Spatial temporal modelling of crop disease data using high-dimensional regression (PhD)

ABSTRACT

Septoria leaf blotch, caused by the fungus Septoria tritici, is one of the most serious foliar diseases of winter wheat across England and Wales, causing considerable reduction in yield quality and quantity. There are increasing pressures (e.g., legislative, environmental protection, public awareness) to control such diseases in a responsible and sustainable fashion. Disease forecasting systems may offer the potential to meet such an aim but previous work has struggled to find a reliable scheme that growers can adopt easily.

This project aims to develop new approaches and methodologies for developing risk prediction schemes, which are reliable and commercially practicable. To achieve this the project attempts to improve the reliability of forecasts for S. tritici through the development of novel methods for analysing weather records and associated historic disease data. The two major advances achieved are the development of high dimensional statistics and the use of a clustering method to define disease forecast zones. Firstly, new regression techniques were developed for the analysis of data where there are many more potential predictors than observations, so called “fat data”. Next, two innovative procedures for selecting the significant predictors in high dimensional models were introduced which included a new methodology to handle data with correlated errors. In addition, the high dimensional models were applied to quantify important environmental factors influencing S. tritici development. This led to the use of cluster analysis to define a set of forecast zones across England and Wales which demonstrated an improvement in prediction accuracy at an early stage in the growing season.

The project results show that epidemics have intrinsic temporal and spatial scales that must be matched by control strategies if they are to be both effective and efficient. In addition to the current application on S. tritici, the high dimensional models developed can be extended to other wheat, barley and arable crop foliar diseases.