Chinese GDP Modeling And Simulating Based On Remote Sensing And Social Pereption Data At Multiple Scales

Gross Domestic Product(GDP)is a comprehensive index to measure the economic development of a country or region,which can effectively reflect the economic development status and development regulars of different regions.The traditional GDP data was usually obtained by sampling survey and economic census.This GDP data was usually summarized in table which it is difficult to reflect the spatial heterogeneity of economic development within administrative regions and have some limitations such as poor timeliness,data missing and so on.The birth of nighttime light data(NTL)provides an effective data source for GDP modeling.However,NTL data mainly reflects the information of human activities at night,and it is difficult to capture the information of social and economic activities during the day.Recently,with the widespread use of smart mobile devices,the location based social media(LBSM)data are increasingly being used as a proxy for real-time human activities.LBSM data is closely related to human activities and can reflect the social and economic activities during the day and night effectively.But little work was carried out to explore the potential of LBSM data in estimating economic development at different scales in China.Therefore,this study filled this gap by evaluating the effectiveness of Tencent user density(TUD)data,a typical type of LBSM data in China,in Gross Domestic Product(GDP)modeling at the provincial,municipal,and county scales by using the Ordinary Least Squares model(OLS)and Geography Weighted Regression model(GWR)compared with NPP/VIIRS nighttime light data.Secondly,Social progress promotes the rapid development of China’s economy,at the same time,it also leads to the imbalance of economic development in different regions of China.However,there are few studies exploring the influence factors of county economic development in China,and the spatial scale effects of indenpendent variables are easily ignored by traditional regression models.So,based on remote sensing data such as land use cover data,DEM data and rainfall data,as well as social perception data such as TUD data,Point of Interest(POI)data and traffic road data,this study use OLS model,GWR model and Multiple Geography Weighted Regression Model(MGWR)respectively to build models to predicted the China’s county GDP.By comparing the results of different models,this study discusses whether the MGWR model can better understand the spatial heterogeneity of the influencing factors of county economic development in China.Then,based on the MGWR model regression coefficients,cluster analysis method was used to achieve the division of China’s seven economic zones.Finally,the relevant GDP spatialization studies usually adopt a unified simulation model,ignoring the spatial heterogeneity of different influence of economic development.Therefore,in this study,the random forest model combined with remote sensing and social perception data was used to simulate China’s GDP at grid scale based on seven economic zones.By comparing the simulated GDP results,this study explores whether the partitioning modeling idea can effectively improve the grid accuracy of GDP.The main conclusions of this study are as follows:(1)TUD data and NPP/VIIRS-NTL data have different GDP modeling ability at different spatial scales.The NTL data has stronger GDP modeling ability at provincial and municipal spatial scales,while TUD data has stronger economic modeling ability at the county scale.Especially in less developed regions with complex terrain and changeable climate and more developed regions with human separation of employment and residence,TUD data has stronger GDP modeling ability and can estimate GDP more accurately.(2)The modeling effect of MGWR model is better than OLS model and GWR model,indicating that there are spatial heterogeneity and spatial scale effects of different factors on the development of county economy in China.In addition,based on the analysis of the multi-scale geographical weighted regression model results,this study found that different factors have significant different impacts on different regional economies in China,such as the Yangtze River Delta,Pearl River Delta and Beijing-Tianjin-Hebei regions have shown different economic development characteristics.(3)Random forest model is better than ordinary linear regression model in GDP simulation at grid scale.The simulated GDP data predicted by random forest model can not only accurately reflect the current situation of China’s economic development at large spatial scale,but also accurately capture the economic differences within small spatial scales.At the same time,it is found that considering economic zones to simulate the GDP can effectively improve the data accuracy and get more accurate GDP data at grid scale.