Finding The Clues Of Geographic Access To Poverty In China

Despite the tremendous progress that has been made in poverty reduction over the last century,poverty is still a persistent worldwide problem that hampers socio-economic development and human progress.Research on poverty has revealed a series of useful geographic poverty indicators.Regional poverty research has sparked significant interest in geographic poverty indicators.Numerous studies have focused on land cover,environmental conditions,and resource endowment,but the relationship between regional poverty and geographical indicators is not yet adequately understood.Previous research has generally focused on a single perspective.Additionally,landscape patterns as visible clues about regional poverty are often overlooked.Further,with the development of high-resolution nighttime images of light use,nighttime light has been widely utilized in poverty measurement as an indicator for human activity.Nighttime light images,however,are limited in their capacity to measure agricultural activity,so their usefulness as an indicator of activity and poverty on the microlevel needs to be investigated further.This thesis aimed to provide valuable insight into the relationship between geography and poverty by using a series of geographical indicators.Based on previous research,a theoretical framework was proposed to analyze regional poverty using land cover,topography,and human activity.An index system was developed to include 77 potential explanatory variables including10 land composition variables,44 landscape pattern variables,3 topographical variables,and 10 human activity variables.Considering the autocorrelation phenomenon among potential explanatory variables,the explanatory variables were selected using a stepwise regression feature selection procedure.Geographically Weighted Regression(GWR)and Random Forest regression(RF)were employed to analyze the reliability of this theoretical framework.Moreover,the spatial non-stationarity phenomenon was explored by varying the coefficient of the explanatory variables in the GWR.A quantitative analysis of the case of Zunyi City and Qiandongnan State in Guizhou Province was conducted to examine the underlying correlation between potential explanatory variables and the poverty headcount ratio in 417 towns.The study yielded various results.The r-square of the GWR and RF models was 0.64 and0.54,respectively,which confirmed the reliability of this regional poverty theoretical framework.The spatial correlation figures(LISA)cluster map derived from the predicted value of the RF has a similar distribution of spatial autocorrelation with observed values.Land cover type,geometric complexity,spatial distribution,and physical connectivity of landscape patterns have a statistically significant correlation to the poverty headcount ratio.The results of the GWR indicate the spatial non-stationarity of explanatory variables across the 417 towns.The results of the stepwise regression reveal that the strength of nighttime light has a non-significant correlation to poverty headcount ratios at the town level,and that topography also has little effect on regional poverty.The results indicate that land cover is closely correlated to regional poverty through land composition and landscape patterns in the two cities.Additionally,the landscape pattern of land cover provides vital insight into regional poverty based on the geometric complexity and spatial distribution of certain land cover types.Moreover,the usefulness of nighttime light images on the microlevel merits further investigation in areas with high poverty rates.The findings provide useful insight into regional poverty through a series of geographical indicators and into the relationship between land and poverty.This insight is critical for tackling poverty in China and for formulating effective targeted poverty reduction policies.