Assessment of Spatial and Temporal Variation of Potential Evapotranspiration Estimated by Four Methods for South Carolina

Abstract

Given South Carolina’s ongoing water planning efforts, in this study, we evaluated seasonal and annual potential evapotranspiration (PET) using measured Class A pan evaporation (PE) and 3 widely used estimation methods for the state with 3 distinct physiographic regions (Coastal, Piedmont, and Mountain). The methods were temperature-based Hargreaves-Samani (H-S), radiation-based Priestley-Taylor (P-T), and process-based Penman-Monteith (P-M). The objectives of the study were to (a) describe seasonal and temporal distribution of PET by all methods, (b) quantify differences among PET methods, and (c) identify relationships between monthly PE and estimated PET by each method. Daily weather variables from 59 National Oceanic and Atmospheric Administration weather stations distributed in the 3 regions of South Carolina (SC) were used to estimate daily PET for an 18-year period (1998–2015). Net radiation was estimated using modeled solar radiation values for weather stations. The average annual H-S PET values adjusted with the empirical radiation factor (KT) and the average annual P-T PET values for 1998–2015 were 1,232 ± 9, 1,202 ± 11, and 1,115 ± 10 mm and 1,179 ± 10, 1,137 ± 11, and 1,082 ± 11 mm, respectively, for the Coastal, Piedmont, and Mountain regions. Both the mean annual H-S and P-T PET for the Mountain region were significantly (α = 0.05) lower than for the Coastal and Piedmont regions. The mean annual P-T PET for the Coastal region was significantly (α = 0.05) greater than that for the Piedmont. Regional differences showed that estimated PET for 1998-2015 was greatest in the Coastal and lowest in the Mountain region. Comparison of all 3 methods using only common 8-year data showed mean annual P-M PET, varying from 1,142 mm in the Piedmont to 1,270 mm in the Coastal region, was significantly higher than both the H-S and P-T PET in both regions. The greatest mean monthly H-S and P-T PET values were observed in June and July. Statistical evaluation using Nash–Sutcliffe efficiency and percent bias showed a slightly better agreement of H-S PET with both the measured PE as well as the P-M method, followed by the P-T. However, the P-T method yielded a close to unity slope and slightly higher R2 than the H-S PET when compared with the PE. The P-T PET method that uses both the temperature and radiation data may be preferred for SC with a humid climate dominated by forest land use, given more rigorous ground-truthing of modeled solar radiation as data become available. Surface interpolation algorithm, inverse distance weighted, was used to spatially map both the distributed H-S and P-T PET for the state. Results from this study can be used to support several components of the ongoing water planning efforts in SC.