The importance of dew to agriculture, together with the absence of dew measurements at standard weather station sites, resulted in the development of predictive models for dew formation in the Umatilla River Basin, Oregon. Meteorological data were obtained at the Pendleton Experiment Station in close proximity to dew-measurement devices, and from the Pendleton National Weather Service Office. Dew measurements, made during the months of April, May and June for the years 1971–76, were compared with values of wind speed, relative humidity and minimum temperature at the Experiment Station. All three variables were found to have significantly different average values when dew occurred as opposed to when dew did not occur. The duration of relative humidity above 90% exhibited the strongest correlation with dew duration (r = 0.60). Using these boundary values beyond which the occurrence of dew was unlikely, dew occurrence was predicted with an accuracy of 81%. With the addition of a boundary value for soil moisture, the predictive accuracy increased to 90%. When the technique for forecasting dew occurrence was combined with a regression model for estimating dew duration, the result was an overall system for the prediction of dew duration (within 3 h of actual) which was accurate 71% of the time. This predictive accuracy rose to 85% for years when soil moisture data were available. Use of climatic data from the Pendleton Weather Service Office led to diminished forecasting accuracy, apparently due to differences in geography and differences in height of the measuring instruments at the two stations. The forecasting system developed for the Pendleton Experiment Station failed to produce a statisfactory level of predictive accuracy when used at a nearby Weston site. The findings of this research, together with the findings of other similar studies, illustrate the difficulty of constructing dew prediction models with general applicability and emphasize the need for widespread dew measurements at standard weather stations.