Web-based information system for predicting infection risk of rice blast disease based on gridded weather forecast data

The future occurrence of plant disease could be predicted by disease forecasting models based on weather forecasts. The prediction may allow crop growers enough time periods for applying crop management practices.

We developed a web-based rice blast forecasting system based on the Digital Weather Forecast (DWF) of the Korea Meteorological Administration (KMA) at 5 km grid resolution. The DWF data is released every 3 hours by KMA, including 16 projections of air temperature, relative humidity, and precipitation probability at 3 hour intervals and 4 projections of precipitation at 12 hour intervals. The system estimates hourly air temperature, relative humidity, and precipitation probability by linear interpolation using the DWF data at 3 hour intervals. Hourly precipitation is estimated using hourly precipitation probability and precipitation at 12 hour intervals. Hourly wetness period is estimated from a classification and regression tree model using air temperature, relative humidity, precipitation, and wind speed as well as a simple relative humidity model. Based on the hourly weather data, the system generates infection risk map of Gyeonggi province for the period from 4 to 27 hours prior to actual occurrence of rice blast at 5 km grid resolution. Users can find the map image of infection risk levels through internet.

The comparisons between the forecast and the observed weather data collected from automated weather stations installed at 19 locations indicated that the forecast data became less accurate as the time increased from the release of weather forecast. Hourly estimation of air temperature and relative humidity resulted in acceptable estimates compared with the AWS data. However, it was not true for precipitation and wetness period. Consequently, the rice blast forecast based on the DWF data did not appear as reliable as when the observed hourly weather data were used. Although the estimation of hourly DWF data, especially precipitation and wetness period, needs to be improved, we expect that the DWF data can be used to extend the time window by 1-2 days for effective and efficient disease control practices by crop growers.