Detecting Predictors And Future Risk Prediction For Malaria Incidences In Northern China Using Remote Sensing Data

Malaria is a parasitic prevalent disease caused by parasites and transmitted with the anopheles bites, featuring in large population at risk and wide risk areas. Since 2010, the trend of malaria increasing back appeared. In China, malaria features in significant regional distribution and Northern China is one of the most prevalent areas.The objective of this research is malaria incidences. In the research, the spatial-temporal data malaria incidences in three provinces in Northern(Shandong, Henan, and Anhui) from 2004 to 2010 was selected to explore the spatial local factors effects on the incidences, and predict its spatial distributions under future climate change scenarios.The detail research objectives include:(1) forming the basic recognitions of the spatial-temporal distribution features of malaria incidences, and summarizing the spatial-temporal regulars on the basis of sample data;(2) analyzing the characteristics of remote sensing data regarded as potential factors to explain malaria incidences;(3) constructing the relationships between climate and malaria incidences within administrative unites and thus exploring factors;(4) predicting spatial distributions of malaria incidences under climate change scenarios in 2030 s, 2050 s, and 2070 s.To address the above problems, four aspects were included in the research and the conclusions were as follows:First, exploratory analysis was performed for malaria incidences using statistics and spatial analysis methods. Significant spatial autocorrelations were detected for the incidences with Moran’s I: 0.522 – 0.712, p < 0.01. For the local spatialautocorrelation test of the 7-year average incidences case, the significant spatial clustering was found in Huaihe River areas. The average incidence was 0.597‰ pa of the 23 counties in hot-spot region(H-H), while that in other 291 counties was 0.013‰ pa.Potential variables especially those derive from remote sensing data could explain the variations of malaria incidences from different angles. The results of correlation analysis showed that the remote sensing variables, NDVI, LST and LST difference, had lagged one month effects on the incidences. Meanwhile, significant negative correlation appeared between the incidence and elevation, significant positive relationship appeared for the incidence and WDI depicting ratios of water regions, and significant negative relationship appeared between the incidence and GDP describing socio-economic conditions.Moreover, geographically weighted regression(GWR) was used to detect the spatial local factors effects on malaria incidences across the space. As the spatial attributes of factors were considered, that is the spatially localized effects, the prediction accuracy was improved(R2 = 0.710) compared with the results of linear regression(R2 = 0.134). In the results, the ratios of counties with significantly positive and negative relationships between the malaria incidences and variables, the spatial distribution of these counties, and the average incidences in the counties were summarized. Further, the above indexes charactering the variation of variables’ effects were also summarized and analyzed for the annual cases, from 2004 to 2010.Finally, considering the sophisticated nonlinear relations between the malaria incidences and variables stem from remote sensing data, genetic programing(GP) method was used for future incidences prediction. The predicting accuracy(R2 = 0.685) was better than that of linear regression(R2 = 0.336). Under the future climatechange scenarios(RCP2.6, RCP4.5, RCP8.5), the malaria incidences were predicted with the future temperatures in 2030 s, 2050 s, and 2070 s. With the increasing temperature, the average incidences of the study area reduced in 2030, and significantly increased in 2050 and 2070. And the maximum incidences in the study area all increased in 2030, 2050 and 2070.