Evaluation of Five Methods for Estimating Class A Pan Evaporation in a Humid Climate

Abstract

Evaporation pans continue to be used extensively throughout the world to measure free-surface water evaporation (E pan ) and to estimate evapotranspiration for irrigation scheduling and water management for agronomic and horticultural crops. E pan is also being used extensively to estimate evaporation rates from lakes, wetlands, rivers, reservoirs, and other water bodies for management of wildlife and ecological habitat. A reliable method is needed to estimate missing daily E pan data. Determination of a reliable method for the estimation of E pan would also be useful in modeling of crop growth, and hydrological and ecological systems. Five methods [Penman (Penman, 1948), Kohler-Nordenson-Fox (KNF) (Kohler et al., 1955), Christiansen (Christiansen, 1968), Priestley-Taylor (PT) (Priestley and Taylor, 1972), and Linacre (Linacre, 1977)] for estimating E pan were compared with the historical (23-year) measured daily values to determine the suc- cess of accurate and consistent E pan estimations under humid climatic conditions in Florida. The root mean square error (RMSE) was used as the criteria to judge the accuracy and reliability of a given method. An RMSE value of <0.5 mm·d -1 (0.02 inches/d) between the measured and estimated E pan was considered as an acceptable error for daily estimations. The standard deviation ( sd ) values, and percent error (%E) between the estimated and measured values were also considered in the performance evaluations. Performance evaluations of the E pan estimates of the methods were made on a daily, monthly, and annual basis. Results indicated that the KNF method provided the best E pan estimations. The Linacre method yielded the poorest estimates. The second, third, and fourth best methods were the Penman, PT, and Christiansen, respectively. The RMSE and sd of E pan estimates were lowest when using KNF method. The mean value of the %E of daily, monthly, and annual estimations were 27%, 27%, and 26% for Christiansen; 6%, 6%, and 4% for KNF; 33%, 32%, and 26% for Linacre; 24%, 24%, and 21% for PT; and 19%, 17%, and 11% for Penman methods, respectively. The weekly, monthly, and annual total of E pan estimates from KNF method were also compared to the measured values of the two selected years of data (1981 and 1983). The annual rainfall totals were significantly lower than the 23-year mean in 1981, and higher in 1983. The %Es of weekly, monthly, and annual total E pan estimates were 9%, 9%, and -1% in 1981; and 11%, 5%, and 4% in 1983, respectively. The KNF method underestimated E pan in 1981 (dry year) and the underestimations were higher in summer months. The underestimations in a dry year, especially in summer months, might be due to the fact that the sensible heat advection is not effectively accounted for in the KNF equation causing underestimations of E pan . Overall results indicated that the KNF method should be the first choice, among the methods tested, for estimating daily E pan for irrigation scheduling and for estimating the missing E pan data in humid areas.