Yunquan Wang

Pedotransfer functions (PTFs) are widely used to estimate soil hydraulic parameters based on easily accessible soil information, playing an important role in the parameterization of earth surface models. However, conventional PTFs, developed using measurements from small volume soil samples, often exhibit significant deviations from field observations and substantial variability when applied to field‐scale hydrological models. Here, we introduce new Site‐Specific Pedotransfer Functions (SPTFs) that combine deep learning with physics‐based modeling of soil hydrological processes. SPTFs differ from conventional PTFs in two aspects: they utilize time‐series data as input and they directly optimize simulated soil water content by the 1‐D Richardson–Richards equation with observations, ensuring improved applicability to field conditions. We trained and tested the model using two years of soil moisture observations from 1,181 sites in the International Soil Moisture Network. Evaluation using field data demonstrates that SPTFs achieve a Nash‐Sutcliffe Efficiency of 0.65 and root mean squared error of 0.072 cm ³ cm ⁻³ in simulating soil water content at the depth of 0.05 m on the test set ( n = 179), which is close to the values predicted by the inverse modeling method, while maintaining the computational efficiency of PTFs. This study highlights the promise of SPTFs as a robust parameterization framework for localized field applications.

Accurately describing bi‐ or multi‐modal soil hydraulic properties (SHPs) is essential for understanding soil water flow processes. Traditional bimodal models, based on capillary theory, often require numerous empirical parameters and fail to adequately represent macropore flow driven by gravity and film flow held by adsorption forces. In this study, we introduce a novel framework to describe bimodal SHPs by shifting from a reference unimodal model. We propose a new bimodal soil hydraulic model based on the widely used unimodal van Genuchten–Mualem (VGM) model and incorporate the additional effects of adsorption forces. The so‐called VGM‐Bi model introduces two additional parameters, which can be easily derived or constrained from the measured soil water retention curve (SWRC), enhancing its usability. We further propose a modified version, the VGM‐Bi‐Macro model, to account for macropore effects by using a hydraulic conductivity value at a slightly negative matric potential as a new matching point. Evaluation with 25 soil samples shows that the VGM‐Bi model accurately captures bimodal SWRCs but struggles with bimodal hydraulic conductivity curves (HCCs) for a few samples. In contrast, the VGM‐Bi‐Macro model effectively captures both. Specifically, the VGM‐Bi‐Macro model achieves a slightly lower mean root‐mean‐square error (RMSE) of 0.200 in estimating HCCs and nearly half the RMSE (0.008 cm ³ cm ⁻³ ) in estimating SWRCs compared to the VGM‐Bi model (0.229 and 0.014 cm ³ cm ⁻³ ) and an existing bimodal model (0.262 and 0.015 cm ³ cm ⁻³ ), underscoring the importance of gravity‐driven macropore flow. These models offer a physically meaningful and efficient tool for simulating water flow in structured soils.

Pedotransfer functions (PTFs) are widely used to estimate soil hydraulic properties (SHPs) from easily measurable soil properties. However, most existing PTFs are based on unimodal hydraulic models, which fail to capture the bimodal behavior of hydraulic properties caused by soil structure. In this study, we developed new PTFs using two bimodal hydraulic models and introduced a soil physics–informed neural network to embed these models into the training process. The results showed that the new PTFs effectively represented bimodality in hydraulic conductivity curves, achieving a root mean square error of 0.531( K in cm/d) on the test set, compared with 0.626 for unimodal models. They also improved predictions of soil water retention curves but struggled with bimodal cases for some samples, likely due to the limited number of bimodal retention curves in the training data set. Evaluation on an independent data set showed that the error for hydraulic conductivity predicted by the new functions was about one‐third that of conventional approaches. In addition, the proposed soil physics–informed neural network, which directly optimizes SHPs, outperformed the conventional approaches that optimize fitted model parameters. We also found that whether water retention and hydraulic conductivity are optimized separately or simultaneously has a large effect on performance. Nonetheless, the lack of explicit soil structure information in the input data, along with limited measurements near saturation, continues to constrain accuracy. This emphasizes the need to develop a more comprehensive soil hydraulic database.

The farming-pastoral ecotone (FPE) of northern China serves as a critical ecological transition zone, in which poplar plantations significantly contribute to afforestation for large-scale ecological restoration projects. Due to concerns about sustainability, precise monitoring of the spatiotemporal dynamics of poplar plantations is needed, but systematic research is lacking. This study investigated the spatiotemporal dynamics of poplar plantation area and growth status from 1989 to 2022, taking the Anguli Nao watershed, a typical region in the FPE of northern China, as an example. Firstly, by utilizing satellite images and the random forest classification algorithm, the poplar plantation areas were well extracted, with a high accuracy over 93% and extremely strong consistency as demonstrated by a Kappa coefficient larger than 0.88. Significant changes in poplar plantation areas existed from 1989 to 2022, with an overall increasing trend (1989: 130.3 km², 2002: 275.9 km², 2013: 256.0 km², and 2022: 289.2 km²). Furthermore, the accuracy of our extraction method significantly outperformed six widely used global land cover products, all of which failed to capture the distribution of poplar plantations (producer’s accuracy < 0.21; Kappa coefficient < 0.18). In addition, the analysis of vegetation growth status revealed large-scale degradation from 2002 to 2013, with a degradation ratio of 24.4% that further increased to 31.1% by 2022, satisfying the significance test via Theisl–Sen trend analysis and the Mann–Kendall test. This study points out the uncertainty of existing land cover products and risk of poplar plantations in the FPE of northern China and provides instructive reference for similar research.

To understand how phreatophytes correspond to groundwater dynamics in arid regions, it is important to examine the specific water use patterns in different habitats. In this study, we investigated whether and how saltcedar (Tamarix ramosissima Ledeb.) responded in its water use patterns to the changing groundwater table during the growing season in two contrasting habitats (i.e., riparian and dune sites). δ¹⁸O and δ²H values of xylem sap and four potential water sources (i.e., shallow, middle, and deep soil water, and groundwater) were measured to determine the water-use pattern. Comparisons of the water sources in different habitats indicated that the depths of water extraction by saltcedar were shallower in the riparian habitat than in the dune habitat. During the growing season, saltcedar in the diparian habitat consistently extracted soil water from a depth of 30−60 cm (volumetric water content: 18.2 ± 3.5%), which was recharged by groundwater. In contrast, the saltcedar in the dune habitat either extracted soil water from a deeper depth (below 100 cm, volumetric water content: 5.8 ± 1.2%) that was also supposed to be recharged by groundwater, or directly used groundwater. These results suggest that the primary water source for saltcedar was from deeper groundwater during the growing season and did not change with the groundwater fluctuation.

Pedotransfer functions (PTFs) are widely used to estimate soil hydraulic properties based on easily accessible soil information, playing an important role in the parameterization of earth surface models. However, traditional PTFs, developed using laboratory-measured hydraulic properties of small-volume soil samples, often exhibit significant deviations from field observations and substantial variability across different PTFs when applied to field-scale soil hydrology modeling. Here, we introduce a new generation of Site-Specific Pedotransfer Functions (SPTFs) that combine deep learning with physical modeling of soil hydrologic processes. SPTFs differ from traditional PTFs in three key ways: they utilize time-dependent observations as input, account for soil structural effects, and directly optimize simulated soil moisture by 1-D Richards equation with observations, ensuring improved applicability to field conditions. Field evaluations demonstrate that SPTFs significantly outperform traditional PTFs in predicting soil moisture dynamics, reducing root-mean-squared-error (RMSE) from 0.122 to 0.080 cm3 cm−3 and increasing R2 from 0.14 to 0.63 in the test set. Additionally, incorporating advanced soil hydraulic model that account for soil structure, capillary flow, and film flow further enhanced SPTF performance, reducing RMSE to 0.078 cm3 cm−3 and raising R² to 0.65. This study highlights the promise of SPTFs as a robust parameterization framework for localized field applications.

Soil hydraulic properties (SHPs) are impacted by various mechanisms such as soil structure, capillarity, and adsorption forces, often showing a bimodal shape. Developing soil hydraulic models (SHMs) that describe SHPs over the entire saturation range often involves balancing the representation of multiple processes while minimizing the number of free‐fitted parameters. Existing SHMs rarely capture bimodal SHPs across the full moisture range or introduce a higher number of free‐fitted parameters. In this study, we propose a novel framework to describe SHPs over the entire moisture range, accounting for the effects of soil structure, capillarity, adsorption forces, and vapor diffusion. In its four free‐fitted parameters form, the proposed models can capture unimodal soil water retention curves and bimodal hydraulic conductivity curves (HCC). This model is well‐suited for situations where small changes in water content near saturation are no longer detectable via measured SWRC, yet soil structure still causes a sharp decline in HCC near saturation. With one additional free‐fitted parameter, the proposed models can capture both bimodal SWRC and HCC. Testing with 355 and 52 soil samples from two public datasets demonstrated that the proposed models performed exceptionally well in describing SHPs across the entire moisture range. The reported lowest root‐mean‐square error values were 0.005 and 0.009 cm³ cm⁻³ for fitting SWRCs, and 0.465 and 0.666 for predicting HCCs, respectively. Due to the minimal introduction of free‐fitted parameters, the proposed framework showed significant application potential.

Understanding the change in soil hydraulic conductivity with temperature is key to predicting groundwater flow and solute transport in cold regions. The most commonly used models for hydraulic conductivity during freeze‒thaw cycles only consider the flow of capillary water in the soil and neglect water flowing along thin films around the particle surface. This paper proposed a new hydraulic conductivity model of frozen soil via the Clausius–Clapeyron equation based on an unsaturated soil hydraulic conductivity model over the entire moisture range using an analogy between freeze‒thaw and dry‒wet processes in soils. The new model used a single equation to describe the conductivity behaviors resulting from both capillary and adsorption forces, thus accounting for the effect of both capillary water and thin liquid film around soil. By comparison with other existing models, the results demonstrated that the new model is applicable to various types of soils and that the predicted hydraulic conductivity is in the highest agreement with the observed data, while reducing the root mean square error by 38.9% compared to the van Genuchten–Mualem model. Finally, our new model was validated with thermal–hydrological benchmark problem and laboratory experiment result. The benchmark results indicated that the advective heat transfer was more significant, and the phase change was completed earlier when considering both capillary and adsorption forces than when only considering capillary forces. Furthermore, the coupled flow–heat model with the new hydraulic conductivity expression replicated well the results from a laboratory column experiment.

Pedotransfer functions (PTFs) are widely used to estimate soil hydraulic properties (SHPs) from easily measurable characteristics. However, most existing PTFs rely on unimodal hydraulic models, which fail to accurately represent the bimodal SHPs caused by soil structure common in field conditions. In this study, we developed new PTFs using two bimodal soil hydraulic models and introduced soil physics-informed neural networks (SPINN) to embed the models into PTF training. The results showed that the new PTFs effectively captured bimodality in hydraulic conductivity curves, achieving an RMSE of 0.578 in the test set, compared to 0.709 for unimodal models. The PTFs also improved soil water retention curve (SWRC) predictions but struggled with bimodal SWRCs for some samples, likely due to the limited number of bimodal SWRCs in the dataset. An independent dataset evaluation revealed that the RMSE for hydraulic conductivity predicted by the new PTFs was approximately one-third of that of classic PTFs. This underscores the significant role of soil structure in SHPs, which classic PTFs fail to capture. Additionally, PTFs developed using the SPINN method outperformed those optimized fitted hydraulic parameters via machine learning, a common approach in the literature. We also found that separate versus simultaneous optimization of water retention and hydraulic conductivity greatly affects PTF performance. Finally, we provided global 1 km-resolution maps of soil hydraulic parameters for the bimodal model.