Early Warning for Septoria tritici in treated crops
Since the late 1970s various attempts have been made in Britain and elsewhere to produce forecasting
schemes for Septoria tritici in wheat crops to give an early warning of risk for disease control. The
success of such work could be described as 'mixed', and efforts to improve risk estimation continue as
weather recording and data analysis techniques have developed. More importantly, insufficient attention
has previously been given to the scale at which predictions are provided - most have been applied at the field scale.
The CropMonitor forecast for Septoria severity in the 2008 season is a novel feature introduced this year, and is based on a twin implementation of Stepwise Partial Least Squares Regression (S-PLSR) and clustering analysis.
S-PLSR is an innovative technique for determining which factors, in this case many hundreds of weather variables (obtained by iterative summary procedure), are best related to a future event. Septoria severity at GS 75 is the future event of interest, and the weather variables are gathered during the previous winter and spring. The S-PLSR technique can identify which factors are important for future Septoria infection and quantify their influence. A further set of weather observations related to spring rainfall will improve forecast accuracy for eventual Septoria severity and the map will be updated accordingly when these become available in early June.
The clustering analysis provides an objective way to divide England & Wales into zones in which the historic severity of Septoria has been statistically similar. In the past arbitrary boundaries (such as counties) have been used to divide the nation into zones but these have been chosen more for convenience than disease occurrence. As a consequence, the aggregate of disease using arbitrary boundaries do not match the actual disease phenomenon under investigation. The clusters used for the CropMonitor map are based on wheat disease survey data from 1991-2004 and are specific for Septoria tritici. This poster (in pdf format) explains the background work, particularly the clustering analysis, in more detail. View Poster: Disease forecasting for Septoria tritici on winter wheat
The CropMonitor forecast for Septoria severity in the 2008 season is a novel feature introduced this year, and is based on a twin implementation of Stepwise Partial Least Squares Regression (S-PLSR) and clustering analysis.
S-PLSR is an innovative technique for determining which factors, in this case many hundreds of weather variables (obtained by iterative summary procedure), are best related to a future event. Septoria severity at GS 75 is the future event of interest, and the weather variables are gathered during the previous winter and spring. The S-PLSR technique can identify which factors are important for future Septoria infection and quantify their influence. A further set of weather observations related to spring rainfall will improve forecast accuracy for eventual Septoria severity and the map will be updated accordingly when these become available in early June.
The clustering analysis provides an objective way to divide England & Wales into zones in which the historic severity of Septoria has been statistically similar. In the past arbitrary boundaries (such as counties) have been used to divide the nation into zones but these have been chosen more for convenience than disease occurrence. As a consequence, the aggregate of disease using arbitrary boundaries do not match the actual disease phenomenon under investigation. The clusters used for the CropMonitor map are based on wheat disease survey data from 1991-2004 and are specific for Septoria tritici. This poster (in pdf format) explains the background work, particularly the clustering analysis, in more detail. View Poster: Disease forecasting for Septoria tritici on winter wheat
This work was funded by the HGCA and ORSAS.
