Application Of OpenStreetMap Data In Urban Land Use Mapping

With the development of Web 2.0,the data production mode of"crowdsourcing"has come into being."Crowdsourcing"uses the local knowledge and expertise of a large number of users to perform tasks at low cost.Ordinary people create geographic information by means of "crowdsourcing" which Goodchild calls Volunteer Geographic Information(VGI).OpenStreetMap is a representative project to produce volunteer geographic information.Fifteen years after its launch,OpenStreetMap has become the largest,most diverse,most complete and up-to-date open source geospatial database in the world.OpenStreetMap data has many advantages,such as updating at any time,free access,high accuracy,and so on.They are widely used,such as cartography,geographic analysis,urban planning and risk prevention.In urban areas where OpenStreetMap volunteers are active,using the data produced by volunteers in urban land use mapping can obtain real-time accurate land use status map at low cost,and provide valuable reference data for policy makers and city administrators.This paper takes the OpenStreetMap data in vector form as the main study data,Sentinel-2 remote sensing images as auxiliary data,and takes Berlin,Hamburg,Munich and Ruhr in Germany as research areas.Firstly,a set of urban land use classification system is established.Then,the methods of using polygon features in OpenStreetMap data,OpenStreetMap data and Sentinel-2 images,OpenStreetMap data and Sentinel-2 images at street block scale to make urban land use map are proposed.Finally,accuracy of the mapping results is evaluated by comparing mapping results of the three methods with the Corine Land Cover data,and accuracy evaluation results of the three methods are compared to analyze the advantages and disadvantages of three mapping methods.The main research contents and achievements are as follows:(1)Establish the classification system of urban land use mapping using OpenStreetMap data at non-street block and street block scales.Considering the attributes and accuracy of OpenStreetMap data and the classification system of Corine Land Cover data,we highlight the characteristics of urban land use in terms of detailed categories of construction land and small area of natural land,and establish a classification system which divides urban construction land in detail and integrates natural land cover categories.The classification system includes eight categories:urban fabric;industrial,commercial and transport units;other construction land;artificial non-agricultural vegetated areas;agricultural areas;forests and semi-natural areas;wetlands;water bodies.In addition,about mapping at street block scale,considering that there are many types of land use in some blocks,four categories of mixed land use are set up by analyzing and summarizing these types,they are zones dominated by water bodies and mixed with other land use types,zones dominated by construction land and mixed with agroforestry land,zones dominated by agroforestry land and mixed with construction land.complex mixed zones.On the whole'there are twelve categories in urban land use mapping at street block scale.(2)Methods of making use of OpenStreetMap data and Sentinel-2 images for urban land use mapping at street block scale and non-street block scale respectively are put forward.Polygon features in OpenStreetMap data are used to determine land use categories in corresponding areas.POIs are used to make training samples for supervised classification of Sentinel-2 images,and classification results are used as land use categories of blank areas in OpenStreetMap data.Road network and water network of OpenStreetMap line features are used to divide the mapping region into street blocks.Then,urban land use mapping is carried out at non-street block scale and street block scale respectively.This method makes full use of point,line and polygon features in OpenStreetMap data,and conforms to the block characteristic of urban land use.(3)The accuracy of urban land use mapping results using OpenStreetMap data,OpenStreetMap data and Sentinel-2 images,OpenStreetMap data and Sentinel-2 images at street block scale relative to Corine Land Cover data is evaluated.Advantages and disadvantages of the three methods are analyzed.Firstly,the results of urban land use mapping using OpenStreetMap data,OpenStreetMap data and Sentinel-2 images.OpenStreetMap data and Sentinel-2 images at street block scale are compared with Corine Land Cover respectively,and the accuracy of mapping results using the three methods are analyzed through completeness,marginal proportion of agreement,total proportion of agreement.Then,accuracy evaluation results of the three methods are compared to analyze the advantages and disadvantages of land use mapping using OpenStreetMap data and Sentinel-2 images at non-street block scale relative to land use mapping using OpenStreetMap data only,as well as the advantages and disadvantages of land use mapping using OpenStreetMap data and Sentinel-2 images at street block scale relative to land use mapping at non-street block scale.Results show that land use mapping using OpenStreetMap data and Sentinel-2 images can improve the completeness of mapping results relative to method using OpenStreetMap data only,and it also improves the accuracy.Land use mapping using OpenStreetMap data and Sentinel-2 images at street block scale can improve the accuracy of mapping results relative to method at non-street block scale,which indicates that land use mapping using OpenStreetMap data and Sentinel-2 images at street block scale has practical value.