Nighttime Light Data Processing Methodologies And Polycentric City Monitoring

The dramatic global urbanization has witnessed a sharp increase in urban extent and urban population,which consequently gives rise to a succession of socio-economic and environmental problems.Thus,it is of critical significance to obtain a comprehensive understanding of the ongoing urban development in a spatial-explicit and timely manner.The prosperity of geo-information and remote sensing technologies,especially nighttime light?NTL?remote sensing,opens up new opportunities to an in-depth insight into urban studies.However,two long-standing technical issues of NTL data,the Pixel Blooming Effect and data inconsistency problem,significantly impairs the data quality and constrains long-term NTL-based applications,including polycentric city monitoring.In addition to these data issues,it lacks an effective method to automatically identify and quantitatively describe the feature of polycentric cities.The overarching goal of this research is to address“the Pixel Blooming Effect”and“data inconsistency”problems of NTL data and to generate high-quality and consistent long-term NTL time series.Based on this NTL time series,this research proposes a holistic method to develop an automatic urban center identification algorithm and quantitative feature description approach to reveal the long-term spatiotemporal dynamic characteristic of polycentric cities.Specifically,?1?This thesis proposes a new concept,the Pixel Blooming Effect?Pi BE?,to delineate the mutual influence of lights from a pixel and its neighbors,and an integrated framework to eliminate the Pi BE in radiance calibrated DMSP-OLS datasets(DMSP_grc).First,lights from isolated gas flaring sources and a Gaussian model were used to model how the Pi BE functions on each pixel through point spread function?PSF?.Second,a two-stage deblurring approach?TSDA?was developed to deconvolve DMSP_grc images with Tikhonov regularization to correct the Pi BE and reconstruct Pi BE-free images.Third,the proposed framework was assessed by synthetic data and VIIRS imagery and by testing the resulting image with two applications.It was found that high impervious surface fraction pixels?ISF>0.6?were impacted by the highest absolute magnitude of Pi BE,whereas NTL pattern of low ISF pixels?ISF<0.2?was more sensitive to the Pi BE.By using TSDA the Pi BE in DMSP_grc images were effectively corrected which enhanced data variation and suppressed pseudo lights from non-built-up pixels in urban areas.The reconstructed image had the highest similarity to reference data from synthetic image?SSIM=0.759?and VIIRS image?r=0.79?.TSDA showed an acceptable performance for linear objects?width>1.5 km?and circular objects?radius>0.5 km?,and for NTL data with different noise levels?<0.6??.In summary,the proposed framework offers a new opportunity to improve the quality of DMSP-OLS images and subsequently will be conducive to NTL-based applications,such as mapping urban extent,estimating socioeconomic variables,and exploring eco-impact of artificial lights.?2?Global-scale NTL data are acquired by two satellite sensors,i.e.,DMSP-OLS and VIIRS,but the data collected by the satellites are not compatible.To address this issue,this thesis proposes a cross-sensor calibration framework to generate long-term and consistent NTL yearly time series.First,a logistic model was employed to estimate and smooth the missing DMSP-OLS data.Second,the Lomb-Scargle Periodogram technique was used to statistically examine the presence of seasonality of monthly VIIRS time series.The seasonal effect,noisy and unstable observations in VIIRS were eliminated by the BFAST time-series decomposition algorithm.Then,this study established a residual corrected geographically weighted regression model?GWRc?to generate DMSP-like VIIRS data.A consistent NTL time series from 1996 to 2017 was formed by combining the DMSP-OLS and synthetic DMSP-like VIIRS data.The assessment shows that the proposed GWRc model outperformed existing methods?e.g.,power function model?,yielding a lower regression RMSE?6.36?,a significantly improved pixel-level NTL intensity consistency?SNDI=82.73,R2=0.986?and provided more coherent results when used for urban area extraction.The proposed method can be used to extend NTL time series,and in conjunction with the upcoming yearly VIIRS data and Black Marble daily VIIRS data,it is possible to support long-term NTL-based studies such as monitoring light pollution in ecosystems,and mapping human activities.?3?Based on the long-term and high-quality NTL time series,this thesis develops a holistic framework to explore the dynamic of polycentric cities.First,the maximum yearly composite NDVI data was derived from Google Earth Engine,and together with NTL data,a Support Vector Machine was employed to obtain the built-up areas.Then,this study proposed an automatic urban center identification algorithm based on region growing operation.Each urban center was fit by a Gaussian Volume Model and the estimated model parameters were used to describe the spatial,morphological and magnitude features of urban centers.As a result,one main center and five sub-centers were identified which were largely in line with the ongoing urban planning strategy of Hangzhou city.The Gaussian Volume Model was found effective in visualizing and analyzing the spatiotemporal dynamic of polycentric cities.Compared with other data sets and methods,the proposed framework provides a more direct and compressive representation of how human activities respond to polycentric urban planning.