High-resolution vector-borne disease infection risk mapping with Area-to-Point Poisson Kriging and Species Distribution Modeling

BackgroundDisease infection data is usually aggregated and shared as a sum of infections in a given area over time. This data is presented as choropleth maps. The aggregation process protects privacy and simplifies decision-making but introduces visual bias for large areas and sparsely populated places. Moreover, aggregated areas of varying sizes cannot be simply used as the input for complex ecological models, which are based on data retrieved at higher resolution on regular grids. The issue is especially painful for vector-borne diseases, e.g. Lyme Disease, where infection risk is closely related to vector species and their ecological niche. MethodsThe paper presents the method of obtaining high-resolution risk maps using a pipeline with two components: (1) spatial disaggregation component, which transforms incidence rate aggregates into the point-support model using Area-to-Point Poisson Kriging, and (2) species distribution modeling component, which detects areas where ticks bite is more likely using MaxEnt model. The first component disaggregates Lyme Disease incidence rates summed over counties in Poland, Central Europe, in 2015. The second component uses ticks occurrence maps, Leaf Area Index, Normalized Difference Vegetation Index, Land Surface Temperature derived from Earth Observation satellites, and Digital Elevation Model. The final weighted population-at-risk map is a product of both components outputs. The pipeline is built upon open source and open science projects, and it is reusable. ResultsThe presented pipeline creates high-resolution risk maps: vector occurrence probability map, population-at-risk map, and weighted population-at-risk map which includes information about local infections and about vector species. The final maps have much better resolution than aggregated incidence rates. Visual bias for population-at-risk maps is removed, and unpopulated areas are not presented on the map. ConclusionsThe pipeline might be used for other vector-borne diseases. The final weighted population-at-risk map might be used as an input for another analytical model requiring high-resolution data placed over a regular grid. The pipeline removes visual bias and transforms aggregated data into a high-resolution point-support layer.