Estimation of evaporative fraction and evapotranspiration from MODIS products using a complemen.....IntroductionReview of complementary modelsProposed methodProposed method to estimate TuImplementation of the proposed ET modelStudy area and MODIS productsStudy regionMODIS productsResultsDiscussionAcknowledgementsReferencesEstimation of evaporative fraction and evapotranspiration from MODISproducts using a complementary based modelVirginia Venturinia, Shafiqul Islamb,⁎, Leticia RodriguezaaFacultad de Ingeniería y Ciencias Hídricas, Universidad Nacional del Litoral, C.C. 217, Santa Fe, 3000, ArgentinabDepartment of Civil and Environmental Engineering, Tufts University, 113 Anderson Hall, Medford, MA 02155, United StatesReceived 16 February 2006; received in revised form 20 April 2007; accepted 21 April 2007AbstractWe present a new formulation to derive evaporative fraction (EF) and evapotranspiration (ET) maps from remotely sensed data withoutauxiliary relationships or site-specific relationships. This formulation is based on Granger's complementary relationship and Priestley–Taylor'sequation. The proposed model eliminates the wind function and resistance parameters commonly applied to ET calculation by including a relativeevaporation parameter (ET/Epot). By combining this relative evaporation parameter, Granger's complementary relationship and Priestley–Taylorequation, we obtain a simple equation to estimate ET. We tested and validated the proposed formulation over the Southern Great Plains (SGP)region of the United States for seven clear sky days during March–October 2003. MODIS Atmospheric and Land products were the only source ofdata used in this study. Estimates of ET show an overall root mean square error and bias of 33.89 and − 10.96 Wm− 2, respectively. Our resultssuggest that the proposed approach is robust and valid for a wide range of atmospheric and surface conditions.© 2007 Elsevier Inc. All rights reserved.Keywords: Evapotranspiration; Evaporative fraction; Complementary relationship; Priestley and Taylor1. IntroductionMany hydrologic modeling, global circulation models andagricultural management applications require estimates ofevapotranspiration (ET) and the evaporative fraction (EF).Reliability of ET estimates will depend on the scale ofapplication. Jiang et al. (2004) evaluated analytically the errorproperty of the traditionally used energy balance residualmethod for latent heat flux estimation in an attempt to identifythe possible existence of an irreducible error bound for latentheat flux measurement and estimation over large areas. Theiranalysis shows that the error is typically on the order of 10–20%or larger for ET. Over the years, several models have beendeveloped to estimate ET for a wide range of surface conditionsand spatial or temporal scales. Most of these models arevariations of Penman's equation (Monteith & Unsworth, 1990)and Priestley–Taylor's equation (Priestley & Taylor, 1972),hereafter referr ed to as P–T. These ET models have been widelyapplied with varying results (e.g., Bastiaanssen e t al., 1996;Granger & Gray, 1989; Holwill & Stewart, 1992; Jackson et al.,1977; Jiang & Islam, 2001; Nishida et al., 2003; Norman et al.,2003; Rivas & Caselles, 2004; Seguin et al., 1989). Few of theseET estimation methodologies have taken advantage of thecomplementary relationship. Bouchet (1963) postulated thatregional ET can be estimated as a complementary function ofthe potential evapotranspiration (Epot) and the wet environ-ment evapotranspiration (Ew) for a wide range of availableenergy. Examples of successful models based on Bouchet'sheuristic relationship include those developed by Brutsaert andStricker (1979), Morton (1983) and Hobbins et al. (2001).These models have been extensively applied for a range ofsurface and atmospheric conditions (Bruetsaert & Parlange,1998; Brutsaert & Stricker, 1979; Granger & Gray, 1990;Hobbins et al., 2001; Morton, 1983; Ramírez et al., 2005; Xu& Singh, 2004).Granger (1989) developed a physi cally based compleme n-tary relationship after a meticulous analysis of potentialevaporation concepts. The author used the term “potentialevaporation” for Epot and Ew concepts and clearly presented thecomplementary behavior of common potential evaporationtheories. Using two potential evaporations, i.e. Epot and Ew,Remote Sensing of Environment xx (2007) xxx – xxx+ MODELRSE-06888; No of Pages 10www.elsevier.com/locate/rse⁎Corresponding author.E-mail address: [email protected] (S. Islam).0034-4257/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.rse.2007.04.014ARTICLE IN PRESSPlease cite this article as: Venturini, V. et al. Estimation of evaporative fraction and evapotranspiration from MODIS products using a complementary basedmodel. Remote Sensing of Environment (2007), doi:10.1016/j.rse.2007.04.014seems to g enera te a universal relationship, and thereforeuniversal ET models. Conversely, attempting to estimate ETfrom only one potential formulation would need site-specificcalibration or auxiliary relationships ( Granger, 1989). In addi-tion, the relative evaporation coefficient introduced by Grangerand Gray (1989) enhances the compl ementary relationship witha dimensionless coefficient that yields a simpler complementarymodel.The common thread among the available complementaryET models is the use of Penman's equation to estimateEpot. Specifically, Morton's CRAE model (Morton, 1983)uses Penman's equation to calculate Epot, and a modified P–Tequation to approximate the wet environment evapotranspira-tion (Ew). Brutsaert and Stricker (1979) developed theirAdvective-Arid model using Penman for Epot and the P–Tequilibrium evaporation to model Ew.When these models were originally developed, networks ofmeteorological stations co nstituted the main source of atmo-spheric data, while the surface tempe rature o r the soiltemperature were available only at some locations around theWorld. The advent of satellite technology provided observationsof the surface temperature (Ts), but the source of atmosphericdata was still ancillary. Thus, many of the current remotesensing approaches were developed to estimate ET with littleamount of atmospheric data (Gillies et al., 1997; Jiang & Islam,1999; Nishida et al., 2003; Price, 1990).The recent introduction of the Atmospheric Profiles Product,derived from MODIS sensors onboard EOS-Terra and EOS-Aqua satellites, provides new opportunities to