RETRIEVING REFERENCE EVAPOTRANSPIRATION FOR IRRIGATION SCHEDULING: FORECAST OR PAST WEATHER DATA?

Nowadays, the increasing availability of numerical weather predictions opens new prospects to retrieve reference evapotranspiration (ET0) forecasts. The reliability of weather simulations provided by COSMO model in predicting ET0 was evaluated during the year 2017 in 7 study sites distributed in 4 countries (Italy, Norway, Romania and Spain) covering a wide range of climate conditions. The main objective of the study was to evaluate the optimal scenario for calculating ET0 by separately assessing the accuracy in the use of past meteorological data and forecasts for estimating irrigation requirements (IR). The ET0 estimates were obtained by incorporating the meteorological variables foreseen in the FAO-56 Penman-Monteith equation. Each weather component (air temperature, Tair; relative humidity, RH; wind speed, u2; solar radiation, Rs; rainfall, P) and ET0 was compared with ground-based observations. Simulated irrigation scheduling was computed every three days using: (i) meteorological data measured during the previous three days and (ii) weather forecast for the next three days; and compared a posteriori with IR obtained from the measured meteorological data. Validations were performed by R2 and RMSE. Results showed a good agreement between measured and estimated meteorological variables. The best performance was obtained for Tair and Rs (R2, RMSE of 0.97, 1.54°C; and 0.89, 37.33 W m-2, respectively), whereas the worst model performance was obtained for P and u2 (R2, RMSE of 0.12, 126.61 mm; and 0.34, 1.20 m s-1, respectively). The comparison between daily ET0 from the measured and predicted meteorological data showed high performance, with R2 and RMSE of 0.90 and 0.68 mm, respectively. IR values have been estimated more accurately using forecast meteorological data (1.7% overestimation), rather than using meteorological data from the past (2.6% underestimation). In conclusion, the use of forecast meteorological data is recommended for the estimation of IR, since the uncertainties due to the weather forecast translate into minor errors compared to those committed when using data from past meteorological conditions.