A comparison of models for estimating potential evapotranspiration for Florida land cover types

We analyzed observed daily evapotranspiration (DET) at 18 sites having measured DET and ancillary climate data and then used these data to compare the performance of three common methods for estimating potential evapotranspiration (PET): the Turc method (Tc), the Priestley-Taylor method (PT) and the Penman-Monteith method (PM). The sites were distributed throughout the State of Florida and represent a variety of land cover types: open water (3), marshland (4), grassland/pasture (4), citrus (2) and forest (5). Not surprisingly, the highest DET values occurred at the open water sites, ranging from an average of 3.3 mm d(-1) in the winter to 5.3 mm d(-1) in the spring. DET at the marsh sites was also high, ranging from 2.7 mm, d(-1) in winter to 4.4 mm d(-1) in summer. The lowest DET occur-red in the winter and fall seasons at the grass sites (1.3 mm d(-1) and 2.0 mm d(-1), respectively) and at the forested sites (1.8 mm d(-1) and 2.3 mm d(-1), respectively). The performance of the three methods when applied to conditions close to PET (Bowen ratio <= 1) was used to judge relative merit. Under such PET conditions, annually aggregated Tc and PT methods perform comparably and outperform the PM method, possibly due to the sensitivity of the PM method to the limited transferability of previously determined model parameters. At a daily scale, the PT performance appears to be superior to the other two methods for estimating PET for a variety of land covers in Florida. (C) 2009 Elsevier B.V. All rights reserved