Automated Mesoscale Pest Risk Forecast Maps for Agricultural Production and Potential Plant Biosecurity threats
Horticultural Crops Research
2013 Annual Report
1a.Objectives (from AD-416):
Incorporate real-time weather data into disease forecasting models for diseases of grapevines, forage grass, and hops.
1b.Approach (from AD-416):
In conjunction with cooperation growers, virtual weather data and weather maps received will be evaluated for running disease forecasting models in the Willamette Valley. Actual weather data will be collected from automated weather stations installed at three field sites for each crop. Weather data to be used include temperature, daily rainfall, and leaf wetness. Virtual and actual weather data will be collected and compared.
This research was conducted in support of NP303 objective 2C of the parent project. The use of pest and disease forecasting is limited by access to quality weather and forecast data that is actionable at the level of individual fields. This project developed and tested new methods and technologies that used knowledge of terrain and climate to generate virtual site-specific hourly weather data and improve weather forecasts for use in disease forecasting models. Weather data collected from public and private sources were interpolated to 800 meter grid cells and these cells were used to generate virtual weather stations at chosen locations. The virtual stations provided weather and forecast data for use in disease forecast models to determine the disease risk for a specific field. New ways to visualize disease risk using real-time risk maps and animations were also developed. The utility of these approaches was assessed for managing diseases in Grape, Grass seed, and Hop production in western Oregon. Over the three year test period, daily temperature measurements were not impacted by time resolution (e.g. 15 minutes versus 1 hour resolution) but were impacted by sensor placement. In-canopy placement had a higher error in grass seed fields than in hop yards and vineyards. Virtual station estimates of daily temperatures were very well correlated with actual temperature measured at each test site. Virtual station estimates of rain and relative humidity had weaker correlations to observed values. These errors caused differences in the disease risk indexes calculated from observed and virtual weather data, especially if the disease model rules used rain. However, there was no difference in management recommendations for powdery mildew of hop or grape with observed verses virtual weather data. The overestimation of rain resulted in more fungicide applications using the virtual data for hop Downy mildew. The underestimate of leaf wetness in grass seed caused by the overestimate of the low temperature resulted in an underestimation of stem rust hazard. Even though there was error in the estimates of the weather parameters, which were translated into error in the disease risk, these errors did not translate into reduced disease management in commercial fields or experiment plots. The errors in the actual disease risk did not result in altered management decisions in Grass seed production. These results indicate the methods and technologies developed for creating virtual weather data generate data that are suitable for running disease forecast models and making management decisions based on the model results without adversely impacting disease management.