Ingestion of AgWeatherNet Data into WRF Modeling System for Accurate Freeze/Frost Prediction in Washington State for Agricultural Use

In the eastern Washington tree fruit agricultural area, availability of weather forecasts to predict for damaging spring and fall frost conditions is important in decision making. The AgWeatherNet program of Washington State University, which is equipped with 150 automated weather stations mostly over the state of Washington, provides current and historical weather data from WSU's weather stations along with a range of models and decision aids for the tree fruit growers. It is also in the development and testing stages of providing online (web-based real-time forecast at http://weather.wsu.edu) daily short-range predictions of air temperature, dew point temperature, wind speed and direction. The Advanced Research core of the Weather Research and Forecasting (WRF-ARWv3.3.1 and v3.4.1) model is the tool implemented to achieve this goal that has been undergoing through a series of performance verification tests. WRF was configured with three nested domains of 27, 9 and 3 km horizontal resolutions with the third nested domain covering only the state of Washington. Vertically, 40 sigma levels were used that extended to 50mb into the free atmosphere as the model top, with the first meteorological variables containing half-sigma level located at about 10m. In this model investigation, three freeze/frost weather events were selected, representing spring and fall frost as well as winter arctic outbreak conditions critical for tree fruit industry. Results showed that the model predicted well over the Columbia basin flat region, while the error increased over the stations located at the complex and high terrain structures. WRF model generally underestimated daytime and overestimated night time temperatures. Averaged statistical results over these freeze/frost cases showed that WRF reproduced observed temperatures that vary from a cold bias of -3.3oC during the day to a warm bias of 3.1oC at night. The overall WRF model temperature average RMSE is thus found to be +/-3.0oC. The WRF forecasting model can be a great early-warning informative tool in combating crop loss due to all weather related adversaries, if further reduction of errors can be achieved particularly over the complex terrain structures by introducing methods of improving the initial and boundary conditions (ICs and BCs) of the model. The use of reliable observational data to improve the ICs and BCs are thus essential method in alleviating the bias amplitude for real-time freeze/frost forecasts. Hence, the AgWeatherNet automated weather station data is used in the evaluation of WRF results when observational data are ingested to investigate the degree of statistical reduction in error for agriculturally significant weather events.