Application of Ground-Penetrating Radar (GPR) for Evaluating the Amelioration of Saline–Alkali Soils in the Yellow River Delta

Abstract

Ground-penetrating radar (GPR) was utilized for subsurface soil investigation in the Yellow River Delta, aiming to provide a scientific basis for the remediation performance of saline soils. The study particularly focuses on the red clay layer, a typical and characteristic soil horizon in this region. GPR antennas with central frequencies of 400 MHz and 900 MHz were adopted to investigate shallow soils within 1 m of the ground surface across three experimental plots (pits, undisturbed soils, and tilled soils) and 18 scattered measurement sites, followed by systematic analysis and interpretation of the acquired GPR profiles. During data acquisition, reasonable survey lines were deployed across the patchy bare areas of cultivated lands covering the experimental plots and measurement points to collect raw GPR data. Meanwhile, subsurface soil data were collected via test pits and borehole sampling along the survey lines. Raw GPR data were further preprocessed and postprocessed to characterize soil horizons and interpret subsurface stratigraphic structures. Finally, the correlations between the relative dielectric permittivity, reflection coefficient, and reflected wave amplitude of each soil layer were systematically analyzed. The results demonstrate that the 400 MHz antenna enables effective identification of soil layers within 1 m depth, while the 900 MHz antenna provides high-resolution detection for soil layers above 0.5 m. The red clay layer presents a distinct strong-amplitude reflection on GPR profiles, and the average relative dielectric permittivity of soils across the study area reaches 30.57. GPR profiles reveal that soil horizons with an absolute reflection coefficient greater than 0.01 yield detectable continuous reflection signals and allow uninterrupted stratigraphic interpretation. An empirical formula was established to calculate soil relative dielectric permittivity from soil moisture content, with a correlation coefficient of 0.9173. However, this formula ignores the influences of soil salinity and other trace soil elements. This study realizes rapid and accurate characterization of the depth and thickness of shallow soil layers, providing technical support for soil remediation of saline–alkali land in the Yellow River Delta. The findings also provide a valuable reference for evaluating the remediation effects, optimizing arable land utilization, preventing and mitigating soil salinization risks, and promoting the sustainable economic development of the study area.

1. Introduction

The modern Yellow River Delta was gradually formed by sedimentation following several minor channel migrations after the Yellow River shifted its course in 1855 [1]. Under the combined influences of sedimentation, erosion, and human disturbance, the land area first increased, subsequently declined, and has maintained a slow growth trend up to the present [2,3]. The development of saline–alkali soils across the Yellow River Delta is driven by a combination of natural and anthropogenic factors, mainly including saline and alkaline constituents in Yellow River sediments [4], seawater intrusion [5], atmospheric precipitation [6], Yellow River irrigation [7], unsuitable cultivation practices [8], and poorly drained terrain [9]. Spatially heterogeneous salt distribution in farmlands induces physiological drought and vegetation degradation when soil salt concentrations exceed plant tolerance limits, ultimately forming patchy bare land areas [10]. Although the total area of saline–alkali land in the Yellow River Delta has exhibited a decreasing trend over the past two decades [11], it still severely restricts local socioeconomic development. Soil productivity deterioration caused by soil salinization and alkalization underscores the urgent necessity of targeted soil amelioration in this region.

Many scholars have conducted extensive studies on saline–alkali soils in the Yellow River Delta. Soft soils are widely distributed in this region, among which clayey soils, silt, and silty soils in the northern part are characterized by high water content, high compressibility, low shear strength, and low permeability [12]. Seawater intrusion constitutes the primary driver of soil salinization in this region, with salt content reaching up to 4.59 g/kg. The local saline soil is predominantly classified as Cl⁻ and SO₄²⁻-Cl⁻, and is mainly characterized by moderate salinization, while severely and extremely severely salinized soils are primarily distributed in the northern and eastern coastal lowlands [13]. The study area features low soil organic matter and substantial spatial variability in soil salinity. Salt distribution exhibits two distinct patterns—surface-accumulated in wastelands and areas covered by natural vegetation, and uniformly distributed salt across farmlands and woodlands—while subsurface salt accumulation primarily takes place during the rainy summer months. Driven by rainfall and evaporation, the salinity and pH of shallow topsoil display pronounced seasonal fluctuations [14]. A series of agricultural sustainable countermeasures, including soil amelioration, crop variety optimization, water resource regulation, and ecological restoration measures, have been put forward for this region [15]. Currently, the primary amelioration techniques proposed for regional saline–alkali soils fall into four categories: physical measures (salt drainage, salt leaching, salt suppression, deep tillage, subsoiling and ridging tillage), chemical amendments (application of chemical conditioners), biological approaches (salt-tolerant vegetation, functional microorganisms), and integrated remediation strategies [16]. BP neural network projections indicate that regional soil salinization will intensify over the next 50 years after 2050 [17]. Long-term dynamic monitoring reveals that groundwater salinity variations are controlled by the coupled effects of irrigation and precipitation, following an initial rise-and-subsequent decline trend. The groundwater table generally remains shallower than 1.2 m between July and September [18]. Seawater intrusion serves as the primary source of groundwater salinity in this region. The concentrations of Cl⁻, F⁻, Na⁺, Mg²⁺, Ca²⁺, and SO₄²⁻ ions generally decline gradually from coastal zones to inland areas [19]. Deep tillage disrupts soil capillary pores, suppresses upward soil water migration, reduces surface evaporation, and thereby effectively inhibits secondary soil salinization [20]. Following saline–alkali soil remediation, scientific evaluation methods are essential to quantify the effectiveness of soil improvement.

GPR is a non-destructive geophysical technique that employs high-frequency electromagnetic waves to detect subsurface media. Featuring simple operation, high precision, extensive detection coverage, and high efficiency, it has been widely applied in engineering, municipal construction, and archeological investigations. Notably, soil moisture and soil salinity significantly attenuate electromagnetic wave signals. GPR images enable accurate identification of the plow layers for typical crops, including rice, wheat, maize, and cotton [21]. Previous soil stratification experiments conducted on farmland soils in Northeast China’s mollisol areas demonstrated that GPR can rapidly and accurately characterize layered soil structure [22]. In addition, a novel and reliable deep feature fusion framework for GPR data was adopted to identify the horizontal variation in compaction in embankment soils [23]. The combined utilization of GPR and Synthetic Aperture Radar (SAR) datasets enables accurate estimation of soil moisture in agricultural fields [24]. Unmanned Aerial Vehicle (UAV) platforms have also been integrated with a GPR system to conduct multi-altitude spatial scanning for soil moisture measurement [25]. GPR can also detect subsurface ponding risks and evaluate the structural integrity of buried seals and subsurface materials [26]. A GPR waveform inversion framework integrated with the Harris Hawks Optimization (HHO) algorithm enables effective inversion of soil water content and stratigraphic thickness [27]. Furthermore, GPR techniques have been applied to characterize structural deformations in the submerged drainage syphon at Ashou Bridge in El Beheira Governorate, Egypt, successfully identifying deformed zones at varying horizontal positions and burial depths on both sides of the structure [28]. GPR exhibits great potential for the real-time, automatic monitoring of soil moisture within potato root zones and supports the implementation of precision automated irrigation [29]; it has also been successfully applied to map and diagnose salt-accumulated layers (SALs) in oasis terraces [30] and, in combination with UAVs, to measure soil permittivity and electrical conductivity [31]. GPR has also been applied for the non-invasive detection of Holocaust-era mass graves in the Vizgiris Forest of Alytus, Lithuania [32]. Combined with geospatial techniques, it supports high-precision archeological surveys of ancient tombs [33]. GPR holds great potential as a rapid in situ soil investigation technique [34]. Under diverse field conditions, GPR measurements can accurately characterize the pavement thickness, structural features, dielectric properties, and signal amplitude of aged road surfaces [35]. Owing to soil heterogeneity and the complexity of data processing, the application of GPR in soil stratification identification is still in an early research stage, with prominent limitations in large-scale field detection [36]. It can effectively retrieve soil moisture content in red clay through waveform inversion [37]. The combination of ground-penetrating radar, CFA, and AEA enables feasible and reliable moisture inversion for oasis saline–alkali soils [38]. Prior to saline–alkali soil improvement, GPR-derived soil conductivity measurements can be used to quantify and classify soil salinization levels [39]. When discussing the identification of stratification in arid, salinized soil using GPR instantaneous attribute fusion [40], GPR exhibits excellent performance in shallow subsurface detection, and it has been widely applied in soil research.

Crop failure in saline–alkali soils can be attributed to the retention of intact red clay layers during soil improvement or the secondary formation of new red clay layers via subsequent sedimentation. Therefore, dynamic monitoring of remediated saline soils is essential for long-term soil stability assessment. To better assess the remediation effects of saline–alkali land in the Yellow River Delta, GPR was employed for shallow subsurface soil investigation. The objectives of this study are threefold: (1) to establish a systematic GPR-based detection methodology for the accurate delineation of the red clay layer; (2) to validate the effectiveness of the proposed method using GPR field survey data; and (3) to establish a dielectric constant–soil water content model tailored to local saline–alkali soils. The research findings provide a valuable reference for evaluating the remediation efficacy of regional saline–alkali soil, optimizing arable land utilization, effectively preventing and mitigating soil salinization, and facilitating the sustainable economic development of the study area.

2. Materials and Methods

2.1. Study Area

This study was conducted in Dongying City, Shandong Province, China (36°55′–38°10′ N, 118°07′–119°10′ E) (Figure 1a). The Yellow River Delta spans a total area of 5400 km², of which 5200 km² falls within the administrative jurisdiction of Dongying City. Administratively, the study area covers the Dongying District, Kenli District, Hekou District, Lijin County, and Guangrao County, with Lijin County and Guangrao County situated outside the Yellow River Delta (Figure 1c). The Yellow River traverses the central part of the study area and empties into the Bohai Sea, as marked by the red dashed line in Figure 1c. The study area has a warm–temperate continental monsoon climate, with a mean annual temperature of 11.3–13 °C, annual precipitation of 530–630 mm (65% of which occurs in summer), and annual evaporation exceeding 1800 mm. The local saline–alkali soils are mainly salinized fluvo-aquic soils and solonchaks formed from Yellow River alluvial sediments. The topsoil primarily consists of light and medium loam and exhibits typical adverse properties, including high salinity, low fertility, poor permeability, and severe compaction.

The study area consists of a fixed experimental field (Figure 1d) and 18 scattered sampling sites distributed across Dongying City (Figure 1c). The experimental field is located in Kenli District and comprises three types of test plots: exploration pits, tilled-soil plots, and undisturbed soil plots. Four rectangular exploration pits (2 m in length, 1.5 m in width, and 1.3 m in depth) are linearly arranged from south to north. The pit area features a weed-covered natural surface, as marked by the red box in Figure 1d. The tilled-soil plots, designated DT80, DT60, DT40, DT20, and DT0 from north to south, were established by excavating soil to depths of 80 cm, 60 cm, 40 cm, 20 cm, and 0 cm below the ground surface, followed by thorough homogenization and in situ backfilling. Maize was subsequently sown in these plots, as delineated by the blue box in Figure 1d. The undisturbed soil plots (designated T80-8, T80-13, T80-24, T50-8, T50-13, and T50-24 from north to south) preserve their intact in situ soil structure, with maize sown on the ground surface, as outlined by the yellow box in Figure 1d. The 18 scattered sampling sites (coded Y01–Y05, Y08–Y11, Y13–Y17, Y24, Y26–Y28) are primarily distributed across northern Dongying City and along the northern reaches of the Yellow River. The dominant crops in the study area, in descending order of planting area, are maize, rice, sorghum, and cotton. At the time of the field study, maize had already been harvested in some farm fields.

A prominent feature of saline soils in the Yellow River Delta is the widespread occurrence of a red clay layer. Characterized by high clay content, fine particles, high bulk density, and the formation of a low-permeability aquitard, this layer impedes downward water infiltration, thereby exacerbating salt accumulation at the soil surface. Given its critical role and influence on saline–alkali land remediation and its common occurrence within the top 1 m of soil, the research scope was confined to the upper 1 m of the soil profile.

2.2. GPR Equipment

GPR surveys were conducted using a SIR-30E system manufactured by Geophysical Survey Systems, Inc. (GSSI, Nashua, NH, USA), equipped with two shielded monostatic antennas operating at center frequencies of 400 MHz and 900 MHz. The SIR-30E features high acquisition speed, high resolution, a wide dynamic range, and a dual-channel design. It delivers a dual-channel scan rate of over 2896 scans per second, supports a maximum field survey speed of 142 Km/h, and achieves a sampling rate of up to 16,384 samples per second.

2.2.1. Data Acquisition

The target detection interval in the study area is confined to the top 1 m below the ground surface. To strike an optimal balance between penetration depth and vertical resolution across the target soil profile, 400 MHz and 900 MHz antennas were selected for the surveys. The study area receives abundant summer rainfall, and soil water content is a key factor governing GPR detection performance. In general, soil moisture recovers to pre-rainfall levels approximately one week after a precipitation event. Accordingly, all GPR surveys were scheduled exclusively for periods with at least one consecutive rain-free week. Based on local historical rainfall records and weather forecasts, GPR data acquisition was conducted at both the experimental field and scattered sampling sites from 9 to 15 September 2025 (Figure 2), ensuring that all measurements were completely unaffected by rainfall interference.

Given the potential of red clay layers to induce surface vegetation mortality, GPR survey lines should be preferentially positioned in unvegetated gaps within cultivated fields to ensure accurate assessment of saline soil amelioration effectiveness. Within the experimental plots, the exploration pits retained natural weed coverage, whereas all other plots were sown with maize. The scattered sampling sites were mostly situated in farmland with maize, rice, cotton, and sorghum, most of which remained unharvested at the time of the survey. Owing to the heterogeneous spatial distribution of salt in saline–alkali soils, portions of the farmland adjacent to the experimental plots and sampling sites exhibited severe salinization. This induced physiological water deficit and subsequent crop mortality, forming patchy bare areas across the cultivated land. To avoid damaging standing crops, GPR survey lines were primarily deployed within or immediately adjacent to these representative patchy bare areas. The survey lines varied in geometry (straight, curved, polygonal, or closed-loop) and length (11–100 m), tailored to the size and shape of each bare patch. For large-scale saline–alkali wasteland or extensive patchy bare patches, multiple parallel survey lines were acquired. For smaller patches, a single profile was repeatedly scanned 2–3 times to ensure high data quality. Prior to data acquisition, prominent surface obstacles (e.g., stones and residual crop roots) along the survey lines were cleared to minimize interference during radar antenna movement. A measuring tape was laid alongside each survey line for precise horizontal positioning.

Prior to formal data acquisition, a pilot test was conducted to calibrate the acquisition parameters and determine the optimal settings. Each field test was repeated at least three times, and the parameter configuration delivering optimal imaging performance was selected for formal data acquisition. Parameter configurations were adjusted promptly whenever suboptimal settings were identified during the field acquisition process. A total of 77 survey lines were acquired. Antennas of different frequencies were selected based on the target layer depth: within the study area, the effective penetration depth was approximately 1.3 m for the 400 MHz antenna and 0.5 m for the 900 MHz antenna (

Table 1. Performance comparison of 400 MHz and 900 MHz GPR antennas.

Antenna	Detection Depth (m)	Vertical Resolution (cm)	Signal Attenuation	Layer Identification Accuracy
400 MHz	1.0–1.3	5	slow	low
900 MHz	0.4–0.5	2	fast	high

). The 400 MHz antenna was triggered by a distance-measuring wheel, while the 900 MHz antenna adopted time-based triggering. Acquisition parameters for the 400 MHz antenna were set as follows: a scan rate of 50 scans/s, 40 scans per unit, 2 units per marker, 512 samples per scan, 32 bits per sample, and a time window of 50 ns. For the 900 MHz antenna, the corresponding parameters were 50 scans/s, 2 units per marker, 512 samples per scan, 32 bits per sample, and a time window of 40 ns. The length of each survey line was determined by field conditions, ranging from 11 m to 100 m.

Simultaneously with GPR data acquisition, ground-truth soil samples were collected adjacent to the survey lines using a hand auger (70 mm in diameter, 1 m maximum drilling depth, with a polished barrel and graduated depth scale). Soil types and stratification layering were identified in situ during drilling. All samples were immediately sealed in airtight plastic bags, transported to the laboratory, and analyzed for gravimetric moisture content and bulk density using the standard oven-drying method.

2.2.2. Data Processing

GPR data processing was performed using RADAN 7 (GSSI, Nashua, NH, USA), the dedicated data processing software bundled with the GPR system. Following completion of field data acquisition, GPR data pre-processing was first conducted, including invalid data removal, survey lines merging, survey line direction adjustment, ground zero-point calibration, and distance normalization. To facilitate comparative analysis across all survey lines, all north–south lines were uniformly oriented from south to north and all east–west from east to west during this pre-processing step. Subsequently, a representative survey line was selected for trial processing following a standard workflow of background noise removal, band-pass filtering, and gain adjustment. The primary objective was to clearly resolve the top and bottom reflection interfaces of the red clay layer. Once the optimal processing parameters were finalized, all pre-processed datasets were processed uniformly with the same parameters, and three of the instantaneous attributes-amplitude, frequency, and phase-were extracted.

2.2.3. Data Interpretation

Subsurface heterogeneities at the survey sites—including surface undulations, residual crop roots, loose topsoil, and gravel—occasionally disrupted signal continuity and induced localized scattering artifacts in the GPR profiles. Nevertheless, the overall lateral trends of the reflection events remained clearly discernible. Sharp stratigraphic boundaries were identifiable at interfaces with strong dielectric contrast between adjacent soil layers, whereas transitions appeared more gradual and attenuated where the dielectric contrast was weak. Accordingly, GPR data interpretation was performed by integrating the processed radar profiles, extracting instantaneous attributes (particularly instantaneous amplitude sections), and documenting in situ stratigraphic characteristics. Horizon tracking and continuous stratigraphic interpretation were conducted following systematic analysis of reflection signatures and lateral continuity of individual layer events. During GPR data interpretation, comparative analysis should be conducted exclusively on profiles of the same type, rather than across profiles with markedly distinct subsurface soil structures along the survey lines.

2.2.4. Soil Dielectric Constant

The soil dielectric constant is a core parameter for GPR soil surveys, as it directly governs the accuracy of subsurface layer depth estimation. The dielectric constant between adjacent soil layers constitutes the fundamental prerequisite for electromagnetic wave reflection and signal detection during GPR data acquisition. The dielectric constant of single-phase homogeneous materials is typically determined via laboratory experiments. In contrast, soil is a three-phase medium composed of solid soil particles, liquid pore water, and gaseous air within soil voids, and soil water exerts a dominant influence on its bulk dielectric constant. GPR surveys universally employ the relative dielectric constant; therefore, the relative dielectric constant of each soil layer must be quantified to enable accurate calculation of the burial depth and thickness of individual strata.

Extensive research has been conducted on the relationship between the relative dielectric constant (ε) and volumetric water content (θ) of soil, yielding a range of well-established models, most prominently those proposed by Topp, Alharathi, Ferre, Ju Zhaoqiang, and Herkelrath [41].

Topp model ε = 3.03 + 9.3 × θ + 146 × θ² − 76.6 × θ³

(1)

Alharathi model ε = ((θ + 0.204)/0.128)²

(2)

Ferre model ε = ((θ + 0.1841)/0.118)²

(3)

Ju Zhaoqiang model ε =((θ + 0.1846)/0.1219)²

(4)

Herkelrath model ε =((θ + 0.285)/0.125)²

(5)

where ε is the relative dielectric constant, dimensionless; θ is the soil volumetric water content, t, dimensionless.

Based on GPR field data and stratification data from soil core samples, the relative dielectric constant of each soil layer can be calculated using Equation (6).

ε = (ct/2h)²

(6)

where c is the propagation velocity of electromagnetic waves in a vacuum, with a constant value of 0.3 m/ns; t is the two-way travel time of electromagnetic waves within each soil layer, in ns; and h is the thickness of the corresponding soil layer, in m.

2.2.5. Reflection Coefficient

The reflection coefficient is a core parameter in GPR surveys for characterizing the reflection strength of electromagnetic waves at interfaces between dissimilar subsurface media. It is defined as the ratio of the reflected wave intensity to the incident wave intensity at a medium boundary, with a theoretical value range of −1 to 1.

$$R={\displaystyle \frac{\sqrt{{\epsilon }_1}-\sqrt{{\epsilon }_2}}{\sqrt{{\epsilon }_1}+\sqrt{{\epsilon }_2}}}$$

(7)

where R is the reflection coefficient at the stratigraphic interface, dimensionless; ε₁ is the relative dielectric constant of the upper medium at the interface, dimensionless; and ε₂ is the relative dielectric constant of the lower medium at the interface, dimensionless.

The amplitude of electromagnetic wave reflections is governed by the reflection coefficient, which quantifies the relative dielectric constant contrast across the interface between adjacent subsurface media. A larger absolute reflection coefficient corresponds to a more pronounced dielectric contrast and thus a stronger received reflection signal; conversely, a smaller absolute value indicates a subtle dielectric contrast and a correspondingly weaker reflected signal. The phase of the reflected wave matches that of the incident wave when the reflection coefficient is positive, whereas phase inversion occurs when the coefficient is negative. Electromagnetic wave energy undergoes progressive attenuation as it propagates through subsurface media. For deep-seated interfaces or high-loss dielectric materials, the returning signal can become extremely faint after extensive propagation paths and cumulative attenuation, even with a moderate initial reflection coefficient. The detection sensitivity and noise floor of a GPR system are intrinsically fixed. When the absolute value of the reflection coefficient is extremely small (e.g., less than 0.03), reflected wave energy may fall below the system’s detection threshold or be obscured by environmental interferences and internal system noise, rendering the reflection signal indiscernible. Strong interfering reflections from near-surface features (e.g., utility poles, retaining walls) or shallow subsurface heterogeneities carry substantially higher energy than weak deep reflections and can mask signals associated with an absolute reflection coefficient below 0.03, rendering them undetectable on radar profiles. Interfaces with an absolute reflection coefficient below 0.03 yield extremely weak reflected energy and are strongly constrained by system sensitivity, signal attenuation, and environmental disturbances. Consequently, distinct reflection events are generally indiscernible on conventional GPR profiles [42].

3. Results

Based on in situ soil observations from exploration pits and boreholes, coupled with GPR survey data, the subsurface soil profile in the top 1 m of the study area can be broadly divided into four sequential layers: a pale-yellow plow layer (loam), an iron-red clay layer, a yellow sand layer, and a pale-yellow sand layer (Figure 3 and

Table 2. Soil physical properties from the exploration pit.

No.	Layer					Dielectric Constant Calculated by Formula							CRC 4
	Layer	Thick (m)	MMC 1 (%)	BD 2 (g/cm3)	VMC 3	Topp	Alharathi	Ferre	Ju, Z.Q.	Malicki	Herkelrath	Fit Form
P1	Plow soil	0.31	31.14	1.43	0.45	29.49	25.84	28.57	26.82	28.75	34.26	36.55	−0.7062
	Red clay	0.23	34.88	1.44	0.50	34.84	30.46	33.85	31.76	34.41	39.68	40.85	−0.0479
	Yellow sand	0.33	28.29	1.52	0.43	28.03	24.61	27.17	25.50	28.79	32.81	33.85	0.0469
	Dark yellow sand	0.43	30.10	1.46	0.44	28.91	25.36	28.02	26.30	28.67	33.69	34.35	−0.0036
P2	Plow soil	0.3	41.01	1.17	0.48	32.45	28.37	31.46	29.53	27.13	37.24	37.82	−0.7203
	Red clay	0.22	35.73	1.42	0.51	35.28	30.85	34.30	32.18	34.46	40.14	41.08	−0.0206
	Yellow sand	0.27	27.96	1.55	0.43	28.37	24.90	27.50	25.81	29.70	33.15	35.34	0.0376
	Dark yellow sand	0.51	27.30	1.56	0.43	27.56	24.22	26.73	25.09	28.94	32.35	33.23	0.0154
P3	Plow soil	0.26	40.21	1.23	0.50	34.20	29.89	33.20	31.16	29.92	39.02	38.11	−0.7212
	Red clay	0.28	29.21	1.50	0.44	28.64	25.13	27.76	26.05	28.98	33.42	34.10	0.0278
	Yellow sand	0.32	29.11	1.49	0.43	28.25	24.80	27.39	25.71	28.43	33.03	33.79	0.0023
	Dark yellow sand	0.28	30.84	1.50	0.46	31.11	27.22	30.15	28.29	31.67	35.88	37.30	−0.0247
	Light red sand	0.16	29.67	1.50	0.45	29.37	25.74	28.46	26.71	29.80	34.14	38.29	−0.0065
P4	Plow soil	0.24	38.45	1.26	0.49	33.22	29.04	32.23	30.24	29.50	38.02	40.64	−0.7288
	Red clay	0.26	26.21	1.60	0.42	26.86	23.64	26.07	24.47	28.83	31.66	31.97	0.0600
	Yellow sand	0.38	29.51	1.54	0.46	30.32	26.55	29.38	27.57	31.57	35.09	35.06	−0.0231
	Dark yellow sand	0.42	31.08	1.47	0.46	30.49	26.69	29.55	27.73	30.43	35.26	36.86	−0.0125

1. MMC: Mass Moisture Content. 2. BD: Bulk Density. 3. VMC: Volumetric Moisture Content. 4. CRC: Calculated Reflection Coefficient.

).

3.1. GPR Images near the Exploration Pits

To evaluate the effects of surface conditions and soil moisture on GPR signal responses, three rounds of GPR data acquisition were conducted at the exploration pits (Figure 2a–c). The first two rounds were performed on the morning of Day 1, and the third on the morning of Day 6. The first acquisition was conducted under undisturbed surface conditions, with the GPR antenna traversing over natural grass cover. No distinct stratigraphic reflection interfaces are observed in the processed profile (Figure 4a) or the instantaneous amplitude section (Figure 4b); only two faint reflection bands are vaguely identifiable within 8–16 ns and 30–50 ns time windows. The second acquisition was performed after removing surface grass. No well-defined stratigraphic reflections are visible in the processed profile (Figure 4c) or the instantaneous amplitude section (Figure 4d), whereas a distinct reflection band appears within the 20–40 ns range. The third acquisition, performed five days later, yielded markedly clearer stratigraphic reflections in both the 900 MHz (Figure 4e,f) and 400 MHz datasets (Figure 4g,h). A strong reflection event between 12 ns and 20 ns corresponds to the red clay layer, and the interface between the upper and lower sand layers is evident near 30 ns. As illustrated in Figure 4, the effective detection range of the 400 MHz antenna extends to 50 ns, whereas that of the 900 MHz antenna is limited to approximately 20 ns.

Notably, the reflection bands observed in the first (30–50 ns) and second (20–40 ns) acquisition rounds correspond to the same red clay layer identified at 12–20 ns in the third-round dataset. This pronounced upward shift in two-way travel time is attributed to the progressive reduction in soil moisture over the five-day interval. The resulting decrease in the soil’s relative dielectric constant increased the electromagnetic wave propagation velocity, thereby significantly shortening the two-way travel time. Additionally, the reflection band spanning 8–16 ns in the first dataset is interpreted as scattering artifacts induced by surface grass cover, as this feature disappeared in the second acquisition following vegetation removal.

3.2. GPR Images of Undisturbed Soil

In the undisturbed soil plots, the GPR profiles and instantaneous amplitude sections (Figure 5) exhibit well-defined stratigraphic reflection sequences. Beneath the strong topsoil reflection lies a zone of weak reflection, followed by an interval of alternating high- and low-amplitude reflections. A distinct lateral boundary is identifiable at approximately 10 m in Figure 5c,d, marking the transition from the maize field (left) to the patchy bare area (right). The overall reflection amplitude in the maize field is significantly stronger than that in the patchy bare area. The prominent reflection event spanning 12–18 ns corresponds to the red clay layer; this interface is readily discernible in the patchy bare area but poorly resolved in the maize field.

3.3. GPR Images of Tilled Soil

GPR profiles from the tilled-soil plots (Figure 6) clearly record artificial tillage disturbances, with the most prominent reflection event corresponding to the bottom boundary of the tilled layer. The overall vertical reflection pattern across these profiles follows a consistent trend: following tillage, surface soil reflections are weaker than those in untilled soil. Immediately below the surface soil, reflection intensity decreases further, then increases to a peak at the bottom boundary of the tilled layer and attenuates again at greater depths. At survey line DT0 (untilled control), a strong reflection event associated with the red clay layer is clearly visible across the range of 12–20 ns. At survey line DT20, the bottom boundary of the tilled layer occurs at approximately 8 ns. Surface soil reflections here are weaker than at DT0, and the red clay layer is poorly resolved. At survey line DT40, the bottom boundary of the tilled layer lies at approximately 15 ns. As this survey line is located immediately south of DT20, an additional stratigraphic reflection interface corresponding to the 20 cm tillage depth is evident near 8 ns. At survey line DT60, the bottom boundary of the 60 cm deep tilled layer is observed at approximately 22 ns. Adjacent to DT40 to the south, the reflection marking the bottom of the 40cm tilled layer remains identifiable near 15 ns. At survey line DT80, the bottom of the tilled layer is located near 30 ns, with uniformly weak reflections occurring between this boundary and the ground surface.

3.4. GPR Images of Dispersed Measuring Points

Data from 18 scattered survey sites demonstrated the regional applicability of GPR for identifying shallow subsurface stratigraphic horizons (Figure 7). Within the upper 1 m below the ground surface, the general reflection pattern consists of a strong surface reflection event, followed by an alternating sequence of weak–strong–weak–strong reflection bands. For survey line Y03, the interface between the overlying loam layer and the underlying sand layer occurs at approximately 18 ns. For survey line Y10, the weak reflection zone beneath the surface reflection is relatively extensive (extending to 30 ns); this is attributed to loose surface soil conditions and is consistent with borehole data indicating sandy sediments within the top 1 m. A distinct dipping high-amplitude reflection event appears near 10 ns between the 40 m and 47 m positions on the right portion of the profile, and drilling verification confirmed this anomaly to be buried corn stalks. For survey line Y08, reflections within the 20–22 ns interval correspond to the red clay layer. For survey line Y26, reflections in the 16–18 ns range represent sand intermixed with a small amount of red clay clasts. For survey line Y05, the red clay layer is similarly identifiable from reflections within the 20–22 ns range. For survey line Y27, the interface between the upper loam and the lower sand layers is located at approximately 9 ns.

4. Discussion

4.1. Comparison of Relative Dielectric Constant

Field-measured relative dielectric constant (ε) of individual soil layers was calculated from stratum thickness determined via exploration pits and boreholes, paired with corresponding two-way travel times extracted from GPR profiles. Combined with volumetric water content measured from soil samples at each survey site, an empirical formula for estimating the relative dielectric constant of saline soils in the Yellow River Delta as a function of water content is shown in Figure 8.

ε = 14.517 + 5.1228 ∗ θ + 137.28 ∗ θ² − 85.774 ∗ θ³

(8)

where ε is the relative dielectric constant, dimensionless; θ is the soil volumetric water content, dimensionless.

As illustrated in Figure 9, the relative dielectric constants calculated using existing empirical models (Equations (1)–(5)) substantially underestimated the field-measured values (red solid line in Figure 9). Even the Herkelrath model (green solid line in Figure 9), which achieves the best agreement with the measured data, yields absolute errors ranging from −0.91 to 7.59. The fitted formula proposed in this study does not account for the influence of additional factors such as soil salinity and thus can only serve as an approximate estimation formula. The relative dielectric constants of the subsurface soil layers in the study area range from 20.25 to 40.64, with a mean value of 30.57. The average relative dielectric constants obtained from exploration pits and scattered survey points were 36.25 and 29.01, respectively.

To validate the accuracy of the proposed formula, three GPR survey lines were deployed for field investigation and paired soil sampling. The maximum relative errors between formula-calculated soil layer thickness and volumetric soil water content and their corresponding field-measured values were 11.76% and 9.12% (

Table 3. Comparison of soil layer thickness and volumetric soil water content: values calculated by the proposed formula versus field-measured values.

Survey Line	Soil Layer	Layer Thickness (m)			Volumetric Soil Water Content
		Field	Formula	Relative Error (%)	Field	Formula	Relative Error (%)
TL1	Plow soil	0.15	11.76	35.23	37.58	6.67	0.17
	Sand	0.15	7.14	29.86	27.17	9.01	0.14
	Red clay	0.23	4.17	26.53	28.95	9.12	0.24
	Sand	0.47	4.44	30.16	32.66	8.29	0.45
TL2	Plow soil	0.13	7.14	29.56	27.62	6.56	0.14
	Sand	0.16	11.11	27.48	29.43	7.10	0.18
	Red clay	0.33	6.45	30.24	28.65	5.26	0.31
	Sand	0.38	2.70	29.78	28.15	5.47	0.37
TL3	Plow soil	0.17	10.53	26.35	24.54	6.87	0.19
	Sand	0.22	4.76	24.52	26.71	8.93	0.21
	Red clay	0.36	0.00	28.65	27.42	4.29	0.36
	Sand	0.25	4.17	31.26	33.58	7.42	0.24

), respectively. The magnitude of relative error is negatively correlated with soil layer thickness, with thinner layers producing larger estimation deviations.

However, the proposed formula accounts only for volumetric soil water content and does not incorporate other influencing factors such as soil salinity. It therefore has inherent limitations and is applicable exclusively to the present study area. For saline–alkali soils in other regions, the formula may serve only as a reference.

4.2. Soil Characteristics and Electromagnetic Wave Energy

For farmland with consistent tillage conditions, variations in soil moisture across different soil layers are controlled by environmental factors (temperature, precipitation, and soil depth) and crop-related variables, including crop species, root depth and distribution, canopy coverage, growth stage, and irrigation management [43,44]. Soil planted with drought-tolerant crops generally exhibits lower moisture content, corresponding to their developed root systems, compared with hydrophilic crops [45,46]. Deep-rooted crops can uptake water from deep soil layers, whereas shallow-rooted crops mainly utilize water in topsoil. Dense canopy coverage effectively reduces surface soil evaporation [47]. Crop water consumption varies dynamically throughout the seedling, vegetative, and maturing stages, with the vegetative stage presenting the highest water demand. Irrigation events cause significant fluctuations in soil moisture [48,49]. High temperatures enhance surface evaporation and reduce soil water content, whereas rainfall is also a vital factor regulating soil moisture variation. Moreover, soil moisture at the depth of 50 cm is generally lower than that at the depth of 20 cm [50]. In the study area, maize, sorghum, and rice are categorized as shallow-rooted crops, while cotton is a deep-rooted crop. Sorghum and cotton are drought-tolerant crops, rice is hydrophilic, and maize is an intermediate type.

As shown in Figure 10, the ground surface of all four test pits was fully covered by vegetation, which suppresses surface soil evaporation and thus maintains relatively high volumetric soil water content. Sites Y10, Y15, and Y24 correspond to harvested maize fields, which feature relatively low volumetric soil water content. Zones with elevated soil moisture display strong reflection amplitude and rapid electromagnetic signal attenuation, accompanied by a shallower effective detection depth. Similarly, patchy bare areas in the study area hold higher moisture and salinity values than the surrounding vegetated farmland. Accordingly, a prominent improvement in the continuity and amplitude of deep subsurface reflections is consistently observed as survey lines extend from field margins into cropped zones.

Quantitatively, volumetric soil water content across the study area varies from 15% to 30%, and the ground surface produces the volumetric reflection coefficient (approximately 0.65–0.73). At the test pits, the average reflection coefficients reach 0.039 at the top boundary of the red clay layer and 0.027 at its bottom boundary, while the interface separating the two underlying sand layers yielded a value of 0.014. Of the 17 identified soil layers, 15 exhibit relative reflection coefficients exceeding 0.01 (Table 1), and all 15 horizons enable unambiguous identification of stratigraphic interfaces on their corresponding GPR profiles. Only two stratigraphic horizons on isolated survey lines show coefficients below 0.01, corresponding to faint reflections and poorly resolved subsurface boundaries. Comparative analysis of relative reflection coefficients against GPR suggests that subsurface stratigraphic horizons with absolute reflection coefficients greater than 0.01 yield distinct reflection events and enable unambiguous identification of stratigraphic interfaces on GPR profiles. This value is applicable exclusively to the present study area and may serve as a reference for other saline–alkali soil regions.

As shown by repeated surveys conducted at the test pits (Figure 4), gradual soil moisture depletion markedly improves the visibility of these weak subsurface reflection events.

The considerable influence of near-surface conditions on electromagnetic wave energy is further demonstrated by comparing natural terrain and artificially paved surfaces (Figure 11). An obvious boundary lies around the 9 m position along the profile, dividing grass-covered soil on the left and paved road on the right. The volumetric soil water content under grass coverage is higher than that beneath the road surface. Driven by elevated subsurface moisture under grass, the grass-covered zone features a shallower effective detection depth and more rapid signal attenuation relative to the road section. The high-amplitude reflection occurring at 14 m corresponds to a manhole cover. Subsurface reflection events below the road display good lateral continuity, while those under grass become discontinuous as a result of grass roots and heterogeneous soil compaction, even though their overall stratigraphic trends can still be distinguished. Stratigraphic interfaces are traceable throughout the whole profile, but stratigraphic interpretation is less straightforward in the grass-covered area. At two-way travel times of 10 ns, 30 ns, 44 ns, and 60 ns, reflections under grass show weaker amplitudes and poorer continuity compared with those under the pavement. Accordingly, stratigraphic interpretation in this study cannot depend merely on the continuity of reflection events. Instead, comprehensive interpretation should combine GPR reflection signatures with actual subsurface soil properties.

A comparable phenomenon occurs at the road-wasteland boundary at the 2 m profile position (Figure 12). Although the wasteland shows stronger shallow-reflection amplitudes relative to the road area, intensified signal attenuation leads to poor continuity of deep subsurface reflection events. Nevertheless, major stratigraphic horizons in both regions can still be effectively recognized through detailed horizon tracing.

For ameliorated saline soils, the presence of a red clay layer may indicate insufficient amelioration efficacy or the need for re-amelioration. GPR-detected burial depth and thickness of the red clay layer can provide technical guidance for targeted amelioration practices, including deep plowing, salt leaching via drainage, and soil profile reconstruction. Notably, the soil amelioration depth must exceed the burial depth of the red clay layer to achieve desired treatment outcomes. To ensure sustained amelioration effectiveness, dynamic monitoring of the remediated soil profile using GPR is essential. The proposed GPR-based approach is suitable for large-scale surveys in the study area, and the deployment of three-dimensional (3D) survey lines enables more accurate delineation of the red clay layer.

5. Conclusions

This study demonstrates the robust feasibility and promising potential of GPR for high-resolution soil investigation in the saline–alkali lands of the Yellow River Delta. GPR effectively identifies subsurface stratigraphic architectures within the uppermost 1 m soil profile. In particular, the reliable detection and mapping of the critical red clay layer provide essential geophysical support for developing precise amelioration strategies for local saline soils. GPR profiles indicate that stratigraphic horizons with an absolute reflection coefficient greater than 0.01 yield distinct, continuously traceable reflection signals. The average relative permittivity of soil layers in the study area is 30.57, with a correlation coefficient of 0.9173 between relative permittivity and soil water content. Compared with conventional empirical formulas, the newly proposed formula exhibits superior performance and enables more accurate estimation of dielectric constants from soil water content data. This approach guarantees highly reliable quantification of the burial depth and thickness of individual saline soil layers. However, this study has inherent limitations. Specifically, the proposed formula accounts only for volumetric soil water content and does not consider the impacts of additional factors such as soil salinity and trace mineral components. The formula performance has been validated exclusively within the present study area, and its general applicability to other saline–alkali soil regions remains to be further verified.

Despite these promising outcomes, this study confirms that soil moisture dominantly controls GPR signal attenuation and reflection continuity. Furthermore, the quantitative modeling proposed herein cannot fully decouple the interactive effects of soil salinity and trace mineral components. In addition, the 2D survey layout failed to cover large-scale basin-level saline–alkali wastelands. To further validate and extend the present findings, future research should prioritize large-scale 3D GPR field surveys during dry post-harvest seasons to mitigate moisture-induced signal scattering. The implementation of multi-parameter inversions, which comprehensively consider the coupled effects of soil moisture, salinity, and soil texture, can produce more robust and detailed subsurface structural models. Such methodological improvements will further enhance the reliability of GPR-based soil investigations in the Yellow River Delta and support the assessment of its general applicability to saline–alkali ecosystems across the globe. This study holds substantial practical value for sustainable farmland management. During saline soil amelioration, GPR surveys of red clay layer burial depth and thickness can guide the implementation of targeted amelioration measures. The presence of red clay layers signals either failed amelioration or salinization recurrence, underscoring the critical need for long-term dynamic monitoring of ameliorated saline soils.