Sarem Norouzi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sarem Norouzi has a total of 78 co-authors. Here are the top co-authors by number of shared publications:
Charles Pesch
Department of Agroecology, Aarhus University
5 shared publications- Emmanuel Arthur
Department of Agroecology Aarhus University Tjele Denmark
5 shared publications Marzieh Zaresourmanabad
Department of Agroecology, Aarhus University
5 shared publicationsPer Moldrup
5 shared publicationsTrine Norgaard
Department of Agroecology, Aarhus University
5 shared publications- Bo Vangsø Iversen
Department of Agroecology Aarhus University Tjele Denmark
4 shared publications
Latest Publications
- Open Access
- 27 April 2026
- Published by Pensoft Publishers in Agricultural and Environmental Modelling
- Vol. 8
Integrating spatial and environmental stressors in a population model of Folsomia candida (Collembola, Isotomidae): a Formal Model within ALMaSS framework
Collembola are key soil invertebrates that contribute to decomposition processes and nutrient cycling in agricultural ecosystems. Folsomia candida , a parthenogenetic euedaphic species, is widely used as a model organism in ecotoxicological testing, yet population-level responses to environmental stressors in spatially structured landscapes remain poorly understood. We present a formal model of F. candida developed using a subpopulation approach within the Animal, Landscape and Man Simulation System (ALMaSS). The model integrates stage-structured population dynamics with spatially explicit landscape processes, incorporating hourly weather data and daily soil moisture dynamics at 10 by 10 m resolution. Growth, development, reproduction, and mortality are driven by temperature, soil moisture, and food quality through empirically parameterised thermal performance curves and dose–response functions fitted to published experimental data. The model explicitly represents egg, juvenile, and adult life stages, with stage transitions and vital rates responding non-linearly to environmental conditions. Soil moisture effects are modelled using a novel integration of ERA5 reanalysis data, soil physical properties, and evapotranspiration to estimate surface soil water potential. Temperature and moisture effects are combined using a limiting-factor approach, reflecting current understanding of physiological constraints. The model is designed for simulations at subpopulation spatial scales of 100 m 2 , within landscapes typically 10 km × 10 km, providing a mechanistic framework for exploring F. candida population dynamics under realistic environmental heterogeneity in agricultural landscapes. This formal model documentation establishes the theoretical foundation and empirical basis for subsequent model implementation, evaluation, and application to questions in landscape ecology and environmental risk assessment.
- 27 April 2026
- Published by Pensoft Publishers in Agricultural and Environmental Modelling
- Vol. 8
Open AccessIntegrating spatial and environmental stressors in a population model of Folsomia candida (Collembola, Isotomidae): a Formal Model within ALMaSS framework
Collembola are key soil invertebrates that contribute to decomposition processes and nutrient cycling in agricultural ecosystems. Folsomia candida , a parthenogenetic euedaphic species, is widely used as a model organism in ecotoxicological testing, yet population-level responses to environmental stressors in spatially structured landscapes remain poorly understood. We present a formal model of F. candida developed using a subpopulation approach within the Animal, Landscape and Man Simulation System (ALMaSS). The model integrates stage-structured population dynamics with spatially explicit landscape processes, incorporating hourly weather data and daily soil moisture dynamics at 10 by 10 m resolution. Growth, development, reproduction, and mortality are driven by temperature, soil moisture, and food quality through empirically parameterised thermal performance curves and dose–response functions fitted to published experimental data. The model explicitly represents egg, juvenile, and adult life stages, with stage transitions and vital rates responding non-linearly to environmental conditions. Soil moisture effects are modelled using a novel integration of ERA5 reanalysis data, soil physical properties, and evapotranspiration to estimate surface soil water potential. Temperature and moisture effects are combined using a limiting-factor approach, reflecting current understanding of physiological constraints. The model is designed for simulations at subpopulation spatial scales of 100 m 2 , within landscapes typically 10 km × 10 km, providing a mechanistic framework for exploring F. candida population dynamics under realistic environmental heterogeneity in agricultural landscapes. This formal model documentation establishes the theoretical foundation and empirical basis for subsequent model implementation, evaluation, and application to questions in landscape ecology and environmental risk assessment.
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Dimensionality reduction of soil vis–NIR spectra: implications for soil health assessment and mapping
The measured soil spectrum in the visible and near-infrared (vis–NIR) range contains information on various soil physical properties, as well as mineralogical and chemical composition. Therefore, this approach can serve as a rapid and cost-effective alternative for assessing soil health. However, the raw soil spectrum is high-dimensional and less suitable for modeling purposes. In this study, we compared the performance of four methods, including PCA, kernel PCA (KPCA), autoencoders (AE), and convolutional autoencoders (CNN-AE), based on their reconstruction error, which directly measures information loss during dimensionality reduction. We used a comprehensive soil dataset from Denmark comprising 7,009 vis–NIR spectra covering a wide range of land uses and soil types. Our analysis shows that reconstruction error decreases as the number of latent variables (LVs) increases, as expected, due to the greater capacity to preserve information. Notably, across all latent dimensionalities, nonlinear methods consistently outperformed PCA. At low latent dimensionalities (5–10 LVs), nonlinear methods achieved approximately 45–55% lower reconstruction error than PCA. As the number of LVs increases, this performance gap decreases. However, PCA still consistently shows lower performance than the other methods. Given the strong representation of soil chemical, physical, and hydrological properties in the spectra, we mapped the latent variables across Denmark using diverse spatial predictors. The mapped latent variables captured complex, nonlinear soil–landscape relationships that reflect key mineral and organic soil components. Overall, our results suggest that spectral representations could serve as a scalable approach to support national soil mapping.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessDimensionality reduction of soil vis–NIR spectra: implications for soil health assessment and mapping
The measured soil spectrum in the visible and near-infrared (vis–NIR) range contains information on various soil physical properties, as well as mineralogical and chemical composition. Therefore, this approach can serve as a rapid and cost-effective alternative for assessing soil health. However, the raw soil spectrum is high-dimensional and less suitable for modeling purposes. In this study, we compared the performance of four methods, including PCA, kernel PCA (KPCA), autoencoders (AE), and convolutional autoencoders (CNN-AE), based on their reconstruction error, which directly measures information loss during dimensionality reduction. We used a comprehensive soil dataset from Denmark comprising 7,009 vis–NIR spectra covering a wide range of land uses and soil types. Our analysis shows that reconstruction error decreases as the number of latent variables (LVs) increases, as expected, due to the greater capacity to preserve information. Notably, across all latent dimensionalities, nonlinear methods consistently outperformed PCA. At low latent dimensionalities (5–10 LVs), nonlinear methods achieved approximately 45–55% lower reconstruction error than PCA. As the number of LVs increases, this performance gap decreases. However, PCA still consistently shows lower performance than the other methods. Given the strong representation of soil chemical, physical, and hydrological properties in the spectra, we mapped the latent variables across Denmark using diverse spatial predictors. The mapped latent variables captured complex, nonlinear soil–landscape relationships that reflect key mineral and organic soil components. Overall, our results suggest that spectral representations could serve as a scalable approach to support national soil mapping.
- Open Access
- 26 February 2026
- Published by IOP Publishing in Environmental Research Letters
- Vol. 21 (4) , 043004
Assessing the impact of nature-based solutions on soil health in sub-Saharan Africa through farmer-centred methods
Soils underpin many ecosystem services, including food production, through functions such as organic matter decomposition. These functions are increasingly threatened by soil degradation, especially in climate-vulnerable regions, such as sub-Saharan Africa, where unstable soils are prone to severe erosion. As soils continue to degrade, farmers face multiple challenges; they cannot afford accurate tests to assess soil, their livelihoods are constrained by demand for food, fuel and water, and competition for valuable resources hampers farming. Hence, there is a pressing need for accessible tools to assess soil health and methods to provide tailored advice on resilient, climate-smart agricultural management and optimal use of resources. This narrative review offers a comprehensive overview of key issues and potential solutions. We highlight tools and approaches that can support farmers to improve soil and secure livelihoods. Practical indicators and field-ready tests are evaluated, with examples from Ethiopia, but tailored to support farmers and advisors across sub-Saharan Africa and other developing countries. A wide range of tests are reviewed, including physical, biological, chemical, function and service-related tests, drawing on scientific and farmers knowledge. Science-based tests require expertise, equipment and incur costs, while locally-derived tests are affordable and seamlessly applicable. We also review Nature-based Solutions for improving soil quality, and assess them against factors such as labour, costs, and crop production. There is no single universally applicable practice; suitability depends on farmers’ priorities and circumstances. Therefore, we explore predictive methods—mechanistic, process-based soil models, data- and knowledge-driven Artificial Intelligence and systems models—to simulate the impact of practices on soil and farm dynamics. Promising approaches include hybrid approaches assimilating data, physics and knowledge through digital soil mapping. Overall, this review emphasizes the need to empower farmers with accessible tools and methods to harness Nature-based Solutions, build climate resilience and secure sustainable futures for generations ahead.
- 26 February 2026
- Published by IOP Publishing in Environmental Research Letters
- Vol. 21 (4) , 043004
Open AccessAssessing the impact of nature-based solutions on soil health in sub-Saharan Africa through farmer-centred methods
Soils underpin many ecosystem services, including food production, through functions such as organic matter decomposition. These functions are increasingly threatened by soil degradation, especially in climate-vulnerable regions, such as sub-Saharan Africa, where unstable soils are prone to severe erosion. As soils continue to degrade, farmers face multiple challenges; they cannot afford accurate tests to assess soil, their livelihoods are constrained by demand for food, fuel and water, and competition for valuable resources hampers farming. Hence, there is a pressing need for accessible tools to assess soil health and methods to provide tailored advice on resilient, climate-smart agricultural management and optimal use of resources. This narrative review offers a comprehensive overview of key issues and potential solutions. We highlight tools and approaches that can support farmers to improve soil and secure livelihoods. Practical indicators and field-ready tests are evaluated, with examples from Ethiopia, but tailored to support farmers and advisors across sub-Saharan Africa and other developing countries. A wide range of tests are reviewed, including physical, biological, chemical, function and service-related tests, drawing on scientific and farmers knowledge. Science-based tests require expertise, equipment and incur costs, while locally-derived tests are affordable and seamlessly applicable. We also review Nature-based Solutions for improving soil quality, and assess them against factors such as labour, costs, and crop production. There is no single universally applicable practice; suitability depends on farmers’ priorities and circumstances. Therefore, we explore predictive methods—mechanistic, process-based soil models, data- and knowledge-driven Artificial Intelligence and systems models—to simulate the impact of practices on soil and farm dynamics. Promising approaches include hybrid approaches assimilating data, physics and knowledge through digital soil mapping. Overall, this review emphasizes the need to empower farmers with accessible tools and methods to harness Nature-based Solutions, build climate resilience and secure sustainable futures for generations ahead.
- Open Access
- 21 January 2026
- Published by Elsevier in Journal of Hydrology
- Vol. 668
A differentiable hybrid modeling approach for learning soil water retention mechanisms from partial knowledge and data
- 21 January 2026
- Published by Elsevier in Journal of Hydrology
- Vol. 668
Open AccessA differentiable hybrid modeling approach for learning soil water retention mechanisms from partial knowledge and data
- Open Access
- 22 April 2025
- Published by Wiley
A Novel Physics-Informed Neural Network for Retrieving the Continuous Soil Particle Size Distribution from Vis-NIR Spectra
We present a novel physics-informed neural network (PINN) for the direct estimation of the continuous soil particle size distribution curve (PSD) from visible–near infrared (vis-NIR) reflectance measurements. Unlike traditional methods that focus on predicting soil fractions (e.g., sand, silt, and clay), the PINN takes in the soil particle size as a predictor and learns a continuous, differentiable, and non-parametric representation of the cumulative PSD by integrating both measurements and physical constraints imposed during training. This approach eliminates the need for harmonization and interpolation of measurements originating from different soil texture classification systems and allows the model to be trained on datasets with varying number of measurements per sample. Performance evaluation on 30% of the 2,777 studied samples showed that the PINN achieved an RMSE of 6.77% and an R² of 0.97 in predicting the cumulative PSD fraction. For the texture fractions, the model achieved RMSE values of 4.72%, 3.06%, and 2.75% for sand, silt, and clay, respectively. A comparison with a physics-agnostic (i.e., physics-uninformed) version of the model revealed that both approaches performed similarly in terms of RMSE and R². However, the physics-agnostic model violated physical constraints even in data-rich scenarios. In contrast, the PSD obtained from the PINN maintained its physical integrity even under data sparsity conditions and consistently produced non-negative, monotonically increasing predictions that sum to 100% at the largest particle size.
- 22 April 2025
- Published by Wiley
Open AccessA Novel Physics-Informed Neural Network for Retrieving the Continuous Soil Particle Size Distribution from Vis-NIR Spectra
We present a novel physics-informed neural network (PINN) for the direct estimation of the continuous soil particle size distribution curve (PSD) from visible–near infrared (vis-NIR) reflectance measurements. Unlike traditional methods that focus on predicting soil fractions (e.g., sand, silt, and clay), the PINN takes in the soil particle size as a predictor and learns a continuous, differentiable, and non-parametric representation of the cumulative PSD by integrating both measurements and physical constraints imposed during training. This approach eliminates the need for harmonization and interpolation of measurements originating from different soil texture classification systems and allows the model to be trained on datasets with varying number of measurements per sample. Performance evaluation on 30% of the 2,777 studied samples showed that the PINN achieved an RMSE of 6.77% and an R² of 0.97 in predicting the cumulative PSD fraction. For the texture fractions, the model achieved RMSE values of 4.72%, 3.06%, and 2.75% for sand, silt, and clay, respectively. A comparison with a physics-agnostic (i.e., physics-uninformed) version of the model revealed that both approaches performed similarly in terms of RMSE and R². However, the physics-agnostic model violated physical constraints even in data-rich scenarios. In contrast, the PSD obtained from the PINN maintained its physical integrity even under data sparsity conditions and consistently produced non-negative, monotonically increasing predictions that sum to 100% at the largest particle size.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Toward Pan-European Mapping of Soil Water Retention Curves Using a Physics-informed Machine Learning Approach: Insights from Denmark
The soil water retention curve (SWRC) is a fundamental soil property that provides information about soil structure, soil texture, plant water availability, drainage, and compaction, and is therefore highly linked to soil functions and soil health. Current large-scale digital maps of the SWRC are typically developed indirectly through a two-step process: i) the development of pedotransfer functions (PTFs) that establish relationships between basic soil properties such as textural fractions, bulk density, organic matter content, and parameters of well-known SWRC models, and ii) the application of these PTFs to basic soil property maps at various scales. This presentation introduces a novel, physically constrained machine learning approach for directly mapping the entire SWRC from saturation to oven-dryness. Unlike previous studies, our new approach neither relies on PTFs nor is limited to a specific SWRC model. Instead, it estimates a non-specific form of the SWRC, learned from both measurements and physical constraints. Applying this method to 1,261 soil profiles across Denmark, encompassing 4,747 measured SWRCs, demonstrates its superior performance compared to established methods. The new approach enables the aggregation of datasets with sparse and incomplete SWRC measurements, which are typically unusable with conventional methods. This capability maximizes spatial coverage and reduces uncertainties in the final predicted maps. Additionally, the new approach addresses the commonly observed imbalance between wet and dry-end measurements in large SWRC datasets. Following a detailed report on the results of our approach for Denmark, we discuss ongoing efforts and progress toward applying this method to SWRC mapping at the European scale.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessToward Pan-European Mapping of Soil Water Retention Curves Using a Physics-informed Machine Learning Approach: Insights from Denmark
The soil water retention curve (SWRC) is a fundamental soil property that provides information about soil structure, soil texture, plant water availability, drainage, and compaction, and is therefore highly linked to soil functions and soil health. Current large-scale digital maps of the SWRC are typically developed indirectly through a two-step process: i) the development of pedotransfer functions (PTFs) that establish relationships between basic soil properties such as textural fractions, bulk density, organic matter content, and parameters of well-known SWRC models, and ii) the application of these PTFs to basic soil property maps at various scales. This presentation introduces a novel, physically constrained machine learning approach for directly mapping the entire SWRC from saturation to oven-dryness. Unlike previous studies, our new approach neither relies on PTFs nor is limited to a specific SWRC model. Instead, it estimates a non-specific form of the SWRC, learned from both measurements and physical constraints. Applying this method to 1,261 soil profiles across Denmark, encompassing 4,747 measured SWRCs, demonstrates its superior performance compared to established methods. The new approach enables the aggregation of datasets with sparse and incomplete SWRC measurements, which are typically unusable with conventional methods. This capability maximizes spatial coverage and reduces uncertainties in the final predicted maps. Additionally, the new approach addresses the commonly observed imbalance between wet and dry-end measurements in large SWRC datasets. Following a detailed report on the results of our approach for Denmark, we discuss ongoing efforts and progress toward applying this method to SWRC mapping at the European scale.
- Top Cited PapersOpen Access
- 14 March 2025
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 61 (3)
Physics‐Informed Neural Networks for Estimating a Continuous Form of the Soil Water Retention Curve From Basic Soil Properties
This paper presents a novel physics‐informed neural network (PINN) approach for developing pedotransfer functions (PTFs) to predict continuous soil water retention curves (SWRCs) based on soil textural fractions, organic carbon content, and bulk density. In contrast to conventional parametric PTFs developed for specific SWRC models, the PINN learns a non‐specific form of the SWRC from both measurements and physical constraints imposed during the training process. This approach allows the estimated SWRC to maintain its physical integrity from saturation to oven‐dry conditions, even in scenarios with sparse data. The new approach is particularly effective for tackling the challenges encountered in developing PTFs on large SWRC data sets, which often have an imbalance toward the wet‐end () and include numerous samples with limited and unevenly distributed measurements, many of which do not meet the requirements to fit traditional SWRC models. We compared the performance of the PINN with that of a conventional physics‐agnostic neural network using a data set of 4,200 soil samples. While both networks performed similarly at the wet‐end where data are abundant, with RMSE values of around 0.041 m3 m−3, the PINN excelled at the dry‐end () where data are sparse and unevenly distributed, achieving a normalized RMSE of 0.172 (RMSE = 0.0045 m3 m−3) compared to a normalized RMSE of 0.522 (RMSE = 0.0136 m3 m−3) for the conventional neural network. The SWRC derived from the PINN is differentiable with respect to matric potential, making it well‐suited for integration into models of water flow in the unsaturated zone.
- 14 March 2025
- Published by American Geophysical Union (AGU) in Water Resources Research
- Vol. 61 (3)
Top Cited PapersOpen AccessPhysics‐Informed Neural Networks for Estimating a Continuous Form of the Soil Water Retention Curve From Basic Soil Properties
This paper presents a novel physics‐informed neural network (PINN) approach for developing pedotransfer functions (PTFs) to predict continuous soil water retention curves (SWRCs) based on soil textural fractions, organic carbon content, and bulk density. In contrast to conventional parametric PTFs developed for specific SWRC models, the PINN learns a non‐specific form of the SWRC from both measurements and physical constraints imposed during the training process. This approach allows the estimated SWRC to maintain its physical integrity from saturation to oven‐dry conditions, even in scenarios with sparse data. The new approach is particularly effective for tackling the challenges encountered in developing PTFs on large SWRC data sets, which often have an imbalance toward the wet‐end () and include numerous samples with limited and unevenly distributed measurements, many of which do not meet the requirements to fit traditional SWRC models. We compared the performance of the PINN with that of a conventional physics‐agnostic neural network using a data set of 4,200 soil samples. While both networks performed similarly at the wet‐end where data are abundant, with RMSE values of around 0.041 m3 m−3, the PINN excelled at the dry‐end () where data are sparse and unevenly distributed, achieving a normalized RMSE of 0.172 (RMSE = 0.0045 m3 m−3) compared to a normalized RMSE of 0.522 (RMSE = 0.0136 m3 m−3) for the conventional neural network. The SWRC derived from the PINN is differentiable with respect to matric potential, making it well‐suited for integration into models of water flow in the unsaturated zone.
- Open Access
- 20 January 2025
- Published by Copernicus GmbH
A Physics-Driven Spectroscopic Approach for Rapid Estimation of the Soil Water Retention Curve
The knowledge about soil hydraulic properties (i.e., water retention and hydraulic conductivity characteristics) is of relevance for accurate determination of land surface fluxes. Yet, the laborious and time-consuming process of measuring the soil water retention curve (SWRC) with conventional laboratory methods poses a challenge. In addition, the measured soil water content and matric potential pairs obtained with standard methods are often fragmentary and consist of only a limited number of measurements across the desired soil water content range. Proximal and remote sensing methods are rapid and cost-efficient alternatives to quantify soil attributes across different scales. However, past studies that centered around proximal and remote sensing of soil hydraulic functions mainly rely on statistical relationships and a physically-based method is still lacking. In this presentation, we introduce an innovative physics-based laboratory method that allows the direct estimation of the complete SWRC across the entire range from saturated to dry conditions. The inputs to the model include measured data pairs of soil water content and reflectance within the shortwave infrared domain. The fundamental hypothesis behind the new method is that the soil reflectance spectra are a function of both soil water content and the pore scale distribution of capillary and adsorbed soil water. The performance of the proposed model was evaluated for 21 soils that vastly differ in physical and hydraulic properties. The RMSE and R-squared between retrieved and measured water contents at various matric potentials were found to be 0.03 m³ m⁻³ and 0.96, respectively, indicating the good performance of the proposed method. The results suggest that the new method is a rapid and efficient alternative to established laboratory measurement methods.
- 20 January 2025
- Published by Copernicus GmbH
Open AccessA Physics-Driven Spectroscopic Approach for Rapid Estimation of the Soil Water Retention Curve
The knowledge about soil hydraulic properties (i.e., water retention and hydraulic conductivity characteristics) is of relevance for accurate determination of land surface fluxes. Yet, the laborious and time-consuming process of measuring the soil water retention curve (SWRC) with conventional laboratory methods poses a challenge. In addition, the measured soil water content and matric potential pairs obtained with standard methods are often fragmentary and consist of only a limited number of measurements across the desired soil water content range. Proximal and remote sensing methods are rapid and cost-efficient alternatives to quantify soil attributes across different scales. However, past studies that centered around proximal and remote sensing of soil hydraulic functions mainly rely on statistical relationships and a physically-based method is still lacking. In this presentation, we introduce an innovative physics-based laboratory method that allows the direct estimation of the complete SWRC across the entire range from saturated to dry conditions. The inputs to the model include measured data pairs of soil water content and reflectance within the shortwave infrared domain. The fundamental hypothesis behind the new method is that the soil reflectance spectra are a function of both soil water content and the pore scale distribution of capillary and adsorbed soil water. The performance of the proposed model was evaluated for 21 soils that vastly differ in physical and hydraulic properties. The RMSE and R-squared between retrieved and measured water contents at various matric potentials were found to be 0.03 m³ m⁻³ and 0.96, respectively, indicating the good performance of the proposed method. The results suggest that the new method is a rapid and efficient alternative to established laboratory measurement methods.
- Open Access
Soil Science-Informed Machine Learning
- Open Access
- 24 June 2024
- Published by Wiley
Physics-Informed Neural Networks for Estimating a Continuous Form of the Soil Water Retention Curve from Basic Soil Properties
The soil water retention curve (SWRC) is essential for describing water and energy exchange processes at the interface between the solid earth and the atmosphere. Despite its importance, measuring the SWRC using standard laboratory methods is challenging and time-consuming. This paper presents a novel physics-informed neural network (PINN) approach for developing pedotransfer functions (PTFs) to predict continuous SWRCs based on soil texture, organic carbon content, and dry bulk density. In contrast to conventional parametric PTFs developed for specific SWRC models, the PINN learns a non-specific form of the SWRC by effectively integrating both measurements and physical constraints into the training process. This approach allows the estimated SWRC to maintain its physical integrity from saturation to oven-dry conditions, even in scenarios with sparse data. The new approach is particularly effective for tackling the challenges encountered in developing PTFs on large SWRC datasets, which often have an imbalance towards the wet-end and include numerous samples with limited and unevenly distributed measurements. We compared the performance of the PINN with that of a conventional physics-agnostic neural network using a dataset of 4200 soil samples. While both networks performed similarly at the wet-end where data are abundant, the PINN excelled at the dry-end where data are sparse and unevenly distributed, achieving a normalized RMSE of 0.172 compared to 0.522 for the conventional neural network. The SWRC derived from the PINN is differentiable with respect to the matric potential and can be seamlessly integrated into the governing equations of water flow in the unsaturated zone.
- 24 June 2024
- Published by Wiley
Open AccessPhysics-Informed Neural Networks for Estimating a Continuous Form of the Soil Water Retention Curve from Basic Soil Properties
The soil water retention curve (SWRC) is essential for describing water and energy exchange processes at the interface between the solid earth and the atmosphere. Despite its importance, measuring the SWRC using standard laboratory methods is challenging and time-consuming. This paper presents a novel physics-informed neural network (PINN) approach for developing pedotransfer functions (PTFs) to predict continuous SWRCs based on soil texture, organic carbon content, and dry bulk density. In contrast to conventional parametric PTFs developed for specific SWRC models, the PINN learns a non-specific form of the SWRC by effectively integrating both measurements and physical constraints into the training process. This approach allows the estimated SWRC to maintain its physical integrity from saturation to oven-dry conditions, even in scenarios with sparse data. The new approach is particularly effective for tackling the challenges encountered in developing PTFs on large SWRC datasets, which often have an imbalance towards the wet-end and include numerous samples with limited and unevenly distributed measurements. We compared the performance of the PINN with that of a conventional physics-agnostic neural network using a dataset of 4200 soil samples. While both networks performed similarly at the wet-end where data are abundant, the PINN excelled at the dry-end where data are sparse and unevenly distributed, achieving a normalized RMSE of 0.172 compared to 0.522 for the conventional neural network. The SWRC derived from the PINN is differentiable with respect to the matric potential and can be seamlessly integrated into the governing equations of water flow in the unsaturated zone.