Application of a simple algorithm to estimate daily evapotranspiration from NOAA–AVHRR images f.....IntroductionMethodsInstantaneous evapotranspirationDaily evapotranspirationDataNOAA–AVHRR imageryHigh resolution dataResultsSensitivity analysisTest of the algorithm by comparing NOAA estimated ET and DAIS estimatesTemporal evolutionConclusionsAcknowledgementsReferencesApplication of a simple algorithm to estimate daily evapotranspirationfrom NOAA–AVHRR images for the Iberian PeninsulaJ.A. Sobrinoa,⁎, M. Gómeza,1, J.C. Jiménez-Muñoza,1, A. OliosobaDepartment of Thermodynamics, Faculty of Physics, University of Valencia, 50 Dr. Moliner, 46100 Burjassot, SpainbINRA Bioclimatologie, Domaine Saint-Paul, Avignon, FranceReceived 6 April 2006; received in revised form 14 February 2007; accepted 17 February 2007AbstractEvapotranspiration (ET) is a key process in land surface–atmosphere studies. It mainly depends on water availability and incoming solarradiation and then reflects the interactions between surface water processes and climate. In this paper, a methodology for retrieving ET from lowspatial resolution remote sensing data is presented. It is based on the evaporative fraction concept, and it has been applied to Advanced Very HighResolution Radiometer (AHVRR) data acquired over the Iberian Peninsula. The methodology does not require other data than the data providedby the satellite and may be applied to areas with almost spatially constant atmospheric conditions and which include wet and dry sub-areas. Thecomparison with high resolution ET estimation shows a root mean square error (RMSE) of 1.4 mm d− 1which is in agreement with the sensitivityanalysis of the method. Finally, the methodology has been applied to temporal NOAA–AVHRR images acquired from 1997 to 2002 in order toanalyze the seasonal evolution of the daily ET. The temporal study of the ET values realized in this paper shows that the highest ET values areassociated with the higher development crops, while the lowest values are related with lower development or null crop. As a conclusion, it isshown that the ET values obtained with the proposed model evolve according to the variations presented in parameters such as surface temperatureor vegetation index.© 2007 Elsevier Inc. All rights reserved.Keywords: Evapotranspiration; S-SEBI; Evaporative fraction; Net radiation flux; Soil heat flux; Latent heat flux; NOAA–AVHRR; DAIS1. IntroductionContinuous evapotranspir ation (ET) measurements at differ-ent scales are important in hydrology and agriculture, and alsoin other environmental studies. ET is a key process in landsurface–atmosphere studies. It mainly depends on wateravailability and incoming solar radiation and then reflects theinteractions between surface water processes and climate. Overthe last years, the scientific community has been interested inestimating evapotranspiration by remote sensing, since it is theunique way to retrieve ET at several temporal and spatial scales.For this reason, different methods have been developed toderive surface fluxes from remote sensing observations, such as:SEBAL (Surface Energy Balance Algorithm for Land,Bastiaanssen, 2000; Bastiaanssen et al., 1998a,b; Jacob et al.,2002), S-SEBI (Simplified Surface Energy Balance Index,Roerink et al., 2000), SEBS (Surface Energy Balance System,Jia et al., 2003; Su, 2002) and TSEB (Two-Source EnergyBalance, French et al., 2003; Kustas & Norman, 1999; Normanet al., 1995).In Sobrino et al. (2005) and Gómez et al. (2005) a simplemethod for retrieving daily ET from high resolution data basedon the S-SEBI model was proposed, which is a simple way toderive evapotranspiration from evaporative fraction concept. Itis based on the estimation of the evaporative fraction from thecontrast between dry and wet areas according to Roerink et al.(2000). The purpose of this paper is to adapt this methodologyto the low spatial resolution data provided by the AdvancedVery High Resolution Radiometer (AVHRR) on board of theNational Oceanic and Atmospheric Administration (NOAA)platform. The paper is organized as follows: Section 2 describeshow instantaneous and daily ET are retrieved from NOAA–Remote Sensing of Environment 110 (2007) 139 – 148www.elsevier.com/locate/rse⁎Corresponding author. Tel.: +34 963543115; fax: +34 96 354 3099.E-mail addresses: [email protected] (J.A. Sobrino), [email protected](M. Gómez), [email protected] (J.C. Jiménez-Muñoz),[email protected] (A. Olioso).1Fax: +34 96 354 3099.0034-4257/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.rse.2007.02.017AVHRR data, Section 3 shows the available imagery andSection 4 presents the results obtained, including a sensitivityanalysis of the method, the test of the method by comparing lowresolution estimations of ET to high resolution ones, and alsomonthly and seasonal analysis of ET over the Iberian penins ulafor several years. The main conclusions drawn from this studyare presented in Section 5.2. Methods2.1. Instantaneous evapotranspirationThe method proposed in this paper for instantaneous ETestimation from NOAA– AVHRR data is based on the Sim-plified Surface Energy Balance Index (S-SEBI) model and theevaporative fraction (Λ) concept proposed by Roerink et al.(2000), in which the latent heat flux (LET) is given by:LETi¼ KiðRni GiÞð1Þwhere Rnis the net radiation flux and G is the soil heat flux. Thesubscript ‘i’ refers to instantaneous values, and units for heatand net radiation fluxes are given in W m− 2. The soil heatflux (G) can be obtained from the Modified Soil AdjustedVegetation Index (MSAVI) (Sobrino et al., 2005), whereasthe evapora tive fraction (Λ) is obtained from the scatterplotbetween surface temperature and albedo (Roerink et al.,2000). The net radiation flux (Rn) is estimated from the radia-tion balance for all incoming and outgoing radiation (seeTable 1 and Fig. 1). Table 1 shows that the net radiation fluxTable 1Equations used to estimate the daily ET from high and low resolution imagesAlbedo α = 0.5 ρNIR+0.5 ρREDAVHRR: ρREDand ρNIRare calculated usingchannel 1 (0.63 μm) and channel 2 (0.91 μm).DAIS: ρREDand ρNIRare calculated usingchannel 10 (0.66 μm) and channel 22 (0.87 μm).Emissivity AVHRR: Land surface emissivity estimated fromNOAA–AVHRR data using the NDVI Thresholdsmethod (Sobrino & Raissouni, 2000).DAIS: Land surface emissivity estimated from DAISdata using the NEM (Normalized