Grapevine powdery mildew, caused by Erysiphe necator, is a cosmopolitan pathogen that threatens grape production worldwide. Accurate disease forecasts would enhance disease management, particularly where severity varies substantially between years, such as in New York State. Ascospore release occurred before local bud break of Vitis vinifera in NY, WA, NC and NJ, and was quantitatively related to accumulation of degree days and rain events during overwintering. Acute cold temperature events (<8°C) induced transient resistance in V. vinifera to E. necator that lasted for 24 to 36 h and reduced infection efficiency and colony development. Acute cold temperatures also damaged existing colonies, with the highest observed levels of mortality occurring in 3-day-old colonies. Occurrence of cold events as defined above was documented for many viticultural areas of the United States, Europe, and Australia. With respect to leaf surface temperatures, acute cold events probably occur more frequently than indicated in historical data bases of air temperature due to radiational cooling of the leaf surface during clear, calm nights, e.g., grape leaf surfaces in a Geneva, NY research vineyard were 0.9 to 6.6 degrees cooler than the air on clear nights. A regression model was developed from 23 years of historical weather and disease data to forecast fruit disease severity on unsprayed vines using the previous autumn's heat accumulation and pan evaporation (Epan) rates (or Epan estimated from evapotranspiration [Eto]) from 2 wks pre- to 6 wks post-bloom; timeframes relevant to production of (i) cleistothecia capable of surviving winter, and (ii) conidial availability for infection of susceptible berries. Monte Carlo simulation coupled historical and forecasted weather data to predict a likely range of future Eto values, which were then used to predict favorability for disease. Powdery Mildew Risk Assessment models were developed using both logistic regression (LR) and recursive partition analysis (RPA) from historical disease and weather records, and classified risk for powdery mildew infection on fruit. The models correctly classified mild and severe years in 18 of 22 and 19 of 22 years in verification studies, and 9 of 10 and 5 of 10 years in validation studies for the LR and RPA models, respectively.