Abstract: Sclerotinia stem rot in oilseed rape occurs every year in the UK but incidence varies by region and year, with occasional epidemics that are hard to predict. Control relies on protectant foliar fungicides ahead of any infection, but it is currently difficult to determine the optimum timing of sprays. The aims of this work were to provide forecasting alerts and reports of risk factors for sclerotinia infection to growers during the oilseed rape flowering phase, and demonstrate that the alerts and risk factors helped to improve predictions of sclerotinia disease.Weather-based alerts were published three times a week for growers (https://cereals.ahdb.org.uk/) 2015-2017, for 15 sites/year, based on 48 hr forecast weather data, crop stage and % petals testing positive for sclerotinia. A Burkard trap was deployed at five of these sites to detect airborne spores, and at three of these sites there was a fungicide timing field experiment. Forecast weather data was used to guide fungicide spray timings, for comparison with sprays at yellow-bud, early-, mid- or late-flower. The number of weather based alerts during flowering from 2015-17 varied from zero to three, with the most alerts in the SW and S area. The main difference across the years was the occurrence of alerts in 2017 in E and NE east areas, whereas in 2015 and 2016 these regions were drier and had few alerts.For analysis of the benefits of risk forecasting, data from fungicide timing experiments in 2015-2017 was combined with data from similar experiments in 2010-2012, to give a total of 23 site-years. Zero weather-based alerts were a correct predictor of no- or low-infection (≤ 1% stem rot incidence). But where there were weather-based alerts (at 15 sites), about half of these sites (7) also had an infection incidence of ≤ 1%. Including the results of sclerotinia inoculum on petals with analysis of weather alerts helped to improve the predictions of infection, by correctly identifying the very-low risk sites. The relationship between % of petals testing positive and untreated disease incidence varied with timing of sample (e. g. R2 = 0.54 for petals sampled at yellow-bud), but when combined with weather-based alerts, petal data improved the accuracy of the risk forecast. In summary, when inoculum was zero, infection risk was zero. Positive inoculum indicated risk, but infection was variable. Spore trap results helped explain where infection did or did not occur, by showing the variation during flowering of daily levels of spores in air samples for each site.