Research On The Spatio-temporal Inconsistency Processing Of GlobeLand30 Global Land Cover Data

Land cover data plays an important role in scientific research and production.GlobeLand30,as the world’s first 30m resolution global land cover data set independently developed by my country,is used in climate change,land use and land cover change monitoring,ecological environment protection,Disaster prevention and mitigation are of great significance.However,due to factors such as differences in imaging time and environment,and the diversity of surface features on a global scale,there are some spatio-temporal inconsistencies in the three-period data of GlobeLand30,which restricts the use of this data to a certain extent.Therefore,this thesis focuses on the characteristics of spatio-temporal inconsistency among the three-period of data,the study uses other multiple sets of global land cover data and auxiliary information of multi-temporal remote sensing images to establish a processing method for the spatio-temporal inconsistency of the three-period of GlobeLand30 data,and obtain a three-period GlobeLand30 global land cover data set with good consistency.The main content and results of the research are as follows:(1)Firstly,preprocessing such as category mapping,reprojection,and cropping was performed on multiple sets of global land cover data collected,and multiple sets of global land cover data consistent with the GlobeLand30 data coordinate system and category system were obtained as auxiliary data for subsequent processing.Then,the spatio-temporal inconsistency among the three-period data of GlobeLand30 was analyzed from two aspects of quantitative statistics and qualitative observation,and the reasons for the spatio-temporal inconsistency among the three-period data were summarized:the spatio-temporal inconsistency caused by similar category classification errors,the spatio-temporal inconsistency caused by natural seasonal changes,and the spatio-temporal inconsistency caused by original image stitching errors;and the inconsistency characteristics between GlobeLand30 data and other auxiliary data were analyzed.(2)Aiming at the characteristics of the inconsistency between three-period GlobeLand30 data and between GlobeLand30 data and other auxiliary data,the data of GLC_FCS30,FROM-GLC,CCI-LC,Global Forest Change and other data are used as auxiliary reference data,reliable category information is obtained through mutual verification between different data,and inference rule sets are established for different types of inconsistencies for correction processing,mainly including:spatio-temporal inconsistency processing of seasonally changing water bodies and wetlands;spatio-temporal inconsistency processing of confusing land types such as forest land,shrubs,grasslands,cultivated land,and wetlands;spatio-temporal inconsistency processing of confusing land types such as tundra,bare land,ice and snow.Finally,the processing of the three-period GlobeLand30 data worldwide was completed,and a set of three-period GlobeLand30 global land cover data with high consistency was obtained.(3)Quantitative and qualitative evaluation of the three-period GlobeLand30 global land cover data after processing.Among them,the qualitative evaluation adopts visual observation and evaluation,selects the typical area of inconsistency treatment,and illustrates the accuracy of inconsistency area correction through the high-resolution simultaneous image comparison on GoogleEarth;quantitative analysis first conducts overall statistical analysis,calculates the statistical information of the difference pixel ratio between the three phases of data to illustrate the effectiveness of consistency processing,and then select the representative regions of consistency correction in different regions of the continents of the world,and carry out accurate accuracy verification through manually determined accuracy checkpoints to prove the validity of the spatio-temporal consistency processing in this study.(4)Due to the limited accuracy of the auxiliary global land cover data used,after the spatio-temporal inconsistency processing based on the inference rule set,there are still some inconsistencies in a few regions.The study carried out spatio-temporal inconsistency processing based on multi-temporal remote sensing images in these areas.After the spatio-temporal inconsistency processing based on multi-temporal remote sensing images,the overall accuracy of the region is increased to more than 80%,and the Kappa coefficient is increased to more than 0.75.The research uses different satellites and different resolutions of remote sensing images to modify and process the data of multiple regions with different climates,geographical locations and ecological environments through this method,which proves the reliability and versatility of the method.