Open AccessValidation of a Commercial System for Remote Estimation of Wetness Duration
Author(s) -
Mark L. Gleason,
Sharon K. Parker,
Ron E. Pitblado,
Richard Latin,
Donna Speranzini,
Ruth V. Hazzard,
Martha J. Maletta,
W. Cowgill,
Darvin L. Biederstedt
Publication year - 1997
Publication title -
plant disease
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.663
H-Index - 108
eISSN - 1943-7692
pISSN - 0191-2917
DOI - 10.1094/pdis.1997.81.7.825
Subject(s) - leaf wetness , dew , environmental science , dew point , magnitude (astronomy) , air temperature , duration (music) , atmospheric sciences , hydrology (agriculture) , meteorology , biology , horticulture , geography , geology , art , physics , literature , geotechnical engineering , astronomy , condensation
To assess the accuracy of remote, real-time mathematical simulations of wetness duration and air temperature, hourly measurements of wetness duration and air temperature at 18 sites in the United States and Canada from May to September 1995 were compared with simulations for these sites provided by SkyBit, Inc. SkyBit simulations of mean, maximum, and minimum daily air temperatures varied from on-site measurements by less than 0.7°C but underestimated the duration of wet periods by an average of 3.4 h/day. At five of six stations tested, SkyBit underestimates of wetness duration were significantly (P < 0.01) larger on days when no rain was measured than on rainy days, indicating that simulations of dew-period duration were much less accurate than simulations of rain-period duration. The vast majority of hours SkyBit misclassified as dry occurred either when entire wet periods were missed (59.3%) or when the onset of a wet period was detected late (28.4%). The results suggest that revision of SkyBit wetness-simulation models should focus on reducing error rates during dew events. In simulations using two disease-warning models, TOM-CAST and Melcast, with mean values of measured and SkyBit-simulated wetness duration, SkyBit-simulated values resulted in fewer and later fungicide spray advisories than did measured values. The magnitude of these impacts varied with the magnitude of the simulation errors and with differences in the models' decision rules.