Spatio-Temporal Prediction of the Epidemic Spread of Dangerous Pathogens Using Machine Learning Methods

Real-time identification of the occurrence of dangerous pathogens is of crucial importance for the rapid execution of countermeasures. For this purpose, spatial and temporal predictions of the spread of such pathogens are indispensable. The R package papros developed by the authors offers an environ...

Full description

Saved in:
Bibliographic Details
Published in:ISPRS international journal of geo-information 2020-01, Vol.9 (1), p.44
Main Authors: Hamer, Wolfgang B., Birr, Tim, Verreet, Joseph-Alexander, Duttmann, Rainer, Klink, Holger
Format: Article
Language:English
Subjects:
Agricultural production
Airborne microorganisms
Biology
Climate
Climate models
Datasets
Disease
Dispersal
Dispersion
Flowers & plants
Fungicides
Humidity
Infections
Infestation
infestation forecast
Learning algorithms
Learning behaviour
Machine learning
Methods
Open data
Pathogens
Pesticides
Powdery mildew
Precipitation
Predictions
Probability theory
random forest
Weather
Yields
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Real-time identification of the occurrence of dangerous pathogens is of crucial importance for the rapid execution of countermeasures. For this purpose, spatial and temporal predictions of the spread of such pathogens are indispensable. The R package papros developed by the authors offers an environment in which both spatial and temporal predictions can be made, based on local data using various deterministic, geostatistical regionalisation, and machine learning methods. The approach is presented using the example of a crops infection by fungal pathogens, which can substantially reduce the yield if not treated in good time. The situation is made more difficult by the fact that it is particularly difficult to predict the behaviour of wind-dispersed pathogens, such as powdery mildew (Blumeria graminis f. sp. tritici). To forecast pathogen development and spatial dispersal, a modelling process scheme was developed using the aforementioned R package, which combines regionalisation and machine learning techniques. It enables the prediction of the probability of yield- relevant infestation events for an entire federal state in northern Germany at a daily time scale. To run the models, weather and climate information are required, as is knowledge of the pathogen biology. Once fitted to the pathogen, only weather and climate information are necessary to predict such events, with an overall accuracy of 68% in the case of powdery mildew at a regional scale. Thereby, 91% of the observed powdery mildew events are predicted.
ISSN:2220-9964
2220-9964
DOI:10.3390/ijgi9010044