T. Vanderzwet et al., EVALUATION OF THE MARYBLYT COMPUTER-MODEL FOR PREDICTING BLOSSOM BLIGHT ON APPLE IN WEST-VIRGINIA AND MARYLAND, Plant disease, 78(3), 1994, pp. 225-230
The MARYBLYT computer model was evaluated for its accuracy in forecast
ing apple blossom infection by Erwinia amylovora and the subsequent ap
pearance of fire blight symptoms. Temperature and rainfall data were c
ollected and disease observations recorded in bearing orchards in West
Virginia and Maryland during 1984 1993. Among the 13 primary infectio
n events identified by the model at all sites in eight of the 10 yr, b
lossom blight symptoms appeared 10 times within +/- 1 day, twice withi
n 2 days, and only once within 3 days of the MARYBLYT prediction. Only
three times in 10 yr did MARYBLYT predict blossom infection without s
ymptom development. In no instance did spurious symptoms appear that w
ould indicate the model failed to identify an infection period. A blos
som sampling procedure conducted during 5 yr (1985, 1987, 1988, 1990,
and 1993) in which blossom blight occurred confirmed the presence of E
. amylovora coincident with the model's threshold calculation of epiph
ytic infection potential. When blossoms were inoculated artificially b
y introducing a bacterial suspension (10(8) cfu/ml) into flower nectar
ies, blossom blight symptoms developed 0, 1-3, and >5 days prior to th
at predicted by the model in one, seven, and three trials, respectivel
y. In 11 trials, an average of 57 degree days >12.7 C was accumulated
between artificial inoculations and symptom appearance, which is consi
stent with the model's algorithm for symptom occurrence. The results o
f our field evaluations of MARYBLYT for predicting blossom infection a
nd subsequent symptom development show that the model is accurate. Tre
atment decisions based on MARYBLYT can be expected to improve the leve
l of control of this destructive disease.