**2.4 SEBAL method**

The SEBAL model developed by Bastiaanssen et al. [38] was used to calculate the actual evapotranspiration from Landsat-8 satellite images. The model's key inputs were the satellite measurements of surface albedo, normalised difference vegetation index (NDVI) and surface temperature (Ts). Also, the DEM and land-use map were used as additional input data. The DEM was applied for topographic and atmospheric correction [39]. However, the land-use map was used mainly to differentiate between LULC types exist in the study area. In addition to the satellite data, the SEBAL model requires minimum inputs of routine weather data (see the Data section). **Figure 2** shows a flowchart that describes the SEBAL model process. The SEBAL model scripts were formed using the Spatial Modeller Tool of the ERDAS IMAGINE 9.2 software, and the ArcGIS 10.2 software was used for data mapping and visualisation.

The SEBAL algorithm computes the latent heat flux as the residue of the energy balance Equation [38, 40, 41]:

$$
\lambda ET = \mathbf{R\_n} - \mathbf{G} - \mathbf{H} \tag{1}
$$

where Rn is the net radiation over the surface (W/m2 ), G is the soil heat flux (W/m<sup>2</sup> ), H is the sensible heat flux (W/m<sup>2</sup> ), λET is the latent heat flux (W/m<sup>2</sup> ) and λ is the latent heat of vaporisation (J/Kg).

### *2.4.1 Estimation of the Main surface parameters*

The main surface parameters of the SEBAL include: the surface albedo, land surface emissivity (ε) and land surface temperature (Ts).

**Figure 2.** *A flowchart explains the SEBAL model process.*

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

Surface albedo is defined as the ratio of reflected radiation from the surface to the incident shortwave radiation [40]. Thus the albedo is a single value that represents the integrated reflectance across the entire shortwave spectrum as represented by bands 2–7 of Landsat-8 data. Accordingly, the surface albedo (α) was estimated using the following formula [42]:

$$a = 0.2453a\_2 + 0.0508a\_3 + 0.1804a\_4 + 0.3081a\_5 + 0.1332a\_6 + 0.05221a\_7 + 0.0011\tag{2}$$

where α2, α3, α4, α5, α6, α7 represent the albedo of band2, band3, band4, band5, band6, and band7, respectively.

Surface emissivity is the thermal energy ratio radiated by the surface to the thermal energy radiated by a blackbody at the same temperature [42]. The NDVI is vital for calculating the land surface emissivity since it is used to estimate the vegetation coverage [43]. The calculation of the NDVI and vegetation coverage was as follows [8, 44]:

$$NNDVI = \frac{(NIR \quad -RED)}{(NIR \quad +RED)} \tag{3}$$

$$P\_v = \frac{(NDVI \quad -NDVI\_{min})}{(NDVI\_{max} \quad + NDVI\_{min})} \tag{4}$$

where NIR is the reflectance in the near-infrared band, which corresponds to band 5 in lansat-8 data, while the RED reflectance corresponds to band 4. Pv is the vegetation coverage.

The surface emissivity was conditionally determined based on the NDVI values using Eq. (5) [45]:

$$
\varepsilon\_0 = \varepsilon\_\mathbf{v} \mathbf{P}\_v + \varepsilon\_\mathbf{r} (\mathbf{1} - P\_v) + \mathbf{C}\_\varepsilon \tag{5}
$$

where ε<sup>0</sup> is the land surface emissivity, ε<sup>v</sup> and ε<sup>s</sup> are the vegetation and soil emissivity, respectively, and C<sup>ε</sup> represents the surface roughness (C<sup>ε</sup> = 0 for homogenous and flat surfaces) taken as a constant value of 0.005 [46].

The Pv values are conditioned with the NDVI ones. The land cover is classified as water when the NDVI ˂0. For the NDVI values range between 0 and 0.2, the land is considered covered with soil. The NDVI values of 0.2–0.5 the land cover are considered mixtures of soil and vegetation. However, when the NDVI >0.5, the land is considered covered with vegetation [43].

The land surface temperature was derived from the thermal bands. This step needs the spectral radiance to be converted into a sufficient brightness temperature. That means it has the black body temperature, assuming that the Earth's surface is a black body. Consequently, brightness temperature was determined using the formula [47]:

$$\mathbf{T\_b = \frac{K\_2}{\ln\left(\frac{K\_1}{L\_k} + 1\right)}} - 273.15\tag{6}$$

where, Tb is the satellite brightness temperature (C°), Lλ spectral radiance at top of the atmosphere, K1 and K2 are satellite calibration constants from the image metadata. The absolute zero value of �273.15C° was added to obtain results in Celsius [48].

*Climate Change in Asia and Africa - Examining the Biophysical and Social Consequence…*

The land surface temperature (T*s*) was computed using Eq. (7) [49]:

$$\mathbf{T}\_s = \frac{\mathbf{T}\_b}{\mathbf{1} + \left(\lambda. \frac{\mathbf{T}\_b}{\rho}\right) . \ln \varepsilon\_0} \tag{7}$$

where, λ is the wavelength of emitted radiance (11.5 μm for Landsat 5 &7; 10.9 <sup>μ</sup>m for Landsat 8), *<sup>ε</sup>*<sup>0</sup> is the land surface emissivity, <sup>ρ</sup> <sup>¼</sup> <sup>h</sup>*:* <sup>c</sup> <sup>σ</sup> ¼ 1*:*438 � <sup>10</sup>�<sup>2</sup> mK, <sup>σ</sup> is the Boltzmann constant 1*:*<sup>38</sup> � <sup>10</sup>�<sup>23</sup> <sup>J</sup>*=*<sup>K</sup> , h is the Planck's constant <sup>6</sup>*:*<sup>626</sup> � <sup>10</sup>�<sup>34</sup> J s , and c is the velocity of light 2*:*<sup>998</sup> � <sup>10</sup><sup>8</sup> <sup>m</sup>*=*<sup>s</sup> ).

#### *2.4.2 Determination of heat fluxes*

The net radiation (Rn) was calculated using surface reflectance and surface temperature (Ts):

$$\mathbf{R\_n = R\_s \downarrow - \infty R\_s \downarrow + R\_L \downarrow - R\_L \uparrow - (1 - \varepsilon\_0) R\_L \downarrow} \tag{8}$$

where Rs↓ is the incoming short wave radiation (W/m<sup>2</sup> ) (solar radiation), ∝ surface albedo (dimensionless), RL↓ is the incoming long wave radiation (W/m2 ), RL↑ is the outgoing long wave radiation (W/m2 ), and ε0 is the surface thermal emissivity (dimensionless). The calculation of these radiations was performed in Eqs. (9)–(11) [50]:

$$RS\,\downarrow = G\_{\kappa}.\cos\theta.r.\tau\_{\kappa w} \tag{9}$$

$$RL\uparrow = \varepsilon\_0.\sigma.T\_s^4\tag{10}$$

$$RL\,\downarrow = \varepsilon\_{\infty}.\sigma.T\_{\infty}^4\tag{11}$$

where Gsc is the solar constant, 1367 W/m<sup>2</sup> and cos θ is the cosine of the solar incidence angle, r is the Earth-Sun distance (dimensionless), and τsw is atmospheric transmissivity. RSI↓ values range from 200–1000 W/m<sup>2</sup> , depending on the image's time and location and the local weather conditions [42]. σ is the Stefan-Boltzmann constant (5.67 � <sup>10</sup>�<sup>8</sup> W.m�<sup>2</sup> .K�<sup>4</sup> ), Ts is the surface temperature (K), εα is the atmospheric emissivity, and T<sup>∝</sup> is the atmospheric temperature (K). The empirical Eq. (6) was used for calculating the ε<sup>∝</sup> [42].

$$
\varepsilon\_{\infty} = \mathbf{0.85} \times \left( -\ln|\tau\_{\text{sw}}| \right)^{0.09} \tag{12}
$$

The soil heat flux (G) is the rate of the heat flux stored or released into the soil and vegetation due to conduction. The ratio G*=*Rn was computed using Eq. (13) developed by [51]:

$$\mathbf{G}/\mathbf{R\_n} = \frac{T\_s}{\infty} \left( 0.0038 + 0.0074 \alpha^2 \right) \left( 1 - 0.98 \text{NDVI}^4 \right) \tag{13}$$

where Ts is the surface temperature (C°), ∝ is the surface albedo, and NDVI is the Normalised Difference Vegetation Index (ranged between �1 and + 1). NDVI values between 0 and 0.2 correspond to bare soil or very sparse vegetation, the values greater than 0.2 represent vegetated areas. The typical estimates of G*=*Rn assumed to be 0.5 for water, 0.05–0.4 for agriculture and 0.2–0.4 for bare soil [42].

The sensible heat flux (H) is the rate of heat loss to the air by convection and conduction, due to a temperature difference. H was determined using the aerodynamic based heat transfer equation as follows [51]:

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

$$H = \frac{\left(\rho \times \mathbf{C\_p} \times dT\right)}{\mathbf{r\_{ah}}} \tag{14}$$

where ρ is air density (kg/m3 ), Cp is air specific heat (1004 J/kg/K), dT (K) is the temperature difference (T1 – T2) between two heights (z1 and z2), and rah is the aerodynamic resistance to heat transport (s/m). The rah is computed for neutral atmospheric stability conditions as:

$$\mathbf{r\_{ah}} = \frac{\ln\left[\frac{Z\_2}{Z\_1}\right]}{u\* \times -k} \tag{15}$$

where Z1 and Z2 are heights in meters above the zero plane displacement of the vegetation, u\* is the friction velocity (m/s) which quantifies the turbulent velocity fluctuations in the air, and k is von Karman's constant (0.41).

#### *2.4.3 Estimating the evapotranspiration*

The instantaneous value of ET in equivalent evaporation depth was computed as:

$$\text{ET}\_{\text{inst}} = \text{3600 } \frac{\lambda \text{ET}}{\lambda} \tag{16}$$

where ETinst is the instantaneous ET (mm/hr), 3600 is the conversion from seconds to hours, λET is the latent heat flux (W/m) consumed by ET, ρw is the density of water (1000 kg/m<sup>3</sup> ), and λ is the latent heat of vaporisation (J/kg) and was computed as:

$$
\lambda = \left[ 2.501 - 0.00236 \left( \text{T}\_s - 273.15 \right) \times 10^6 \right] \tag{17}
$$

The reference ET fraction (ET0F) or crop coefficient (kc) was calculated based on ETinst for each pixel and ET0 was obtained from local ground weather stations.

$$\text{ET}\_0\text{F} = \text{ET}\_{\text{inst}} / \text{ET}\_0 \tag{18}$$

The daily values of ET (ET24) (mm/day) for each pixel was calculated as follows:

$$\rm ET\_a = ET\_0 \rm F \times ET\_0 24 \tag{19}$$

where ET0F is the reference ET fraction, ET024 is the cumulative alfalfa reference for the day (mm/day), and ETa is the actual evapotranspiration for the entire 24-hour period (mm/day).

The actual monthly and annual ET was calculated using daily ET data as follows [42]:

$$\text{ET}\_{\text{a,period}} = \sum\_{i=m}^{n} \text{ET}\_{0} \mathbf{F} \times \text{ET}\_{0} \mathbf{24} \tag{20}$$

$$\text{ET}\_{\text{a,annual}} = \sum \text{ET}\_{a,\text{period}} \tag{21}$$

Allen et al. [18] showed that one cloud-free satellite image per month is sufficient to develop ET0F curves for seasonal ETa estimations.

#### *2.4.4 Validation of the SEBAL model evapotranspiration*

The produced ETa from Landsat-8 images and SEBAL model was validated using the FAO P-M method [52]. The FAO P-M was used to calculate the reference crop evaporation (ET0) from the actual climate data in the study area based on Eq. (22):

$$\text{ET}\_0 = \frac{\mathbf{0.408\Delta(R\mathbf{n} - \mathbf{G}) + \gamma \frac{900}{\mathbf{T} + 2\mathcal{D}3} \mathbf{U2(es - e\mathbf{a})}}{\Delta + \gamma(\mathbf{1} + \mathbf{0.34U2})} \tag{22}$$

where ET0 is reference evapotranspiration (mm/day), Δ is slope vapour pressure curve (kPa/°C), γ is psychrometric constant (kPa/°C-1), T is mean daily air temperature at 2 m height (°C), U2 is wind speed at 2 m height (m/s), es is saturation vapour pressure (kPa), ea. is actual vapour pressure (kPa), es ð Þ � ea represents the saturation vapour pressure deficit (kPa).

The crop coefficient (Kc) for the different croplands and the open water determined based on Allen et al. [52]. The ET0 obtained from the FAO P-M method and the kc were used to calculate the ETa depending on actual weather data as follows:

$$\rm ET\_a = ET\_0 \times kc \tag{23}$$

The ETa resulted from the FAO P-M method was used to validate the ETa obtained from SEBAL model.

A linear correlation and the root mean square error (RMSE) between the measured (FAO P-M) and the SEBAL daily ETa was computed [13]. The RMSE was calculated as follows [26]:

$$\text{RMSE} = \sqrt{\frac{\sum\_{i=1}^{N} (O\_i - P\_i)^2}{N}} \tag{24}$$

where Oi represents the observed values of the FAO P-M method as the standard model; Pi represents the estimated values from the SEBAL algorithm; and Oi and Pi are the mean values from the FAO P-M method and SEBAL model, respectively.

### **3. Results and discussions**

#### **3.1 LULC mapping**

The LULC map of the study area showed that the main identified classes were the date palm, cropland, bare land, urban land, aquatic vegetation, and water (**Figure 3**). The area occupied by each LULC type within the oasis boundaries is shown in **Table 2**. The date palm covers about 131 km2 of Al-Ahsa Oasis, and it is the most important land-use class for the local and national economy. Croplands used only 144 km2 of the oasis area, and it is dominated by rice and vegetables. The bare land class occupies most of the oasis area with 4759 km2 . Bare lands dominated by desert and rock outcrops also occurred in this class. Most of the urban land occurs on the oasis periphery, as most oasis land is under agricultural use. The aquatic vegetation and water classes occupy together an area of about 17 km2 . Al-Dakheel [53] reported that date palm covered about 92% within the oasis boundary.

The overall classification accuracy of the LULC map was 90%, with a kappa index of 88%, while the user's and producer's accuracies differed along with LULC types (**Table 2**). This accuracy level indicates that the classification method adopted in this study effectively produced a compatible LULC map over the study area.

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

**Figure 3.** *LULC map of the study area.*


#### **Table 2.**

*Areas and accuracy assessment of the LULC classes.*

#### **3.2 Analysis of land surface parameters**

The statistical means values of land surface albedo show that it was raining between 0.46 and 0.51 during Apr. 2017-Mar. 2018 (**Figure 4a**). However, the spatial distribution of land surface albedo indicates higher in the bare lands areas

**Figure 4.**

*Mean values of land surface parameters: (a) albedo; (b) emissivity and (c) surface temperature. Bars denoted standard error.*

than the other LULC types (**Figure 5a**). Nevertheless, the seasonal variation shows that the surface albedo was higher during the summer (April, July) compared to the winter (November, February). The surface albedos levels were found lower in the vegetation-covered areas than the exposed soil [8].

The land surface emissivity means values clearly show that it does not vary along the study period (**Figure 4b**). However, the land surface emissivity's spatial patterns indicated higher in the open water and lowered in the date palm and croplands classes (**Figure 5b**). The lowest values of the land surface emissivity were observed in bare lands. These results are inconsistent with that reported by Kong et al. [8].

The land surface temperature statistics showed that the average minimum value of 299 K was observed in January and February 2018, while the maximum mean value of 332 K occurred on 10 August 2017 (**Figure 4c**). **Figure 5c** indicated that the highest land surface temperature values were shown in summer, and they associated with the bare lands. Nevertheless, during the winter, the difference in the land surface temperature between the bare lands and the other LULC types was 10–20 K. The land surface temperature is an essential parameter for quantifying the ET process among the different LULC types of the study area. Accordingly, the land surface temperature estimation is essential for land classification, energy budget estimations, and crop production [54].

#### **3.3 Heat fluxes analysis**

The surface heat fluxes estimated over the different LULC types can affect the ET amount measured along the study period at varying scales.

The net radiation flux's statistical values show that the values in April, May and June are clearly higher than those in October, November, December and January (**Figure 6a**). As shown in **Figure 7a**, the open water received the highest amount of

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

**Figure 5.**

*Spatial distributions of land surface parameters. (a) Albedo; (b) emissivity and (c) surface temperature.*

net radiation followed by the aquatic vegetation, date palm, and croplands. However, the lowest net radiation found in the bare lands. Moreover, the net radiation flux increased significantly in April and July compared to November and February.

The increase of the net radiation in April and May might be due to rising groundair temperature and the gradual death of the sparse vegetation over bare lands. The change in the surface energy budgets due to irrigation results in increased net radiation over agricultural lands [55]. Also, the significant variation of the net radiation flux during the study period could be due to the region's heterogeneous LULC types.

The soil heat flux tendency showed higher in the summer recording average value of 91 W/m<sup>2</sup> on 23 June 2017, while the lowest mean value was 5 W/m2 observed on 16 December 2017 (**Figure 6b**). The spatial distribution of the soil heat flux for the vegetation cover classes was higher than that of the bare lands during the summertime, while the difference was slight in winter (**Figure 7b**).

The sensible heat fluxes mean values do not vary consistently along with the summer and wintertime. The average highest value of 203 W/m<sup>2</sup> detected on 04 April 2017, while the lowest was on 10 August 2017 (**Figure 6c**). According to the spatial distribution in **Figure 7c**, the bare and urban lands'sensible heat flux show relatively high values compared to the vegetation lands.

**Figure 6.** *Mean values of heat fluxes: (a) net radiation; (b) soil heat and (c) sensible heat. Bars denoted standard error.*

The increasing trend in sensible heat flux during April and May can be attributed to increased net radiation flux under persistent irrigation land used for agriculture and urban areas. However, Amatya et al. [56] indicated that the increase in wind speed and the ground-air temperature difference could increase sensible heat flux.

#### **3.4 Analysis of the actual evapotranspiration**

The spatial distribution of the daily ETa over the study area during 20 April 2017–2022 March 2018 is shown in **Figure 8**. The temporal patterns of the daily ETa showed that the highest values were observed during the peak summertime in July and August. The mean daily ETa values for the different LULC types in the oasis are shown in **Figure 9**. It is clear that the daily ETa for the water bodies and aquatic vegetation was between 5.6–8.7 mm.day�<sup>1</sup> in summer and about 2.3–5.6 mm.day�<sup>1</sup> in winter. However, the date palm and croplands showed daily ETa of 3.5–8.0 and 2.0–3.6 mm.day�<sup>1</sup> for the winter and summer, respectively.

The variability of the daily ETa values for the Date palm and cropland during the summer and winter times mainly attributed to the irrigation water distribution, soil salinity, drainage, and agricultural practices and their impact on moisture and salinity in the root zone. Under Saharan oasis conditions, soil texture, plot size, and farmers' practices in particular irrigation found to have significant effects on the daily ETa [57]. In Saudi Arabia, the daily ETa of date palm was observed to decrease during winter and increased during summer, depending on the crop's growth stage [58]. Also, the daily water consumption of major cropping systems in Saudi Arabia varied spatially depending on cropping practises and climatic conditions [22].

The mean daily ETa for the urban land ranged from 1.3 to 4.5 mm.day�<sup>1</sup> throughout the study period (**Figure 6**). The urban land is covered with some lanes and parks that make seasonal variations of the daily ETa within the study area.

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

**Figure 7.**

*Spatial distributions of heat fluxes. (a) Net radiation; (b) soil heat and (c) sensible heat.*

Nevertheless, the low values of the daily ETa showed in the bare land mainly resulted from the sparse vegetation in this land-use system.

The variation of the daily ETa estimates for the different LULC types under the oasisarid ecosystem indicates that they change significantly throughout the seasons [24].

**Figure 10** shows the spatial pattern of the monthly and annual ETa across the study area. The high rates of the ETa found to be between 80 to 200 mm.month�<sup>1</sup> during July, August and September for the water, aquatic vegetation, date palm and cropland. However, at the beginning of the summer in April, May and June, the ETa rates were 60–100 mm.month�<sup>1</sup> for the same LULC types. Nevertheless, in the winter during Oct 2017–Mar 2018 the ETa ranged between 40 to 140 mm.month�<sup>1</sup> for the water, aquatic vegetation, date palm and cropland land-use systems.

The mean annual ETa produced by SEBAL model for the different LULC types in the study area is shown in **Figure 11**. The ETa rates of date palm trees ranged from 800 to 1400 mm.year�<sup>1</sup> during Apr. 2017 to Mar. 2018. The annual water consumption for date palm was highly variable. This might be attributed to the type of irrigation system and the age variations of date palm trees along the oasis.

The open water evaporation lost was around 2000 mm.year�<sup>1</sup> , while an average of 1600 mm.year�<sup>1</sup> was evaporated from aquatic vegetation. Nevertheless,

**Figure 8.** *The spatial distribution of the daily ETa.*

**Figure 9.**

*Temporal variation of mean daily ETa for different LULC systems.*

croplands showed lower annual ETa of 800 mm.year�<sup>1</sup> compared to the date palm. The main crops like rice and vegetables can be cultivated during a particular time of the year in the oasis; therefore, croplands showed relatively low annual ETa compared to date palm areas.

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

**Figure 10.**

*The spatial patterns of the monthly and annual ETa.*

The annual ETa of urban lands was 400 mm.year�<sup>1</sup> . Urban lands are affected by the irrigation of trees, lanes and parklands, which resulted in consumption of a large amount of the oasis groundwater. The annual evaporation from bare lands was very low (200 mm.year�<sup>1</sup> ) and less than the long-term average rainfall. Bare lands equipped most of the study region areas, and they covered with sand dunes (**Figure 11**). Moreover, the bare lands characterised by low vegetation coverage levels and low water contents in the soil surface [41]. Also, very low rates of ETa from bare soil observed in the western and southern parts of Saudi Arabia [22].

#### **3.5 Validation, limitation and uncertainty of SEBAL model**

The FAO P-M was used as a standard method to validate the SEBAL model [8, 59]. The validation measurement for ETa between the SEBAL model and FAO P-M method for the different land-use system is shown in **Figure 12**. A significantly high level of agreement can be observed between the two methods for the selected LULC types. The RMSE for the most validated LULC system in the study area found to be less than 1.0 mm.day�1. However, the RMSE was slightly higher for cropland areas (**Figure 12d**), mainly due to the method used for calculating the crop coefficient (kc) for the different crop types within the croplands system. The SEBAL model does not require kc information because the model biophysical properties estimated kc as part of the SEBAL process. However, the FAO P-M computed kc based on the characteristics and climatic regions for the different crops. Accordingly, the kc of vegetables (tomato and cucumber) ranged between 0.5–1.15, and for the rice, it was 1.0–1.35 [52]. Nevertheless, the date palm's kc was in the range of 0.9–0.95, while it was 1.05 for the open water [52].

Moreover, it seems that SEBAL underestimates the ETa for croplands since they were diverse in terms of crop type and growing season. Also, the kc used to calibrate the ETo was made only for a few experimental plots. Therefore, the variability in the ETa values predicted by SEBAL and measured by the FAO P-M method was slightly higher. However, the low values of ETa in winter do not affect the outputs of seasonal ETa over croplands, as farmers use a little water for irrigation. The kc of

**Figure 11.**

*Mean annual ETa produced by SEBAL for the different LULC types in the study area during April 2017–March 2018. Bars denoted standard error.*

**Figure 12.**

*Linear correlation between the FAO P-M and SEBAL model for the different LULC types. (a) Open water; (b) aquatic vegetation; (c) date palm; (d) cropland.*

*Mapping and Assessment of Evapotranspiration over Different Land-Use/Land-Cover Types… DOI: http://dx.doi.org/10.5772/intechopen.96759*

crops can vary during the growing season, depending on their growth stage [60]. Rahimi et al. [3] reported that the application of SEBAL for estimating the ETa over agricultural land in the Tajan catchment of Iran resulted in RMSE of 1.49 mm.day-1. However, the daily estimated ETa for both cropland and grassland in the Midwestern United States using SEBAL contributed to RMSE ranging between 1.74 and 2.46 mm.day�<sup>1</sup> [6].

SEBAL model utilises satellite imagery and a small set of surface weather information, including wind speed and air temperature at a reference height of 2 m to solve the energy balance [38]. However, the steps used for the SEBAL process are complicated. The model requires an experienced modeller and a substantial number of working hours for image processing. Also, the large model number of mathematical formulas increases the possibility of human errors [42]. Furthermore, the procedure used by SEBAL for selecting the "hot" and "cold" pixels endmembers involves not only an analysis of the surface temperature but also an understanding of other biophysical properties, such as vegetation indices, surface albedo, and LULC type [16, 61]. Therefore, identifying these endmembers requires modeller intervention and knowledge of the biophysical parameter values and ranges.

### **4. Conclusions**

This study demonstrates the power of remote sensing data and the biophysical modelling for quantifying the ETa process over an arid ecosystem in Saudi Arabia. The estimated mean annual ETa was 2000 mm.year�<sup>1</sup> for open water and varied between 800 and 1400 mm.year�<sup>1</sup> for date palm. However, it was 1600 mm.year�<sup>1</sup> for the aquatic vegetation, while an average of 800 mm.year�<sup>1</sup> was observed in croplands. The validation measure showed a significant agreement level between the SEBAL model and the FAO P-M method with RMSE of 0.62, 0.84, 0.98 and 1.38 mm.day�<sup>1</sup> for aquatic vegetation, date palm, open water and cropland respectively.

The obtained ETa information will help Saudi Arabia formulate strategies to reduce the gap between the water supply and demand in the irrigated areas. Furthermore, the ETa patterns mapped over the diverse LULC systems can be used as a baseline framework for sustainable water resources management and agrometeorological services in the different regions of Saudi Arabia. However, conducting long-term ETa studies using remote sensing data coupled with the implementation of different models and field tools may improve the assessment of the ETa dynamic process in arid regions.

## **Acknowledgements**

The author's acknowledge the Deanship of Scientific Research at King Faisal University, for the financial support under the annual research project (Grant No. 186002).

*Climate Change in Asia and Africa - Examining the Biophysical and Social Consequence…*
