Realization of daily evapotranspiration in arid ecosystems based on remote sensing techniques

Daily evapotranspiration is a major component of water resources management plans. In arid ecosystems, the quest for an efficient water budget is always hard to achieve due to insufficient irrigational water and high evapotranspiration rates. Therefore, monitoring of daily evapotranspiration is a key practice for sustainable water resources management, especially in arid environments. Remote sensing techniques offered a great help to estimate the daily evapotranspiration on a regional scale. Existing open-source algorithms proved to estimate daily evapotranspiration comprehensively in arid environments. The only deficiency of these algorithms is the course scale of the used remote sensing data. Consequently, the adequate downscaling algorithm is a compulsory step to rationalize an effective water resources management plan. Daily evapotranspiration was estimated fairly well using an Advance Along-Track Scanner Radiometer (AATSR) in conjunction with (MEdium Resolution Imaging Spectrometer) MERIS data acquired in July 2013 with 1 km spatial resolution and 3 days of temporal resolution under a surface energy balance system (SEBS) model. Results were validated against reference evapotranspiration ground truth values using standardized Penman–Monteith method with R2 of 0.879. The findings of the current research successfully monitor turbulent heat fluxes values estimated from AATSR and MERIS data with a temporal resolution of 3 days only in conjunction with reliable meteorological data. Research verdicts are necessary inputs for a well-informed decisionmaking processes regarding sustainable water resource management.

Abstract.Daily evapotranspiration is a major component of water resources management plans.In arid ecosystems, the quest for an efficient water budget is always hard to achieve due to insufficient irrigational water and high evapotranspiration rates.Therefore, monitoring of daily evapotranspiration is a key practice for sustainable water resources management, especially in arid environments.Remote sensing techniques offered a great help to estimate the daily evapotranspiration on a regional scale.Existing open-source algorithms proved to estimate daily evapotranspiration comprehensively in arid environments.The only deficiency of these algorithms is the course scale of the used remote sensing data.Consequently, the adequate downscaling algorithm is a compulsory step to rationalize an effective water resources management plan.Daily evapotranspiration was estimated fairly well using an Advance Along-Track Scanner Radiometer (AATSR) in conjunction with (MEdium Resolution Imaging Spectrometer) MERIS data acquired in July 2013 with 1 km spatial resolution and 3 days of temporal resolution under a surface energy balance system (SEBS) model.Results were validated against reference evapotranspiration ground truth values using standardized Penman-Monteith method with R 2 of 0.879.The findings of the current research successfully monitor turbulent heat fluxes values estimated from AATSR and MERIS data with a temporal resolution of 3 days only in conjunction with reliable meteorological data.Research verdicts are necessary inputs for a well-informed decisionmaking processes regarding sustainable water resource management.

Introduction
Evapotranspiration is the principle process in defining mass and energy relationship in the surrounding hydrosphere (Allen et al., 2007a, b;Cruz-Blanco et al., 2014).The consumptive use of irrigational water in agriculture is the fundamental component of a balanced estimation of evapotranspiration (Bastiaanssen et al., 2011;Cammalleri and Ciraolo, 2013).
The concept of water use efficacy basically depends on the reliable estimation of evapotranspiration and surface water evaporation (Berengena and Gavilán, 2005;Elhag et al., 2011).Weather and wind conditions induce a regional and seasonal variation of evapotranspiration estimation (Hanson, 1991;Cristóbal and Anderson, 2013).
Conventional techniques of field scale evapotranspiration estimations are achieved, especially over homogenous surfaces using ordinary techniques: lysimeter systems, eddy covariance (EC) and Bowen ratio (BR).Nevertheless, conventional methods of evapotranspiration estimations are incapable of fulfilling the quest for regional evapotranspiration estimation, specifically in harsh climatic conditions (Gavilán et al., 2006;Ghilain et al., 2011).Therefore, remote sensing evapotranspiration models are adequate techniques for obtaining satisfactory estimates (Allen et al., 2007a, b;De Bruin et al., 2010).
Remote sensing evapotranspiration models are numerous.Several algorithms are already in practice with different complexity levels to estimate evapotranspiration based on different climatic conditions and land use variability (Elhag et al., 2011;Espadafor et al., 2011;Cristóbal and Anderson, 2013).
The first concept is to use the surface reflectance in different visible (VIS), near-infrared (NIR) and even extended to thermal infrared (TIR) portions of the electromagnetic spectrum to rationalize the surface energy balance (SEB).The other concept it to use vegetation indices derived from canopy reflectance to conceptualize remotely sensed crop coefficient (K cr ).
Ground truth data collection exercised at less than 1 m canopy height, at which all related surfaces fluxes and atmospheric surface variables of the vegetation cover takes place in an arid environment (Beljaars and Holtslag, 1991;Zwart and Bastiaanssen, 2004).Based on Brutsaert (1991Brutsaert ( , 1999)), Monin-Obukhov similarity (MOS) and bulk atmospheric boundary layer (ABL) functions were calculated.Brutsaert (1999) suggested sets of criteria to estimate MOS or ABL if scaled down appropriately for a given circumstances.Brutsaert criteria are valid only for unstable conditions.
Therefore, van den Hurk and Holtslag (1995) adjusted and validated Brutsaert criteria using atmospheric surface layer scaling according to Brutsaert (1982) to be used in stable conditions.Generic estimation of surface albedo for vegetated land covers is based on the red (R) and NIR band reflectance (Brest and Goward, 1987) model.
The aim of the current study is to monitor turbulent heat fluxes in Wadi ad-Dawasir to estimate the daily evapotranspiration rate and relative evaporation ratio using Advance Along-Track Scanner Radiometer (AATSR) and MEdium Resolution Imaging Spectrometer (MERIS) sensors.The final step is to identify the regression coefficient between the estimated evapotranspiration's rates and the actual ground truth data.

Study area
The study area, the town of Wadi ad-Dawasir, is located on the plateau of Najd at Lat 44 • 43 and Lon 20 • 29 ; about 300 km south of the capital city Riyadh (Fig. 1).This study area comprised of gravelly tableland disconnected by insignificant sandy oases and isolated mountain bundles.Across the Arabian Peninsula the tableland slopes toward the east from an elevation of 1360 m in the west to 750 m at its easternmost limit.Wadi ad-Dawasir and Wadi al-Rummah are the most important remaining riverbeds in the study area.Wadi ad-Dawasir and Najran regions are the major irrigation water abstraction from Al-Wajid aquifer.Agriculture in the Wadi ad-Dawasir area consists of technically highly de-veloped farm enterprises that operate modern pivot irrigation system.The size of the centre pivot ranges from 30 to 60 ha with farms managing hundreds of them with the corresponding number of wells.The main crop grown in winter is wheat and occasionally potatoes, tomatoes, or melons.All year fodder consists of alfalfa, which is cut up to 10 times a year for food.Typical summer crops for fodder are sorghum and Rhodes grass, which is perennial, but dormant in winter.The shallow alluvial aquifers could not sustain the high groundwater abstraction rates for a long time and groundwater level declined dramatically in most areas.The meteorological features of the area are speckled.Five elements of meteorology are constantly recorded through fixed weather stations located within the study area.Temperature varies from a minimum temperature of 6 • C to maximum temperature of 43 • C. Relative humidity is mostly stable at 24 %.Solar radiation of average sunrise duration is generally 11 h day −1 .Average wind speed is closer to 13 km h −1 and may reach up to 46 km h −1 in thunderstorm incidents.Finally, mean annual rainfall is about 37.6 mm (Al-Zahrani and Baig, 2011).

Methodological framework
The current research work is based on assessing a regression correlation between estimated evapotranspiration data conducted from AATSR and MERIS sensors and its corresponding ground truth evapotranspiration data conducted through standardized Penman-Monteith.Therefore, accurate synchronization of remote sensing data bypassing and ground truth data collection was exercised.

SEBS model fundamentals
Remote sensing data were acquired from AATSR and MERIS sensors on the 8th of July 2013.The satellite data were georeferenced to WGS-84 datum, atmospherically corrected using the Simplified Model for Atmospheric Correction (SMAC) (Rahman and Dedieu, 1994).Several meteorological data were collected from a stationary station located within the designated study area (2004-2014, average meteorological data).
The surface energy balance system (SEBS) was initiated by Su (2002) based on further surface energy balance index improvements.SEBS dynamicity works for regional and local evapotranspiration (ET) estimation.Regional ET estimation uses Monin-Obukhov similarity (MOS), bulk atmospheric similarity and thermal roughness principles.On the other hand, local ET estimation uses only atmospheric surface layer (ASL) scaling fundamentals (Brutsaert, 1999;Su, 2001;Su et al., 2001).The boundary conditions (wet and dry) are essential components in ET estimation using the SEBS model.According to the water availability limitation, H dry is considered to be equal to the available energy AE if evaporation is assumed to be "zero".Following the Penman-Monteith parameterization (Monteith, 1965(Monteith, , 1981)), wet boundary condition (H wet ) is calculated as follows: where e is the actual vapour pressure (kP a ), e s is the saturation vapour pressure (kP a ), c is the psychrometric constant (kP a • C −1 ), γ is the rate of change of saturation vapour pressure with temperature (kP a • C −1 ) and r ah is the bulk surface external or aerodynamic resistance (s m −1 ) .
Consequently, an evaporative fraction ( ) and relative evaporative fraction ( r ) are calculated per image pixel as follows: Daily evaporation is estimated based on the estimation of the evaporative fraction only when the daily net energy (G) and the net radiation (R n ) are available.Therefore, the amplitude variation of the diurnal energy cycle is sky clarity dependent.
where is the evaporative fraction, r is the relative evaporative fraction, , H is the actual sensible heat flux, determined by the bulk atmospheric similarity approach.P ω is the density of water measured in kg m −3 .

Validation
Using a standardized Penman-Monteith method, 50 ground truths data were collected and used to validate the implemented model.The sampling locations were consistently distributed over the designated study area.The lysimeter technique for the estimation of daily evapotranspiration was carried out following Liu and Wang (1999) with calibrated accuracy equal to ±0.025.The calibration procedure was principally based on placing double infiltrometers (Taylor, 1981).
The corrected Penman equations for estimating the daily evapotranspiration was conducted according to Jensen et where ET 0 is reference evapotranspiration (mm day −1 ), is the slope of saturation vapour pressure-temperature curve (kPa • C −1 ), γ is the psychrometric constant (kPa • C −1 ), R n is the net radiation (mm (mbar)), c is the adjustment factor, f (U ) is the wind function, e s is the saturation vapour pressure (mbar), e a is actual vapour pressure.Consequently, the wind function was conducted following to Doorenbos and Pruitt (1977) as where U 2 is the wind-speed-measured surface at 2 m height (km day −1 ).
Meanwhile, e a was calculated according to Allen et al. (1998) as follows: where e o (T min ) is the saturation vapour pressure at a daily minimum temperature (kPa), e o (T max ) is the saturation vapour pressure at a daily maximum temperature (kPa), RH max is the maximum relative humidity (%), RH min is minimum relative humidity (%).
Linear regression model was used to find the correlation coefficient between the estimated and the actual evapotranspiration values.Root mean square error (RMSE) was used to signify the inequality of variance and correlation of the linear regression model (Box, 1954).The RMSE was calculated as follows: where N is the number of observations, P i is the predicted ET values (mm day −1 ), O i is the calculated ET values (mm day −1 ).

Results and discussion
The SEBS model implementation over the designated study area results in 10 different turbulent heat fluxes thematic maps.The histogram and the scatter plot of SEBS output thematic maps were plotted against the daily evapotranspiration values.The estimated daily evapotranspiration values ranged from zero to 6.61 mm day −1 .The spatial distribution of the highest evapotranspiration area is the peripheral of the agricultural pivots as demonstrated in Fig. 2. The circular movement of the pivots drives the excess irrigated water towards the prereferral of the irrigated fields leading to higher salinity levels (Zwart and Bastiaanssen, 2004;Cruz-Blanco et al., 2014).The mean actual evapotranspiration value is almost 5 mm day −1 (Fig. 3), which is considered a high value for arid conditions (Elhag et al., 2011).Such an evapotranspiration value supports the hypothesis of the mismanagement of irrigational water in Wadi ad-Dawasir.Principally under extreme dry climate conditions, relative evaporation may reach unity (Lockwood, 1999).The relative evaporation thematic map, demonstrated in Fig. 4, confirms high correspondence between the actual and potential evapotranspiration, especially in the peripheral of the agricultural pivots.Normal distribution of the relative evaporation is demonstrated in Fig. 5. Mean relative evaporation ratio is counted for 0.91.Only within the pivots, the relative evaporation decreases to 0.45 indicating the wet condition of the agricultural land (De Bruin et al., 2010).Fifty points of ground truthing data of daily evapotranspiration were collected during July 2013.The points were consistently distributed over the designated study area.Daily evapotranspiration estimation was conducted according to Liu and Wang (1999) using the Lysimeter with a calibrated accuracy of ±0.025.The actual evapotranspiration data were intersected with the estimated raster image under a GIS environment.A linear regression model with an R 2 value of 0.83 was conducted to assess the association between the estimated and actual measured values (Fig. 6).
Implementation of the SEBS model over the designated study area showed higher daily evapotranspiration values than projected.Higher daily evapotranspiration values were noticed because the sensible heat flux is the major part of the energy, while the latent heat flux is dominating only over the agricultural area (Frey et al., 2011;Elhag, 2014a, b).The behaviour of the SEBS model could be explained by its tendency to simulate the potential daily evapotranspiration rather than the actual daily evapotranspiration, which is identified as the lack of leaf area index value over desert areas (Li et al., 2009;Elhag et al., 2011).The application of the SEBS model over the designated study area showed an insignificant difference to the Nile Delta case in term of accuracy assessment (Elhag et al., 2013).

Conclusions
Projected evapotranspiration data using a surface energy balance system model and multiple remote sensing imageries demonstrated robust association with the ground truth data.The application of the surface energy balance system model mapped the daily evapotranspiration and evaporative frac-tion objectively over Wadi ad-Dawasir region.The findings of the current research will help the decision makers towards a modification of the agriculture activities in similar areas in terms of conservative irrigational water regulations.The model shows consistent results in the estimation of daily evapotranspiration in Nile Delta region and in Wadi ad-Dawasir.Accordingly, the surface energy balance system model can be considered a reliable and effective tool in the estimation of daily evapotranspiration, explicitly in arid environments.

Figure 3 .
Figure 3. Normal distribution of actual daily evapotranspiration data.

Figure 5 .
Figure 5. Normal distribution of relative evaporation data.

Figure 6 .
Figure 6.The relationship between actual and simulated daily evapotranspiration.