Spatiotemporal Assessment of Desertification in Wadi Fatimah

Abstract

Over the past four decades, Wadi Fatimah in western Saudi Arabia has undergone significant environmental changes that have contributed to desertification. High-resolution spatial and temporal analyses are essential for monitoring the extent of desertification and understanding its driving factors. This study aimed to assess the spatial distribution of desertification in Wadi Fatimah using satellite and climate data. Landsat imagery from 1984 to 2022 was employed to derive land surface temperature (LST) and assess vegetation trends using the Normalized Difference Vegetation Index (NDVI). Climate variables, including precipitation and evapotranspiration (ET), were sourced from the gridded TerraClimate dataset (1980–2022). LST estimates were validated using MOD11A2 products (2001–2022), while TerraClimate precipitation data were evaluated against observations from four local rain gauge stations: Wadi Muharam, Al-Seal Al-Kabeer, Makkah, and Baharah Al-Jadeedah. A Desertification Index (DI) was developed based on four variables: NDVI, LST, precipitation, and ET. Five regression models—ridge, lasso, elastic net, polynomial regression (degree 2), and random forest regression—were applied to evaluate the predictive capacity of these variables in explaining desertification dynamics. Among these, Random Forest and Polynomial Regression demonstrated superior predictive performance. The classification accuracy of the desertification map showed high overall accuracy and a strong Kappa coefficient. Results revealed extensive land degradation in the central and lower sub-basins of Wadi Fatimah, driven by both climatic stressors and anthropogenic pressures. LST exhibited a clear upward trend between 1984 and 2022, especially in the lower sub-basin. Precipitation and ET analysis confirmed the region’s arid climate, characterized by limited rainfall and high ET, which exacerbate vegetation stress and soil moisture deficits. Validation of LST with MOD11A2 data showed reasonable agreement, with RMSE values ranging from 2 °C to 6 °C and strong correlation coefficients across most years. Precipitation validation revealed low correlation at Al-Seal Al-Kabeer, moderate at Baharah Al-Jadeedah, and high correlations at Wadi Muharam and Makkah stations. These results highlight the importance of developing robust validation methods for gridded climate datasets, especially in data-sparse regions. Promoting sustainable land management and implementing targeted interventions are vital to mitigating desertification and preserving the environmental integrity of Wadi Fatimah.

1. Introduction

Desertification represents a complex form of land degradation occurring specifically in arid, semi-arid, and dry sub-humid regions, characterized by the reduction or loss of biological productivity and ecosystem integrity [1]. Losing vegetation cover in the drylands leads to land degradation, which results in desertification [2]. This phenomenon results from multiple interacting factors, including climate variability and anthropogenic pressures such as overgrazing, deforestation, and unsustainable agricultural practices [3,4]. Desertification affects approximately 10–20% of drylands globally [5], threatening the livelihoods of over three billion people [6]. Recently, it has been found that climate change is accelerating desertification processes through increased temperature variability and precipitation unpredictability [7,8]. The biophysical manifestations include soil erosion, salinization, vegetation degradation, and altered hydrological cycles, which collectively diminish ecosystem services and agricultural productivity [9].

The Saudi Arabian arid desert landscape provides unique challenges due to its harsh environment [10], which is vulnerable to desertification. As it is characterized by an arid climate with limited and unpredictable rainfall, the scarcity of water, combined with extreme temperatures and high evaporation rates, accelerates the drying of soils and vegetation, making the land susceptible to degradation [11]. These climate fluctuations affect plant life and contribute to soil erosion [12]. Such factors are coupled with the changes in land use and urban expansion that have occurred over the last four decades [13]. Rapid urbanization and infrastructure expansion can disrupt natural land cover, altering drainage patterns and increasing the risk of erosion. In the drylands of Saudi Arabia, valleys (wadis) serve as critical zones for vegetation growth due to their ability to retain moisture and accumulate fertile sediments. These valleys also support agricultural activities by providing more favorable soil and microclimatic conditions compared to the surrounding arid landscape.

The geographical location of Wadi Fatimah has gained significant importance due to its proximity to three major cities in the western part of Saudi Arabia: Jeddah, Makkah, and Al-Taif [14]. Wadi Fatimah basin stretches from Al-Taif in the east through Makkah in the center and terminates in the Red Sea south of Jeddah. Historically, Makkah relied heavily on the agricultural productivity of Wadi Fatimah [15]. These agricultural activities utilized water springs distributed across the basin for irrigation [14]. Subsequently, the government constructed a number of hand-dug water wells to supply water to Jeddah and Makkah for domestic purposes [16]. Individual farm owners across the Wadi Fatimah basin have established hand-dug water wells for agricultural irrigation and livestock purposes. Due to increasing demand, some farmers have also adopted the practice of selling water. Heavy trucks transport fresh water from the basin for both domestic and industrial use. These practices exert significant pressure on the natural resources of the Wadi Fatimah basin and have adverse effects on the local environment, especially for dryland. Most of the initial morphological features have undergone remarkable changes over the last four decades. These changes can be largely attributed to desertification processes, which have been triggered by shifts in anthropogenic activities within the context of the pre-existing arid conditions.

Wind erosion and abandoned agricultural lands play a significant role in accelerating desertification in Wadi Fatimah [17]. The region’s arid climate and sparse vegetation cover make the landscape highly susceptible to wind-driven soil degradation. When farms in Wadi Fatimah are abandoned, which is often due to significant water scarcity, the absence of cultivation and vegetation exposes the soil surface to the full force of prevailing winds [12]. This results in the removal of topsoil, a process that not only reduces soil fertility, but also contributes to the spread of sand dunes and the loss of arable land. Over time, these dynamics lead to permanent land degradation, reinforcing the desertification process and threatening the ecological balance of the area [17]. Sustainable land management and targeted rehabilitation efforts are essential to mitigate these effects and restore degraded areas.

Desertification can be monitored using two primary approaches: in situ field measurements and remote sensing techniques. While field-based observations offer detailed ground truth data, they are limited in their capacity to cover large areas or capture long-term trends [18]. Fieldwork is also time-consuming, labor-intensive, and costly. In contrast, remote sensing provides a practical and efficient alternative due to its broad spatial coverage, temporal consistency, and ease of data accessibility. Recent advancements in remote sensing data and techniques have significantly enhanced our ability to assess and monitor desertification and land degradation [19,20]. Satellite imagery enables the analysis of various indicators such as land use and land cover changes, vegetation health and stress, and surface condition dynamics [21]. The availability of long-term satellite datasets, such as Landsat, facilitates time-series analysis, which is critical for understanding the progression of desertification, particularly in arid and semi-arid regions [22,23,24]. Such advantages overcome the limitations of in situ data acquisition, especially in arid regions.

Among remote sensing datasets, Landsat imagery is one of the most widely used for monitoring desertification and land degradation [19]. Since its inception in 1972, Landsat has provided a continuous and reliable record of Earth observations, making it highly suitable for studying long-term environmental changes. Since desertification monitoring requires long-period evolution [25], Landsat data are the most effective Earth observation data for monitoring desertification and land degradation. Its fine spatial and temporal resolution further enhances its applicability in desertification research. Complementing Landsat data, gridded climate datasets such as TerraClimate offer consistent monthly climate information—including precipitation and evapotranspiration (ET)—across wide spatial scales. These data are useful for several hydrological and ecological applications [26], including desertification. These data offer monthly climate data for a wide range of climate variables including precipitation and ET [27]. Such datasets are particularly valuable in regions lacking ground-based meteorological observations, such as desert areas. However, analyzing large-scale geospatial datasets, especially when integrating multiple variables and resolutions, requires high-performance computing resources for effective processing and analysis.

Google Earth Engine (GEE), a cloud computing platform, enables the processing of big geospatial datasets over large areas [28]. The platform hosts a variety of geospatial data, including remote sensing imagery, digital elevation models, vector data, and weather/climate data [29]. The availability of these vast datasets, coupled with powerful processing tools and algorithms, allows users to analyze geospatial data on the cloud without downloading it locally [28]. GEE’s computing system reduces processing time through application programming interfaces (APIs) written in JavaScript and Python [30]. These capabilities enable users to leverage different geospatial datasets, especially satellite imagery, for time-series analysis to assess environmental issues such as desertification and land degradation using multi-temporal remote sensing data [31].

Integrating spatial data from different sources and spatial resolutions requires careful attention to consistency. The direct analysis of such datasets may introduce errors and uncertainties. Downscaling gridded climate data is essential for various applications [32], including the assessment of desertification. A range of statistical and dynamical methods have been employed for this purpose [33]. Among these, spatial interpolation is one of the most widely used techniques for generating high-resolution climate datasets [34]. The application of spatial interpolation methods enhances the reliability of comparisons between datasets of differing spatial resolutions, thereby supporting more accurate assessments of desertification.

Several studies have assessed desertification and land degradation in Saudi Arabia across various regions [35,36]. Many of these studies have utilized remote sensing data to evaluate desertification either across the entire country or within specific regions. For example, Qi, et al. [37] analyzed desertification across the Asian continent, including Saudi Arabia, using remote sensing data, and found that most of the Saudi Arabian territory exhibited severe desertification. Salih, et al. [38] employed the NDVI to map desertification and its severity over the Al-Ahsa Oasis in eastern Saudi Arabia. Hasan, et al. [39] investigated the severity of desertification in the Jazan Province of southwest Saudi Arabia between 2001 and 2022, using remote sensing data combined with statistical analysis. Within the Wadi Fatimah basin, Al-Mutiry, et al. [17] analyzed the causes and consequences of desertification in the middle zone of Wadi Fatimah using various satellite images, including CORONA, Landsat, and SPOT, as well as climatic data.

However, all previous studies that examined desertification in Wadi Fatimah relied either on low-spatial or temporal-resolution remote sensing data or ground-based meteorological data. Low-resolution imagery lacks the detail needed to accurately capture the spatial extent of desertification in the region. Moreover, weather stations are limited to specific parts of the Wadi Fatimah basin, and their climate records are often incomplete or unavailable for certain periods, making them unreliable for long-term desertification assessment.

The aim of this study was to monitor desertification in the Wadi Fatimah basin using Landsat imagery and gridded TerraClimate data through the Google Earth Engine (GEE) platform. The spatiotemporal characteristics of desertification in the basin were analyzed for the period between 1984 and 2022. A trend analysis was conducted using the Normalized Difference Vegetation Index (NDVI), while spatial interpolation techniques were applied to land surface temperature (LST), precipitation, and evapotranspiration (ET) data to generate time-series analyses. The specific objectives of this research were (1) to monitor the spatial extent of desertification in the Wadi Fatimah basin, and (2) to assess the desertification process by analyzing its driving factors, including climate variables and anthropogenic influences.

2. Study Area

Wadi Fatimah, situated north of Makkah City (Figure 1), experiences an arid climate characterized by mean annual temperatures ranging from approximately 25° to 38.2°. Rainfall is low, infrequent, and localized, while evaporation rates significantly surpass precipitation rates. The total area of Wadi Fatimah’s drainage basin is 5223 km², encompassing three distinct zones within its fluvial system [40] (Figure 1). The first zone, identified as the sediment production area, serves as the upstream erosion zone. This zone comprises three sub-basins: Al-Yamaniah, Horah, and Al-Shamiyah sub-basins, which cover 2750 km², representing more than half of the entire basin of Wadi Fatimah. The second zone, designated as the sediment transfer zone, is situated in the middle section of the system. This zone includes two sub-basins, Bany Omear and Allaff, with a combined area of 650 km². The Wadi Fatimah Dam marks the end of the transfer zone (Figure 1). The third zone, the depositional zone, extends from Abu Hasanyiah village to the Khomrah district in Jeddah, covering an area of 1823 km² (Figure 1). This region hosts a majority of human activities, including agriculture, residential areas, and industrial operations. Numerous villages are dispersed throughout the sub-basins, along with two cities, Al Jammum and Bahara, contributing to a total population of 184,178 as of 2022.

Geologically, Wadi Fatimah is located within the Makkah Quadrangle and contains Cambrian plutonic rocks (Figure 2). Some volcanic structures date back to the Tertiary period, while certain clastic deposits are associated with the Quaternary period. All of these geological formations are shaped by structural elements such as faults and various dykes. The Pre-Cambrian rock formations include metamorphic rocks of volcanic origin scattered with plutonic gabbro and diorite dykes. The Tertiary rock formations are characterized by volcanic lava flows that spread across the Horah sub-catchment upstream of the Wadis. These layers have significant water potential beneath the basalt. The Quaternary rock units consist of more recent sediments with a range of grain sizes, from larger grains in the upstream regions to smaller grains downstream. These sediments sit atop the primary and secondary wadis, creating conditions conducive to unconfined aquifers [41]. As stated by Sharaf, et al. [42], Wadi Fatimah is ideal for groundwater conservation due to its bedrock made up of severely eroded and fractured metamorphic and igneous rocks from the Arabian Shield. Moreover, the average transmissivity of the shallow aquifer is 140 m²/day, with porosity ranging from 14% to 30%. As reported by Es-Saeed, et al. [43] and Jamman [44], the average storativity is also 0.1.

The degree of slope greatly influences the amount of surface water that seeps into the soil. Areas with gentle slopes can hold rainwater and promote its infiltration into the ground, thereby contributing to aquifer replenishment. Steep slopes hasten runoff, which shortens the time available for water to soak into the ground, consequently lowering the potential for aquifer recharge. To understand the variations in slope with elevation, it is essential to recognize the topographical characteristics. In the examined region, slope values varied between 0° and 38°. The slopes in Wadi Fatimah primarily fell between 0° and 9.1°, while those in the highlands ranged from 9.1° to 38°. The terrain generally inclines toward the Red Sea [45].

Figure 2. A section of the geologic map of Wadi Fatimah within the Makkah quadrangle, Sheet 21D (after Moore and Al-Rehaili [46]).

Figure 2. A section of the geologic map of Wadi Fatimah within the Makkah quadrangle, Sheet 21D (after Moore and Al-Rehaili [46]).

The geomorphological landscape of Wadi Fatimah can be systematically categorized into several distinct units, each characterized by its geological composition, topographical features, and hydrological importance. The most prominent unit comprises a high mountain range, primarily formed of ancient Proterozoic rocks. These mountains serve as the main catchment area for the basin and play a vital role in shaping regional climatic cond itions. Their elevation induces an orographic effect, wherein moist air masses are forced to ascend, cool, and condense, resulting in enhanced precipitation. This orographic lifting promotes atmospheric convection and acts as a natural thermal trap. Additionally, the convergence of air masses in the lower atmosphere further amplifies rainfall in this region, thereby significantly influencing the hydrological dynamics of the Wadi system.

The second geomorphological unit comprises a mountainous zone located predominantly in the eastern and central parts of Wadi Fatimah. This area is characterized by a complex terrain of hills and ridges shaped by prolonged weathering and erosional processes. The landscape is extensively dissected, featuring numerous gullies and valleys that indicate active geomorphic dynamics. The underlying rock formations exhibit substantial evidence of both physical and chemical weathering, which contribute to sediment production and transport. These processes play a critical role in shaping the basin’s morphology and influencing its geomorphological evolution.

The third geomorphological unit is the Piedmont Plain, which serves as a transitional zone between the rugged mountainous terrain and the coastal margin of the Red Sea. This low-lying plain is characterized by broad, gently sloping surfaces that accommodate the main drainage channels of the Wadi system. It encompasses several morphotectonic depressions—structural features formed by tectonic activity—that function as important conduits for surface runoff and sediment transport. The Piedmont Plain plays a critical role in channeling water and sediments from the upland areas toward the coastal zone, thereby representing a key component of the Wadi’s fluvial system and its hydrological connectivity with the Red Sea [47].

3. Materials and Methods

3.1. Datasets

A variety of satellite data were employed in this research, including Landsat images and climatic data (

Table 1. The data used in this research.

No	Data Type	Range (Period)	Purpose	Used Bands	Provider
1	Landsat surface reflectance (SR) data	1984–2022 (All available data)	LST	Thermal infrared bands	USGS
2	Landsat Top-of-Atmosphere (TOA) data	1984–2022 (All available data)	NDVI	Near-infrared bands and red bands	USGS
3	TerraClimate	1980–2022 (Monthly)	Actual evapotranspiration	aet	University of California Merced
4	TerraClimate	1980–2022 (Monthly)	Precipitation accumulation	pr	University of California Merced
5	MOD11A2	2001–2022	LST evaluation	LST_Day_1km	USGS

). Landsat data facilitated the calculation of the LST and the trend analysis. The analysis incorporated all available data between 1984 and 2022, sourced from the United States Geological Survey (USGS). The data encompass three sensor types: Thematic Mapper (TM) for Landsat 4 and 5, Enhanced Thematic Mapper Plus (ETM+) for Landsat 7, and Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) for Landsat 8 and 9.

Landsat Surface Reflectance (SR) data were utilized for LST retrieval, while Landsat Top-of-Atmosphere (TOA) data were employed for the trend analysis. Two filters were applied to these data collections. The first filter was based on acquisition dates, aligned with the operational period of each sensor. TM images covered the study area from 1 January 1984 to 31 December 2002, and ETM+ images were available from 1 January 2002 to 31 December 2013. Lastly, OLI and TIRS data were filtered for the study area from 1 July 2013 to 31 December 2022.

The second filter was applied to the metadata of the datasets, selecting images with a cloud cover of less than 5. In total, 832 images from 1984 to 2022 (Figure 3) were prepared for LST and trend analysis. These images covered two paths, 169 and 170, and one row, 45. A scale factor was applied to the SR data using the image function. The scale factor for the red and near-infrared bands was 0.00341802 + −0.2, while the scale factor for the thermal infrared band was 0.00341802 + 149.0 (Figure 4).

The second dataset used in this research is TerraClimate [26], which provides a range of climatic elements such as actual ET, climate water deficit, precipitation accumulation, minimum and maximum temperature, vapor pressure, and wind speed at 10 m. The TerraClimate dataset is a composite of climatically aided interpolation and high-spatial-resolution climatological normals from the WorldClim dataset [26]. While the dataset has a coarser spatial resolution, it incorporates time-varying data from Climatic Research Unit (CRU) Time-Series (TS) Version 4.0 and the Japanese 55-year reanalysis (JRA55) [26].

With a high spatial resolution (1/24°, ~4 km) [26], TerraClimate produces a monthly dataset from 1958 to December 2022. In this research, only ET and precipitation data (Table 1) were used as climatic indicators. The datasets were filtered by date, covering the period from 1 January 1980 to 31 December 2022, with the AET band selected for ET and the precipitation band selected for precipitation. A total of 516 images were obtained for each climatic element, ET, and precipitation, covering a 43-year period.

3.2. Sampling and Spatial Interpolation

The data used in this research have different spatial resolutions. Since the climate data (ET and precipitation) have a spatial resolution of ~4000 m, the Landsat thermal bands have a spatial resolution of 100 m resampled to 30 m. In order to fairly compare various datasets with different spatial resolutions, these datasets need to be rescaled to a similar resolution. Therefore, sampling points were carried out to regrid the datasets within a similar spatial resolution. A set of 185 points was created randomly over the study area and imported into GEE as a feature collection. The values of each dataset were extracted to these points. Spatial interpolation was implemented on the samples using empirical Bayesian kriging (EBK). EBK is a robust method of spatial interpolation due to its ability to smooth the outliers in the filtered pixels [48,49]. It also uses many semivariogram models rather than a single semivariogram, which can reduce the errors that occur when using a single semivariogram [50].

3.3. Land Surface Temperature (LST) Retrieval and Validation

3.3.1. LST Retrieval

LST plays a crucial role in various environmental and climate studies, including the investigation of urban heat island effects, drought severity detection, climate change, energy balance, soil moisture estimation, and desertification monitoring [51,52,53,54,55,56]. It reflects the amount of radiation emitted from the Earth’s surface and can be estimated from thermal remote sensing data at different scales [57].

Various methods have been developed to retrieve the LST using Landsat thermal bands (e.g., [58,59,60,61]). Among these methods is a single-channel algorithm developed by Jiménez-Muñoz and Sobrino [62]. This algorithm is simple and utilizes only one thermal channel, making it suitable for Landsat data, particularly Landsat TM and ETM+. However, it requires prior knowledge of land emissivity and atmospheric profiles [63].

The single-channel algorithm uses atmospheric transmittance/radiance codes as input data for the atmospheric profile, correcting the radiance for residual atmospheric attenuation and emission [64]. This algorithm was applied to estimate the LST using the thermal band 6 of Landsat 5 TM, Landsat ETM+, and the thermal band 10 of Landsat 8 TIRS exclusively. The thermal band 11 of Landsat 8 TIRS was disregarded due to significant data bias and scatter [65]. Equation (1), as presented by [57,66], was used for LST estimation.

T = γ [1/ε (ψ₁ L_λ + ψ₂) + ψ₃] + δ

(1)

Planck’s function-dependent parameters γ and δ were calculated based on Equations (2) and (3) [57,66]:

γ = {c₂ L_λ/T² [λ⁴ L_λ/c₁ + 1/λ]}⁻¹

(2)

δ = γ L_λ + T

(3)

where T is the at-sensor brightness temperature, ε is the land surface emissivity, c1 (1.19104 × 108 W µm⁻⁴ m⁻² sr⁻¹) and c2 (1.438773 × 104 µm k) are the physical constants of Planck’s radiation and λ is the effective wavelength of the thermal channel (for Landsat TM, the λ is 11.457; for Landsat ETM+, the λ is 11.270; and for Landsat TIRS band 10, the λ is 10.907).

The single-channel method requires an atmospheric correction; therefore, different emission parameters should be considered when calculating the LST. The atmospheric functions ψ1, ψ2, and ψ3 are related to the water vapor content of the atmosphere [57]. The calculation of such parameters was performed using Equation (4) for Landsat TM and ETM+ [57,66]:

$$\left\{\begin{array}{c}{\psi }_1=0.14714w^2-0.15583w+1.1234\\ {\psi }_2=1.1836w^2-0.37607w-0.52894\\ {\psi }_3=0.04554w^2-1.8719w-0.39071\end{array}\right\}$$

(4)

and Equation (5) for Landsat TIRS [57]:

$$\left\{\begin{array}{c}{\psi }_1=0.04019w^2+0.02916w+1.01523\\ {\psi }_2=0.38333w^2-1.50294w-0.20324\\ {\psi }_3=0.00918w^2-1.36072w-0.27514\end{array}\right\}$$

(5)

The atmospheric water vapor data were obtained using the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) reanalysis project [67] within GEE. This is a joint project between the NCEP and the NCAR. The data are available from 1948 to the present and have a temporal resolution of 6 h and a spatial resolution of 2.5° [67].

Land surface emissivity was obtained using the NDVI. The NDVI is a combination of red and near-infrared bands and is calculated using the following equation [68]:

NDVI = (p_nir − p_red)/(p_nir + p_red)

(6)

where pnir is the surface reflectance (P) of the near-infrared band and pred is the surface reflectance of the red band.

The fractional vegetation cover (FVC) was used to differentiate between bare soil and vegetation cover [69] and calculated using the following equation:

$$\mathrm{FVC}={\left({\displaystyle \frac{\mathrm{NDVI}-{\mathrm{NDVI}}_{\min }}{{\mathrm{NDVI}}_{\max }-{\mathrm{NDVI}}_{\min }}}\right)}^2$$

(7)

where NDVImax represents the vegetation cover, and NDVImin represents the bare soil. Then, the FCV was used to calculate land surface emissivity (ε) [70,71].

The process chain in GEE began by computing the NDVI for each image in the collections. The NDVI was added to each collection as a single band. Subsequently, the maximum and minimum values of the NDVI were determined using the “ee.Number” and reducer functions. The FVC was calculated based on these maximum and minimum values of the NDVI bands. A conditional statement was then employed to compute land emissivity using the FVC bands.

Finally, the LST was retrieved using the emissivity bands, thermal bands, and the coefficients of the atmospheric functions. The coefficients of the atmospheric functions were implemented using an Earth Engine feature collection. Additional details about the codes used for LST computation directly in GEE can be found in Ermida, et al. [72], while information on the single-channel algorithm using the Python API is available in Nill, et al. [73]. The final step involved mosaicking all LST bands from 1984 to 2022 to create the LST collection.

3.3.2. LST Validation

The Terra Moderate Resolution Imaging Spectroradiometer (MODIS) Land Surface Temperature/Emissivity 8-Day (MOD11A2) Version 6.1 products were used to evaluate the LST results obtained from Landsat. The MOD11A2 products were accessed through GEE and provided by USGS. These data have 1 km spatial resolution and are collected over 8 days [74]. All LST MOD11A2 data obtained between 2001 and 2022 in the study area were used. The digital number (DN) of the LST was rescaled and converted to degrees Celsius using the following equation:

LST = DN × 0.02 − 273.15

(8)

where 0.02 is the scale factor of the LST MOD11A2 products [74]. The mean values for the LST were calculated for each year between 2001 and 2022. Then, The LST bands were mosaicked to create the LST collection. The mean values of the LST collection were extracted to the 185 points that were created over the study area. Root mean square error (RMSE) and Pearson correlation coefficient (r coefficient) were calculated to evaluate the results’ reliability. The following equation was used to calculate the RMSE:

$$\mathrm{RMSE}=\sqrt{{\displaystyle \frac{1}{n}}\sum _{\mathrm{i}=1}^n{(X_i-Y_i)}^2}$$

(9)

where X_i is the LST from Landsat, Y_i is the LST MOD11A2, and n is the number of observations. The following equation was used to calculate the r coefficient:

$$\mathit{r}={\displaystyle \frac{\sum (X_i-\overline{X}\left)\right(Y_i-\overline{Y})}{\sqrt{{\sum (X_i-\overline{X})}^2}\sqrt{{\sum (Y_i-\overline{Y})}^2}}}$$

(10)

where X_i is the LST from Landsat, $\overline{X}$ is the mean of Landsat LST, Y_i is the LST MOD11A2, and $\overline{Y}$ is the mean of MOD11A2 LST.

3.4. Precipitation and ET Evaluation

The precipitation results obtained from reanalysis climate data were compared with the meteorological station data in the study area. There are only four rain gauge stations within Wadi Fatimah: Wadi Muharam, Al-Seal Al-Kabeer, Makkah, and Baharah Al-Jadeedah. Some data were missing for certain years at some of these stations, while other data were unavailable for different years at other stations. Therefore, monthly averages were calculated for the period from 1987 to 2013, based on available data. A 15 km buffer zone was created for each station and all points were selected inside each buffer zone.

The ET data were not available from Umm Al-Joud meteorological station, which is only the comprehensive weather station within Wadi Fatimah.

3.5. Trend Analysis

The non-parametric Theil–Sen (TS) slope estimator [75,76] was used to compute the trend, and the Mann–Kendall method [77] was used to test the significance of the TS slope. The TS technique is the median of the slopes calculated between observation values (xj and xi) using pairwise time steps (tj and ti) [78].

The Mann–Kendall method was employed to evaluate the significance of the TS slope estimator. Much like the TS technique, this method is non-parametric and resistant to outliers [79]. The Mann–Kendall method was applied to the NDVI images in GEE.

3.6. Desertification Index

To quantitatively assess land degradation and desertification dynamics, a composite Desertification Index (DI) has been developed based on four key indicators: NDVI, LST, precipitation, and actual ET. As these variables have different units, all inputs were normalized using minimum and maximum normalization to ensure comparability among variables with differing units and scales [80,81]. The scales of each variable ranged from 0 to 1 using this equation:

$$X_{new}={\displaystyle \frac{x-x_{min}}{x_{max}-x_{min}}}$$

(11)

where x is the original value of the variable, x_min and x_max are its minimum and maximum values across the study period and area. This transformation preserves relative variability while eliminating unit dependence.

Following normalization, we constructed the DI as a weighted linear combination of the normalized indicators:

DI = w₁ × (1 − NDVI) + w₂ × (1 − LST) + w₃ × (1 − Precipitation) + w₄ × (1 − ET)

(12)

where wi is the weighted linear combination. Initially, equal weights wi = 0.25 were used for all components, following standard practice in multivariate index development [82].

This formulation assumes that high LST and low NDVI, precipitation, and ET are indicative of greater desertification severity. The DI was calculated for each year between 1984 and 2022. Then, the average DI was computed to create four desertification classes: severe, high, moderate, and low desertification. The desertification classification system was formulated based on the standards of previous studies [22,23,25,83,84,85].

3.7. Model Evaluation and Comparison

To assess the predictive capacity of NDVI, LST, precipitation, and ET on desertification dynamics, five regression models were evaluated: ridge, lasso, elastic net, polynomial regression (degree 2), and random forest regression. Ridge, lasso, and elastic net regression models are effective for handling multicollinearity and applying regularization, with lasso and elastic net additionally providing feature selection to enhance model interpretability [86]. Polynomial regression (degree 2) is capable of capturing non-linear trends in the data [87], while random forest regression is particularly effective for modeling complex, non-linear relationships and is highly robust to outliers and noise [88]. All models utilized the long-term mean values (1984–2022) of the four predictor variables. Model training and validation were performed using a 70/30 train-test split.

3.8. Accuracy Assessment

An accuracy assessment was conducted to evaluate the spatial distribution of the DI. A total of 499 sample points were generated across the study area using a stratified random sampling approach. High-resolution imagery from Google Earth was employed to assist in the accurate placement and classification of these sample points. The sampling design was informed by key land characteristics corresponding to each desertification class. The classification criteria were defined as follows:

• Severe desertification: Areas with no vegetation cover and extremely high LST, typically consisting of sand dunes and bare soil.
• High desertification: Areas with sparse, seasonally growing vegetation and high LST.
• Moderate desertification: Areas characterized by partial vegetation cover and moderate LST.
• Low desertification: Areas with substantial vegetation cover and relatively low LST.

This classification scheme ensured that the reference data reflected the full spectrum of land degradation conditions present in the study area. To evaluate the classification accuracy, a confusion matrix was generated, and the Kappa coefficient was calculated to quantify the agreement between the classified results and the reference data [89].

4. Results

4.1. Precipitation and Evapotranspiration (ET)

While the precipitation is generally low in the Wadi Fatimah basin, it varies between the upper, middle, and lower zones throughout the year (Figure 5). During January, February, November, and December, the average precipitation is relatively high in the middle and lower zones and low in the upper zone. From March to October, the average precipitation is low in the middle and lower zones and relatively high in the upper zone of Wadi Fatimah (Figure 5). The ET is correlated with the precipitation. Since the average ET is high in the middle and lower zones of Wadi Fatimah during the rainiest months, which are January, February, November, and December, the average ET is low in the upper zone (Figure 6), and the precipitation is low during the same period. Similarly, the average ET is high in the upper zone from March to October, and the average precipitation is also relatively high.

The average accumulated rainfall and the average accumulated ET show annual variations between 1980 and 2022 for the sub-basins of Wadi Fatimah (Figure 7 and Figure 8). In 1993, the highest accumulated rainfall was recorded in the Al Shamiyah and Al Yamaniyah sub-basins, reaching 250 mm. Conversely, 1980, 2005, and 2007 exhibited the lowest accumulated rainfall across all sub-basins. The Down sub-basin had the lowest precipitation record, whereas the Al Shamiyah and Al Yamaniyah sub-basins had the highest precipitation record. Only 2004, 2015, and 2022 demonstrated higher precipitation records compared to other sub-basins, with an average increase of approximately 110 mm in the Down sub-basin. The accumulated ET is correlated with precipitation, showing an increase in ET with increasing precipitation.

4.2. Land Surface Temperature (LST)

The middle and lower zones showed higher surface temperatures, while the upper zone indicated a lower surface temperature for all time periods (Figure 9). The average surface temperature showed that 1984, 2004, 2007, and 2010 were the hottest years. Most areas in the middle and lower zones had a high surface temperature. In contrast, the average surface temperature of the 2000 period indicated that this period experienced the lowest average surface temperature.

The average LST across the sub-basins of Wadi Fatimah exhibits annual variations between 1984 and 2022 (Figure 10). The Down sub-basin recorded the highest average LST, whereas the Al Shamiyah and Al Yamaniyah sub-basins had the lowest average LST. The highest LST records were observed in 2001, 1984, and 2013, reaching 48.5 °C, 48 °C, and 47 °C, respectively, while the lowest LST record was noted in 1985 at 29 °C. Other sub-basins, such as Horah, Bany Omear, and Allaff, showed marginal variations from 1984 to 2022.

4.3. Climate Data Evaluation

A comparison between TerraClimate precipitation data and rain gauge stations within Wadi Fatimah showed varying results across the four stations (Figure 11). The correlation between TerraClimate data and ground measurements was low (R² = −0.06) at Al-Seel Al-Kabeer station, while a moderate correlation (R² = 0.35) was observed at Baharah station. In contrast, both Makkah and Wadi Mahram stations exhibited a high correlation (R² = 0.80) between TerraClimate data and ground observations. The low correlation may be attributed to the shorter period of available weather station data (1987–2013). Additionally, missing data for certain years may have contributed to the weaker performance, particularly at Al-Seel Al-Kabeer station.

The evaluation of Landsat-derived LST showed that the highest RMSE occurred in 2001, reaching 6 °C, while the lowest RMSE was recorded in 2020 at 2 °C (

Table 2. RMSE and r coefficient values of Landsat LST and MOD11A2 products.

Year	RMSE	r Coefficient	Year	RMSE	r Coefficient
2001	6	0.7	2012	4.8	0.3
2002	2.9	0.8	2013	5.1	0.6
2003	5.5	0.8	2014	3.6	0.7
2004	3.6	0.7	2015	2.4	0.8
2005	4.1	0.5	2016	3.1	0.8
2006	3.4	0.7	2017	4	0.8
2007	3.8	0.6	2018	2.6	0.8
2008	3.9	0.7	2019	3	0.8
2009	3.3	0.6	2020	2	0.9
2010	3.5	0.8	2021	3.4	0.7
2011	3	0.7	2022	3.4	0.6

). The RMSE values for the other evaluated years ranged between 2 °C and 6 °C. The correlation coefficient (r) indicated strong agreement between Landsat LST and MOD11A2 data in most years (Table 2). However, the 2012 LST data showed a notably low r value of 0.3, whereas the 2020 data exhibited a high correlation with an r value of 0.9.

4.4. Land Use and Land Cover Changes and Land Degradation

The land features in the middle and lower zones of Wadi Fatimah underwent significant changes between 1984 and 2022 (Figure 12). The area experienced extensive urban expansion, with rural regions transforming into urban landscapes. For instance, Al Jammum, which was a small rural settlement surrounded by agricultural fields in 1984, evolved into a large urban area by 2022. Similar transformations occurred in other rural areas, converting them into urban settlements. The extent of agricultural land decreased, with many agricultural regions transitioning to either urban settlements or bare soil. In 1984, bushes were prevalent in numerous areas surrounding the main valley trunk (Figure 12a). However, by 2022, these bushes had been replaced by bare soil (Figure 12b). Furthermore, urban and industrial areas in Haddah and Bahra utilized the lower part of the valley trunk for wastewater discharge, exerting harmful impacts on the natural ecosystems. Overall, significant modifications occurred over the last four decades in the middle and lower zones of Wadi Fatimah (Figure 13). The field photographs show the drought conditions (Figure 13). The increase in surface temperatures has affected the infrastructure (Figure 12a), and wind erosion has produced a significant sand budget that has been transported into the area of sand accumulation [12] (Figure 13d,f). These changes are consequences of long-term land degradation in the region.

The spatial distribution of degradation levels shows clear changes in the upper and lower parts of the Wadi Fatimah basin (Figure 14). The downstream sub-basin exhibits significant land degradation. Most areas of the downstream sub-basin, extending from Al Khayf village northeastward to Bahra southeastward, have experienced high negative changes. Areas around Daf Khaza’ah and Al Murshidiyah also exhibit negative changes. Previously, these lands were covered by bushes and natural vegetation; these areas now show signs of high degradation, with vegetation replaced by dunes (Figure 14b). Agricultural fields in Al Khayf, Abu Urwah, Ayn Shams, and the vicinity of Al Jammum also display significant degradation (Figure 14a). Furthermore, the majority of the areas south of Bahra indicate land degradation (Figure 14c), with reduced vegetation cover replaced by sand dunes.

The areas in square kilometers and percentages for each sub-basin of Wadi Fatimah exhibit varied levels of land degradation for each land type (

Table 3. Area in square kilometers and percentage in each level of land type for Wadi Fatimah sub-basins.

Type	Horah		Al Shamiyah		Al Yamaniyah		Bany Omear		Allaff		Down
	km2	%	km2	%	km2	%	km2	%	km2	%	km2	%
High positive change	117.3	17.2	156.5	22.9	91.1	13.4	48.1	7.0	25.2	3.7	244.0	35.8
Positive change	403.0	22.1	417.9	22.9	205.1	11.2	127.7	7.0	114.1	6.3	557.2	30.5
No change	424.7	20.6	442.9	21.5	179.9	8.7	115.9	5.6	146.7	7.1	746.7	36.3
Negative change	181.0	20.1	181.7	20.2	86.0	9.6	39.9	4.4	57.5	6.4	352.2	39.2
High negative change	24.3	21.6	16.4	14.6	7.7	6.8	12.4	11.1	5.9	5.2	45.7	40.6

The percentage (%) of each type is from the entire area’s basin.

). The Down sub-basin recorded the highest degree of land degradation, with 39.2% and 40.6% for negative change and high negative change, respectively. Following closely, the Horah sub-basin showed significant degradation levels, exceeding 21%. High negative change percentages ranged between 5% and 15% in the remaining sub-basins. Positive changes or land restoration were likely associated with seasonal fluctuations observed in the valleys’ flanks and the margins of the uplands. Most of the restoration was observed in the uplands of the Al Yamaniyah and Al Shamiyah sub-basins. Additionally, certain areas within the main basin, particularly in the Down sub-basin, have recently undergone afforestation activities, which have replaced some abandoned agricultural lands.

4.5. Desertification Classification

The spatial distribution of desertification in the Wadi Fatimah basin reveals that most of the downstream sub-basin is affected by severe desertification (Figure 15). Other parts of the basin exhibit varying levels of desertification. The upstream sub-basins, such as Horah and Al Shamiyah, are characterized by lower levels of desertification, predominantly falling within the low to moderate classes. In contrast, the Bany Omear and Al Yamaniyah sub-basins show areas classified as severely desertified. Overall, the downstream sub-basin is the most affected by extreme desertification (severe and high levels) among all Wadi Fatimah sub-basins (Figure 15).

4.6. Accuracy Assessment of Desertification Classification

The accuracy assessment of the desertification classification, as presented in

Table 4. Confusion matrix and Kappa coefficient of the desertification classification.

Classified Data	Reference Data
	Severe	High	Moderate	Low	Total	User Accuracy
Severe	132	4	0	0	136	97.1
High	7	120	3	0	130	92.3
Moderate	0	5	79	12	96	82.3
Low	0	0	8	129	137	94.2
Total	139	129	90	141	499
Producer accuracy	95.0	93.0	87.8	91.5
Overall accuracy	91.5		Kappa coefficient		0.89

, demonstrates a high level of agreement between the classified data and the reference data. The overall accuracy of the classification reached 91.5%, with a Kappa coefficient of 0.89, indicating strong reliability beyond chance agreement. Among the classified categories, the severe class showed the highest user accuracy (97.1%) and a producer accuracy of 95.0%, suggesting excellent detection with minimal omission and commission errors. The high and low classes also exhibited high levels of accuracy, with user accuracies of 92.3% and 94.2%, respectively. The moderate class showed comparatively lower accuracies, with a user accuracy of 82.3% and a producer accuracy of 87.8%, likely due to its transitional nature and spectral confusion with adjacent classes. These results confirm the robustness of the classification approach in mapping desertification levels across the study area.

5. Discussion

5.1. Land Degradation in Wadi Fatimah

5.1.1. Causes of Land Degradation

The temperature in the study area is high, leading to higher rates of evaporation from the soil surface and transpiration from vegetation cover. This rapid water loss also exceeds the amount of rainfall in the study area, resulting in a net water deficit in the soil. Arid conditions dominate most of Wadi Fatimah’s areas. Limited precipitation and increased temperatures speed up soil dryness when the available water is insufficient to replenish the water loss through evaporation and transpiration processes. The LST results indicate that the lower and middle zones of Wadi Fatimah experience high surface temperatures, with most of its areas exceeding 45 °C, while the LST is slightly lower in the upper zone of Wadi Fatimah (less than 45 °C). The results of this study are similar to those of other studies that have analyzed the temperature in Wadi Fatimah. For example, Al-Harbi [90] used climatic stations within the Wadi Fatimah watershed and found that the highest temperature values were located in the middle zone of Wadi Fatimah. Similarly, Al-Al-Mutiry, et al. [17] emphasized that the middle zone of Wadi Fatimah experiences high temperatures. It is also suggested that the surface temperature trend has been remarkably increasing in recent years while the precipitation trend is declining, indicating that the area within Wadi Fatimah is vulnerable to more desertification in the future.

The middle and lower zones have undergone significant modifications over the last four decades. Urban development expanded massively between 1984 and 2022. Anthropogenic activities have shifted from rural communities to urban areas, driven by the country’s economy, which is heavily dependent on oil revenues [10]. Since the 1970s, most of Wadi Fatimah’s areas have been predominantly agricultural. However, the agricultural dominance has since shifted, with extensive built-up areas observed in 2022, especially in the lower zone of the basin, suggesting a significant land use change. The results indicate that urban areas increased significantly between 1984 and 2022, highlighting a substantial and rapid urbanization trend in the basin. Such dramatic changes can have profound effects on the local ecosystem, potentially leading to challenges such as habitat loss, altered hydrology, and increased vulnerability to land degradation in the middle and lower zones.

In addition to climatic and anthropogenic impacts, inappropriate water management has further accelerated the vulnerability to land degradation in the middle and lower zones of the Wadi Fatimah basin. These areas receive very low and infrequent rainfall, relying heavily on runoff from upstream catchments as the primary source for recharging downstream alluvial aquifers. In 1985, the Wadi Fatimah Dam was established with the purpose of annually storing 20 million cubic meters of fresh water derived from surface runoff generated by the upstream catchments [91]. The dam has effectively impeded the recharge of shallow alluvial aquifers in the lower zone of the Wadi Fatimah basin by disrupting processes linked to intermittent surface runoff from upstream catchments.

Furthermore, there has been a substantial decline in the number of water springs over the past few decades, decreasing from a few hundred in the 1920s to only five during the 1980s. This reduction may be linked to an extensive discharge strategy implemented during the 1970s to meet freshwater demands for the domestic needs of Makkah and Jeddah. Additionally, activities involving hand-dug water wells have recently increased for agricultural, residual, and industrial purposes [16,92]. These practices have significantly contributed to the reduction in shallow groundwater in the downstream area of the Wadi Fatimah basin.

5.1.2. Land Degradation Implications

Desertification in the Wadi Fatimah basin is primarily driven by the combined impacts of long-term climatic changes and intensified human activities. Between 1984 and 2022, the region witnessed a marked increase in land surface temperature, indicating a trend toward heightened aridity. This rise in temperature, along with reduced moisture availability, has likely accelerated soil degradation and vegetation stress. Concurrently, substantial land use and land cover changes have taken place, particularly the expansion of urban and built-up areas. The conversion of vegetated lands into impervious surfaces has disrupted natural land functions, reduced infiltration, and increased surface runoff and erosion. Additionally, shifts in agricultural practices, such as land abandonment, have further degraded the land and allowed for wind erosion. These synergistic pressures from both climate and human-induced factors have amplified the basin’s vulnerability to desertification.

Over the past four decades, Wadi Fatimah has experienced the significant degradation of natural vegetation, which has played a crucial role in accelerating desertification processes. The region, characterized by desert vegetation such as shrubs, grasses, and scattered perennial trees, has seen much of this cover replaced by bare land or urban development [12,93]. The decline in vegetation is a clear indicator of increasing water scarcity and persistent soil moisture deficits, which have negatively affected plant survival and regeneration [12]. As vegetated areas transition into non-vegetated land, the exposed soil becomes more vulnerable to wind erosion, leading to the mobilization of sand and the formation of extensive sand accumulations throughout the basin [12,17,94]. This loss of protective vegetation cover not only reduces biodiversity, but also weakens the ecological stability of the landscape, making it increasingly prone to further desertification.

The middle and lower zones of Wadi Fatimah are the areas most severely affected by desertification. Wind erosion has significantly contributed to the expansion of sand accumulations in these zones. The morphology of these sand deposits has changed rapidly over the past five decades, with sand dunes advancing toward fertile lands that include abandoned agricultural farms, shrubs, and scattered plant communities [17]. The increasing spatial extent of sand accumulation indicates a continuous and sufficient supply of loose sand [95], which is closely associated with rising aridity in the region. Such changes significantly affected other geomorphological features within the Wadi Fatimah basin and contributed to the desertification process.

5.1.3. Driving Factors of Desertification

The results highlight the varying contributions of NDVI, LST, precipitation, and ET to the DI, underscoring the importance of integrating vegetation, thermal, and hydrological indicators in land degradation assessment (Figure 16). Among the predictors, NDVI emerged as the most influential variable, consistent with previous studies emphasizing vegetation health as a primary indicator of desertification severity [25,93]. NDVI captures spatial and temporal changes in green biomass, which are directly impacted by degradation processes such as overgrazing, drought, and soil erosion.

LST exhibited a moderate influence, reflecting the relationship between surface heating and soil moisture availability. Elevated LST values are often associated with exposed soils and reduced ET, conditions characteristic of degraded drylands [96]. The influence of precipitation was also moderate, supporting the notion that rainfall variability, especially in arid and semi-arid regions, can trigger vegetation stress and soil instability when not followed by sufficient infiltration or retention.

ET showed the lowest importance among the four variables, which may be attributed to its complex interaction with vegetation and soil moisture. While ET is a key indicator of ecosystem function, its spatial patterns are often modulated by both biophysical (e.g., canopy cover) and meteorological (e.g., radiation) factors, making it less directly interpretable in isolation. Nevertheless, its contribution, although smaller, still reinforces the multidimensional nature of desertification processes. These results support the conclusion that vegetation health, as captured by NDVI, remains the most critical indicator of desertification in the studied region.

5.2. Remote Sensing Data and Validation

5.2.1. Remote Sensing-Based Desertification Monitoring

Desertification is a result of climate and human-induced degradation [97]. Therefore, the impact of such factors requires long-term observation over large-scale modifications. Remote sensing is able to provide essential data for monitoring and assessing land degradation and desertification, especially for long-period observation. The integration of multiple remote sensing variables, such as NDVI, LST, precipitation, and ET, provides a comprehensive approach for monitoring and assessing desertification in Wadi Fatimah. This multi-sensor approach offers a more holistic understanding of the environmental conditions contributing to land degradation. A number of studies have used remote sensing data to monitor desertification in Wadi Fatimah. For example, Al-Mutiry, et al. [17] used CORONA and SPOT data and climatic data to determine the causes and consequences of desertification in the middle zone of Wadi Fatimah. Hermas, et al. [12] utilized multi-source satellite imagery to characterize sand accumulation as an indicator of desertification in the middle zone of Wadi Fatimah. However, research using remote sensing data to measure the spatial extent of desertification is limited. This research integrated multi-source remote sensing data using the GEE platform to map the spatial extent of land degradation in the Wadi Fatimah basin.

The NDVI is a vegetation index widely used in various applications, including desertification assessment [19]. It effectively captures the spatial distribution of vegetation cover [98]. However, several factors—such as climate conditions, spatial scale, and the ecosystem type of the study area—can influence vegetation dynamics and, consequently, affect NDVI results [98]. In arid and semi-arid regions, temperature and precipitation are the most significant influencing factors [99,100]. These limitations can reduce the effectiveness of the NDVI, particularly for sub-basin level analyses in desert environments. Therefore, it is essential to consider the environmental factors that influence NDVI distribution in further analyses of the Wadi Fatimah basin.

5.2.2. Model Validation

Figure 17 presents the performance of five regression models—ridge (a), lasso (b), elastic net (c), random forest (d), and polynomial regression (degree 2) (e)—in predicting the DI derived from NDVI, LST, precipitation, and ET. Among the linear models, ridge regression (a) achieved the best performance with R² = 0.375, R² = 0.375, R² = 0.375, and RMSE = 0.0347, reflecting its capacity to manage multicollinearity among correlated predictors by shrinking coefficients without eliminating variables [101]. Lasso regression (b) and elastic net (c), which include L1 regularization, showed lower performance (R² = 0.258, R² = 0.258, R² = 0.258, and R² = 0.263 m, R² = 0.263, R² = 0.263, respectively), possibly due to over-penalization and exclusion of moderately informative predictors [102,103].

In contrast, the random forest model (d) demonstrated superior predictive performance (R² = 0.472, R² = 0.472, R² = 0.472; RMSE = 0.0319), reflecting its ability to capture nonlinear relationships and interactions among climatic and vegetation variables [104]. The polynomial regression (degree 2) model (e) also performed well (R² = 0.470, R² = 0.470, R² = 0.470; RMSE = 0.0320), indicating that incorporating interaction and quadratic terms can enhance model flexibility without relying on black-box techniques.

The scatter plots reveal an overall positive alignment between predicted and observed DI values, with more accurate predictions at mid-range values and underestimation at higher DI values—particularly evident in linear models. These results suggest that while linear models offer interpretability, nonlinear methods are more effective for modeling complex land degradation processes in arid environments [105].

5.3. Advantages and Limitations

The advantages of remote sensing data include the ability to continuously monitor and observe desertification and land degradation. The long-term data availability and consistency, such as Landsat imagery that exceeds 50 years of Earth observations, allows us to track changes and determine the causes of land degradation. The results of this study provide a systematic procedure for assessing land degradation and desertification in arid areas using Landsat imagery. The use of time-series analysis enables the detection of trends of land degradation or restoration. Such analysis provides an opportunity to predict the areas that are vulnerable to desertification. Furthermore, time-series analysis is valuable for tracking the changes and identifying the trend of land degradation or restoration. This type of analysis is useful for modeling dynamic processes [106] and predicting the future direction.

Desertification assessment requires accurate and consistent long-term climate data. In previous studies, ground-based meteorological data were heavily utilized for this purpose (e.g., [107,108]). However, the limited availability of such data presents a major challenge for desertification research that relies on meteorological stations. Gridded climate datasets, such as TerraClimate, provide a reliable alternative for monitoring desertification and other environmental applications [27]. Nevertheless, two key challenges arise when using gridded climate data. First, these datasets often require downscaling to be applicable at the regional level. Second, evaluating the accuracy of gridded data depends on the availability of high-quality ground station measurements, which are often sparse or incomplete in arid regions. Spatial interpolation techniques can help address the downscaling issue, but the evaluation of gridded datasets remains problematic due to inconsistencies and data gaps in ground-based observations. The analysis conducted in this study revealed variability in the performance of gridded precipitation data when compared to different stations in Wadi Fatimah. The reliability of the evaluation was highly dependent on the availability and completeness of ground station data, highlighting the issue of data scarcity. Therefore, developing robust methods to evaluate gridded climate data in the absence of consistent ground observations is crucial for future desertification studies in the Wadi Fatimah basin.

Desertification is a complex process influenced by various factors. Changes in climate variables, including precipitation, surface temperature, and ET, coupled with anthropogenic activities, are widely recognized as primary drivers of desertification [7,109,110,111]. However, the impact of other ecological factors, such as soil characteristics, erosion rates, moisture content, and wind velocity, cannot be overlooked.

Numerous studies have established correlations between these ecological factors and desertification processes. For example, Li, et al. [111] demonstrated that meteorological conditions and soil properties are the principal determinants of grassland desertification patterns. In another study, Han, et al. [108] found that soil type exerts a moderate influence on desertification, while factors such as evaporation, precipitation, and vegetation type have more pronounced effects.

In the present study, certain ecological variables—including soil conditions (type, erosion susceptibility, and moisture content) and wind velocity—were not incorporated into the analysis due to data limitations for the study area. Instead, the spatial trend of land degradation using time-series satellite images was the primary scope of this research. It additionally focused on elucidating the relationships between land degradation and both climatic and anthropogenic changes in Wadi Fatimah, utilizing remote sensing techniques.

It is important to note that comprehensive investigations into the intricate interplay among ecological, climate, and anthropogenic factors in driving desertification processes extend beyond the scope of this study. Such in-depth analyses would require additional resources and methodologies not employed in the current research framework.

6. Conclusions

This study has demonstrated the effectiveness of integrating multi-source remote sensing data, including Landsat imagery, precipitation, and ET data, for comprehensively monitoring and assessing desertification and land degradation processes in Wadi Fatimah, Saudi Arabia. Landsat image data from 1984 to 2022, along with gridded TerraClimate data spanning 1980 to 2022, were utilized to analyze the spatiotemporal characteristics of desertification in the basin.

The analysis of spatial extent showed that the middle and lower zones of the Wadi Fatimah basin are the most severely affected by desertification, as evidenced by increased land surface temperatures, vegetation loss, and extensive sand accumulation between 1984 and 2022. The findings underscore the fact that desertification in this region is driven by a complex interplay of climatic factors, such as rising aridity, and anthropogenic influences, including urban expansion and agricultural land abandonment. These interacting forces have significantly altered the landscape, making it more susceptible to soil degradation and wind erosion.

Wadi Fatimah represents a critical ecological and socio-economic zone in western Saudi Arabia, where increasing anthropogenic pressures and climatic stress have intensified the risk of desertification. Assessing desertification in this basin is essential due to its unique environmental sensitivity, agricultural potential, and ongoing land use changes. By monitoring the spatial extent of desertification between 1984 and 2022, this study has provided valuable insights into the temporal evolution and spatial distribution of land degradation across the region. The findings highlight areas most vulnerable to desertification, such as the middle and lower zones of the basin, which have experienced substantial vegetation loss, soil degradation, and sand accumulation. These spatial patterns can inform local land management authorities, urban planners, and environmental agencies by identifying priority areas for restoration, guiding sustainable land use planning, and supporting targeted interventions to reduce further degradation. Ultimately, the results of this study contribute to a science-based framework for combating desertification in Wadi Fatimah and can serve as a model for monitoring similar environments in arid regions.

To enhance the assessment and management of desertification, it is recommended that future studies adopt an integrated approach combining high-resolution remote sensing data, climatic datasets, and field-based validation. Continuous monitoring using time-series satellite imagery should be prioritized to detect changes early and inform mitigation strategies. Additionally, the development of spatial risk maps highlighting vulnerable zones can support targeted land management interventions. Establishing community-based awareness programs and enforcing land use regulations can further help reduce anthropogenic pressures. Such efforts are essential for promoting sustainable land use and combating desertification in Wadi Fatimah and other vulnerable arid regions within Saudi Arabia.