Mapping of Evapotranspiration and Determination of the Water Footprint of a Potato Crop Grown in Hyper-Arid Regions in Saudi Arabia

Abstract

Seasonal quantification of a crop’s evapotranspiration (ET) and water footprint (WF) is essential for sustainable agriculture. Therefore, this study was conducted to estimate the ET and WF of an irrigated potato crop using satellite imagery of Landsat and Sentinel-2 sensors. The Simplified Surface Energy Balance (SSEB) algorithm was used to evaluate the crop water use (ET_a) for potato fields belonging to the Saudi Agricultural Development Company, located in the Wadi-Ad-Dawasir region, Saudi Arabia. Normalized difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), and land surface temperature (LSD) were computed for Landsat and Sentinel-2 datasets, which were used as inputs for mapping the potato tuber yield and, subsequently, the WF. The results indicated that the NDVI showed the best accuracy for the prediction of the potato tuber yield (R² = 0.72, P > F = 0.021) followed by the SAVI (R² = 0.64, P > F = 0.018), compared to the field harvested actual yield (Y_A). A comparison between the satellite-based ET_a and the actual amount of water applied (W_A) for irrigation showed a good correlation (R² = 0.89, RMSE = 4.4%, MBE = 12.9%). The WF of the potatoes in the study area was estimated at values between 475 and 357 m³ t⁻¹ for the early (September–December) and late (December–April) growing periods, respectively. A major portion (99.2%) of the WF was accounted for from irrigation with variations of 18.5% and 3.5% for early- and late-planted potatoes, respectively, compared to the baseline (crop planted in season). In conclusion, the results showed the possibility of satisfactorily estimating the WF using the SSEB algorithm by integrating the Landsat-8 and Sentinel-2 datasets. In general, the high rates of ET in the early planting season led to higher WF values compared to the in-season and late planting dates; this will help in selecting suitable planting dates for potato crops in the study area and areas with similar environments, which enhances the opportunities for sustainable management of irrigation water.

1. Introduction

Population growth and rapid changes in people’s lifestyles have led to a significant increase in the demand for water all over the world, and this requires paying special attention to the quantities of water used in irrigating agricultural crops and those used in associated industries [1]. The water footprint (WF), which was introduced to assess irrigation efficiency [2], is defined as the ratio of the volume of water used (m³ ha⁻¹) to the crop yield (t ha⁻¹) [3,4,5]. The scarcity of freshwater resources and harsh environmental conditions have forced Saudi Arabia to make greater efforts to adopt sustainable agricultural practices [6]. In this regard, the importance of using effective agricultural methods in estimating the water requirements of different crops has emerged as an essential element for improving decision-making strategies relating to agricultural production and the consequent positives in the area of food security [7].

Multidate remotely sensed images and mapping algorithms has become a useful tool for the precise administration of irrigation water and scheduling, as they provide comprehensive data of agricultural fields at appropriate time intervals that help in estimating crop water consumption during the different growth stages of crops [8,9,10]. The use of various applications of remote sensing and GIS in irrigation management practices has been extensively studied by many researchers, as several protocols used for estimating crop biophysical parameters under different environmental conditions have been developed and tested [9,11,12,13]. Remote sensing applications are used in agriculture to estimate biomass and yield, monitor vegetation activity and drought stress, and map agricultural land uses, as well as a wide range of uses in precision agriculture and management of irrigation water [14,15,16].

Advancements in remote sensing techniques and algorithms have been proven to achieve estimates of crop yield, actual evapotranspiration (ET_a), crop water use, and water footprint with the use of multispectral data, such as MODIS, Landsat, Sentinel, IRS, VIIRS, AVHRR, etc. [17]. The use of time-series data with appropriate temporal and spatial resolution effectively contributes to the monitoring of crop health, phenology, vigor, surface temperature, and associated factors in the mapping of ET_a, crop water use, and water footprints (WF) as per the length of the crop growth period [18]. On the other hand, crop water-use mapping is also achieved with the integration of multispectral images from multiple sensors [19,20].

The WF (m³ t⁻¹) of a certain crop is calculated by dividing the total volume of the applied irrigation water (m³ ha⁻¹) by the quantity of the production (t ha⁻¹) [21,22,23,24]. As per the source of the irrigation water, the WF is quantified as a green component (WF_green), where the irrigation water for crops is the rainwater resource. Regarding the blue component of WF_blue, the source of irrigation is groundwater. The two components WF_green and WF_blue together indicate the consumptive water footprint (WF_B+G) of a crop in a given season, while the WF_grey indicates the amount of water used to absorb the pollutants from the used chemicals as well as the quality standards of irrigation water [25].

Advancements in remote-sensing image analysis techniques have proven effective in accurate estimation of ET_a through several algorithms that have been successfully used and reported in the literature [26]. The most widely used algorithms include the SEBAL (Surface Energy Balance Algorithm over Land), the METRIC (Mapping Evapotranspiration with Internalized Calibration), the SEBS (Surface Energy Balance System), the SAFER (Simple Algorithm For Evapotranspiration Retrieving), the SSEB (Simplified Surface Energy Balance), the SSEBI (Simplified Surface Energy Balance Index), the TSM (Two-Source Model), and the TSTIM (Two-Source Time-Integrated Model), etc. [27,28]. The SSEB algorithm is widely used for estimating the ET_a of agricultural fields due to its simplicity, its dependence on minimum data, and its ease of implementation without affecting the final accuracy [29,30]. Many previous studies on WF in crop production focused on the use of either global scales [22,31] or local scales [32,33,34,35,36,37,38]. The crop coefficient (K_c) is a key factor in determining ET_a based on the reference ET (ET_o) of a certain crop [39]. Several studies have successfully substituted K_c values with the fraction of ET (ET_f) in ET_a mapping [40,41,42].

Understanding the time-based and spatial variations in crop water demand at a regional level is essential for better management of available water resources by formulating reliable irrigation schedules [43]. Irrigation management mainly depends on planting dates and respective evapotranspiration rates, which in turn vary in the volume of irrigation water application to the crops for early- and late-planted potatoes, compared to the baseline (crop planted in season). Therefore, this study aimed to estimate the WF of a potato crop grown in three different planting periods, namely early, on-season, and late planting, in the arid climates of Wadi-Ad-Dawasir, Saudi Arabia, using the SSEB algorithm by integrating the Landsat-8 and Sentinel-2 datasets. The specific objectives included the following: (i) estimating the ET fraction (ET_f) of the potato crop, (ii) understanding the relationship between the obtained ET_f and the FAO-crop coefficient (K_c) values for the replacement of K_c with ET_f in ET_a mapping using the SSEB algorithm, and (iii) mapping and evaluating the WF of the potato crop against the field-measured actual yield (Y_A) and actual applied water (W_A) for better management of water resources by selecting the appropriate planting dates.

2. Materials and Methods

2.1. Study Site

The experimental work has been conducted on six agricultural fields in a commercial farm belonging to the Saudi Agricultural Development Company, located in the Wadi-Ad-Dawasir area in Riyadh province, Saudi Arabia, between latitudes of 19.90° and 20.33° N and longitudes of 44.81° and 44.95° E (Figure 1). The soil in the study farm is characterized as sandy loam, and the major cultivated crops are potato, wheat, melon, and other vegetable crops.

The climatic conditions of the study area are characterized by hot summers with an average daily maximum temperature of 49 °C and cold winters with an average daily minimum temperature of 7 °C. The relative humidity is normally stable at 24%, while the daytime is, on average, around 11 h. The average wind speed is 13 km h⁻¹; however, it can reach as high as 46 km h⁻¹ during thunderstorms, with an average annual rainfall of 37.6 mm [44,45]. The major cultivated crops in the study farm are potatoes, corn/maize, durum wheat, watermelon, and fodder (alfalfa), due to sparse and low rainfall with arid climates and higher evapotranspiration rates which exceed rainfall. Hence, supply of irrigation water is mandatory to achieve the crop evaporative demands (Figure 2). Most of the irrigation water (92%) is extracted from groundwater resources. During the study, the potatoes are grown under full irrigation.

2.2. Sampling and Field Data

The experimental farm (INMA commercial farm) contains 90 fields irrigated by center-pivot systems, of which 17 fields were cultivated with a potato crop (Figure 1). The potato fields were cultivated in three planting periods, namely early planting, in-season planting, and late planting (

Table 1. Details of experimental fields.

Season	Irrigation Method	Field ID	Area (ha)	Time	Planting Date	Harvest Date
Autumn	Center Pivot	72 N	29	Early Planting	5 September	29 December
Autumn	DRIP	82 F	56	Early Planting	17 September	10 January
Winter	DRIP	45 F	25	On-season	19 October	13 March
Winter	Dragon Line	60 F	40	On-season	16 November	18 Mar
Winter	Center Pivot	49 N	25	Late Planting	23 December	24 April
Winter	DRIP	30 N	21	Late Planting	20 December	21 April

). Of the 17 potato fields, 6 fields (2 experimental fields per planting date/period) were allocated for sampling, and a total of 120 sampling locations (30 from each field) were randomly identified as permanent plots, with the help of a handheld GNSS receiver (GeoXH6000, Trimble, Westminster, CO, USA). Growth/phenology stages and biophysical parameters (LAI, canopy surface temperature) were recorded from emergence to harvest. Periodic ground truth data recorded during the crop growth included phenology stages and biophysical properties such as plant height, leaf area index (LAI), chlorophyll content (SPAD), NDVI and canopy temperature. All samples were divided into three groups as follows: (i) training data (50%) for developing classification models, (ii) validation data (25%) for hyper parameter tuning, and (iii) test data (25%) for evaluation of classification/prediction map accuracies. Subsequently, the generated models were applied to all the 17 potato cultivated fields. Soil samples were collected at a depth of 0 to 10 cm from the soil surface and then analyzed using standard laboratory methods to determine soil bulk density (g m⁻³), water retaining capacity (%), and wilting point (%). LAI measurements were performed using the Plant Canopy Analyzer (PCA-2200, Li-cor Biosciences, USA). The information on agricultural practices, irrigation schedules, volume of water applied to the fields, and crop yield data was obtained from the records of the experimental farm. Meteorological data were collected from the weather station installed in the experimental farms. Field-based crop water use was assessed based on the soil water budget method.

2.2.1. Potato Yield

Yield samples were collected from selected potato fields by harvesting an area of 2.5 square meters from each of the randomly specified sampling locations. The net weights of the potato yield samples were converted to a standard yield unit (t ha⁻¹) after excluding damaged, small, or irregularly shaped potato tubers.

2.2.2. Irrigation Water Demand—CROPWAT Method

CROPWAT is a decision support tool developed by the Land and Water Development Division of the FAO. CROPWAT (ver. 8.0) for windows is a computer program for the calculation of crop water requirements based on soil, climate, and crop data. In addition, the program facilitates the development of irrigation schedules for different management practices. The crop water requirement (W_A) and irrigation schedules were developed using the CROPWAT program, as described in Allen et al. [46]. To match the volume of irrigation water with crop water requirement, readily available water for the plants was assessed.

2.3. Satellite Data and Image Analysis

Free-of-cost time series satellite images of Sentinel-2 (A/B) and Landsat 8/9 (L8/L9) were downloaded from the United States Geological Survey (USGS) facility (https://earthexplorer.usgs.gov/ (accessed on 20 February 2023)) coinciding with the growth period of the potato crop according to the presented details in

Table 2. Details of Sentinel-2 and Landsat-8 data used to map the water footprint (WF) of potato in Wadi-Ad-Dawasir, Saudi Arabia.

Sensor	Spectral Bands	Temporal Resolution	Spatial Resolution	Start Date	End Date
Landsat-8/9	Optical Bands 5, 6, 8	16 days	30 m	14 September 2020	19 April 2021
Landsat-8/9	TIR Bands 11 and 12	16 days	30 m
Sentinel 2	Optical Bands 5, 7, 8A and 11	5 days	10–20 m	7 September 2020	15 April 2021

. The downloaded images were subjected to atmospheric correction for clouds and shadows and then used to calculate vegetation indices (NDVI, SAVI, LAI), land surface temperature (LST), vegetation indices, canopy temperature, potato yield, evapotranspiration (ET), and water footprint (WF) according to potato planting periods, using the QGIS (version 3.28) and SNAP (Version 9.0.0) and ESRI ArcGIS (Version 10.7) software programs. The datasets were standardized to the projected coordinate system of WGS84/UTM zone 38 N, and then regions of the experimental fields (region of interest—ROI) were extracted.

2.4. Mapping of Potato Water Footprint

Satellite images of Sentinel-2 (A/B) and Landsat 8/9 (L8/L9) were acquired. The water footprint (WF) of potato crop was determined (Equation (1)) by dividing the total volume of water use (m³ season⁻¹) by the quantity of the production (t ha⁻¹), i.e., potato yield [21,22,23,24]. Both the green and blue WF components were assessed as per potato planting periods.

$$\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{F}\mathrm{o}\mathrm{o}\mathrm{t}\mathrm{p}\mathrm{r}\mathrm{i}\mathrm{n}\mathrm{t}=\mathrm{C}\mathrm{r}\mathrm{o}\mathrm{p} \mathrm{w}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{u}\mathrm{s}\mathrm{e}/\mathrm{c}\mathrm{r}\mathrm{o}\mathrm{p} \mathrm{y}\mathrm{i}\mathrm{e}\mathrm{l}\mathrm{d}$$

(1)

where water footprint (WF) is in m³ t⁻¹, crop water use (m³ ha⁻¹) is derived from satellite images based on the SSEB algorithm, and the crop yield (t ha⁻¹) is predicted using potato yield prediction models developed by Al-Gaadi et al. [47] in the same experimental farm.

2.4.1. Crop Water Use

The SSEB algorithm, proposed by Senay et al. [27], was adopted to estimate the actual evapotranspiration (ET_a), which was considered as the CWU, as illustrated in the systematic flow presented in Figure 3. The ET fraction (ET_f) was computed to substitute the crop coefficient (K_c) value in calculating the ET_c (Equation (2)).

Initially, the land surface temperature (LST) was calculated using the TIR bands of Landsat-8/9. The ET_f, was calculated from the LST and vegetation indices layers [48,49]. Vegetation indices (NDVI, SAVI, and LAI) were calculated from Sentinel-2 data. Then, the daily ET was calculated using a methodology similar to the classic method of the crop coefficient (K_c), using Equation (2) provided by Allen et al. [46]. Equation (3) confirms that the ET_f is similar to the K_c [40,50]. Since Sentinel-2 does not contain thermal data, the SSEB model cannot be applied for ET mapping alone [19]., the ratios of ET_f to NDVI of both Landsat and Sentinel-2 were integrated as described in Singh et al. [19]. The ET_f of each pixel was assessed by combining the cold (LST_CP) and hot (LST_HP) pixels of the studied potato fields. Then, the obtained ET_f of individual pixels was used to calculate the ET_a for each field. The reference ET (ET_o) was calculated according to the FAO-56 Penman–Monteith formula (Equation (4)).

$${\mathrm{E}\mathrm{T}}_{\mathrm{c}}={\mathrm{E}\mathrm{T}}_{\mathrm{o}}\times {\mathrm{K}}_{\mathrm{c}}$$

(2)

$${\mathrm{K}}_{\mathrm{c}}={\mathrm{E}\mathrm{T}}_{\mathrm{f}}={(\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{H}\mathrm{P}}-{\mathrm{L}\mathrm{S}\mathrm{T})/(\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{H}\mathrm{P}}-{\mathrm{L}\mathrm{S}\mathrm{T}}_{\mathrm{C}\mathrm{P}})$$

(3)

where

LST_HP = the average temperature of three hot pixels selected from each scene;

LST = the land surface temperature in a composite scene;

LST_CP = the average of three cold pixels selected in each Landsat scene.

$${\mathrm{E}\mathrm{T}}_{\mathrm{o}}=\frac{0.408∆\left({\mathrm{R}}_{\mathrm{n}}-\mathrm{G}\right)+\mathsf{\gamma }\frac{900}{\mathrm{T}+273}{\mathrm{u}}_2\left({\mathrm{e}}_{\mathrm{s}}-{\mathrm{e}}_{\mathrm{a}}\right)}{∆+\mathsf{\gamma }\left(1+0.34{\mathrm{u}}_2\right)}$$

(4)

where ET_o is the reference evapotranspiration (mm d⁻¹), R_n is the net radiation at the crop surface (MJ m⁻² d⁻¹), G is the soil heat flux density (MJ m⁻² day⁻¹), T is the daily air temperature at 2 m height (°C), u₂ is the wind speed at 2 m height (m s⁻¹), e_s is the saturation vapor pressure (kPa), e_a is the actual vapor pressure (kPa), e_s–e_a is the saturation vapor pressure deficit (kPa), Δ is the slope of the vapor pressure curve (kPa °C⁻¹), and γ is the psychrometric constant (kPa °C⁻¹).

Subsequently, the blue (CWU_blue) and green (CWU_green) components of the CWU were calculated as per Equations (5) and (6).

$${\mathrm{C}\mathrm{W}\mathrm{U}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}=10\times \sum _{\mathrm{d}=1}^{\mathrm{l}\mathrm{g}\mathrm{p}}{\mathrm{E}\mathrm{T}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}$$

(5)

$${\mathrm{C}\mathrm{W}\mathrm{U}}_{\mathrm{b}\mathrm{l}\mathrm{u}\mathrm{e}}=10\times \sum _{\mathrm{d}=1}^{\mathrm{l}\mathrm{g}\mathrm{p}}{\mathrm{E}\mathrm{T}}_{\mathrm{b}\mathrm{l}\mathrm{u}\mathrm{e}}$$

(6)

where ET_green is the green water evapotranspiration, and ET_blue is the blue water evapotranspiration from day one (d = 1) until the specified length of the growth period (lgp). The factor 10 is to convert water depths in millimeters into water volumes per unit area (m³ ha⁻¹).

2.4.2. Water Footprint

The crop water footprint (WF, m³t⁻¹) was determined as the sum of the blue (WF_blue) and green (WF_green) components of the WF, as shown in Equation (7) and provided by Hoekstra et al. [23]. The CWU (m³ ha⁻¹) for each component was calculated using Equations (8) and (9) according to Aldaya et al. [51].

$$\mathrm{W}\mathrm{F}={\mathrm{W}\mathrm{F}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}+{\mathrm{W}\mathrm{F}}_{\mathrm{b}\mathrm{l}\mathrm{u}\mathrm{e}}$$

(7)

Finally, the green and the blue WF of potato crop was obtained as per Equations (8) and (9) as described in Hoekstra et al. [23]. Both the WF_blue and WF_green components of a crop’s WF was computed by dividing the CWU by potato yield (CP) (t ha⁻¹).

$${\mathrm{W}\mathrm{F}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}=\frac{{\mathrm{C}\mathrm{W}\mathrm{U}}_{\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}}}{\mathrm{C}\mathrm{P}}$$

(8)

$${\mathrm{W}\mathrm{F}}_{\mathrm{b}\mathrm{l}\mathrm{u}\mathrm{e}}=\frac{{\mathrm{C}\mathrm{W}\mathrm{U}}_{\mathrm{b}\mathrm{l}\mathrm{u}\mathrm{e}}}{\mathrm{C}\mathrm{P}}$$

(9)

where CWU_green and CWU_blue are the green (rainfall) and blue (surface and groundwater) water used in irrigating the crop.

2.5. Water Footprint Based on Field Data

For accuracy assessment, field data-based WF was computed using the collected field observations and agri-meteorological parameters. The blue WF, which represents the volume of irrigation water applied to the crop, was calculated using Equation (10) according to Scarpare et al. [52].

$${\mathrm{W}\mathrm{F}}_{\mathrm{M}\mathrm{b}\mathrm{l}\mathrm{u}\mathrm{e}}=\frac{10\times \left(\mathrm{I}\mathrm{r}-\left(\mathrm{D}\mathrm{P}+\mathrm{R}\mathrm{O}\right)\right)}{{\mathrm{Y}}_{\mathrm{f}}}$$

(10)

where WF_Mblue (m³ t⁻¹) is the blue WF estimated from measured irrigation data, Y_f is the potato tuber yield, Ir is the irrigation water applied during the growth period (mm), DP is the deep percolation water leaving the root zone (mm), and RO is surface runoff water. Since it is difficult to measure the losses from such large fields, an average irrigation efficiency of 70% was considered in the center pivot irrigation systems [53] to account for the losses due to surface runoff and deep percolation.

2.6. Statistical Analysis

The crop yield and WF estimated based on remote sensing approaches were assessed using selected accuracy indicators such as the root mean square error (RMSE), mean bias error (MBE), and co-efficient of regression (R²). The statistical analysis was performed using the SPSS (Ver.20) software program.

3. Results

The basic climatic datasets required for the CROPWAT model, such as the average values of monthly rainfall, temperature, wind velocity, water vapor content, and day-light duration, was obtained from the weather station installed in the experimental farm. The average daily maximum air temperature ranged between 37 °C in May and 47 °C in August, while the average daily minimum temperature ranged between 14.2 °C in January and 38 °C in August, with an average daily temperature of 34 °C. The maximum (22.6%) water vapor content was observed in the period from November to February, and the minimum (6.9%) values were found in the period from May to September. The lowest levels of the wind speed in the study area were in October, and the highest were in July. The duration of sunshine hours ranged between 9.5 and 13.5 h, with an average daily sunshine of about 11 h. The effective rainfall was calculated using the CROPWAT model. Similarly, the reference evapotranspiration was calculated using the Penman–Monteith method (Equation (4)) according to Allen et al. [46]. The monthly values of evapotranspiration are presented in

Table 3. Reference ET (ET_o) of experimental farm as per the potato planting periods.

Month	Evapotranspiration (ETo)
	Early		On-Season		Late
	mm	mm d−1	mm	mm d−1	mm	mm d−1
September	245.8	12.3
October	332.4	11.1	133.0	11.1
November	211.0	7.0	154.7	5.2
December	180.5	5.8	207.2	6.7	60.2	6.7
January	71.5	7.2	221.8	7.2	150.2	4.8
February			268.2	9.6	268.2	9.6
March			190.5	10.6	328.1	10.6
April					221.9	9.6
Sum/Mean	1041.2	8.7	1175.4	8.2	1028.7	7.7

. The recorded rainfall across the experimental period was 68 mm, most of which occurred during the period between February and April. The reference evapotranspiration (ET_r) reached its maximum (12.3 mm d⁻¹) in September. The average daily ET_o was found to be 8.7 mm d⁻¹, 8.2 mm d⁻¹, and 7.7 mm d⁻¹ for early, on-season, and late plantings of potato seasons, respectively.

The total yield of potato tubers in the experimental fields ranged between 51.1 t ha⁻¹ and 55.4 t ha⁻¹. An average commercial harvest was estimated at 45.2 t ha⁻¹ for the on-season crop, followed by early- (39.6 t ha⁻¹) and late-planted (34.4 t ha⁻¹) crops. Based on the actual yield (Y_F) data and SAVI, the predicted yield (Y_P) map was generated (Figure 4) with an accuracy of 10.9 ± 2 (

Table 4. Field-measured actual (Y_F) and Sentinel-2 satellite-predicted (Y_P) potato yields (t ha⁻¹).

Potato Yield	Early		On-Season		Late
	Total	Commercial	Total	Commercial	Total	Commercial
Actual (YF)	55.4	39.6	54.1	45.2	51.1	34.4
Sentinel-2 (YP)	64.4	43.8	58.8	49.3	58.2	38.4
R2	0.65	0.67	0.71	0.69	0.63	0.67
RMSE (%)	8.0	4.6	5.9	6.0	5.4	5.2
MBE (%)	16.2	10.6	8.7	9.1	13.9	11.7

). There is a variation in the predicted yield, where the in-season potato tuber yield was 83% of the actual yield, whereas the yield of the early- and late-planted potato crops was estimated at 71% and 69% of the actual yield, respectively. Linear regression analysis was performed between the predicted and actual field-measured potato yield, and a significant correlation was found with R² values ranging between 0.63 and 0.71 and p values ranging between 0.020 and 0.036.

The crop water use of the studied potato planting periods varies between 1237 mm and 1814 mm for late- and early-planting periods (Figure 5). An approximately 17.3% increase and a 20% reduction in applied irrigation water for the early- and late-planting period, respectively, was noticed, compared to on-season plantation. The SSEB-based ET_a also showed a similar trend (

Table 5. Field-estimated and satellite image-predicted crop productivity *, crop water use, and water footprint of potato crop.

Season	Commercial Yield (CP)		Crop Water Use (mmha−1)		Water Footprint (WF, m3t−1)
	YF (tha−1)	YP (tha−1)	Field Based (WA)	Predicted (SSEB)	Field Based	Predicted (SSEB)
					WF	WFG	WFB	WFG+B
Early	39.6	48.6	1814	2308	458	14	461	475
On-season	45.2	46.9	1546	1675	342	14	343	357
Late	34.4	41.1	1237	1494	360	12	352	364
Mean	39.7	45.5	1532	1826	387	13	385	382

* Commercial yield.

).

The WF_G+B of potato fields based on plantation period varied between 357 m³ t⁻¹ for the on-season and 475 m³ t⁻¹ for the early sowing period (Figure 6). This accounts for a 24.5% (early) and a 9% (late) increase in water consumed for the production of potato compared to the on-season plantation period.

4. Discussion

The water footprint (WF) of potato crop was estimated with the use of free-of-cost satellite images, viewed by Sentinel-2 and Landsat satellites. The potato yield was predicted from S2 satellite datasets. The predicted yield, crop water use (ET_a), and WF were subjected to linear regression analysis against the actual field measurements. All the studied parameters showed significant correlations with R² values of 0.67–0.69 (Yield), 0.78–0.81 (CWU), and 0.74–0.76 (WF), as presented in

Table 6. Accuracy assessment of generated maps and field data.

#	Early			On-Season			Late
	Yield	CWU (ETa)	WF	Yield	CWU (ETa)	WF	Yield	CWU (ETa)	WF
Samples	256	256	256	168	168	168	424	424	424
R2	0.67	0.78	0.74	0.69	0.81	0.76	0.67	0.80	0.75
RMSE (%)	3.44	3.89	4.61	3.71	4.21	4.34	3.58	4.05	4.48
MBE (%)	10.9	13.7	19.6	12.6	15.9	17.9	9.2	10.8	12.9

.

On the other hand, the CWU (ET_a) was successfully determined using the SSEB algorithm with an accuracy of 89.2%. A variation of 10.6% and 12.6% was noticed for CROPWAT and the actual amount of water applied to the crop, respectively. The average total WF of potato was 684.6 m³ t⁻¹, in which the green and blue WF components were represented by approximately 8% and 92%, respectively. A comparison between the values of the blue WF, which is the major component of the WF in this study, estimated by the SSEB and those calculated on the basis of actual field data showed good agreement (MSE = 8.4%, and MBE = 12.9%). The effect of planting date on the WF estimates was also investigated in this study, where slight variations of 3.5% (late planting) to 18.5% (early planting) for a deviation of about two months from the optimum recommended planting dates were noticed. It can be concluded that estimation of the WF could be performed satisfactorily using the SSEB model for better irrigation management.

Regarding the crop growth stage, ET_f values of the potato crop were assessed against the K_c values (Figure 7). In the initial stage, ET_f values are low (0.48) for all planting periods. During the development stage, it reached a peak value of 2.1, and decreased to 0.7 at mature stages (Figure 7). The results are in agreement with published studies, such as de Sousa Junior et al. [54]. Badr et al. [55] developed K_c maps based on NDVI and TIR bands of Landsat imagery for wheat, potato, and sugar beet crops to understand the spatial and temporal distribution of K_c value in a field. Jayanthi et al. [56] and Kumar et al. [57] standardized the SSEB-based ET_c estimation as a promising method. The obtained WF values of the potato crop were compared with the published data (

Table 7. Water footprint (WF) of potato crops compared to published data.

#	WF(G+B)	Region	Reference
1	382	Present study
2	548	Saudi Arabia	[58]
3	224	WaterStatglobal	[59]
4	279	WaterStatSA	[60]
5	150	Egypt	[61]
9	286	South Africa	[62]

). The WF_a maps prepared at the field-scale level with a resolution of 10 m are useful in quantifying the actual water requirement corresponding to the different crop growth stages.

The SSEB-based ET_f values of the studied potato fields ranged between 0.3 and 1.0, which were found to coincide with the FAO-recommended K_c values. As an example, the K_c value at the initial crop growth stage (0.4 to 0.5) nearly matched the ET_f values of 0.30 to 0.42, whereas during the mid-stage of the crop growth, the ET_f reached a peak value of 1.0, which is slightly lower by 0.2 compared to the FAO-recommended value (1.2). A significant linear relationship (R² = 0.96) was observed between the ET_f and K_c values, which proves the effectiveness of using the predicted ET_f and, subsequently, the WF of the potato crop.

5. Conclusions

The Simplified Surface Energy Balance (SSEB) algorithm was successfully used to evaluate the crop water use (ET_a) for potato fields belonging to the Saudi Agricultural Development Company, Wadi-Ad-Dawasir, Saudi Arabia. NDVI and LST were computed by integrating Landsat and Sentinel-2 datasets, which in turn were used for the ET_a and WF mapping of potato crops. A comparison between satellite-based ET_a and field-based estimates showed a good correlation (R² = 0.89, RMSE = 4.4%, and MBE = 12.9%). The WF of potato in the study area was estimated at values between 475 and 357 m³ t⁻¹ for the early (September–December) and late (December–April) growing periods, respectively. Due to sparse and low rainfall, a major portion (99.2%) of the WF was accounted for by irrigation, with variations of 18.5% and 3.5% for early- and late-planted potatoes, respectively, compared to the baseline (crop planted in season). These results will help in achieving sustainable management of available groundwater resources by selecting suitable planting dates for potato crops in the study area and associated areas.