Crop Water Requirements with Changing Climate in an Arid Region of Saudi Arabia

Abstract

Agriculture is critical for a country’s population growth and economic expansion. In Saudi Arabia (SA), agriculture relies on groundwater, seasonal water, desalinated water, and recycled water due to a lack of surface water resources, a dry environment, and scanty rainfall. Estimating water consumption to plan crop water requirements (CWR) in changing environments is difficult due to a lack of micro-level data on water consumption, particularly in agricultural systems. High-resolution satellite data combined with environmental data provides a valuable tool for computing the CWR. This study aimed to estimate the CWR with a greater spatial and temporal resolution and localized field data and environmental variables. Obtaining this at the field level is appropriate, but geospatial technology can produce repeatable, time-series phenomena and align with environmental data for wider coverage regions. The CWR in the study area has been investigated through two methods: firstly, based on the high-resolution PlanetScope (PS) data, and secondly, using the FAO CROPWAT model v8.0. The analysis revealed that evapotranspiration (ET_o) showed a minimum response of 2.22 mm/day in January to a maximum of 6.13 mm/day in July, with high temperatures (42.8). The humidity reaches a peak of 51%, falling to a minimum in June of 15%. Annual CWR values (in mm) for seven crops studied in the present investigation, including date palm, wheat, citrus, maize, barley, clover, and vegetables, were 1377, 296, 964, 275, 259, 1077, 214, respectively. The monthly averaged CWR derived using PS showed a higher correlation (r = 0.83) with CROPWAT model results. The study was promising and highlighted that such analysis is decisive and can be implemented in any region by using Machine Learning and Deep Learning for in-depth insights.

1. Introduction

Due to expanding demand, population, and economic expansion, agriculture is critical in SA. Agriculture has been stressed for enhancing food security and self-sufficiency due to macro-climate change, declining rainfall (80–140 mm), limited water resources, huge aridity, and scarce cultivable lands. SA, a country with limited water resources, may face substantial consequences due to climate change and water insufficiency [1].

SA’s total runoff (2200 million m³/year) contributes to shallow groundwater inflows [2]. Water use has recently fluctuated dramatically due to population increase, industrialization, and urban development. Water use has climbed from 2352 million m³/year in 1980 to over 20,000 million m³/year in 2004 (irrigation needs > 88 percent) [3]. Groundwater accounts for 75–85 percent of the country’s water resources [4]. SA’s aquifers have been replenished at a 1.28 billion m³/year rate, whereas 394 million m³/year has been drained [5]. The shallow aquifer is renewable, with a capacity of 950 billion m³, whereas the (non-renewable) deeper aquifer has a capacity of 500,000 billion m³ [6]. The consumption of drinking water in Saudi Arabia has increased yearly with the increasing population [3]. Desalinated water represented 63% (2.14 billion m³) of the water distributed per year, while groundwater represented 37% or 1.26 billion m³. According to the report, the total water demand for different uses in 2017 and 2018 was 23.350 and 25.99 billion m³ (rises of 8%), respectively [5]. The two main environmental factors that are primarily responsible for the country’s meagre surface water resources are high aridity and low precipitation; this poses the need to improve knowledge and planning of CWR in the country.

Seasonal water and shallow/deep aquifer water supplies are essential for agriculture in arid and semi-arid regions like SA. Overexploitation of water from aquifers has severe environmental consequences and imposes pressure on the aquifer budget in several SA areas. Estimating CWR is a unique and vital question for researchers assessing efficient water use and sustainable irrigation practices, especially in rainfall deficit areas. Agriculture and water storage can be calculated and monitored using Remote Sensing (RS) and Geographic Information Systems (GIS) techniques and also using the CROPWAT tool. RS/GIS provides improved accessibility, repetition, and larger area coverage for estimating CWR. With a spatial resolution of 30 and 500 m, Landsat, and MODIS, respectively, provide good long-term vegetation monitoring. RS-based Normalized Difference Vegetation Index (NDVI) is one of the methods used to investigate vegetation in several studies. The difference in plant cover reflectance between visible and near-infrared light is described by the NDVI [6].

Earlier studies used RS-based NDVI for crop information, such as [7,8,9,10]. El-Shirbeny et al. [10] calculated the crop coefficient (Kc) = 2 × NDVI − 0.2 using Landsat data for cultivated areas in the Nile valley and delta. The major issue of this study was fixing the dynamic link between Kc and NDVI. Reyes-González et al. [11] estimated crop evapotranspiration (ETc) from Kc and NDVI using 27 Landsat images from 2013 to 2016 for northern Mexico. They used the AmeriFlux data to model the NDVI-Kc relationship. Kc values were calculated using the American Society of Civil Engineers (ASCE) Manual and adjusted for different phases of maize growth throughout the growing season. Climatic change has impacted vegetation and CWR practices, necessitating the use of updated long-term climate datasets.

The Food and Agriculture Organization (FAO) developed the CROPWAT tool to provide valuable recommendations for CWR and irrigation schedules for the crops. Many researchers utilize the CROPWAT tool to estimate CWR for sustainable agricultural practices [12]. Khan et al. [12] used CROPWAT software to study four crops in the Qaseem region, utilizing FAO climate data updated till 2013. Mahmoud and Gan [13] used MODIS NDVI and Kc in SA; however, the spatial resolution of MODIS is an issue for small pivot agriculture field estimation.

There were major limitations in earlier studies including the use of a static Kc relationship with NDVI employed by de Oliveira et al. [14], El-Shirbeny et al. [10], and Reyes-González et al. [11]; however, all months have a varied response for the diverse crops, so fixing a single relationship is error-prone. Additionally, different crops have different rates of evapotranspiration resulting in different amounts of CWR; therefore, fixing the Kc-NDVI relationship is a major gap. Previous studies such as [13,14,15] utilized either CROPWAT or the Remote Sensing approach. However, earlier studies did not discuss the link between the Remote Sensing-based CWR and the CROPWAT model-based CWR. Additionally, the present investigation reaped the benefits of a high spatial resolution of the PS dataset to precisely estimate the NDVI and CWR. The present investigation also aimed to fill the earlier gaps of static CWR estimation in SA, which lacks points coverage and has low temporal assessment. The novelty of the present study was to use high-resolution PS data based CWR estimation with environmental change assessment and the comparison with the CWR estimated using the FAO CROPWAT model.

2. Materials and Methods

2.1. Study Area

Al Majmaah city is located at 25°54′14″ N 45°20′44″ E in Riyad Province of SA (Figure 1). Al Majmaah has a total area of about 30,000 km² and a population of 97,349. The geology of the study area is diverse, including sand, clay, and limestone [16]. The climate in the region is hot and dry, with chilly winters. The climatic parameter values for the Majmaah region for the year 2021 are shown in

Table 1. Climate parameters for Majmaah region in 2021.

Mean Daily Maximum Temperature (°C)	Mean Daily Minimum Temperature (°C)	Mean Humidity (%)	Wind Speed (km/Day)	Sun Hours/Day	Mean Annual Rainfall (mm)
32.7	18.6	30	72	8.7	101.3 mm

. The Majmaah region’s main crops are dates, clover, and wheat (Figure 2).

2.2. Data Sources

PS high-resolution data, GRACE data, and climate data were used in this study. Except for the PS data, all these datasets are open to the public.

2.2.1. High-Resolution PS Data

PS reflectance imagery of 3 m spatial resolution with NIR and red bands was used to create the NDVI time series from January 2018 to December 2021 (

Table 2. List of PS Satellite images used in the current investigation.

Month	Scene ID	Spatial Resolution (m)	Month	Scene ID	Spatial Resolution (m)
18 January	20180121_065911_0f28	3	20 January	20200122_070410_0e20	3
18 February	20180218_065906_1029	3	20 February	20200219_071627_1038	3
18 March	20180327_070135_0f43	3	20 March	20200329_071506_103c	3
18 April	20180426_070208_103a	3	20 April	20200412_072711_1105	3
18 May	20180529_070354_0f4e	3	20 May	20200504_071729_103c	3
18 June	20180620_070422_102e	3	20 June	20200621_071817_1039	3
18 July	20180724_070527_102e	3	20 July	20200718_071551_1003	3
18 August	20180818_070608_101e	3	20 August	20200831_070155_0e20	3
18 September	20180920_070618_1002	3	20 September	20200929_071802_103b	3
18 October	20181013_070718_103c	3	20 October	20201031_071710_1010	3
18 November	20181129_071100_1014	3	20 November	20201125_071535_1002	3
18 December	20181211_070937_0f51	3	20 December	20201216_071648_0f17	3
19 January	20190119_071002_1029	3	21 January	20210126_045134_104e	3
19 February	20190221_071021_100c	3	21 February	20210216_071400_1009	3
19 March	20190322_071038_100c	3	21 March	20210320_071540_1013	3
19 April	20190429_071201_100e	3	21 April	20210420_071720_1008	3
19 May	20190523_071253_101f	3	21 May	20210518_071452_1005	3
19 June	20190628_071208_1035	3	21 June	20210625_071517_1011	3
19 July	20190722_070741_0e16	3	21 July	20210728_071538_100a	3
19 August	20190818_071405_1011	3	21 August	20210827_071350_1034	3
19 September	20190927_070523_0e20	3	21 September	20210927_071721_0f22	3
19 October	20191022_071451_1027	3	21 October	20211008_065131_0e0f	3
19 November	20191120_072215_1105	3	21 November	20211126_073108_44_2307	3
19 December	20191229_071744_1038	3	21 December	20211203_073208_41_227a	3

). The false color composite of PS data for 3 December 2021, is shown in Figure 1. The vegetation is readily visible in red patches, whereas the other land cover classes or no vegetation are shown in cyan color. High spatiotemporal resolution of 3 m and one day makes PS an ideal dataset to analyse near real-time precision agriculture decision-making.

2.2.2. GRACE Data

GRACE and GRACE-FO satellites provided data on terrestrial water storage (TWS) from April 2002 to April 2021 at 0.5° geographical and monthly temporal resolutions. In this study, GRACE release 06 mascon products were used. They were obtained from CSR (the University of Texas at Austin’s Center for Space Research) and Jet Propulsion Laboratories (JPL).

2.2.3. Climate Data

The monthly minimum and maximum temperature data in °C (T_min, T_max), humidity (%), wind (km/day), and rainfall dataset (in mm) of four AWS were collected from January 2018 to December 2021 (Figure 3). The sun hours data were collected from the National Renewable Energy Laboratory (NREL) website (https://www.nrel.gov/gis/solar-geospatial-data-tools.html; accessed on 10 June 2022). The high resolution of the AWS dataset’s temporal availability provides an opportunity to calculate the reference ET_o and radiation. The data is interpolated using the kriging method for monthly climate anomalies from four AWS. The CROPWAT model uses USDA and dependable rain methods to obtain effective rainfall estimates. In water-scarce areas, the USDA method is better, while the dependable rain method is better for water-sufficient areas; therefore, the monthly effective rainfall was calculated based on the USDA method [3].

2.2.4. Soil and Crop Information Data

The soil moisture data were collected using Lutron PMS-714 m from January 2021 to December 2021 between the 20th and 30th day of each month. Except for dates acquired from the FAO dataset [17], the rooting depth of each crop was evaluated using five random samples. Equation (1) was used to compute the Soil Moisture Depletion (SMD) using the FAO model.

$$SMD=0.026 RD$$

(1)

The FAO catalogue provided soil texture, maximum infiltration rate, and crop planting and harvesting dates. The mean vegetation harvesting area was determined using NDVI data from January 2018 to December 2021.

2.3. Methodology

Two methods were used to estimate the CWR in the study area: first, high-resolution PS data, and second, the FAO CROPWAT model v8.0. Figure 4 depicts the recommended methodology.

2.3.1. Calculation of Reference Evapotranspiration (ET_o)

The calculation of ET_o was the most important criterion for both techniques. Equation (2) shows the Penman–Monteith method, which was used to calculate reference evapotranspiration.

$$\frac{0.408∆\left(R_n-SH\right)+\gamma \frac{900}{T+273}w\left(e_{sv}-e_{av}\right)}{∆+\gamma \left(1+0.34w_2\right)}$$

(2)

where, SH = density of soil heat flux, T = air temperature, R_n = net radiation, w = wind speed, ∆ = slope of vapor pressure curve, e_sv − e_av = saturation vapor pressure deficit, γ = psychrometric constant.

The effective rainfall ($R_e$) was obtained from the FAO method [3] using Equation (3).

$$R_e=R_t\frac{125-0.2\left(R_t\right)}{125}$$

(3)

where $R_t$ is the total rainfall (mm). This method is valid where $R_t<250 \mathrm{mm}$ for a rainfall of Ptot < 250 mm.

2.3.2. CWR Estimation Using PS High-Resolution Remote Sensing

The ET_o calculated in Section 2.3.1 was used with PS data to calculate the CWR. The availability of 4 AWS stations in the study area with recent climatic data provided a good opportunity to calculate the CWR. To obtain the CWR based on PS data requires the NDVI and Kc estimation, which is given in the following sections.

NDVI

The NDVI index measures plant “greenness” or photosynthetic activity [15,18,19]. The NDVI method was developed on the assumption that plants absorb the red light, which is required for photosynthesis, but reflect near-infrared (NIR) light. In non-vegetated areas or stressed vegetation, the process is reversed, with red light reflected more and NIR light reflected less.

The NDVI is calculated by Equation (4) as follows:

$$\mathrm{NDVI}=\frac{NIR-RED}{NIR+RED}$$

(4)

Total vegetation cover, soil moisture, and vegetation stress are all factors that influence the NDVI. Because it is a ratio, NDVI has several advantages, including topographic lighting correction and multi-temporal image analysis for comparing images obtained at different seasons.

Estimation of Crop Coefficient (K_c)

The crop coefficients, K_c represent the crop type and the development of the crop. The K_c also relates to the crop evapotranspiration (ET_c) and ET_o ratio, which is critical when calculating ET_o. The ground-truth data for measured K_c values are required to establish an NDVI-K_c connection. From January 2018 to December 2021, the extent of crops was determined using NDVI data. Based on the approach of Kamble et al. [20] and Mahmoud and Gan [5], K_c is determined using NDVI averaged data between 2018 and 2021. Twenty sampling locations from various fields and crops were chosen to check the NDVI computed from the PS dataset. A database of default K_c was created to use in the GIS environment. The database comprised seven crop types: date palm, wheat, clover, maize, barley, citrus, and vegetables. The FAO’s database [17] contains K_c values for crops as well as dates for the growing season. K_c is calculated using NDVI time-series data and linear regression given in Equation (5).

$$K{c}_x={\alpha }_0+{\displaystyle \sum }_{y=1}^m{\alpha }_{y}NDV{I}_{x,y}+{\epsilon }_x$$

(5)

where ${\mathrm{Kc}}_{\mathrm{x}}$ is the response variable for xth case, ${\mathrm{NDVI}}_{\mathrm{x},1}$……${\mathrm{NDVI}}_{\mathrm{x},\mathrm{m}}$ were monthly NDVI and ${\mathsf{\epsilon }}_{\mathrm{x}}$ error term with mean zero value. The values ${\mathsf{\alpha }}_0$ and ${\mathsf{\alpha }}_{\mathrm{y}}$ in Equation (5) are values that were calculated using least squares regression.

The coefficients in Equation (5) were calculated using seven different crop datasets and the date of the year during the growing season. The coefficient of determination (r²) denotes the percentage of variance assigned to the linear combination of the independent variables. For the linear model stated in Equation (5), the coefficient of determination was used to determine the K_c and NDVI relationship, as shown in Equation (6).

$$r^2={\left[\frac{{{\displaystyle \sum }}_{x=1}^m\left(K_c-\overline{K_c}\right)\left(NDVI-\overline{NDVI}\right)}{{{\displaystyle \sum }}_{x=1}^m{\left[{\left(K_c-\overline{K_c}\right)}^2{\left(NDVI-\overline{NDVI}\right)}^2\right]}^{0.5}}\right]}^2$$

(6)

PS data’s high resolution (3 m) makes it an excellent alternative to computing K_c values at a high spatial resolution, which was a crucial concern in previous research that used the MODIS dataset for relatively small fields and had resolution issues [5,20].

ET_c Calculations

The K_c values and ET_o values were multiplied to get the ET_c values for the study area based on Equation (7).

$$E{T}_{c }=E{T}_o X K_c$$

(7)

The monthly $E{T}_{c }$ maps were developed from January 2018 to December 2021 to obtain the CWR for the study period.

Site Selection

Twenty crop fields were selected and extracted from NDVI maps for each month from January 2018 to December 2021 to study the pattern of NDVI, K_c, and ET_c.

2.3.3. CWR Estimation using CROPWAT 8.0 Model

Smith (1991) developed the CROPWAT software, which can calculate crop water needs, and irrigation requirements for various crops, and reference ET_o. The FAO developed the most recent version of CROPWAT v.8. It includes a sample water balance model that includes yield reduction calculations and water stress condition simulations and is based on well-established ET_c approaches [21]. CROPWAT is a helpful tool for planning irrigation under various management systems and scheme supplies and assessing irrigation application efficiency and rain-fed production. Both meteorological and crop data are required as inputs to CROPWAT. CLIMWAT is a climatic database that works with the CROPWAT to calculate crop water requirements, with climatological stations for 144 nations; however, the spatial grid size is too large for sparse and small crop areas. The local AWS data were utilized to calculate the ET_o, CWR, and irrigation water requirement (IWR). Monthly ET_o was calculated using Equation (2). The monthly ET_o values and K_c values were multiplied to get the CWR. Equation (8) was used to calculate the irrigation water requirement (IWR).

$$IWR={{\displaystyle \sum }}_{c=1}^n{A}_c\left(\mathrm{E}{T}_c-R_e\right)\times {10}^4$$

(8)

where E$T_c$ = actual crop evapotranspiration; $R_e$ = the effective rainfall; c = crop index; $\mathrm{E}{T}_a$ = actual evapotranspiration and A_c = cultivated area. The value of the effective rainfall was obtained using Equation (3).

The program includes a cropping pattern with the planting date, K_c data, growing and harvesting dates, and the area planted (0–100% of the total area). The total area of crops was derived using the NDVI model.

The soil moisture deficit threshold is required to compute an irrigation schedule. The limit of Readily Available Moisture (RAM) in the soil is the critical level when the crop gets stressed, see Equation (9).

$$RAM=TAM * F$$

(9)

where TAM is total available moisture in the soil, measured in millimeters of water per meter in soil thickness, see Equation (10).

$$TAM=\left(Root depth\right)*\left(Soil moisture capacity in mm/meters\right)$$

(10)

An ideal irrigation plan will never strain the crop, and no water will be lost due to over-irrigation. Water use will be lowered while crop yield will not be affected. The “F” threshold is the upper limit with a 40–60% typical value. Because the crop root depth increases as the season advances and the crop grows, the TAM and RAM will likewise increase.

2.3.4. GRACE Processing

GRACE is unable to distinguish anomalies from numerous TWS elements on its own. The soil moisture content (SMC) must be deducted from the change in terrestrial water storage (TWS) to calculate the change in groundwater storage (GWS) [22]. See Equation (11).

$$GWS=TWS-SMC$$

(11)

2.3.5. Trend and Correlation Analysis

Mann–Kendall trend tests were carried out for trend analysis to detect trends and changes in vegetation, GWSC, and climate variables over the years of analysis. Sen’s slope values were used to understand the trend of all parameters from January 2018 to October 2021. Equations (12)–(15) define the trend tests and slope value computation.

$$S={{\displaystyle \sum }}_{i=1}^{n-1}{{\displaystyle \sum }}_{j=i+1}^{n}Sgn\left(X_j-X_i\right)$$

(12)

$$Sgn\left(X_j-X_i\right)=\{\begin{array}{c}+1,>\left(X_j-X_i\right)\\ 0,=\left(X_j-X_i\right)\\ -1,<\left(X_j-X_i\right),\end{array}$$

(13)

$$\mathrm{VAR}\left(\mathrm{S}\right)=\frac{1}{18}\left[n\left(n-1\right)\left(2n+5\right)-{{\displaystyle \sum }}_{p=1}^q{t}_p\left(t_p-1\right)\left(2t_p+5\right)\right]$$

(14)

$$Z \{\begin{array}{c}\frac{S-1}{\sqrt{VAR\left(S\right)}} if S>0 \\ 0 if S=0\\ \frac{S+1}{\sqrt{VAR\left(S\right)}} if S<0.\end{array}$$

(15)

3. Results and Discussion

3.1. CWR Estimation Using PS High-Resolution Remote Sensing

The CWR estimated using the remote sensing method is given stepwise in NDVI, K_c, CWR, and IWR results analysis.

3.1.1. NDVI Analysis

The largest vegetation area in the observation period was recorded in December 2018, and the lowest vegetation area was detected in April 2020 (Figure 5 and Figure 6).

Vegetation area peaks were seen for 2019 and 2021 for December and January. The year 2020 had a lower annual mean vegetation area than the other three years, namely 2018, 2019, and 2021. Low vegetation is present from March 2019 to August 2020. The COVID-19 pandemic effect and reduced rainfall in 2019 were the fundamental causes of this low agricultural output. For better visualization, the NDVI images were divided into four categories: no vegetation from −1 to 0.15, low vegetation from 0.15 to 0.30, medium vegetation from 0.30 to 0.45, and high vegetation from 0.45 and above (Figure 7). The NDVI values in May 2020 were the highest of any month. For all three years of research, the NDVI of November, December, and January were more significant than other months. However, there was less vegetation in April, May, and September. The wheat cycle was shown to be highly linked with the NDVI data.

Previous studies revealed that NDVI curves were found to have low values in the early parts of the season, then steadily climbed from the beginning to the middle of the season, then remained constant until the end. The high resolution of PS data allows more exact measurements that are less influenced by pixel spectrum mixing. In SA, the date is a significant crop, particularly in Al-Majmaah; it is a perennial crop. Harvesting occurs between July and August, and the dates are highly connected with the NDVI images. The CWR for November 2018 exhibited a negative CWR for the river region (in black) with water after rainfall; thus, this RS-based CWR can capture environmental pattern changes such as rainfall and is highly connected with temperature. The NDVI curves of 20 selected fields of Al Majmaah for the observation period are shown in Figure 8. The NDVI curves for 2018 and 2021 are almost similar for all 20 fields.

3.1.2. K_c Analysis

Crop factors such as crop area and planting dates were used to create monthly K_c maps for Al Majmaah (Figure 9). Figure S1 shows the average monthly K_c for all crops for the 20 sites.

The monthly averaged K_c value curves for all crops and the monthly NDVI curves showed a similar pattern. Monthly linear regression models were developed to find the link between NDVI and averaged K_c for the selected 20 sites to quantify this relationship (Figure 10).

The R² values for monthly regression models of K_c and NDVI ranged from 0.75 to 0.94. The highest R² value of 0.94 was observed for May and July, with RMSE values of 0.003 and 0.002, respectively. The least R² value of 0.75 with an RMSE value of 0.145 was observed for Jan. The change in R² values was due to the vegetation seasons for different crops, which exhibit low relation during more vegetation and comparatively higher R² values in low vegetation time. The regression models developed in the current investigation were comparable with previous studies.

3.1.3. CWR Analysis

Monthly CWR maps for Al Majmaah were created using the monthly Kc estimated in the earlier section and ETo calculated using the environmental dataset (Figure 11). CWR maps were categorized into five categories: ≤40, 40–60, 60–80, 80–100, and ≥100. According to the CWR maps, the summer months of June, July, and August require the most water for crops, while the winter months require the least. It was clear that the higher temperature and lack of rainfall in summer months necessitated more water for crops, whereas winters get rainfall and the relatively lower temperatures mean that less water is required for crops.

In July, the CWR for the maximum area was greater than 100 (Figure 11). Al-Majmaah received very high rainfall in November 2018 This is also indicated in Figure 11, which depicts the water channel in black as a negative CWR due to water-filled channels and surrounding agricultural areas. Figure S2 shows the monthly CWR trends for 20 sites from January 2018 to October 2021. CWR peaks were recorded in all four years throughout the summer months, with less CWR in April 2020 due to less vegetation during COVID-9-related lockdowns.

3.1.4. IWR Analysis

Figure S3 depicts the monthly IWR trends for 20 sites from January 2018 to October 2021. The difference between CWR and effective rainfall was used to calculate the IWR. IWR curves followed a similar pattern to CWR curves, with peaks in the summer months and less irrigation required in the winter months.

3.2. CWR Estimation Using CROPWAT 8.0 Model

CROPWAT 8.0 was used to calculate the CWR values of seven crops for Al-Majmaah, including date palm, wheat, citrus, maize, barley, clover, and vegetables (Figure 12).

The annual CWR (in mm) for the seven crops evaluated in this study were 1377, 296, 964, 275, 259, 1077, and 214 for date palm, wheat, citrus, maize, barley, clover, and vegetables, respectively. Due to their perennial nature, date palm, citrus, and clover can be found throughout the year (Figure 12). CWR requirements were found in wheat, maize, barley, and vegetables between November and March. CROPWAT model-based CWR confirms the PS-based dataset showing the highest CWR requirements for July. In July, the date, clover, and citrus plants required 170 mm, 164 mm, and 115 mm of CWR, respectively. Figure 13 depicts a separate CWR for all seven crops.

Optimal Irrigation Schedules

Efficient use of available water resources needs management systems that adapt to changing environments in the semi-arid region of Al-Majmaah, SA. The CROPWAT model was used to calculate the best irrigation schedules for all seven crops (Figure 14). The best irrigation schedule saves water while also maximizing the irrigation rate in a timely and cost-effective manner. It recommends the best irrigation practices for a given region based on meteorological and soil factors.

3.3. Groundwater Change Using GRACE Dataset

An attempt has been made to examine GWS change using the GRACE and GRACE-FO datasets for Al-Majmaah (

Table 3. Change analysis using Kendall’s tau and Sen’s slope with (95% confidence interval and α = 0.05).

Parameter	Kendall’s Tau	Sen’s Slope
NDVI	−0.01/year	0
Kc	0.05/year	0.02
CWR	1.3 mm/year	0.38
IWR	1.1 mm/year	0.34
GWS CSR	−2.6 m/year	−0.59
GWS JPL	−0.3 m/year	−0.08
Tmax	0.038 °C/year	0.01
Tmin	0.037 °C/year	0.01
Humidity	0.12/year	0.04
Wind speed	No Change	0
ETo	0.1/year	0.28

). The CSR-based GWS trend revealed a rapid rate of water depletion. GWS based on JPL, on the other hand, showed far less depletion (Figure 15). Similar observations were made by others [15,22]. The important aspect to note is that GRACE’s spatial resolution is relatively coarse (~400 km) and has a combined multitude of parameters. Thus, it may be overestimating the values for a region such Al-Majmaah.

3.4. Trend and Correlation Analysis of CWR, K_c, Climatic Variables, and GRACE Dataset

The analysis revealed that ET_o showed a minimum response of 2.22 mm/day in January to a maximum of 6.13 mm/day in July, with high temperatures (42.8 °C). The humidity reaches its peak of 51%, then falling to a minimum in June of 15% (Figure S4). The long-term trend of rainfall and temperature for Al Majmaah from January 1981–December 2020 based on CRUTS version 4.0.5 data shows decreasing and increasing trends, respectively (Figure 16 and Figure 17).

3.5. Correlation Analysis

The effect of climatic variables such as temperature and rainfall on vegetation has been investigated (Figure 18).

All of the parameters have significant correlations with each other, with a strong correlation between them. For the combined four years, an interesting association (r = 0.81) was observed between CWR computed using PS Satellite data and CROPWAT 8.0. The CROPWAT model results have a higher correlation (r = 0.83) with the monthly averaged CWR calculated using PS (Figure 19 and Figure 20). The most recent identical climate data to estimate ET_o and the efficacy of high-resolution PS data to extract the finer vegetation characteristics could be the main reasons for this good match.

4. Comparison with Other Studies

There are five major aquifers in SA, namely Saq, Wajed, Minjur, Um Rudimah, and Wasea [22]. The Saq aquifer is one of the largest and most well-documented aquifers in the SA [23,24,25,26,27,28,29,30]. The area monitored in the present study used groundwater from Minjur aquifer for agriculture. Haq [22] observed a decreasing trend of GWS for the Minjur aquifer with a value of −5.9 + 1.5 mm/year based on JPL-GWS and CSR-GWS from 2003 to 2020. However, utilizing JPL-GWS and CSR-GWS from 2003 to 2020, the present study found a much smaller drop of −0.331.4 mm/year and −2.661.4 mm/year. It depicts the sustainable use of water for agriculture in the Majmaah region. Haq [15] observed a decreasing trend of GWS for the Qaseem region with a value of greater than 10.3 + 1.4 mm/year based on JPL-GWS and CSR-GWS from 2003 to 2020.

Previous studies [13,14,20,31] utilized either CROPWAT or the Remote Sensing approaches for CWR estimation. However, earlier studies did not discuss the link between the Remote Sensing-based CWR and the CROPWAT model-based CWR. The novelty of the present study was to use high-resolution PS data and compute the CWR based on the remote sensing method, comparing it with CROPWAT based CWR.

The regression models developed in this study were equivalent to those used in previous studies. According to Mahmoud and Gan [13], monthly R² values between Kc and NDVI ranged from 0.72 to 0.95. Mahmoud and Gan [13] employed soil characteristics, gridded water-holding capacity, and MOD13A2 data to create Kc and NDVI regression models with spatial resolutions of 40 × 40 km, 28 × 28 km, and 1 × 1 km for the Al-Qaseem region. The mean area of pivot agricultural fields in Al-Qaseem is roughly 0.5 km², which is substantially less than the dataset used by Mahmoud and Gan [13], making it subjective for precise modeling of the relationship between Kc and NDVI (Jia et al., 2021). On the other hand, to cover a greater area, coarse resolution data can be used for computation. Precision agricultural analysis using high-resolution data is very useful; however, the complexity of details was noted, especially for higher resolution datasets [32].

A well-known study [20] found that combining MODIS NDVI and K_c yielded a combined regression model with an R² value of 0.82. The combined model has a similar R² value in the current investigation. Equations (16) and (17) show the integrated model of Kamble et al. [20] with the current investigation.

$$y=1.457x-0.1725$$

(16)

$$y=1.374x+0.0933$$

(17)

where x = NDVI and y = K_c.

Kamble et al. [20] used MODIS data for NDVI and AmeriFlux data for K_c for three different locations in the USA, including desert and mountain agriculture.

Another study was performed by El-Shirbeny et al. [10]. They used Landsat 8 data and developed a regression model between Kc and NDVI as per Equation (18).

$$y=2x-0.2$$

(18)

Landsat 8’s 30 m spatial resolution makes it an excellent choice for analyzing vegetation and crop requirements for scattered and smaller fields, which are common in dry and semi-arid environments. The only problem with the Landsat data is the 16-day temporal resolution. However, the PS data used in this study has a 1-day temporal resolution and a spatial resolution of 3 m, making it a viable candidate for precision agriculture.

Another study conducted by Reyes-González et al. [11]. They used 27 Landsat 7 and 8 images acquired between 2013 and 2016 to analyze Kc and NDVI for CWR in the Northern Mexico area, as in Equation (19).

$$y=1.2358x+0.0245$$

(19)

Reyes-González et al. [11] found a good relationship between K_c and NDVI, with an R² value of 0.97. The number of fields considered for the regression model was five in total, which was a key point of subjectivity in the Reyes-González et al. [11] study. Another significant difficulty was using Landsat 7 data, which has a problem with off SLC. The normalization of this effect requires a mean filter to fill the no data values. Due to a shortage of clear sky images during the growing seasons, the NDVI curves were incomplete [11].

Aragon et al. [33] combined a high-resolution CubeSat dataset with the Priestley–Taylor’s Jet Propulsion model developed by Fisher et al. [34]. This model is based on the Penman–Monteith model, with some downscaling of evapotranspiration into the different canopy and soil subcomponents. The choice of parameters such as sensitivity β = 3, A = 0.31, k, Rn = 0.6, m² = 1, and b2 = −0.05 makes this model a little more complicated to use for calculating all parameters obtained for the global context utilizing AVHRR (1 km).

Khan et al., [12] used the CROPWAT 8.0 model to estimate CWR for five crops in the Al-Qaseem region: rice, citrus, maize, wheat, and barley. The climate data used in Khan et al. [12] is similar to that of Abbas Abdullah [35], which was taken from CLIMWAT dataset till 2011. CLIMWAT was a large worldwide climate dataset with over 5000 stations from 1971 to 2000 [21]. The climatic parameters are a dynamically changing phenomenon, and the current characteristics such as ET_o, temperature, and rainfall used from historical data from 1971–2000 are prone to subjectivity when applied to the current context. Another problem identified by Khan et al. [12] was the rice crop, which was essentially non-existent outside of the Al-Ahsa region, known for its red rice. Saudi Arabia’s rice production has decreased from 2800 tons in 1961 to 0 tons in 2019. This is due to water conservation guidelines.

The COVID-19 pandemic allows us to analyze system-wide changes in water, energy, and food (WEF) [23]. The examination of COVID-19’s impacts on the CWR showed a decrease from April 2020 due to less vegetation during COVID-9-related lockdowns.

5. Conclusions

The goal of this study was to figure out how to calculate CWR using remote sensing and the FAO’s CROPWAT model. The annual CWR (in mm) for the seven crops evaluated in this study were 1377, 296, 964, 275, 259, 1077, and 214 for date palm, wheat, citrus, maize, barley, clover, and vegetables, respectively. From January 2018 to November 2021, the current study found a slight rise in vegetation, CWR, and IWR. The increase in tree growth, which necessitates more water for date crops, is the primary cause. The GWS trend indicated a downward trend, whereas the temperature trend was upward.

The monthly averaged CWR calculated using PS and the CROPWAT model had a greater correlation (r = 0.83) in the current study. The efficiency of high-resolution PS data to get the finer vegetation details may be the fundamental explanation for this good correlation. We utilized the most recent identical climatic data to estimate ET_o and the efficiency of high-resolution PS data to get the finer details of vegetation. CWR was generated using four years of high-resolution data in this study. Furthermore, the availability of daily PS satellite data with a very high temporal resolution can provide CWR at a daily resolution, which is particularly useful for near real-time water management decision-making in sustainable agriculture. The future scope of the present investigation is to use advanced deep learning models [36,37] to predict the CWR based on high-resolution satellite data and climate data.