Barley yellow dwarf virus (BYDV) damage to winter cereals and population dynamics of the aphid Rhopalosiphum padi during fall were monitored in fields during 10 years at various locations in the northern half of France. Logistic regression was used to examine whether a simple risk probability algorithm based only on the autumnal population dynamics of R. padi can accurately predict yield losses caused by BYDV and, therefore, the need for insecticide treatment. Results showed that the area under the curve of the percentage of plants infested by R. padi during autumn was highly significantly related to BYDV yield losses. Then, a cost/benefit analysis was performed to estimate the optimal decision threshold resulting in the lowest annual average costs of BYDV damage and control. A "model use" strategy allowed a reduction in the annual average costs of BYDV disease and control of up to 36% when compared with a "prophylactic spraying" strategy. The optimal decision threshold was highly sensitive to variation in disease prevalence. This property was used to propose an easy way to adapt the model to any production situation through the determination of the most accurate decision threshold.

Barley Yellow Dwarf (BYD) disease is one of the most severe viral diseases in autumn sown cereals. In western Europe, crop losses are mainly due to the PAV species of BYD viruses transmitted by Rhopalosiphum padi, the most abundant aphid in autumn. The proportion of migrant winged aphids that carry viruses in autumn is considered a major epidemiological factor for determining the disease incidence. In the main French cereal areas, during a 6 week period in autumn 1999 to 2002, the proportion of viruliferous R. padi assessed using a TAS-ELISA technique was on average of 4.98 % (range 2.01-9.91 %). Variations according to trap location were correlated with land-use at the regional scale, annual variations being correlated with the climate of the year. Implementations of these results to improve BYD disease management program are discussed.