Open AccessA Stochastic Optimization Method to Estimate the Spatial Distribution of a Pathogen from a Sample
Author(s) -
Stephen Parnell,
T. R. Gottwald,
Mike Irey,
Wenjian Luo,
Frank van den Bosch
Publication year - 2011
Publication title -
phytopathology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.264
H-Index - 131
eISSN - 1943-7684
pISSN - 0031-949X
DOI - 10.1094/phyto-11-10-0311
Subject(s) - kriging , geostatistics , sample (material) , statistics , spatial distribution , sample size determination , statistic , distribution (mathematics) , range (aeronautics) , sampling (signal processing) , biology , spatial variability , mathematics , computer science , engineering , mathematical analysis , chemistry , filter (signal processing) , chromatography , computer vision , aerospace engineering
Information on the spatial distribution of plant disease can be utilized to implement efficient and spatially targeted disease management interventions. We present a pathogen-generic method to estimate the spatial distribution of a plant pathogen using a stochastic optimization process which is epidemiologically motivated. Based on an initial sample, the method simulates the individual spread processes of a pathogen between patches of host to generate optimized spatial distribution maps. The method was tested on data sets of Huanglongbing of citrus and was compared with a kriging method from the field of geostatistics using the well-established kappa statistic to quantify map accuracy. Our method produced accurate maps of disease distribution with kappa values as high as 0.46 and was able to outperform the kriging method across a range of sample sizes based on the kappa statistic. As expected, map accuracy improved with sample size but there was a high amount of variation between different random sample placements (i.e., the spatial distribution of samples). This highlights the importance of sample placement on the ability to estimate the spatial distribution of a plant pathogen and we thus conclude that further research into sampling design and its effect on the ability to estimate disease distribution is necessary.