Remote sensing applied to schistosomiasis control: The Anning River project.

摘要

This dissertation presents the use of remote sensing in identifying habitat of the Oncomelania hupensis robertsoni snail, the intermediate host for schistosomiasis in mountainous regions of China. This work is motivated by the construction of the Three Gorges Dam on the Yangtze River, which will be the largest dam in the world, resulting in considerable ecological change, and potentially introducing snail habitat and subsequent schistosomiasis into a previously non-endemic area. Field data from 1994 snail surveys were used to develop a two-tiered classification approach using Isodata clustering and maximum likelihood classification to discriminate between snail habitat and non-habitat from Landsat TM satellite data for the Anning River valley in southwest Sichuan province. A ranking scheme was devised to prioritize areas as to their potential for being snail habitat. A field validation study showed that the classification performed very well in identify potential snail habitat from landscape that was clearly not potential habitat. Within areas identified as potential habitat, sites with higher snail priority rankings generally corresponded to positive snail sites. A large source of potential misclassification was attributed to flooding in the south of the study area the year of the validation study. Measurements of water, soil, and landscape ecology associated with snail habitat were conducted in the field. These data were used to create a classification that required only 4 soil variables (sulfur, phosphorus, magnesium, and silt) to create an accurate distinction between snail habitat and non-habitat This classification was used to classify marginal habitat sites—sites that were thought to be able to support snails by field ecologists, despite there being no snails found. Over 70% of these marginal sites were correctly classified as no-snail sites based on soil data. An ecological interpretation of the remote sensing classification was determined by using 6 land cover variables and 6 soil variables to predict the results of the remote sensing classification.