Previous View
APSnet Home
Plant Disease Home



Development and Validation of an Empirical Model to Estimate the Duration of Dew Periods. M. L. GLEASON, Department of Plant Pathology, Iowa State University, Ames 50011. S. E. TAYLOR, Department of Agronomy, and T. M. LOUGHIN and K. J. KOEHLER, Department of Statistics, Iowa State University, Ames 50011. Plant Dis. 78:1011-1016. Accepted for publication 11 July 1994. Copyright 1994 The American Phytopathological Society. DOI: 10.1094/PD-78-1011.

An empirical model to estimate the occurrence and duration of dew periods was developed using hourly data for relative humidity (RH), air temperature, and wind speed from June to September 1990 for Ames, Iowa. After using a nonparametric classification procedure called CART to eliminate periods in which dew occurrence was unlikely, stepwise linear discriminant (SLD) analysis was performed with categories of measured dew (0 = no dew, 1 = dew) as the dependent variable. The resulting CART /SLD model and an alternative model that assumed dew was present when RH > 90% were validated by using hourly data from 13 weather stations in Iowa, Kansas, Nebraska, and Illinois during April through October 1992. For 17,487 potential dew hours, both models predicted the mean duration of dew periods within 1 hr, but mean square error was considerably larger for the RH > 90% model than the CART/SLD model. The CART/SLD model estimated presence or absence of dew correctly for 83.5% of the potential dew hours compared to 78.6% for the RH > 90% model. Similarly, the CART/SLD model predicted the duration of dew periods within 2 hr on 76.0% of 1,502 nights compared to 67.2% of nights for the RH > 90% model. For both models, size distribution of errors in estimating dew duration was approximately normal for three weather stations, skewed toward overestimation for eight stations, and skewed toward underestimation for two stations. After further modification, the CART/SLD model could provide dew-period estimates over broad geographical areas for disease-warning systems that are driven by wetness duration and temperature.

Keyword(s): disease prediction, integrated pest management, weather models