| Prevention and control of Vector-borne diseases are major public health pr-oblems at present . As the global change, human impact on ecological environment and constructing large-size hydraulic works,the risk of Vector-borne diseasesoutbreak will increase in local regions.One key issue of prevention and control such infectious diseases is how to evaluation the level of these infectious disea-ses reasonablely.Vector density is a common used index and its quality of monitor and statistical inference,will affect the evaluation of Vector-borne diseases directly, thereby affecting the decision-making results. At present,although conside-ring geographical differences when lay out monitoring points, the statistical inference still using traditional methods. Traditional statistical inference methods und-er the assumption of independent samples, but vector density data has spatial aut-ocorrelation and spatial heterogeneity, which is not meet the assumption of traditional methods. Therefore, using the traditional statistical inference methods to estimate the density of vector will produce a larger deviation. In addition,if we want to study the relationship between geographical distribution of rat density and environmental factors, we must ensure the same sacle of environmental fac-tors and rat density,so we need to estimate the rat density of geographic units and then analysis these data. It will increase the monitoring costs greatly when change the basic research unit(such as from the administrator to watershed or ecological),because using traditional sampling methods must redesign samping sche-me.To cope with these problems, this paper used rat density data as a case based the project of Vetor-borne diseases risk assessment in Three Gorges Reservoir,used Sandwich spatial sampling model to estimate rat density,which can resolve the problems when using traditional statistical inference models. Specific-ally, this paper include the following:Firstly, this paper described the problems of traditional monitoring methods and statistical inference of vector density, and pointed out there is large deviati-on in statistical inference of rat vector density without considering spatial autoc-orrelation and spatial heterogeneity.Secondly, used rat density monitoring data of Three Gorges Reservoir to do a case study.We used Sandwich spatial sampling model, combined with prior kn-owledge, choosed soil, landforms, and NDVI as the auxiliary factor to estimate average rat density for each county.Compared the variance of knowledge strat-ification and choosed the classification of NDVI as the knowledge stratificatio-n of Sandwich model.We compared the result of average rat density for each county of Three Gorges Reservoir before and after water impounded using two report units such as administrator and grid cells.We think that grid units has the great advantages than administrator's, for instance, it could show us more details of rat density spatial distribution in the small scale. Moreover, when we want to study the relationship between the geographical distribution of environmental factors and rat density, grid report units also show a powerful advantage. Thirdly, we compared the estimation accuracy among geostatistical models, hierarchical Bayesian model and Sandwich model. As a result, although geostat-istical model could estimate average rat density for every county of Three Gorg-es reservoir, it couldn't meet the application requirements as there was too lar-ge errors; Hierarchical Bayesian model had the same accuracy as Sandwich spatial sampling model, but it couldn't estimate rat density when there was no monitoring points in the county area; Sandwich model could estimate average rat density for every county, but also can guarantee a smaller estimation error which could meet the practical application requirements.Finally, this paper also raised some worth considering questions under sum-mary of the research, that will provide references for further studies. |
PDF Full Text Request