Urban Impervious Surface Extraction Driven By Multi-source Satellite Remote Sensing Data

Urban areas are at the core of sustainable development.Urban areas produce more than 75% of global GDP,consume about 75% of global final energy use,and generate approximately 2 billion tons of waste each year.More scientific decision-making knowledge are needed to help urban have a sustainable future.Now,the need for this scientific decision-making knowledge has become urgent for governments and decision makers at the local,national,and even international levels.Over the past fifty years,advances in remote sensing technology have greatly improved our understanding and awareness of urban areas.The main contributions of urban remote sensing include,but are not limited to,the extraction of urban impervious surfaces,urban land cover changes,and thermal remote sensing of urban climate.Today,the proliferation of new sensors,archival of long-sequence satellite records,joint analysis of Earth observation data and ancillary data sets,the widespread availability of high-performance computing facilities,and the gradual use of data and methods beyond remote sensing provide new opportunities to create scientific knowledge for the urbanized earth.Urban impervious surfaces data is important for urban planning and environmental and resource management.The extraction of impervious surfaces in urban areas has also recently attracted unprecedented attention.The coordinated observation of multi-source satellite remote sensing data will significantly increase the spatio-temporal frequency of impervious surface observations and provide a data basis for accurate extraction.At the same time,with the development of cloud computing and artificial intelligence technology,a cloud processing platform represented by Google Earth Engine(GEE)has emerged,which provides a new opportunity for impervious surface extraction.At present,there is an need to effectively evaluate how to extract accurate urban impervious surfaces based on massive multi-source time series remote sensing data.This paper uses Landsat 8 OLI,VIIRS,and LJ-1 multi-source time series remote sensing data to conduct research and experiments on urban impervious surface extraction based on the random forest algorithm and GEE.The research mainly evaluates and analyzes the extraction method for the medium-resolution(30m)impervious surface by multi-source time series remote sensing data.The study area is in Beijing,China.The main research results are as follows:1)Evaluate and compare the differences in the characteristics of the spectrum,principal component analysis(PCA),independent component analysis(ICA),and texture of the Landsat8 OLI image in a single temporal in the extraction of urban impervious surfaces.And the feature importance evaluation and accuracy evaluation were performed separately.2)Based on the annual dense time series Landsat 8 OLI data,image synthesis can effectively reduce the impact of data quality on the extraction of urban impervious surfaces.The research uses the annual time series Landsat8 OLI image data of Beijing,China,combined with the multi-dimensional array theory for geographical big data and random forest algorithm,and discusses how to effectively use dense time series Landsat images for annual synthesis for the study of urban impervious surface extraction.3)Combine the annual Landsat 8 OLI dense time series data with VIIRS,LJ-1and other data to extract the impervious surface of the urban.The performance of the medium-resolution(30m)urban impervious surface extraction based on multi-source time-series remote sensing images was evaluated.The study found that compared with single temporal remote sensing data,the use of multi-source time series remote sensing data can significantly improve the accuracy of the impervious surface extraction.