Use of an ecologically relevant modelling approach to improve remote sensing-based schistosomiasis risk profiling

Walz, Yvonne, Wegmann, Martin, Leutner, Benjamin, Dech, Stefan W., Vounatsou, Penelope, N'Goran, Eliézer K., Raso, Giovanna and Utzinger, Jürg, (2015). Use of an ecologically relevant modelling approach to improve remote sensing-based schistosomiasis risk profiling. Geospatial Health, 10(2), 271-279

Document type:
Article
Collection:
UNU-EHS Articles

Metadata
Links
Versions
Statistics
  • Sub-type Journal article
    Author Walz, Yvonne
    Wegmann, Martin
    Leutner, Benjamin
    Dech, Stefan W.
    Vounatsou, Penelope
    N'Goran, Eliézer K.
    Raso, Giovanna
    Utzinger, Jürg
    Title Use of an ecologically relevant modelling approach to improve remote sensing-based schistosomiasis risk profiling
    Appearing in Geospatial Health
    Volume 10
    Issue No. 2
    Publication Date 2015
    Place of Publication Pavia
    Publisher PAGEPress
    Start page 271
    End page 279
    Language eng
    Abstract Schistosomiasis is a widespread water-based disease that puts close to 800 million people at risk of infection with more than 250 million infected, mainly in sub-Saharan Africa. Transmission is governed by the spatial distribution of specific freshwater snails that act as intermediate hosts and the frequency, duration and extent of human bodies exposed to infested water sources during human water contact. Remote sensing data have been utilized for spatially explicit risk profiling of schistosomiasis. Since schistosomiasis risk profiling based on remote sensing data inherits a conceptual drawback if school-based disease prevalence data are directly related to the remote sensing measurements extracted at the location of the school, because the disease transmission usually does not exactly occur at the school, we took the local environment around the schools into account by explicitly linking ecologically relevant environmental information of potential disease transmission sites to survey measurements of disease prevalence. Our models were validated at two sites with different landscapes in Côte d’Ivoire using high- and moderate resolution remote sensing data based on random forest and partial least squares regression. We found that the ecologically relevant modelling approach explained up to 70% of the variation in Schistosoma infection prevalence and performed better compared to a purely pixelbased modelling approach. Furthermore, our study showed that model performance increased as a function of enlarging the school catchment area, confirming the hypothesis that suitable environments for schistosomiasis transmission rarely occur at the location of survey measurements.
    Copyright Holder PAGEPress
    Copyright Year 2015
    Copyright type Creative commons
    DOI http://dx.doi.org/10.4081/gh.2015.398.
  • Versions
    Version Filter Type
  • Citation counts
    Google Scholar Search Google Scholar
    Access Statistics: 928 Abstract Views  -  Detailed Statistics
    Created: Wed, 09 Dec 2015, 00:59:33 JST