On Methods For Rapid Change Detection Of Geospatial Variables And Their Preliminary Application

Since mutation detection methods of time series have been studied deeply and broadly,many useful results have been obtained and more effective mutation detection techniques have been put forward to.However,because these researches fail to regard geospatial variables,the temporal-spatial relationship between geographic variables is splited.The most of geographic variables are characteristic of non-linearity and non-stationarity,which not only hide important geographical phenomena,but also implicitly contain dynamic information of geographic changes due to non-linearity and non-stationarity.Therefore,it is of great significance to develop geospatial mutation detection based on spatial resolution further so as to improve the theoretical system of spatial mutation,and to mine the geographic information of system mutation and to promote the development of geostatistics.In this paper,the methods of theoretical analysis and case verification are combined to study the spatially rapid change detection approaches of geospatial variables based on geospatial resolution.The research paradigm is from practice to theory and then returning to practical verification,that is to say,firstly spatially rapid change examples in daily life are summaried and discussed again,and then the concept and relevant theoretical methods of spatially rapid change are extracted and refined.In the end,approaches of spatially rapid change are proposed and their validities are verified using known phenomena of spatially rapid change in practical application.We first recalled and summarized the theoretical framework of time series mutation,and then parallelly extended its relevant research ideas and methodologies to geo-space,and debated the concept,classification and characteristics of geospatial mutation.Through the analysis of some spatial mutation examples in daily life,the understanding of spatial mutations in geospatial variables based on geospatial resolution was deepened and broadened.Afterwards,we proposed the spatial mutation detection methods of spatial value and spatial trend variation,which were verified by two real-world examples.The main research methods and topics are as follows:(1)A theoretical framework system for geospatial mutations was proposed.Using the theoretical analysis method,the theoretical systems of the existing time series mutation,including their concepts,mutation classifications and detection methods and known mutation cases in nature,were summarized.In order to understand spatial mutation deeply and broadly,the concept of spatial mutation was proposed and some known spatial mutation cases in daily life and the literature were debated again.The classification of spatial mutation was showed and its characteristics were studied.The relationship between spatial scale and mutation was discussed.(2)The detection method for rapid change of spatial value was proposed.We analyzed the concept of spatial value mutation,and clarified the relationship among range,gradient and edge,and studied the characteristic of Canny edge detection method,and discussed the applicability of Canny edge detection method in spatial value mutation detection.The some limitations of this detection method were analyzed.(3)The detection method for rapid change of the spatial trend variation was proposed.We analyzed the concept of spatial trend variation,and generated the simulated data of spatial trend variation and debated simulated data characteristics.In order to fulfill rapid change detection of the spatial trend variation,we first "sliced"the spatial gridded dataset into several one-dimensional dataset in X and Y direction,respectively,to reduced data dimensionality,and then Mann-Kendall non-parameter test method was used to detect the spatial trend variation of the one-dimensional sliced data in X or Y direction,respectively,and finally the spatial mutation points were connected to show the detection result of the spatial variable trend variation.Additionally,the some limitations of this method were analyzed.(4)Experiment verification.The two experiments were provided.The one is to verify the correctness of detection method of the spatial value mutation by using the simulated data and the gridded data of daily precipitation in China.The other is to verify the correctness of detection method of the spatial trend variation mutation using daily precipitation datasets in the summer half year in 2009 in China.The result showed that the two proposed methods are able to detect the situation of spatial mutation,and the mutation location is basically consistent with the known results.It is proved that the proposed two mutation detection methods are feasible,correct,reliable and available.(5)Preliminary application of the two methods.The detection methods of spatial value mutation and spatial trend variation have been used in the mean precipitation of spatial points in China during the past 30 years spanning from 1987 to 2017.The possible geographic mechanism of the abrupt change has been analyzed and debated.Concluding,the proposed spatial mutation concept based on scale in this study is clear.The proposed detection methods of spatial value mutation and spatial trend variation can detect the types of rapid change mutation,and the spatial position of the mutation occurs more accurately.At the same time,the Canny edge detection of image processing is introduced into geography,which provides a new and promising idea for the solution of geography problems.