,this paper studies the compression time and error distribution characteristics of different data blocks.This paper stratifies the time dimension,analyzes the data error stability and trend,and realizes error evaluation of the climate model data.This paper defines the data organization of the climatic mode block tensor based on the mathematical tensor theory and the data feature distribution.In this paper,through tensor decomposition theory,block data management of climate model data and error controllable data analysis.The content studied in this paper has certain reference significance for data information mining caused by error accumulation in long-term mode simulation.