Evaluation of evaporation-pan networks

A methodology to assess the relative merits of using evaporation-pan networks and semiempirical functions to estimate reference-crop evapotranspiration (ET) is presented. Estimation-error variance is proposed as a basis for comparison. In the case of pan networks, ordinary kriging and universal kriging are viable options for determining estimation-error variance. In the case of semiempirical ET functions, a formulation is developed that expresses the estimation-error variance in terms of the uncertainties in extrapolating meteorological data from measurement locations. The proposed methodology is demonstrated within a 44,000 [km.sup.2] area of southern Florida. The results show that universal kriging is the preferred method for estimating reference-crop ET from pan measurements, and the Penman-Monteith method is the preferred estimation method from meteorological data. The pan-based universal kriging approach yields a lower estimation-error variance than the Penman-Monteith function, indicating that pan-based estimates of reference-crop evapo-transpiration are preferable to semiempirical evapotranspiration functions in south Florida.