Study On The Evolution Of Spatial-temporal Pattern And Sustainability Evaluation Of Typical Chinese Cities Based On Remote Sensing Monitoring

The urbanization level in China has increased rapidly since the beginning of the reforming and opening.The urbanization has promoted the rapid development of the national economy and the improvement of the living standards of people.However,there are also some problems,such as uneven regional development,excessive urban growth,extensive use of land,enormous waste of resources and environmental degradation.All these problems restrict the city sustainable development seriously.The precondition to solve these problems is to figure out the situation of urban development.The dynamic changes of urban areas can reflect the urban expansion process and the pattern of urban development.Urban expansion monitoring,spatial-temporal evolution,and sustainability assessment play vital roles in the study of city development.It can provide important imformation for urban planning and construction,and promote sustainable urban development.However,there are some problems,such as inconsistency defination of "urban area",large sample demands for urban area extraction,uncomprehensive consideration for urban area prediction,and bias for developed cities in sustainability evaluation.To solve these problems,the researches of urban expansion monitoring,spatial-temporal evolution,and sustainability assessment are done in this paper.The main work is summarized as follows:(1)A new method of urban area extraction based on a progressive learning model is proposed.By combining apriori knowledge and image features,the number of training samples required for machine learning is reduced.The problem of using high-level semantic information expression in the process of urban area in classification is avoided by using road network data to segment,and the accuracy of urban area extraction is improved.The road network data is used to divide the city into blocks.Thus,the scene classification method can be used to extract the urban area,and the space constraint of urban extraction principles is integrated into this method by using layer-by-layer pyramid method,which can greatly reduce the number of sample and get the results similar to mannual results.The proposed method is a solution for automatic extraction of urban areas.(2)The LSTM-RBF model for urban area prediction is proposed and the spatio-temporal evolution analysis and simulation prediction of the provincial capital city expansion are carried out.Taking 31 provincial-level cities as study area and high-resolution remote sensing images as main data source,the urban boundaries of 2000,2005,2010 and 2015 are manually corrected based on the results of progressive learning extraction and the urban areas of 31 capital cities in 2016 are predicted.The comparison between the LSTM-RBF model and the results of the LSTM network,the RBF network and multiple linear regression demonstrates the superiority of the LSTM-RBF model in urban area prediction.After the analysis of urban scale distribution,expansion characteristics,and expansion types,the following results are obtained:(a)From 2000 to 2015,China's provincial capitals maintained a high-speed growth trend with a total area increase of 90%;(b)the provincial capital system is near the scale distribution of the Zipf,and it gradually becomes more balanced over time;(c)the situations of urban expansion in different regions are significantly different.The expansion rate of urban in Eastern China is gradually slowing,but the expansion rates of urban in Western China and Northeast China are accelerating,and the the cities in Central China expand steadily;(d)61% of provincial capital cities are expanding in an extended manner;(e)China's provincial capital development is highly correlated with national urban development policies and regional development strategies.(3)A comprehensive evaluation model,named Urban Geographic Environment Index(UGEI),is constructed to evaluate the environment of Chinese capital cities in 2015.From the perspective of limited natural resources,the model uniformly adopts the area-averaged(used urban areas from high resolution imagery)indicators instead of original indicators,per capita indicators or multiple types of indicators.Land use/cover indicators are more considered than the social economic indicators for carrying capacity to support the pressure–state–response model,and the average index is weighed through the analytic hierarchy process and the entropy method.The results are compared with the original indicators,the per capita indicators and the area-averaged indicators using the urban administrative area and the objectivity and reliability of the model are verified.The major findings are as follows:(a)about half of the provincial capital cities in China are in bad geographical environment situation;(b)the geographical environment of China's provincial capital cities can be divided into four regions,and the Qinling-Huaihe River is an important dividing line between the north and the south.The southwestern region is the best,and the northwestern region is the worst;(c)the top 3 cities with the best geographical environment in China's capital cities are Haikou,Nanning,and Changsha,and the bottom 3 are Zhengzhou,Taiyuan,and Shijiazhuang.