A Global To Local Quantitative Transformation Model(GTLQTM) For Socioeconomy By DMSP/OLS Nighttime Imagery

The visible and infrared imagery from Defense Meteorological Satellite Program's(DMSP)Operational Linescan System(OLS)instruments have been being using to monitor the global distribution of clouds and cloud top temperatures,of which the telescope pixel values are replaced by Photo Multiplier Tube(PMT)values at night.For this reason,the imagery obtained at night is able to detect city lights even low-intensity lights such as small-scale residential areas,traffic flows,which makes city distinct from dark rural background.This provides a new way to study large-scale urbanization.Currently,DMSP/OLS nighttime light remote sensing images have been widely used to estimate urban development,social development,economic growth,energy and trade.However,due to the spatial resolution of DMSP/OLS data is 2.7km,the current related researches based on DMSP/OLS data is only focused on medium and large scale,the issues at a small-scale area such as a county,town,village with applying DMSP/OLS data are rarely involved.The issue of regional quantitative modeling at small-scale based on DMSP/OLS data is rarely discussed by relevant scholars.Additionally,the existing large-scale modeling based on DMSP/OLS data is also not performed rigorous theoretical proof,the rationality and scientific of model construction is lack in existing research.These questions are needed to be solved.Therefore,aiming at these problems this paper proposes a new global to local quantitative transformation model(GTLQTM)based on consistency assumption by DMSP/OLS data to model socio-economic indicators(SEI)at small scale.According to the consistency assumption,a sub-region,as a sub-sample in the large area(or large sample),usually meets the same statistical law as that of large area.Further the probability theory and mathematical statistics theory was utilized to analyze the uncertainty of GTLQTM in both simulate experiments verification and real data verification.The main contents are as follows:1)A global to local quantitative transformation model(GTLQTM)The Modeling of GTLQTM is proposed upon the existing large and medium-scale target areas modeling between night lights and socio-economic parameters.Taken the sum of lights(SOL)of global area as an index of nighttime lights and SEI as study object,the GTLQTM was built.2)Disturbance(practical)GTLQTMsIn reality,SEI may be influenced by random error.Aiming at this imperfection,a series random disturbance terms are added as noise to linear relationship between SEI and SOL.In order to identify how disturbance term influence ideal GTLQTM,random disturbance term is designed to be added at slope,intercept,and both slope and intercept respectively in GTLQTM.3)Quality controlIn this paper,Random Sample Consensus(RANSAC)algorithm is employed to remove outliers during experiments.Since simple least squares method optimally fits all points including the outliers to a line,it fails to detect and reject gross errors.RANSAC,on the other hand,can produce a model which is only computed from the inliers so that to promote the goodness of model fitting.4)Uncertainty analysisBy forecasting through point and interval prediction method and residual analysis of a model,the prediction interval is given to realize quantitative analysis of forecast uncertainty.Theoretically,when the predicted point closer to the mean,the higher the accuracy of its forecast,on the contrary,the more deviation from the mean value,the lower the prediction precision.The experimental results show that:(1)In the simulation experiments,the correlation coefficients between the SOL of Wuhan and simulated GDP of Wuhan port are followed as0.9115,0.9105,0.9233,this means in the case of ideal condition and three random disturbance added simulation the GDP of Wuhan port and SOL of Wuhan city has a strong correlation.Furthermore,the determination coefficient of modeling Wuhan port GDP by SOL from Wuhan city are 0.8321,0.8208,0.8641,which presents the model established basing on GTLQTM can well explain observations;(2)In real data verification,the correlation coefficient between GDP of Wuhan city and SOL of Hubei province is 0.9332 after excluding outliers,which compared with the correlation coefficient 0.8033 before has increased.And the determination coefficient of modeling between GDP of Wuhan city and SOL of Hubei province is 0.8709,indicating goodness of fit is good.The second real data experiment,validation from Guangdong province and Guangzhou city,has obtained an accordant result.The correlation coefficient between Guangzhou population and Guangdong SOL is 0.8992,and the coefficient of determination of modeling between Guangzhou population and Guangdong SOL is 0.8086.Above all,when local SOL and local SEI has significant correlation and correlation is(or approximately)equal to correlation between global SOL and global SEI,the proposed GTLQTM can well explain the relationship of global SOL and local SEI.Namely,when SOL and SEI have a strong correlation,the socio-economic parameters of the local area can be converted from the global SOL to local SEI equivalently.GTLQTM is a good solution to estimate local area SEI using SOL.It also explains the possibility to solve small-scale problems using DMSP/OLS data.This paper will broaden the application range of scales of DMSP/OLS nighttime light imagery.Moreover,GTLQTM is not just limited to the DMSP/OLS nighttime light application,but also provides a new idea for geographic and other research disciplines issues from the global to the local problem.