Accuracy of reference evapotranspiration forecast

The ability to forecast reference evapotranspiration (ET_o) accurately greatly enhances our ability to manage the irrigation of high-frequency irrigation systems. If the synoptic and mesoscale models are accurate, then ET_o predictions from the forecasts are likely to be more accurate than statistical models. In addition, accurate ET_o forecasts can eliminate or reduce the size of automated weather networks that are currently used to provide near-real time ET_o data for irrigation scheduling. This will reduce costs and provide ET_o data in a more timely fashion. In this paper, the inclusion of ET_o in an existing mesoscale weather forecast model is presented. The mesoscale model inputs data from a NWS synoptic database and predicts rainfall, solar irradiance, surface temperature, surface humidity, and surface wind speed hourly for up to 72 hours for any location in California using a 20 km grid resolution. Hourly ET_o forecasts for six locations representing a wide range of climatic conditions in California were computed using hourly forecast weather data from the mesoscale model. For comparison, measured hourly weather data from the same locations were used to calculate ET_o using the UN_FAO Penman-Monteith equation. The slope and coefficient of determination were calculated using a regression of observed versus predicted ET_o through the origin, and the root mean square error was used to compare the hourly forecast with the observed ET_o. To evaluate the importance of individual weather variables on the ET_o forecast, ET_o was recalculated after substituting observed weather variables, one at a time, for each predicted variable, while retaining the other three predicted variables. The ET_o calculated with one observed and three forecast variables were plotted versus the ET_o calculated using all forecast data to investigate the importance of individual variables. In general, the forecast underestimated observed ET_o by 2 to 10%. The regression slope and the root mean square error were nearly the same regardless of the time of the year. Because the effect on ETo estimates with the Penman-Monteith equation is small, the error in ET_o predictions cannot be attributed to differences in observed and predicted temperature, wind speed, or water vapor pressure. Inaccuracy in predicting solar irradiance seems to be the biggest factor affecting the ET_o forecast accuracy.