Assessment of evapotranspiration estimation models for use in semi-arid environments

Limited information exists on reliable estimates of evapotranspiration (ETo), to be used as forecasts, to achieve high irrigation water use efficiency in semi-arid environment, particularly under microirrigation (MI) system. The objectives of this study were to: (i) assess the estimates of ETo obtained using six models against experimentally determined values in a semi-arid environment, Karaj in Iran; (ii) assess the usefulness of short-term weather data in the computation of ETo estimates for forecasting purposes; and (iii) compare ETo computed for a semi-arid environment with that of a humid temperate environment, Tottori in Japan. In-field lysimeter experiments were conducted in 1993 and 1994 in Karaj to compute daily ETo from water-balance data and a similar experiment was conducted in 1972 and 1973 in Tottori. The ETo estimates were obtained using the Penman (PE), Penman-Monteith (PM), Wright-Penman (WP), Blaney-Criddle (BC), Radiation balance (RB), and Hargreaves (HG) models. The ETo forecasts for Karaj and Tottori were obtained using 5- or 8-year weather data and land 2-year return period approach. Compared with the lysimeter values, over- and underestimations of ETo, by all six models, were a norm in Karaj. Nevertheless, the PM model produced the best ETo estimates as assessed by root mean square, mean bias error, and t-test statistics for the semi-arid Karaj, whereas the PE model performed best for the temperate Tottori. In general, the ETo forecasts obtained using 8-year weather data were better than 5-year weather data. The 2-year return period ETo forecasts were better than 1-year return period, regardless of the duration of weather data duration. The results show that: (i) PM model produced best ETo estimates for semi-arid environment whereas the PE model produced the best ETo estimates for humid temperate environment; and (ii) 8-year weather data and 2-year return period approach produced most reliable ETo estimates for forecasting purposes. (C) 2003 Elsevier B.V. All rights reserved.