Defra - Department for Environment, Food and Rural Affairs.

Science Search

Science and Research Projects

Return to Science Search homepage   Return to Project List

Risk prediction algorithms for diseases in combinable crops. - AR0513


The project will develop disease risk prediction algorithms to underpin forecasting and live-intelligence schemes, and improve the biological understanding of foliar diseases in combinable crops. This will be achieved using mathematical and statistical approaches. Specifically, data-mining and mechanistic modelling will be used to describe the effects of meteorological variables on disease development and epidemic progress. These relationships will be developed to provide predictive models for disease dynamics. The over-use of fungicides is largely attributable to the absence of reliable risk prediction schemes. Hence, progress from this project will support DEFRA’s Science and Innovation Strategy, which aims to develop and test more resource-efficient and less polluting farming systems. In addition, the project outputs can contribute to DEFRA’s objectives of understanding risks of plant diseases and developing methodologies for surveillance, detection and monitoring of risk.

The project will be closely aligned to other DEFRA funded research. In particular, the project will collaborate closely with the ‘Fellowship in Crop Environmental Interactions with Pathogenic Fungi’ led by N. Paveley (ADAS). Further the project will establish a link and information flow from the DEFRA project (AR0503) ‘Combinable crop health status and protective practice,’ to provide underpinning research on risk prediction algorithms, which could be exploited through seasonal live reporting for growers.

Statistical data mining techniques will be developed and applied to existing/archival data sets and data sets generated by new DEFRA funded research. The work will concentrate on the fungal wheat leaf diseases Mildew (Erysiphe graminis), Yellow rust (Puccinia striiformis), Brown rust (Puccinia hordei) and Glume blotch (Stagonospora nodorum). This research will yield risk algorithms to underpin seasonal live reporting and will also be available for exploitation through forecasting and decision support schemes (e.g. DESSAC).

A barrier to the uptake of disease prediction and warning schemes is that risk algorithms are rarely tested against criteria that are acceptable to the end user. Therefore, a method will be developed to evaluate the performance of risk prediction algorithms with statistical rigour, but with emphasis on the needs of the end-user. A method that has proved to be successful for this purpose in medical epidemiology will be the starting point for this work. Risk prediction algorithms developed from data mining studies will be evaluated using this tool.

Mechanistic models for the relation between environmental variables and epidemic progress will be developed, parameterised and analysed. The models will incorporate both plant and pathogen characteristics as dependent on environmental conditions. The insight from this analysis will be combined with the results from the data mining research to quantify the effects of environmental variables on disease increase throughout the growing season. This work will underpin decision support systems such as the DESSAC Wheat Disease Manager.

1. Develop risk prediction algorithms using data mining methodology.
(a) Extend data mining techniques to include cultivar, thermal time and crop history.
(b) Construct data sets for the four target diseases from extant experimental records.
(c) Analyse the data sets using the extended data mining techniques.
(d) Define risk prediction algorithms on basis of the analysis in 1(c).
(e) Submit a publication of the results of this work.

2. Further operationalise model evaluation techniques and apply them to example risk algorithms.
(a) Develop the method of the Receiver Operating Characteristic to test disease models practicable for disease management.
(b) Construct a data set to be used for the evaluation of two risk algorithms.
(c) Evaluate the risk algorithm using the methods developed.
(d) Submit a publication of the results of this work.

3. Develop, parameterise and analyse mechanistic models for environment-epidemic interaction.
(a) Develop a model for the interaction between environmental variables and plant and pathogen characteristics.
(b) Parameterise this model for a splash dispersed and for a wind-dispersed disease (Septoria tritici and Puccinia striiformis).
(c) Develop a method for sensitivity analysis of this type of model.
(d) Apply sensitivity analysis to the models developed.
(e) Submit a publication of the results of this work.

Time-Scale and Cost
From: 2003

To: 2007

Cost: £406,913
Contractor / Funded Organisations
Rothamsted Research (BBSRC)
Arable Farming              
Climate and Weather              
Climate Change              
Crop Diseases              
Sustainable Farming and Food              
Sustainable Production              
Fields of Study
Arable Crops