A comparative analysis of reference evapotranspiration from the surface of rainfed grass in Yaounde, calculated by six empirical methods against the penman-monteith formula

Abstract

Five reference evapotranspiration (ETo) methods and one adjusted method were compared with the Penman-Monteith formula, standardized by the Food and Agriculture Organization (FAO-56PM), using rainfed grass data from an agricultural weather station at the Radmilovac experimental site. The methods compared include four widely and commonly used empirical equations (Piestley-Taylor, Turc, Makkink, and Hargreaves-Samani), as well as a new equation recently proposed by Alexandris et al. (2006), namely the “Copais” equation. An adjusted Hargreaves equation for Southeast Europe proposed by Trajkovic (2007) was also included in the comparison. Daily ETo values, obtained during two growing seasons (2005 and 2006) were compared, using linear regression and statistical indices of quantitative approaches to model performance evaluation. All the statistical indices were calculated on a daily basis. The soil water content of the rainfed grass field was adequate during the investigated period, as shown by soil moisture profiles (Petkovic et al., 2006). It was, therefore, assumed that rainfed grass evapotranspiration closely resembled hypothetical reference evapotranspiration. Because solar radiation (Rs) was the only radiation measured, the net radiation (Rnet) variable was derived empirically, following the procedure outlined in the FAO-56 paper (Allen et al., 1998). The mean bias errors (MBEs), for estimates of grass crop evapotranspiration applying the Copais, Priestley-Taylor, Hargreaves-Samani, Turc, Makkink and Adjusted Hargreaves methods, and compared to FAO-56 PM estimations, were 0.019, -0.037, 0.741, -0.620, -0.186 and 0.158 mmday-1, respectively, and the average FAO56- PM ETo was 2.857 mmday-1. In general, Priestley-Taylor and Copais models performed well for the study region and yielded results closest to the FAO56-PM method. Systematic overestimation of Hargreaves-Samani ETo values was noted, while the other two methods, Turc and Makkink, produced underestimated results. The results of statistical comparisons provided a confident statistical justification for the ranking of the compared methods, based on performance indices.