期刊
EUROPEAN JOURNAL OF PLANT PATHOLOGY
卷 137, 期 4, 页码 689-706出版社
SPRINGER
DOI: 10.1007/s10658-013-0280-x
关键词
FHB; DON prediction; All-subsets regression; Resampling
资金
- European Commission
- Quality of Life and Management of Living Resources Programme (QOL)
- Key Action 5 on Sustainable Agriculture [QLK5-CT-2000-01517 (RAMFIC)]
- OECD Fellowship
- BBSRC [BBS/E/J/000CA310] Funding Source: UKRI
- Biotechnology and Biological Sciences Research Council [BBS/E/J/000CA310, BBS/E/J/00000606] Funding Source: researchfish
Fusarium head blight (FHB) of wheat, caused by several Fusarium species, is a damaging disease, resulting not only in yield reduction but also accumulation of mycotoxins in grain. Epidemiology and management of FHB has been extensively studied worldwide. Data on FHB development and accumulation of mycotoxins were obtained in four European countries during 2001-2004 to study the effect of FHB development and environmental conditions on accumulation of deoxynivalenol (DON). The occurrence of DON was highly correlated with presence of one or more toxigenic Fusarium species. Hourly weather data recorded at each sampling site were summarised over several periods of different lengths (5-30 days) during the anthesis and pre-harvest period. All-subsets regression was used to determine the extent to which the probability of DON occurrence is related to weather variables and also the consistency of such a toxin-weather relationship. Combined with a re-sampling technique, all-subsets regression analysis showed the difficulties in identifying a single 'best' model of relating the probability of toxin a parts per thousand yen90 mu g kg(-1) to weather predictors. A wide range of inter-related weather predictors based on time windows around anthesis and pre-harvest were selected in different models. There were many alternative models based on weather predictors only with similar predictive power because of high correlation among weather predictors. The performance of these alternative models was generally poor, particularly in terms of the high proportion of false positive predictions (specificity was only around 0.60-0.65). Inclusion of the number of toxigenic Fusarium species at harvest into models did not improve the model sensitivity (ca. 0.75-0.80) but appreciably improved the specificity (ca. 0.70-0.75). On balance, weather summarised over a 15-day window frame led to models with better predictions than other three window frames (5, 10 and 30 days).
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据