Integration of VIS–NIR Spectroscopy and Multivariate Technique for Soils Discrimination Under Different Land Management

Abstract

Proximal sensing has become increasingly popular due to developments in soil observation technologies and the demands of timely information gathering through contemporary methods. By utilizing the morphological, physical, and chemical characteristics of representative pedogenetic profiles established in various soils of the Sohag governorate, Egypt, the current research addresses the characterization of surface reflectance spectra and links them with the corresponding soil classification. Three primary areas were identified: recently cultivated, old cultivated, and bare soils. For morphological analysis, a total of 25 soil profiles were chosen and made visible. In the dark room, an ASD Fieldspec portable spectroradiometer (350–2500 nm) was used to measure the spectrum. Based on how similar their surface spectra were, related soils were categorized. Ward’s method served as the basis for the grouping. Despite the fact that the VIS–NIR spectra of the surface soils from various land uses have a similar reflectance shape, it is still possible to compare the soil reflectance curves and the effects of the surface soils. As a result, three groups of soil curves representing various land uses were observed. Cluster analysis was performed on the reflectance data in four ranges (350–750, 751–1150, 1151–1850, and 1851–2500 nm). The groups derived from the soil surface ranges of 350–750 nm and 751–1150 nm were not the same as those derived from the ranges of 1151–1850 nm and 1851–2500 nm. The last two categories are strikingly comparable to various land uses with marginally similar features. Based on the ranges of 1151–1850 nm and 1851–2500 nm in surface spectral data, the dendrogram effectively separated and combined the profiles into two separate clusters. These clusters matched different land uses exactly. The results can be used to promote the widespread usage of in situ hyperspectral data sets for the investigation of various soil characteristics.

1. Introduction

Agriculture and natural resources are under tremendous pressure due to mismanagement of land resources, rising population growth, and rising food consumption (OECD/FAO, 2016). Numerous detrimental effects of this evolution can be linked to the growth and intensification of agriculture [1]. Given these developments, it is imperative to improve our knowledge of soil and determine which types are most suited for farming, a task that can only be accomplished by precise mapping. Soil has a direct impact on human health and is crucial for the safety of food [2]. Differentiating and characterizing the physicochemical properties of soils is the most crucial phase. However, it has been shown that the conventional laboratory techniques for conducting soil surveys are costly, time-consuming, and require the use of specialized researchers. Although the method offers comprehensive details on particular places, it has limitations in terms of data volume, number, and spatial coverage [3]. Currently, a different approach to soil examination and quantification is visible and near-infrared (VIS–NIR) spectroscopy. According to [4], this method offers a repeatable analytical technique, is quicker, non-destructive, environmentally benign, and requires little sample preparation. A promising technique for soil analysis is soil spectroscopy, which has the ability to predict several soil parameters from a single spectral measurement, speed in acquiring data from large numbers of samples, and no need for chemicals or extractions [5]. Additionally, measurements can be carried out in a completely non-destructive manner—without contact when necessary, or with minimal intrusion in the case of taking soil cores or modifying sample presentation on-site without removing samples—and handling is straightforward and flexible thanks to modern instrumentation [6]. Hence, spectral measurements provide a quick and effective technique to determine certain characteristics of materials and objects, like soil, without causing damage or invasiveness.

Spectral reflectance studies can be used to identify changes in the physical and chemical characteristics of soil. This is useful for creating algorithms for each parameter based on the statistical correlations between the data from the remotely sensed lab and its spectral reflectance [7]. Chemometric analysis, which predicts various soil properties using the visible, near-infrared, and shortwave infrared reflectance domains, is one of the most promising techniques for quickly making conclusions about the qualities of soil. The technique is becoming more and more popular because of its capacity to do rapid, non-destructive examination with little to no sample preparation. It provides a chemical-free, real-time assessment of soil quality that may be applied immediately at a field location, thereby overcoming the limitations of wet chemistry. It produces a fingerprint of the sample in response to an infrared (IR) source that is strongly associated with its physicochemical composition [8,9,10,11]. This technique has gained popularity for a number of nutrients, including soil organic carbon, nitrogen, phosphorus, potassium, and other nutrients [11], because of its rapid predictability when combined with statistical multivariate modeling. In comparison to more labor-intensive procedures, spectroscopic techniques for evaluating soil parameters are a desirable option because they minimize laboratory effort and are quick to perform. Furthermore, spectroscopic methods offer constant high density spatial and temporal information. However, when soil spectral reflectance interferes with other targets, the precision of predictions made using spectroscopic approaches declines [7].

These soil surface reflectance spectra may be connected to conventional soil classification schemes, which are based on the morphological, physical, and chemical characteristics of typical pedogenetic profiles. By using it in both the lab and the field, measurements may be repeated and the ranges of different soil parameters, including CaCO₃, soil texture, and organic carbon content, can be estimated [12,13]. Numerous studies have attested to the great potential of VIS–NIR spectroscopy in identifying and differentiating across soil profiles, as well as offering insights into the impact of soil parameters on soil reflectance [14]. Poppiel et al. [15] discovered that, because the two methods created similar groups, the surface reflectance data were related to the soil surface and subsurface parameters found by conventional soil sample studies. The best bands for differentiating between different types of soil, according to El-Zeiny et al. [10] are the infrared bands, spectral reflectance, and statistical analysis. Additionally, for use in future applications, they created a hyperspectral library for the topsoil of several sub-great groups of soil. Lazaar et al. [16] assessed the ability of VIS–NIR spectroscopy to distinguish between the topsoil of various soil types located in the Triffa plain and to determine the impact of soil OM and CaCO₃ levels on topsoil reflectance curves. They came to the conclusion that there was a noticeable and distinct effect of soil OM content on the reflectance curves of the topsoil in the Triffa plain in the visible spectrum, specifically in the range of 580 to 750 nm. In the meantime, there was a noticeable fluctuation across the whole 350–2500 nm spectral range in the reflectance intensity curves as a function of CaCO₃ content. Using the CovSel technique, linear discriminant analysis, and spectral data from VIS–NIR spectroscopy, it is possible to distinguish between topsoil of different soil types. It was proven that the Triffa plain’s topsoil of various soil types could be distinguished using the spectral bands 1999, 686, 1280, 2340, and 1951 nm. According to Demattê et al. [17] soil reflectance is a cumulative attribute that results from the naturally diverse composition of soils. This includes differences in their chemical composition, organic matter content, particle distribution, and mineralogical constitution. Soil spectral curves can be used to illustrate this composition, which is the consequence of pedogenetic processes and causes. According to Nocita et al. [18] these signatures are kept as references in soil spectral libraries, where they can be utilized for a range of soil research. Significant connections between soil surface and subsurface were also found when the parameters of the soil were determined using reflectance spectroscopic techniques [18,19,20,21]. Reflectance spectroscopy can also be used to infer soil classification, as it can link the reflectance spectra of each horizon in a profile to the corresponding classification. Processes can be rationalized by this technique; however subsurface soil data are needed. Accordingly, a few researchers [14,22,23,24,25,26,27] revealed a close association between their surface spectra and categorized profiles. The amount of crop residue cover is a critical factor in determining the degree of soil tillage. It is closely related to managing crop systems and adapting to crop rotations, tillage practices, and harvesting methods. Based on the characteristics of their spectrum absorption, crop residues can be distinguished from other materials and vegetation by their reflectance spectra. Their main constituents, cellulose and lignin, have absorption characteristics in the 2100 to 2300 nm range [28]. It was shown that the Normalized Difference Tillage Index (NDTI) was the best of the Landsat-based tillage indices for estimating residue cover, using the difference in reflectance between the two Landsat shortwave infra-red (SWIR) bands centered at 1600 nm and 2300 nm [29].

Chemical analyses performed in the laboratory can be used to determine various soil characteristics. While this approach has the benefit of great precision, it has drawbacks such as requiring a lot of time, possibly resulting in harmful pollutants, and producing results that are delayed [30]. Recent advances in visible and near-infrared (VIS–NIR) technology have made it widely used in classification and quantitative analysis. These advances are due to the technology’s strong wavelength continuity, high spectral resolution, and nondestructive nature [31]. The overtones and combinations of fundamental molecular vibrations, such as pronounced reactions to functional groups like C=O, N=H, and O=H, can be reflected by VIS–NIR. The physical and chemical properties of soil and the composition of minerals, such as total nitrogen (TN) [32], SOM [33], soil moisture [34], organic carbon [35], soil exchangeable cations [11,36], and the soil adsorption coefficient of glyphosate [37], have been studied by many researchers with positive results [34]. Due to variations in surface cover, climate, soil physicochemical characteristics, and soil types, VIS–NIR prediction models vary by region. The accuracy of predictions made using different models and treatments for the same soil parameter can vary [38]. Although Wang and Wang [39] discovered that the limit learning machine outperformed the others in the research of lime-cured black soil. El-Sayed et al. [40] reported that PLSRs were the best model when using VIS–NIR to predict the pH and salinity in arid agricultural areas of Egypt. Comparably, for the hyperspectral prediction of the same parameters (SOM), there are frequent variations in preprocessing and modeling techniques [41,42] et al., 2022. In this context, the current work discusses surface reflectance spectrum characterization and its relationship to the corresponding soil classification using the morphological, physical, and chemical characteristics of representative pedogenetic profiles that were developed in a few soils in the Egyptian governorate of Sohag. The main objectives of this research are to: (1) characterize the surface reflectance spectra of the soils being studied; (2) relate the spectra to the corresponding morphological, physical, and chemical properties of representative pedogenetic profiles; (3) investigate the VIS–NIR spectral range’s potential for soil classification; and (4) ascertain whether cluster analysis can effectively group the soils based on spectral reflectance. Using remote sensing technologies from various sensors and platforms—both imaging and non-imaging; broadband multispectral or narrow-band hyper-spectral data acquisition—helps to understand the properties of the soil. Additionally, data was acquired from various platforms, including ground-based field and laboratory sensors, airborne sensors aboard manned and unmanned aircraft, and space-borne satellite-based sensors.

2. Materials and Methods

2.1. Description of Study Area

The research region spans 715.14 km² and is located between 26.51° and 26.9° N and 31.24° and 31.57° E (Figure 1). It is a portion of the Sohag Governorate in Egypt, which stretches from the southern border of Assiut Governorate at latitude 26°57′ N to the northern border of Qena Governorate at latitude 26°07′ N. Elevation variations are seen, ranging from roughly 0 to 425 m above mean sea level (Figure 2). The research location is located in an arid region of North Africa, which is typically defined by scorching summers and mild winters with little precipitation. The summer temperature ranges from 36.5 °C (summer) to 15.5 °C (winter), whereas the winter relative humidity ranges from 51% to 61%, the spring relative humidity ranges from 33% to 41%, and the summer relative humudity ranges from 35% to 42%. Rainfall often occurs infrequently and sporadically throughout the region. The Nile River, irrigation canals, and drains hydrologically constitute the surface water of the Sohag region. The alluvial deposits of the Nile created the Quaternary aquifer system in the research area, which is divided into two strata with different hydraulic characteristics. The clay–silt component, which is the top layer, has little permeability both vertically and horizontally. High permeability, both vertically and horizontally, is found in the major aquifer, which is the graded sand component at the bottom [43].

The research area’s soils include those from recently reclaimed land, old farmed land, and barren land. The farmed lands have loam, sandy clay, clay, and sandy loam soils. While alluvium soils have been added to the surface of freshly reclaimed soils in varying proportions to enhance their qualities, the soils of the recently reclaimed fields include sandy clay, sandy loam, clay loam, sandy, and clay. In the barren regions, sandy and sandy loam were ultimately discovered [44]. Soil conditions in the study area facilitate the transport of contaminants through the soil profile, which affects groundwater quality. The Sohag region’s economy is based primarily on agriculture. Except for populated areas, the whole surface area of the Nile Valley is utilized for agricultural purposes. Newly planted margins delineate the valley’s east and west flanks. Winter, summer, and Nilotic (fall) crops are grown in a two- or three-year crop cycle. The conventional flooding method, also known as basin irrigation, is used to apply irrigation water twice or three times every month. The main landforms observed were Nile alluvium (NA), low recent river terraces (LR), high recent river terraces (HR), and wadi bottom (WB). Soil fertility limitations and condition modifiers characterizing different units were identified. The fertility of the NA and LR soils was good with fewer limitations. Meanwhile, both HR and WB soils have many limitations that render them under either poor or very poor fertility capability [45]. Three regions were identified based on the soil quality index (SQI) of the study area: the first region, which is primarily found in old cultivated soils and some new ones, has a fair quality index; the second region is found in some newly reclaimed soils and desert soils that have low values of favorable properties, which negatively affect the SQI that is defined as poor; and the third region, which is mostly found in barren lands, has very poor quality [46].

2.2. Remote Sensing Data

For identifying and understanding the dynamic causes of land-use land-cover change (LULC) on any terrain, remotely sensed and spatial data are reliable sources of information [47]. The geocoded digital data without any cloud storage was acquired from the United States Geological Survey’s (USGS) Landsat OLI (operational land imager) sensor repository (https://earthexplorer.usgs.gov, accessed on 20 March 2024). Landsat data is stored more often than other forms of remote sensing data and has been available for a longer period of time [48,49].

Table 1. The principal specifications of the used remotely sensed Landsat satellite imagery.

SN	Satellite Image	Year of Acquisition	Path/Row	Resolution (m)	Image Type
4	Landsat 9 (OLI)	2024	175/42, 176/42 and 176/41	15–30	Level-1

OLI = operational land imager.

displays the main attributes of the image that were utilized.

2.2.1. Pre-Processing of Satellite Image

Remotely sensed data must be pre-processed to minimize noise, make the data more suitable for subsequent analysis, and increase its preparation. Initially, all spectral bands were layer-stacked in the ENVI 4.8 software [50] after the multi-temporal Landsat image was imported. In terms of the geometric correction procedure, all of the used datasets were resampled and projected with a root mean square error (RMSE) of less than 0.5 pixels onto the Universal Transverse Mercator (UTM) using the World Geodetic System 1984 (WGS84) [51,52]. Throughout the geometric rectification process, the image of interest was represented using a standard geographic coordinate system. Additionally, for geometric rectification, selecting ground control points (GCPs) is essential. Using 50 GCPs that were clearly distinguishable on satellite and ground images, the digital data for 2024 were registered (map-to-image registration). Furthermore, the pixel size for image registration was chosen at 30 m using the nearest neighborhood resampling technique and the first polynomial order methods. In addition, the closest neighborhood approach was used to resample the image while maintaining the original pixel brightness magnitudes. The temporal image needs to be radiometrically corrected to normalize brightness changes brought on by shifting atmospheric conditions before image classification can begin. Pseudo-invariant features (PIFs) are spatially well-defined, spectrally stable, and radiometrically stable features that were used to perform radiometric normalization on the image.

2.2.2. Image Classification and Accuracy Assessment

Under the ENVI 4.8 software environment, support vector machines (SVM) were used for supervised image categorization. According to Lin et al. [53], accuracy evaluation procedures determine the validity of the spatial information derived from remote sensing photos for precise image categorization. When utilized in conjunction with ground control points that act as references, remotely sensed spatial data is accurate and dependable [54]. For evaluating accuracy, a variety of statistical techniques involving the error matrix can be applied [14]. As a result, using Equations (1)–(4) [54], the correctness of the categorized image was ascertained and assessed by computing the producer’s accuracy, user’s accuracy, overall accuracy, and Kappa coefficient values.

$$A_p=p_{ii}∕p_{+i}$$

(1)

$$A_u=p_{ii}∕p_{i+}$$

(2)

$$A_o={\displaystyle \frac{{\sum }_{i=1}^m{p}_{ii}}{n}}\times 100$$

(3)

$$k={\displaystyle \frac{n{\sum }_{i=1}^m{p}_{ii}-{\sum }_{i=1}^c{p}_i+{p+}_i}{n^2-{\sum }_{i=1}^m{p}_i+{p+}_i}}$$

(4)

where: k is the kappa coefficient, A_o is the overall accuracy, A_p is the producer’s accuracy, and A_u is the accuracy user. Zones categorized and reference categories make up p_ii, while zones categorized and ground truth categories make up p_+i = p_i+ = ${\sum }_{i=1}^m{p}_{ii}$.

2.3. Field and Laboratory Activities

Using the guidelines provided in the Soil Survey Manual [55], 25 representative soil profiles (Figure 3) were chosen and exposed for morphological analysis. After inspection, soil samples were gathered and prepared for analysis. A handheld GPS device captured every profile’s location (Germin e-trix Lenexa, Kansas, MA, USA). The soil samples were allowed to air dry at room temperature; subsequently, the dried soil samples were crushed using a wooden mortar and then sieved through a 2 mm sieve. Particle size distribution [56] was conducted using sodium hexametaphosphate for dispersion in case of calcareous soils [57], calcium carbonate, cation exchange capacity (CEC) and exchangeable sodium [58], (ECe) with an EC meter (Orion, Model 150, USA); the soil pH with a pH meter was supplied from HANNA Instruments (Leighton Buzzard, UK) for organic matter content [59].

2.4. Spectroscopic Analyses

Using an ASD Spectro radiometer contact probe in a dark room, the spectral reflectance patterns of the topsoil of the samples under investigation were recorded. Two calibrated tungsten quartz halogen lamps were used in the dark to collect the VIS–NIR laboratory-based spectral measurements using the ASD FieldSpec 3 (ASD-Analytical Spectral Devices, Boulder, CO, USA), which has a spectral range of 350 to 2500 nm and a spectral resolution of 10 nm, with 1 nm intervals for all spectral ranges. Soil samples were put onto Petri dishes containing around 95 cm³ of soil. An average of 100 scans was determined by the equipment for each soil sample. A white Spectralon with 100% reflectance (LabSphere, North Sutton, NH, USA) was used to calibrate the spectroradiometer before it was scanned [60].

2.5. Grouping Soil Profiles from Surface Reflectance

Using surface spectra as a guide, comparable soils were grouped together. The relationship between soil class-specific changes in the underlying horizons and throughout the soil profile, and the patterns of the soil surface made this possible. Accordingly, as Demattê et al. [61,62,63] and Nanni et al. [23,24] have shown, it is conceivable to connect surface spectral curves with soil classes. These correlations should be carefully considered, though, as the soil surface data may not always be enough to classify the soil. Hierarchical agglomerative clustering (HCA) is a grouping approach that relies on the Euclidean distance between every observation using Statistica V10 software. Principal component analysis (PCA) and HCA had been used widely in soil studies such as [64,65,66,67]. The Euclidean distance between the groups is negatively correlated with the similarity in spectral behaviors of samples belonging to different soil classes. Using Ward’s approach [68], a variance method, clusters are created in an effort to reduce within-cluster variance. The means of all the variables are calculated for every cluster. The squared Euclidean distance to the cluster means is then determined for each object. All of the objects’ distances are added together. The two clusters that increase in the overall sum of squares, within cluster distances, the least at each stage are joined [69]. Ward’s approach was used in this study to categorize the varied spectral responses of the various soils since it provided a more accurate grouping.

2.6. Generation of Thematic Maps

Using the Inverse Distance Weighted (IDW) interpolation feature of ArcGIS 10.8 software, thematic soil maps were produced for every soil characteristic. By combining a set of sample points in a linearly weighted manner, inverse distance weighted interpolation (IDW) calculates cell values. The significance of known points on the interpolated values can be adjusted by the user using the IDW technique, taking into account their distance from the output point. Furthermore, the following Equation (5)—which uses weights of the measured surrounding points—is how the IDW predicts unknown values for every geographic point:

$$Z\left(x,y\right)={\displaystyle \frac{\sum _{i=1}^n\left(\frac{Zi}{d_{ij}^p}\right)}{\sum _{i=1}^n\left(\frac{1}{d_{ij}^p}\right)}}$$

(5)

where:

Z(x,y): is the predicted value of variable z at point i.

Zi = The measured value at point i.

d_ij: is the distance between point i and j.

n: The number of points used in the predictions.

p: power parameter.

3. Results and Discussion

3.1. Land Use/Land Cover (LULC) of the Study Area

In the study area, four LULC categories were identified: the Nile River; cultivated fields (old and recently cultivated soils); barren soils (bare lands, sand sheets, and rocky land); and urban areas (residential, commercial, industrial, and road constructions). These results are consistent with previous research [70,71,72,73,74], which discovered the same LULC categories in other regions of Egypt, including the Nile Delta and the Northwestern Coast. The data gathered for each of the four LULC groups is shown in

Table 2. LULC classification for 2024.

Land Use	2024 Areas
	km2	%
Nile River	12.3	1.72
Barren soils	226.18	31.63
Cultivated soils	403.15	56.37
Urban Areas	73.51	10.28
Total	715.14	100.0

. Additionally, the spatial distribution of LULC types (classified maps) for the analyzed date of 2024 is depicted in Figure 4. The percentages of LULC classes across the whole study area indicated that cultivated and barren soils were consistently the most extensively represented at 56.37, and 31.63% of the study area, respectively, with urban soils (10.28) and Nile River (1.72%) being far less extensive (Table 2). Urban areas had the lowest user accuracy (57.34%) among the land cover and land use classes, whereas all other units showed higher than 90%. Every class achieved a producer accuracy level above 90%. According to Cohen [75], there is nearly perfect agreement between fieldwork and remote sensing data, with an overall accuracy of 95.49% and a kappa value of 0.94 (

Table 3. Accuracy assessment of each classified image 2024.

LULC/Classes	Landsat OLI 2024
	Producer’s Accuracy (%)	User’s Accuracy (%)
Old cultivated soils	99.41	98.92
Recently cultivated soils	97.76	99.23
Barren soils	91.71	86.37
Urban Areas	96.87	57.34
Overall accuracy	95.49
Kappa coefficient	0.94

).

3.2. Soil Characteristics Within Study Area

3.2.1. Barren Soils

Table S1 provided the morphological properties of the different profiles. According to the results, moist samples ranged in color from yellowish brown (10YR 5/6) to very light brown (10YR 7/4), while dry samples ranged in color from greyish brown (10YR 5/2) to very pale brown (10YR 7/4). Less mottles, varying in size, amount, and contrast, were seen in the substratum of some profiles in comparison to older cultivated soils. Furthermore, larger-sized calcium carbonate concretions were seen, particularly in the subsurface horizons, exhibiting strong effervescence when diluted HCl was added. Every profile is well-drained and deep. The soil structure was single-grain except for the subsurface of profile 3 and all of profile 5′s layers, which were subangular blocky of extremely fine to medium size and of a weak grade. This is because the majority of these soils have low clay and organic carbon contents. The soils range in consistency from loose to highly friable (moist), soft to somewhat hard (dry), and not sticky to not plastic (wet). All profiles’ low levels of clay and organic carbon were the cause of their consistency. While the soils exhibited a sandy texture class, the subsurface of profile 3 and all of profile 5′s horizons had loamy sand and sandy loam texture classes. Every profile has a boundary that varies from distinct to gradual and from smooth to wavy in topography. Nearly every profile’s horizon displayed less growth in terms of uniformity, texture, and structure. Observations of sporadic natural vegetation have resulted in the presence of very small and sparse roots. Although these soils have not yet been farmed, they could be a potential site for farming. When compared to the soils that were previously mentioned, these soils have the coarsest components (the sandy texture class is prominent). The range of the clay content was shown by the particle size distribution, which was between 3% and 15%. Table S2 shows that the silt and sand contents ranged from 2% to 19% and 66% to 94%, respectively. The dominate soil texture is sandy (Figure 5a).

The pH range of the moderately alkaline soils was from 7.7 to 8.3 (Figure 5b). High pH soils are prevalent in semi-arid and arid regions [76]. One of the most crucial parameters in plant growth is the pH of the soil. Numerous interactions exist between soil pH, such as those involving ion exchange capacity and nutrient availability. For instance, the solubility of iron compounds decreases with rising pH; hence, high soil pH (soil alkalinity) sometimes results in an iron deficit that is detrimental to plant growth. The main effects of decreased infiltration brought on by dispersion caused by sodium are lower plant water availability, increased runoff, and soil erosion. In addition to lowering the amount of water that penetrates the soil, soil dispersion also modifies the soil’s hydraulic conductivity [77]. The rate at which water percolates through soil is known as hydraulic conductivity. Hydraulic conductivity also decreases when soil dispersion brought on by sodium leads the soil to lose structure [77]. These soils range in salinity from 6.4 to 13.4 dS m⁻¹, which is extremely high (Figure 5c). Plant productivity is restricted by salinity, a common abiotic stressor. Important plant functions are slowed down or stopped as a result of ionic toxicity and water deficiency in plants. Salinity causes an increase in the synthesis of active oxygen species and a decrease in chlorophyll, which in turn causes an accumulation of osmotic solutes and an increase in both enzymatic and nonenzymatic antioxidants [78]. Furthermore, there is relatively little organic carbon in the mixture (Figure 5d). These soils have a calcium carbonate level that ranges from 14.8% to 36.5%, making them calcic. Addition of P fertilizer to calcareous soils can cause extremely hard layers that are impervious to water and crop roots [79] (Figure 5e). Table S2 illustrates the poor cation exchange capacity, which ranges from 1.1 to 4.1 cmolc kg⁻¹. Since clay and organic matter have a substantial positive correlation with CEC, the results of CEC are dependent on the amount of these two substances in the soil (Figure 5f) [80].

3.2.2. Recently Cultivated Soils

These soils occurred in between the old cultivated and Barren soils. In addition, these soils are subjected to different reclamation processes. Hence, the soils exhibit both characteristics of the aforesaid soils. Morphological features of various profiles were given in Table S3. The data indicated that the color of dry samples varied from dark gray (10YR 4/1) to yellow (10YR 7/6) while the color of moist samples varied from very dark gray (10YR 3/1) to yellow (10YR 7/8). As compared to previously mentioned soils, abundant mottles, which differ in quantity, size and contrast, were observed in the surface and substratum of some profiles. In addition, calcium carbonate concretions in various sizes were observed that gave strong to violent effervescence with diluted HCl. The decrease of CaCO3 may be due to the effect of reclamation processes that include the addition of organic and bio-fertilizers, organic amendments, and different agricultural activities in these soils that lead them to solubilize the existing insoluble fraction of CaCO₃ and improve some of the bad properties of these soils [81,82]. All profiles are deep and well drained. Various amounts of alluvium soils and shale was added to the topsoil to enhance the soil’s characteristics and increase its productivity. The soil structure was single-grain to subangular blocky, with a very fine to coarse size and weak to strong grade. The consistency of the soil varied from soft to slightly hard (dry), loose to slightly firm (moist), not sticky to sticky and not plastic to plastic (wet). These wide variations in consistency were due to variations in agricultural activities that occurred in these soils. The soil showed wide textural variation, i.e., sandy, sandy clay loam, sandy loam, loam, and loamy sand. The boundary of all profiles is clear in distinctness and smooth to wavy in topography. Almost all horizon profiles showed less development in structure, texture, or consistency. As different crops were cultivated, some horizons of profiles contain some roots that are very fine to medium in size and few to very few in quantity. Granulometric data (Table S4) revealed that the percentage of sand, silt, and clay varied from 43% to 94%, 3% to 32% and 2% to 30%. The sandy texture dominated followed by sandy loam, loamy sand, and loam. Sandy clay loam texture class, however, occurred in the surface layers of P10 and P12. These soils are slightly saline to moderately saline and varied from 1.4 to 6.3 dS m⁻¹. The pH values ranged from 7.5 to 8.6, indicating slight to moderate alkalinity. The cation exchange capacity of these soils is moderate and varied from 8.5 to 11.6 cmolc kg⁻¹. The organic carbon content of these soils ranges from very low to moderately high in some soils that received different amounts of alluvium soils and organic amendments. These soils are calcic, and calcium carbonate content ranges from low to high, which ranges from 3.1% to 11.6% (Table S4).

3.2.3. Old Cultivated Soils

Table S5 provided the morphological properties of the different profiles. According to the results, the color of the moist samples ranged from extremely dark grey (10YR 3/1) to dark greyish brown (10YR 4/2). The color of the dry samples varied from dark yellowish brown (10YR 4/6) to brown (10YR 5/3). According to Ibáñez Asensio et al. [83], the chemical and mineralogical composition, soil texture, and moisture regime all seem to affect the color of the soil. Additionally, certain profiles’ substratum revealed some mottles that varied in size, number, and contrast. Particularly in the subsurface horizon, different-sized calcium carbonate concretions were seen. Every profile is well-drained and ranges from modest to deep. The amount of clay contained in the soil determines its structure, which can range from massive to subangular with medium size, and from moderate to strong grade. The soils range in consistency from somewhat sticky to very sticky, slightly plastic to very plastic (wet), firm to very firm (moist), and slightly hard to very hard (dry). The significant differences in consistency were caused by variances in the profiles’ clay content. A wide range of textures, including clay, clay loam, loam, sandy clay, and sandy clay loam, were present in the soils. The age of the soils, in situ weathering, translocation of clay, and parent material composition are the main causes of this large diversity in soil texture [84]. Every profile has a recognizable boundary that is topographically smooth. When compared to surface horizons, every profile’s subsurface horizon displayed development in terms of uniformity, texture, and structure. The subsurfaces met the criteria for the cambic (Bw) diagnostic subsurface horizon, as a result. The roots varied in size from extremely fine to coarse and in abundance from few to numerous in the various profile horizons. Moreover, the number of roots dropped with depth and was greater in the surface layer. Table S6 displays the specific soil particle size distribution. The data show that the percentage of clay ranged from 27% to 55%. Variability in the parent material’s in situ weathering and the vertical migration or translocation of clay may be the main causes of the rise in clay content [85]. The silt content varied from 3% to 38% and showed an uneven pattern with depth in all pedons. Generally, subterranean horizons have larger concentrations of silt and clay than surface horizons, which might be explained by the illuviation process that takes place throughout soil development. The percentage of sand varied between 20% and 67%. The dominant texture was clay, which was followed by clay loam, sandy clay, and clay loam. However, the P18 subsurface layers were of the loam texture class

Regarding chemical features, Table S6 shows that a variety of soil chemical parameters can affect soil pH, which can vary seasonally based on pedogenic processes and external inputs employed [86]. The pH of the somewhat alkaline soils ranged from 7.5 to 8.2. The data on soil EC of saturated extract ranged from 0.27 to 1.34 dSm⁻¹, suggesting that there are no salinity problems. Between 17.4 and 32.4 cmolc kg⁻¹ was the range of the cation exchange capacity in the profiles. There is low calcium carbonate present—between 0.8 and 3.6%. The distribution of calcium carbonate in the soil profile consistently displayed an increasing pattern with soil depth, suggesting that a high pH level may be the cause of calcium leaching down and subsequent precipitation at lower depths [87]. The organic carbon content of the soil was found to be low, with a range of 0.1 to 1.4%. The research area’s organic matter (OM) content is comparatively low because of the adverse effects of the dry and semi-arid climate, which include high temperatures that accelerate the pace at which organic matter decomposes in the soil [88,89].

3.3. Effects of Different Land Uses on Soil Reflectance

Different methods of land management have an impact on soil parameters including moisture content, organic matter, and organic fertilizers, among others, which in turn has an impact on the spectral fingerprints of different soils. Higher soil organic matter content and smaller particle sizes [90,91], intense mineral fertilization, increased microbial activity, and lower soil pH [92] could be the cause of the reduced reflectance of soils. Compared to soils with lower OC level, those with higher OC content are darker and have lower spectral reflectance [93]. Soil reflectance is highly influenced by OM, and it decreases as OM content rises [94]. Similarly, soils seem darker as soil moisture levels rise because incoming visible light reflectance steadily decreases with soil moisture [94,95]. Red soils have more iron oxides and less organic matter than black soils. Thus, soils with higher iron content exhibit higher reflectance than soils with lower organic carbon content [94]. Throughout the spectrum, sandy loam soils had the highest reflectance, followed by clay soils at 400–1000 nm and loam soils at 1000–2400 nm. Smaller soil particles, therefore, have a higher reflectivity. This conclusion is in line with the findings [94] which showed that clay soils had lower reflectance and sandy soils had higher reflectance. These all have an impact on the sustainability and health of the soil.

Figure 6 displays the spectra of soil sample measurements made using the ASD Fieldspec spectroradiometer. Surface soils with varying land uses often show comparable VIS–NIR spectra, with decreased reflectance in the VIS range (350–700 nm) and higher reflectance in the NIR range (701–2500 nm). That visible region’s spectra did not considerably alter in form, and the reason for this is most likely due to variations in water content. Around 1400, 1900, and 2200 nm, the water absorption from –OH bonds is apparent. According to Stenberg et al. [20], reflectance in the visible range (350–780 nm) is mostly linked to iron from minerals and soil chromophores from organic materials. Around 2200 and 2300 nm are a few tiny peaks that may be seen; these are connected to clay minerals with ratios of 1:1 and 2:1 [96,97]. The principal causes of the variations in the baselines and peak intensity were the type and primary features of the soil samples, including their texture, CaCO₃ content, and OM content [16]. Upon initial glance at Figure 6, it was easy to compare the soil reflectance curves and assess the impact of the surface soils. As a result, three sets of curves were observed.

The first group (Figure 7) is made up of eight profiles: P1, P2, P3, P4, P5, P6, P7, and P8. These profiles are connected with desert and less developed soils, and they had higher intensity reflectance than other soil types, with a maximum of 0.56 in the NIR region at 2135 nm.

The soil types in this zone are characterized by four factors: (1) a high content of calcium carbonates due to their development from calcareous rocks, which facilitated the enrichment of the soil with calcium carbonates; (2) coarse fractions in the textures (dominantly sandy texture class); (3) high salinity level; and (4) low organic matter (OM) content. The latter may be attributed to reduced agricultural activities and the high content of calcium carbonate (CaCO₃), which slows down the processes of alteration and thus accelerates the decomposition of soil OM and prevents the process of humification [16]. It is important to note that there was a strong correlation between the predominant salt type found in the soils and their reflectivity. The soils with the highest levels of red and near-infrared reflectance are those that have white salt crust. Dark hydrated salts work to absorb water molecules from the atmosphere to the chemical composition of the salt, making it appear darker. In addition, salts have the ability to form complexes with organic compounds that were already present in the soil, which is why the lowest reflectance values were found in these soils. Because of this, the soil becomes darker, and the soil surface absorbs the majority of incident light on the soil. On the other hand, white salt crusts like NaCl, MgCl₂, and Na₂SO₄ have a pale appearance and function to reflect incident light, increasing their reflectance in comparison to absorbed light [98], which causes the soil’s surface to absorb the majority of incident light.

Between the first and third groups and comprising nine profiles taken from recently cultivated soils (P9, P10, P11, P12, P13, P14, P15, P16, and P17), is the second group (Figure 8). At 1805 nm, these soils’ moderate intensity reflectance peaked at 0.43 in the near-infrared spectrum. The results show that these soils’ moderate reflectance is caused by their medium soil texture, higher organic carbon content, and lower CaCO₃ concentration than arid soils. The reflectance curves of several topsoils revealed that those with high OM content had low reflectance intensities over the visible spectrum, and if their OM level was more than 2%, they might have masked other soil elements. Nonetheless, the high-intensity reflection in the VIS band was a characteristic of the topsoils with low OM concentration. Furthermore, the change in topsoil reflectance with respect to CaCO₃ content demonstrated that all spectral domains between 350 and 2500 nm had higher reflectance in the reflectance curves of the topsoil with higher CaCO₃ content. On the other hand, reflectance decreased when topsoil CaCO₃ content was low [98,99]. According to Asgari et al. [60], the group with the highest carbonate content showed the highest reflectance intensity. They stated that the incidence of soil carbonates was indicated by a specific absorption wavelength of 2338 nm, which increased with the number of carbonates present. This was consistent with the concave feature’s ascent. This wavelength was utilized by Rossel et al. [100] and Shahrayini et al. [101] as the carbonate’s unique characteristic.

The third group (Figure 9) is linked to eight profiles (P18, P19, P20, P21, P22, P23, P24, and P25), one of which is an alluvial plain made of historically farmed agricultural soil. These soils displayed reduced intensity reflectance, peaking at 0.37 in the NIR region at 2136 nm, due to their higher concentration of OM and clay and lower CaCO3 content. In actuality, samples with lower OM contents typically had greater reflectance intensity than those with higher OM contents, and the reflectance of the soil decreased as OM concentration increased. Two absorption characteristics at 850–900 and 2320–2370 nm that are associated with OM were identified by Asgari et al. [60] in their investigation.

3.4. Grouping Soils Based on Surface Spectra

Based on the Euclidean distance (Figure 10a–d), cluster analysis was applied to the reflectance data in four ranges (350–750, 751–1150, 1151–1850, and 1851–2500 nm). The main objective is to observe the best range that discriminates between different soils and group them based on surface reflectance. The groups obtained from the ranges (350–750 nm and 751–1150 nm) of the soil surface differed from those based on the ranges 1151–1850 nm and 1851–2500 nm). The latter two ranges are very similar to different land uses that are slightly similar in their characteristics. The cluster based on the ranges (650–750 nm) failed in correctly grouping the soils profiles as presented in Figure 10a. On the other hand, the clustering based on the 751–1150 nm range is far superior to the previous one. Only profile 144 (which belongs to recently cultivated soils but was clustered under old cultivated soils in the clustering analysis) was not successfully placed in its proper category by this cluster analysis (Figure 10b).

The dendrogram obtained from surface spectral data based on the ranges (1151–1850 nm and 1851–2500 nm) successfully classified and merged the profiles into two distinct clusters. These clusters completely corresponded to various land uses, e.g., old cultivated, recently cultivated soils, and desert soils. The profiles from 1 to 8 that correspond to barren soils make up the single cluster that makes up the first cluster. In contrast, the second cluster is made up of two subclusters, one of which included profiles P6 through P17, that were gathered from recently cultivated soils. The profiles P18–P125 that were taken from old cultivated soils are linked to the second subcluster. The various land uses are well-aligned with the analysis results. Given that the two methods produced comparable groupings, it is notable that the surface reflectance data were connected to the soil surface characteristics identified by the conventional analysis of the soil samples.

Spectroscopic data of the soil surface horizons can be used to classify individual soils with an accuracy of more than 75%, according to Zeng et al. [14,102]. This is consistent with Xie et al. [103], who found that surface reflectance could identify between five soil classifications with 92.2% accuracy. Nanni et al. [24] reported that an overall accuracy of 52–70% was achieved in the discrimination of soil classes using spectral information of the soil surface as indicators. Furthermore, using soil surface reflectance as a basis [15] effectively clustered oxisols, entisols, and incisols. Since the two methods created similar groups, the surface reflectance data were connected to the subsurface and surface attributes of the soil, as determined by conventional examinations of soil samples.

4. Conclusions

Accurate mapping is the only way to address the pressing need to better understand soil and determine which types are most suited for agriculture. Differentiating and characterizing the physicochemical properties of soils is the most crucial phase. However, it has been shown that the conventional laboratory techniques for conducting soil surveys are costly, time-consuming, and require the use of specialized researchers. Currently, a different approach to soil examination and quantification is visible and near-infrared (VIS–NIR) spectroscopy. Three groups of surface soils representing various land uses were studied, as the study demonstrated that the VIS–NIR spectra of the various land uses produced distinct reflectance curves. The eight profiles that comprise the first group are P1, P2, P3, P4, P5, P6, P7, and P8. These profiles were associated with less developed and bare soils; they also exhibited higher intensity reflectance (maximum of 0.55 in the NIR region at 2155 nm) than other soil types. The second group consists of nine profiles (P9, P10, P11, P12, P13, P14, P15, P16, and P17) derived from recently cultivated soils and is situated between the first and third groups. In the near-infrared spectrum, the moderate intensity reflectance of these soils peaked at 0.42 at 2137 nm. Finally, eight profiles (P18, P19, P20, P21, P22, P23, P24, and P25) are associated with the third group, one of which is an alluvial plain composed of agricultural soil that has historically been cultivated. Reduced intensity reflectance was seen in these soils, with a peak at 0.36 in the near-infrared spectrum at 2132 nm. Additionally, the outcomes amply demonstrated the usefulness and efficiency of the clustering strategy. Furthermore, it proved effective to identify soil groups with comparable characteristics by clustering data based on soil surface reflectance. The profiles were effectively categorized and combined into two separate clusters using the dendrogram created from surface spectral data based on the ranges (1151–1850 nm and 1851–2500 nm). The different land uses, such as recently cultivated soils, old cultivated soils, and barren soils, were all fully correlated with these clusters. Meanwhile, information pertinent to soil genesis may be ignored by soil spectral data. Understanding and differentiating between the true reasons of the trend of changing soil properties can be facilitated by knowing if the parent material’s homology or heterogeneity is evident in the study of soil genesis. Parent material discontinuity is universal for the soil genesis process. Determining the discontinuity of parent material is not simple, though. Distinguishing the discontinuity of parent material becomes simpler when there are noticeable variations in the soil texture and structure in the profile morphology. But, if the profile features are all the same, it is hard to spot. Overall, due to its limitations, soil spectroscopy cannot fully replace conventional soil classification, but it has a high potential to aid soil classification once the soil spectral libraries are well established. It also should be noted that the prediction of soil classification based on soil spectra still remains an academic research topic and more efforts should be pursued for its practical use in decision making for farmers and stakeholders. More investigations are still needed on this subject in the future and recent work will build upon this research with the addition of spatial environmental variables and the incorporation of stratification into the modeling framework.