Combination of Physical and Geostatistical Models for Assessing Surface Moisture in Semiarid Agricultural Soils with Sentinel-1 Through Remote Sensing

1. Introduction

Surface soil moisture (SSM) is one of the most significant edaphic parameters for understanding the hydrological, ecological, and agronomic processes occurring in terrestrial systems. This parameter directly controls infiltration, evapotranspiration, surface runoff, and the energy balance between the soil surface and the atmosphere; it also influences vegetation development, agricultural productivity, and the ecological stability of the land (Jackson et al., 1996; Porta Casanellas et al., 2019). In agricultural contexts, surface moisture levels near field capacity improve water-use efficiency and reduce crop water stress, whereas prolonged deficits can trigger irreversible soil degradation processes (Porta Casanellas et al., 2019). In semi-arid Mediterranean regions, such as the southeast of the Community of Madrid, SSM exhibits high spatial and temporal variability due to irregular precipitation, climate seasonality, and the progressive increase in temperatures associated with climate change (IPCC, 2021). Under these conditions, monitoring soil moisturethrough traditional in situ techniques is insufficient, both due to high costs and the limited spatial representativeness of point sampling (Hillel, 2004). Radar remote sensing has established itself as an effective alternative for the indirect and multitemporal estimation of surface soil moisture. SAR (Synthetic Aperture Radar) sensors provide spatially distributedinformation regardless of cloud cover or lighting conditions, demonstrating high sensitivity to the dielectric properties of the soil, which are directly related to water content (Zribi et al., 2020; Salgado et al., 2021). Within this framework, the Sentinel-1 satellite, part of the Copernicus program, provides C-band imagery with a spatial resolution of 10 m and a revisit frequency of 6 days—features that are particularly suited for the regular monitoring of SSM in agricultural landscapes (Buah et al., 2020; El Hajj et al., 2017). This study stems from the need to develop a validatable and repeatable methodology for estimating surface soil moisture in Mediterranean agricultural landscapes using Sentinel-1 imagery, following approaches similar to those proposed by Ma et al. (2021) and Graldi et al. (2023). The primary objective is to integrate semi-empirical radar backscatter models with advanced spatial analysis techniques to generate high-resolution SSM mapping with practical utility for sustainable soil management.

2. Study Area

The study area is located in the municipality of Villaconejos, situated in the Las Vegas region in the southeast of the Community of Madrid. This territory is representative of the traditional rain-fed (secano) agricultural systems of the central Iberian Peninsula, characterized by the presence of olive groves, herbaceous crops, Mediterranean scrubland, anextensive areas of bare soil. From a pedological perspective, loamy-clay soils predominate, with moderate to low infiltration capacity, which influences surface moisture retention (Departamento de Edafología, Universidad de Granada, n.d.). The local hydrography consists mostly of temporary streams and ravines that are active only after heavy rainfall, exhibiting a clearly seasonal behavior. At an agro-economic level, Villaconejos maintains an economy historically linked to agriculture, with a predominance of traditional olive groves, family-run farms, and a majority use of rain-fed farming (SIGPAC). This territory shows a marked susceptibility to soil degradation, erosion, and desertification processes, making it an especially suitable setting for the application and evaluation of surface moisture estimation methodologies through remote sensing (Salgado, 1998; European Commission, 2019; Ministry for the Ecological Transition and the Demographic Challenge, 2025).

Figure 1. Location of the pilot plot (A), the municipality of Villaconejos (B), and the Community of Madrid (C).

Figure 1. Location of the pilot plot (A), the municipality of Villaconejos (B), and the Community of Madrid (C).

Figure 2. Land use of the study plot (sioseAR 2016-2018 CC-BY 4.0).

Figure 2. Land use of the study plot (sioseAR 2016-2018 CC-BY 4.0).

Figure 3. Land use of the municipality of Villaconejos (sioseAR 2016-2018 CC-BY 4.0).

Figure 3. Land use of the municipality of Villaconejos (sioseAR 2016-2018 CC-BY 4.0).

3. Materials and Methods

3.1. Data Sources

For the development of this study, 12 Sentinel-1 radar images and 12 Sentinel-2 multispectral images were used, downloaded via the Copernicus Browser platform. The Sentinel-1 images correspond to the GRD (Ground Range Detected) product, featuring VV polarization, incidence angles of approximately 39°–40°, and C-band frequency (Sentinel-1 Team, 2013). Additionally, digital terrain models, PNOA orthophotographs, SIGPAC data, SIOSE AR land cover layers, and edaphic properties from the SoilGrids platform were employed. Sand (335 g/kg) and clay (208 g/kg) values were obtained from SoilGrids at a spatial resolution of 250 m, based on digital soil mapping and machine learning techniques (Poggio et al., 2021; ISRIC, 2020).

Figure 4. Workflow diagram of the study.

Figure 4. Workflow diagram of the study.

3.2. SAR Images Preprocessing

SAR images exhibit a characteristic type of noise known as speckle noise, which appears as a random granular pattern and complicates both visual interpretation and quantitative analysis (Santana, 2016). To mitigate this effect, a Single Product Speckle Filter was applied using the SNAP software. Furthermore, Range Doppler Terrain Correction was performed using SRTM digital elevation models to correct geometric distortions resulting from topography and the sensor’s acquisition angle (ESA, 2023). Finally, the backscatter values were converted from linear units to decibels (dB) to facilitate their analysis, comparison, and cartographic representation.

Figure 5. Workflow diagram for Sentinel-1A image preprocessing in SNAP. Source: Author.

Figure 5. Workflow diagram for Sentinel-1A image preprocessing in SNAP. Source: Author.

3.3. Selection of Wet and Dry Periods

The image acquisition dates were selected by distinguishing between a wet period (February–May) and a dry period (June–August), taking into account the absence or presence of prior precipitation based on historical data from AEMET (Annex 9.2).

3.4. Surface Moisture Estimation

The estimation of SSM was carried out by combining the Hallikainen and Dubois semi-empirical models. The Hallikainen model allows for relating the soil dielectric constant to its moisture content, while the Dubois model links radar backscatter to soil moisture and surface roughness. Given that moisture is an unknown variable, the Hallikainen model was applied inversely through an iterative process.

ε^′ = (a₀ + a₁S + a₂C) + (b₀ + b₁S + b₂C)m_v + (c₀ + c₁S + c₂C)m²

(1)

σ⁰(VV) = 10^−2.35 cos³θ 10^{0.046ε tan θ} (ks sin θ)^1.1λ^0.7

(2)

sin³ θ

3.5. Geostatistical Analysis and Downscaling

To improve the spatial resolution of the estimates, regression and geostatistical techniques were applied. First, an Ordinary Least Squares (OLS) regression was used as a diagnostic analysis. Subsequently, the Empirical Bayesian Kriging Regression Prediction (EBK-RP) method was implemented, incorporating explanatory variables such as slope, aspect, DEM, TWI, and NDVI (Krivoruchko, 2012).

4. Results

The results obtained show high spatial and temporal variability of surface soil moisture across the study area. The estimated values ranged from 0.02 to 0.45 m*/m*, which is consistent with the typical ranges found in semi-arid agricultural soils (Santana, 2016).

4.1. Radar Backscatter Analysis

Backscatter values in decibels were mostly within low to moderate ranges (below -10 dB), which is consistent with rain-fed agricultural surfaces and soils of medium roughness. The comparison of the observed values with the thresholds proposed by Santana (2016) allowed for the validation of the consistency of the results obtained.

Table 1. Backscatter coefficient levels for typical scenarios (Santana, 2016).

Backscatter Level	Typical Scenario
Very High (> −5 dB)	Anthropic structures, mountains, slopes facing the sensor.
High (−10 to 0 dB)	Rough surface. Dense vegetation (forests).
Moderate (−20 to −10 dB)	Intermediate vegetation. Agricultural crops.
Low (< −20 dB)	Flat surfaces, calm water, roads, dry sand.

Table 1. Backscatter coefficient levels for typical scenarios (Santana, 2016).

Backscatter Level	Typical Scenario
Very High (> −5 dB)	Anthropic structures, mountains, slopes facing the sensor.
High (−10 to 0 dB)	Rough surface. Dense vegetation (forests).
Moderate (−20 to −10 dB)	Intermediate vegetation. Agricultural crops.
Low (< −20 dB)	Flat surfaces, calm water, roads, dry sand.

Table 2. Observed backscatter ranges for each captured image.

Image Acquisition Date	Observed Backscatter Range (dB)
2015-03-22	(-14,17 a 1,28)
2015-07-26	(-17,42 a 0,27)
2017-05-04	(-19,88 a -2,93)
2017-08-08	(-16,95 a -1,29)
2019-04-24	(-13,96 a 1,02)
2019-06-29	(-16,08 a 0,3)
2021-04-25	(-16,96 a -1,9)
2021-08-05	(-16,19 a -1,34)
2023-05-27	(-16,49 a -2,08)
2023-08-19	(-16,71 a -0,68)
2025-03-11	(-16,43 a -1,09)
2025-06-03	(-17,04 a -1,61)

Table 2. Observed backscatter ranges for each captured image.

Image Acquisition Date	Observed Backscatter Range (dB)
2015-03-22	(-14,17 a 1,28)
2015-07-26	(-17,42 a 0,27)
2017-05-04	(-19,88 a -2,93)
2017-08-08	(-16,95 a -1,29)
2019-04-24	(-13,96 a 1,02)
2019-06-29	(-16,08 a 0,3)
2021-04-25	(-16,96 a -1,9)
2021-08-05	(-16,19 a -1,34)
2023-05-27	(-16,49 a -2,08)
2023-08-19	(-16,71 a -0,68)
2025-03-11	(-16,43 a -1,09)
2025-06-03	(-17,04 a -1,61)

4.2. Spatial Estimation of Surface Moisture

The application of the Hallikainen and Dubois models enabled the generation of continuous SSM maps for each analyzed date. During wet periods (February–May), mean values ranged between 0.33 and 0.45 m*/m*, whereas in dry periods, they decreased to values between 0.02 and 0.20 m*/m*. In the pilot plot, a direct response of SSM to precipi- tation events was observed; however, when scaling the analysis to the municipality level, complex spatial patterns were detected, with higher moisture values in topographically convergent areas and lower values on exposed slopes.

Figure 6. Cartographic composition of surface moisture during the dry period in the study plot.

Figure 6. Cartographic composition of surface moisture during the dry period in the study plot.

Figure 7. Cartographic composition of surface moisture during the wet period in the study plot.

Figure 7. Cartographic composition of surface moisture during the wet period in the study plot.

4.3. Multitemporal Analysis

The multitemporal analysis revealed a general trend toward progressively drier conditions between 2015 and 2025, especially during the summer months. The years 2023 and 2025 showed mean reductions of 0.15–0.20 m*/m*, associated with prolonged drought episodes and increased evaporation, in line with the findings described by Singh et al. (2020).

4.4. Geostatistical Analysis

The OLS analysis confirmed that all explanatory variables exert a statistically significant influence on SSM, although with low overall explanatory power (maximum adjustedR² of 0.03). The Moran’s I analysis evidenced the presence of spatial autocorrelation in the residuals, indicating that the spatial structure of the phenomenon was not adequately represented. The application of the EBK-RP model allowed for a substantial improvement in predictive capacity, generating SSM maps with a spatial resolution of 2 m, capturing local variability with greater precision (Krivoruchko, 2012).

Figure 8. Cartographic composition of surface moisture maps at the municipal level during the wet period.

Figure 8. Cartographic composition of surface moisture maps at the municipal level during the wet period.

Figure 9. Enter CaptioCartographic composition of surface moisture maps at the municipal level during the dry period.

Figure 9. Enter CaptioCartographic composition of surface moisture maps at the municipal level during the dry period.

5. Discussion

The results obtained confirm the suitability of radar remote sensing, and specifically Sentinel-1 imagery in VV polarization, for estimating surface soil moisture (SSM) in semiarid agricultural environments under Mediterranean climates. The sensitivity of radar backscatter to the soil’s dielectric properties has enabled the generation of continuous moisture maps with high spatial and temporal consistency, aligning with the findings of Zribi et al. (2020) and Salgado et al. (2021). The positive correlation between backscatter values (dB) and estimated surface mois- ture supports the validity of the semi-empirical models employed, particularly on bare soil surfaces or areas with sparse vegetation cover, where the SAR signal responds directly to variations in water content (Buah et al., 2020; El Hajj et al., 2017). However, a systematic underestimation of soil moisture was observed in areas with significant vegetation presence, 156 due to the attenuation and volume scattering of the radar signal, as noted by Solorza (2013) 157 and Singh et al. (2020). This result highlights the necessity of applying NDVI masks or 158 models that explicitly integrate the contribution of the vegetation canopy. From a methodological standpoint, the combination of the Hallikainen and Dubois models allowed for moisture estimates within physically plausible ranges (0.02–0.45 m*/m*), consistent with semi-arid agricultural soils (Santana, 2016). Nevertheless, the inherent limitations of both models—such as the dependence on coefficients calibrated for specific contexts and the restriction of the Dubois model to moisture levels below 35%—must be considered when interpreting the results, especially during wet periods (Quan et al., 2014). The multitemporal analysis conducted between 2015 and 2025 reveals a general trend toward progressively drier conditions, particularly during the summer months, with decreases in surface moisture associated with declining precipitation, increased evaporation, and potential soil degradation processes, in line with the scenarios described by the IPCC (2021). Nonetheless, the high interannual variability observed demonstrates that soil moisture dynamics respond to the complex interaction of climatic, edaphic, and topographic factors (Gaur & Mohanty, 2016; Scholz et al., 2024). The statistical analysis using OLS regression showed a significant influence of the to- pographic and spectral variables considered, albeit with limited overall explanatory power and the presence of spatial autocorrelation in the residuals. In this context, the application of the Empirical Bayesian Kriging Regression Prediction (EBK-RP) method represented a substantial improvement by capturing the spatial structure of surface moisture and gener- ating high-resolution maps (2 m), in accordance with the proposals of Krivoruchko (2012) and Bauer-Marschallinger et al. (2018). Overall, the discussion emphasizes that, despite the inherent limitations of the models and the indirect nature of radar estimation, the proposed methodology constitutes a robust and transferable tool for monitoring surface soil moisture in semi-arid agricultural environments, with high potential to support sustainable land management and climate change adaptation (European Commission, 2019).

6. Conclusions

This study has developed and validated an integrated methodology for the estimation and multitemporal mapping of surface soil moisture (SSM) in semi-arid Mediterranean agricultural environments. The combination of physical radar backscatter models (Hallikainen and Dubois) with advanced geostatistical techniques (EBK-RP) has enabled the generation of SSM maps with spatial resolutions of up to 2 m, overcoming the limitations of traditional in situ and optical methods. The results evidence significant spatial and tempo- ral variability of SSM and an interannual trend toward drier conditions between 2015 and 2025, associated with climatic factors and land-use patterns. The proposed methodology constitutes a valuable tool for sustainable soil management and climate change adaptation in vulnerable agricultural regions.