Application Of Spatial Information Techniques To Study The Spatiotemporal Trend And Ecological Risk Factors Of Japanese Encephalitis In Guangxi Zhuang Autonomous Region, China

Part one:Application of spatial regression techniques to study spatiotemporal trend and ecologic risk factors of Japanese encephalitis in Guangxi.ObjectivesTo describe the spatiotemporal trend of Japanese encephalitis (JE) in Guangxi Zhuang Autonomous Region (Guangxi), from the year2001to2010. And to explore spatial regression techniques to identify JE-related climatic factors and socio-economic factors.Methods1. Date resources were collected from Guangxi Bureau of Meteorology, Agency of Statistics, as well as CDC, including cases of JE, population sizes, area, average annual temperature, yearly maximum temperatures, yearly minmum temperature, average annual relative humidity, average annual rainfall, average annual sunshine hours, average annual airpressure, GDP and sown area, from2001to2010.2. Exploratory spatial data analysis was used to pre-process. Specificly speaking, histogram was used to study distribution characteristics of spatial data, variance inflation factor was used to diagnose multicollinearity of models, spatial empirical bayes model was used to smooth incidence of JE.3. According to Lagrange Multiplier, AIC, SC, R2and residuals, the optimized global spatial regression model was choosed from Spatial Lag Model, Spatial Erorr Model, Spatial Durbin Model and Ordinary Least Squares Regression.4. Based on the optimized global spatial regression model, JE-related ecologic risk factors were identified, on the scale of global level. Moreover, the spatial heterogeinetiy of ecologic factors on JE were detected by Geographical Weighted Regression.Results1. The thematic maps of JE incidence revealed that JE was significantly clustered in northwest of Guangxi, from2001to2007. However, the spatial pattern of JE was uniformity, between2008and2010.2. Lagrange multiplier of Spatial Lag Model was more significant than Spatial Error Model, from2001to2007. Additionaly, the residuals were independent, R2increased, and AIC dropped for Spatial Lag Model. On the contrary, Ordinary Least Squares Regression was better fitted to data features, from2008to2010.3. Overal, average annual rainfall, annual average relative humidity, annual average sunshine hours, Per Capita GDP and proportion of sown area were significantly associated with JE incidence as follows:(1)In2001, proportion of sown area was related with JE (β=-1.30×100, P=0.0005),(2) In2002, Per Capita GDP (β=-2.04×10-5, P=0.036) and proportion of sown area (β=-1.39×100, P=0.000) were associated with JE, (3)In2003, proportion of sown area (β=-7.39×10-1, P=0.035) and Per Capita GDP (β=-3.51×10-5, P=0.000) were associated with JE,(4) In2004, annual average sunshine hours was related with JE (β=-4.75×10-4, P=0.026),(5)In2005, annual average relative humidity (β=2.79×10-2, P=0.016) and Per Capita GDP (β=-2.60×10-5, P=0.002) were significantly related with JE,(6) In2006, average annual rainfall ((β=-2.22×10-4, P=0.022) and annual average sunshine hours (β=-3.61×10-4, P=0.014) as well as proportion of sown area (β=-1.00×100, P=0.000) were associated with JE,(7) In2007, Per Capita GDP (β=-1.30×10-5, P=0.002) and proportion of sown area (β=-8.51×10-1, P=0.001) was related with JE,(8)In2008, proportion of sown area (β=-8.91×10-1, P=0.001) was related with JE,(9)In2009, Per Capita GDP (β=-6.74×10-6, P=0.029) and proportion of sown area (β=-1.21×100, P=0.000) were related with JE,(10)In2010, the incidence of JE was influenced by average annual rainfall (β=-2.56×10-4, P=0.010) and proportion of sown area (β=-1.31×100, P=0.000).4. There was no spatial autocorrelation for residuals, after Geographical Weighted Modelling. Values of AIC of GWR were smaller than Ordinary Least Squares and Spatial Lag Model. The impacts of ecologic risk factors on JE were spatial heterogeneous, as well as β and R2.ConclusionsIn terms of spatial autocorrelation and spatial heterogeneity, Spatial Regression Techniques were more powerful in processing spatial data, compared with traditional regression models. Average annual rainfall, average annual relative humidity, average annual sunshine hours, Per Capita GDP and proportion of sown area were the predominant risk factors of JE in Guangxi, with spatial heterogeneity. Part two:Study on the ecologic risk factors of Japanese encephalitis in Guangxi, integrated RS with GIS.ObjectivesTo identify ecologic risk factors of JE by RS/GIS techiniques.MethodsNnormalized difference vegetation index, surface temperature, land use type and elevation were extrated from MODIS Image. Then, Inverse Distance Weighted and Spearman’s correlation were employed to analyse the relationship between JE and ecologic risk factors.ResultsNormalized difference vegetation index (rs=0.30, P=0.005), average elevation (r,=0.40, P=0.000) and forest (rs=0.23,P=0.030) were positively correlated with JE. However, land surface temperature (rs=-0.32, P=0.002), Cropland (rs=-0.26, P=0.013), built-up land (r,=-0.32, P=0.002) had negative correlation with JE. There was no correlationship between grassland (rs=0.19, P=0.081), water body (rs=-0.03, P=0.785) as well as unused land (r,=-0.01, P=0.933) and JE.ConclusionsRS/GIS techniques could be applied to extract ecologic risk factors from remote image. The spatial pattern of JE was consistent with spatial pattern of ecologic risk factos.