Combining Machine Learning Research Articles

Computation of Real-Fluid Thermophysical Properties Using a Neural Network Approach Implemented in OpenFOAM

Machine learning based on neural networks facilitates data-driven techniques for handling large amounts of data, either obtained through experiments or simulations at multiple spatio-temporal scales, thereby finding the hidden patterns underlying these data and promoting efficient research methods. The main purpose of this paper is to extend the capabilities of a new solver called realFluidReactingNNFoam, under development at the University of Perugia, in OpenFOAM with a neural network algorithm for replacing complex real-fluid thermophysical property evaluations, using the approach of coupling OpenFOAM and Python-trained neural network models. Currently, neural network models are trained against data generated using the Peng–Robinson equation of state assuming a mixture’s frozen temperature. The OpenFOAM solver, where needed, calls the neural network models in each grid cell with appropriate inputs, and the returned results are used and stored in suitable OpenFOAM data structures. Such inference for thermophysical properties is achieved via the “Neural Network Inference in C made Easy (NNICE)” library, which proved to be very efficient and robust. The overall model is validated considering a liquid-rocket benchmark comprised of liquid-oxygen (LOX) and gaseous-hydrogen (GH2) streams. The model accounts for real-fluid thermodynamics and transport properties, making use of the Peng–Robinson equation of state and the Chung transport model. First, the development of a real-fluid model with an artificial neural network is described in detail. Then, the numerical results of the transcritical mixing layer (LOX/GH2) benchmark are presented and analyzed in terms of accuracy and computational efficiency. The results of the overall implementation indicate that the combined OpenFOAM and machine learning approach provides a speed-up factor higher than seven, while preserving the original solver accuracy.

A Fourier neural operator-based lightweight machine learning framework for topology optimization

The combination of machine learning and topology optimization has preliminarily mitigated the enormous computational costs incurred during optimization processes. Its powerful design capabilities are warmly welcomed by many scholars, but this strategy still faces many problems, such as unclear structural boundaries and high dependence on big data. Here, we propose a new machine learning topology optimization paradigm that learns the mapping relationship between the initial and optimized Level set functions in the infinite-dimensional function space under the Fourier neural operator to truly parameterize the structural model and its mechanical information, thereby achieving purely data-driven topology optimization. A base model trained under the proposed approach could obtain a topological structure with clear and smooth structural boundaries within 0.4 seconds, making it faster than traditional methods by tens of times. In addition, a lightweight retraining strategy under the base model was proposed. Four additional numerical examples were used to demonstrate that this strategy requires only 100 or 200 new samples and 3-6 seconds of retraining time to attain prediction capabilities under different boundary conditions, different meshes (resolution), various initial level set functions, and multiple loads. In addition, we discussed all the experiments and found that most of the mechanical properties of the results were close to the ground truths, while some results were even better than those of the traditional method. Although we only used the maximum structural stiffness as an example, this method can be extended to different objective functions.

Robust human locomotion and localization activity recognition over multisensory.

Human activity recognition (HAR) plays a pivotal role in various domains, including healthcare, sports, robotics, and security. With the growing popularity of wearable devices, particularly Inertial Measurement Units (IMUs) and Ambient sensors, researchers and engineers have sought to take advantage of these advances to accurately and efficiently detect and classify human activities. This research paper presents an advanced methodology for human activity and localization recognition, utilizing smartphone IMU, Ambient, GPS, and Audio sensor data from two public benchmark datasets: the Opportunity dataset and the Extrasensory dataset. The Opportunity dataset was collected from 12 subjects participating in a range of daily activities, and it captures data from various body-worn and object-associated sensors. The Extrasensory dataset features data from 60 participants, including thousands of data samples from smartphone and smartwatch sensors, labeled with a wide array of human activities. Our study incorporates novel feature extraction techniques for signal, GPS, and audio sensor data. Specifically, for localization, GPS, audio, and IMU sensors are utilized, while IMU and Ambient sensors are employed for locomotion activity recognition. To achieve accurate activity classification, state-of-the-art deep learning techniques, such as convolutional neural networks (CNN) and long short-term memory (LSTM), have been explored. For indoor/outdoor activities, CNNs are applied, while LSTMs are utilized for locomotion activity recognition. The proposed system has been evaluated using the k-fold cross-validation method, achieving accuracy rates of 97% and 89% for locomotion activity over the Opportunity and Extrasensory datasets, respectively, and 96% for indoor/outdoor activity over the Extrasensory dataset. These results highlight the efficiency of our methodology in accurately detecting various human activities, showing its potential for real-world applications. Moreover, the research paper introduces a hybrid system that combines machine learning and deep learning features, enhancing activity recognition performance by leveraging the strengths of both approaches.

Enhancing Maize Yield Simulations in Regional China Using Machine Learning and Multi-Data Resources

Improved agricultural production systems, together with increased grain yield, are essential to feed the growing global population in the 21st century. Global gridded crop models (GGCMs) have been extensively used to assess crop production and yield simulation on a large geographical scale. However, GGCMs are less effective when they are used on a finer scale, significantly limiting the precision in capturing the yearly maize yield. To address this issue, we propose a relatively more advanced approach that downsizes GGCMs by combining machine learning and crop modeling to enhance the accuracy of maize yield simulations on a regional scale. In this study, we combined the random forest algorithm with multiple data sources, trained the algorithm on low-resolution maize yield simulations from GGCMs, and applied it to a finer spatial resolution on a regional scale in China. We evaluated the performance of the eight GGCMs by utilizing a total of 1046 county-level maize yield data available over a 30-year period (1980–2010). Our findings reveal that the downscaled models created for maize yield simulations exhibited a remarkable level of accuracy (R2 ≥ 0.9, MAE < 0.5 t/ha, RMSE < 0.75 t/ha). The original GGCMs performed poorly in simulating county-level maize yields in China, and the improved GGCMs in our study captured an additional 17% variability in the county-level maize yields in China. Additionally, by optimizing nitrogen management strategies, we identified an average maize yield gap at the county level in China ranging from 0.47 to 1.82 t/ha, with the south maize region exhibiting the highest yield gap. Our study demonstrates the high effectiveness of machine learning methods for the spatial downscaling of crop models, significantly improving GGCMs’ performance in county-level maize yield simulations.

MicroRNA-guided drug discovery for mitigating persistent pulmonary complications in critical COVID-19 survivors: A longitudinal pilot study.

The post-acute sequelae of SARS-CoV-2 infection pose a significant global challenge, with nearly 50% of critical COVID-19 survivors manifesting persistent lung abnormalities. The lack of understanding about the molecular mechanisms and effective treatments hampers their management. Here, we employed microRNA (miRNA) profiling to decipher the systemic molecular underpinnings of the persistent pulmonary complications. We conducted a longitudinal investigation including 119 critical COVID-19 survivors. A comprehensive pulmonary evaluation was performed in the short-term (median = 94.0days after hospital discharge) and long-term (median = 358 days after hospital discharge). Plasma miRNAs were quantified at the short-term evaluation using the gold-standard technique, RT-qPCR. The analyses combined machine learning feature selection techniques with bioinformatic investigations. Two additional datasets were incorporated for validation. In the short-term, 84% of the survivors exhibited impaired lung diffusion (DLCO < 80% of predicted). One year post-discharge, 54.4% of this patient subgroup still presented abnormal DLCO . Four feature selection methods identified two specific miRNAs, miR-9-5p and miR-486-5p, linked to persistent lung dysfunction. The downstream experimentally validated targetome included 1473 genes, with heterogeneous enriched pathways associated with inflammation, angiogenesis and cell senescence. Validation studies using RNA-sequencing and proteomic datasets emphasized the pivotal roles of cell migration and tissue repair in persistent lung dysfunction. The repositioning potential of the miRNA targets was limited. Our study reveals early mechanistic pathways contributing to persistent lung dysfunction in critical COVID-19 survivors, offering a promising approach for the development of targeted disease-modifying agents.

Hybrid river stage forecasting based on machine learning with empirical mode decomposition

The river stage is certainly an important indicator of how the water level fluctuates overtime. Continuous control of the water stage can help build an early warning indicator of floods along rivers and streams. Hence, forecasting river stages up to several days in advance is very important and constitutes a challenging task. Over the past few decades, the use of machine learning paradigm to investigate complex hydrological systems has gained significant importance, and forecasting river stage is one of the promising areas of investigations. Traditional in situ measurements, which are sometime restricted by the existing of several handicaps especially in terms of regular access to any points alongside the streams and rivers, can be overpassed by the use of modeling approaches. For more accurate forecasting of river stages, we suggest a new modeling framework based on machine learning. A hybrid forecasting approach was developed by combining machine learning techniques, namely random forest regression (RFR), bootstrap aggregating (Bagging), adaptive boosting (AdaBoost), and artificial neural network (ANN), with empirical mode decomposition (EMD) to provide a robust forecasting model. The singles models were first applied using only the river stage data without preprocessing, and in the following step, the data were decomposed into several intrinsic mode functions (IMF), which were then used as new input variables. According to the obtained results, the proposed models showed improved results compared to the standard RFR without EMD for which, the error performances metrics were drastically reduced, and the correlation index was increased remarkably and great changes in models’ performances have taken place. The RFR_EMD, Bagging_EMD, and AdaBoost_EMD were less accurate than the ANN_EMD model, which had higher R≈0.974, NSE≈0.949, RMSE≈0.330 and MAE≈0.175 values. While the RFR_EMD and the Bagging_EMD were relatively equal and exhibited the same accuracies higher than the AdaBoost_EMD, the superiority of the ANN_EMD was obvious. The proposed model shows the potential for combining signal decomposition with machine learning, which can serve as a basis for new insights into river stage forecasting.

Machine Learning for Polaritonic Chemistry: Accessing Chemical Kinetics.

Altering chemical reactivity and material structure in confined optical environments is on the rise, and yet, a conclusive understanding of the microscopic mechanisms remains elusive. This originates mostly from the fact that accurately predicting vibrational and reactive dynamics for soluted ensembles of realistic molecules is no small endeavor, and adding (collective) strong light-matter interaction does not simplify matters. Here, we establish a framework based on a combination of machine learning (ML) models, trained using density-functional theory calculations and molecular dynamics to accelerate such simulations. We then apply this approach to evaluate strong coupling, changes in reaction rate constant, and their influence on enthalpy and entropy for the deprotection reaction of 1-phenyl-2-trimethylsilylacetylene, which has been studied previously both experimentally and using ab initio simulations. While we find qualitative agreement with critical experimental observations, especially with regard to the changes in kinetics, we also find differences in comparison with previous theoretical predictions. The features for which the ML-accelerated and ab initio simulations agree show the experimentally estimated kinetic behavior. Conflicting features indicate that a contribution of dynamic electronic polarization to the reaction process is more relevant than currently believed. Our work demonstrates the practical use of ML for polaritonic chemistry, discusses limitations of common approximations, and paves the way for a more holistic description of polaritonic chemistry.

Using photographs and deep neural networks to understand flowering phenology and diversity in mountain meadows

AbstractMountain meadows are an essential part of the alpine–subalpine ecosystem; they provide ecosystem services like pollination and are home to diverse plant communities. Changes in climate affect meadow ecology on multiple levels, for example, by altering growing season dynamics. Tracking the effects of climate change on meadow diversity through the impacts on individual species and overall growing season dynamics is critical to conservation efforts. Here, we explore how to combine crowd‐sourced camera images with machine learning to quantify flowering species richness across a range of elevations in alpine meadows located in Mt. Rainier National Park, Washington, USA. We employed three machine‐learning techniques (Mask R‐CNN, RetinaNet and YOLOv5) to detect wildflower species in images taken during two flowering seasons. We demonstrate that deep learning techniques can detect multiple species, providing information on flowering richness in photographed meadows. The results indicate higher richness just above the tree line for most of the species, which is comparable with patterns found using field studies. We found that the two‐stage detector Mask R‐CNN was more accurate than single‐stage detectors like RetinaNet and YOLO, with the Mask R‐CNN network performing best overall with mean average precision (mAP) of 0.67 followed by RetinaNet (0.5) and YOLO (0.4). We found that across the methods using anchor box variations in multiples of 16 led to enhanced accuracy. We also show that detection is possible even when pictures are interspersed with complex backgrounds and are not in focus. We found differential detection rates depending on species abundance, with additional challenges related to similarity in flower characteristics, labeling errors and occlusion issues. Despite these potential biases and limitations in capturing flowering abundance and location‐specific quantification, accuracy was notable considering the complexity of flower types and picture angles in this dataset. We, therefore, expect that this approach can be used to address many ecological questions that benefit from automated flower detection, including studies of flowering phenology and floral resources, and that this approach can, therefore, complement a wide range of ecological approaches (e.g., field observations, experiments, community science, etc.). In all, our study suggests that ecological metrics like floral richness can be efficiently monitored by combining machine learning with easily accessible publicly curated datasets (e.g., Flickr, iNaturalist).

Combining machine learning and remote sensing-integrated crop modeling for rice and soybean crop simulation.

Machine learning (ML) techniques offer a promising avenue for improving the integration of remote sensing data into mathematical crop models, thereby enhancing crop growth prediction accuracy. A critical variable for this integration is the leaf area index (LAI), which can be accurately assessed using proximal or remote sensing data based on plant canopies. This study aimed to (1) develop a machine learning-based method for estimating the LAI in rice and soybean crops using proximal sensing data and (2) evaluate the performance of a Remote Sensing-Integrated Crop Model (RSCM) when integrated with the ML algorithms. To achieve these objectives, we analyzed rice and soybean datasets to identify the most effective ML algorithms for modeling the relationship between LAI and vegetation indices derived from canopy reflectance measurements. Our analyses employed a variety of ML regression models, including ridge, lasso, support vector machine, random forest, and extra trees. Among these, the extra trees regression model demonstrated the best performance, achieving test scores of 0.86 and 0.89 for rice and soybean crops, respectively. This model closely replicated observed LAI values under different nitrogen treatments, achieving Nash-Sutcliffe efficiencies of 0.93 for rice and 0.97 for soybean. Our findings show that incorporating ML techniques into RSCM effectively captures seasonal LAI variations across diverse field management practices, offering significant potential for improving crop growth and productivity monitoring.

AI4OPT: AI Institute for Advances in Optimization

AbstractThis article is a short introduction to AI4OPT, the NSF AI Institute for Advances in Optimization. AI4OPT fuses AI and optimization, inspired by societal challenges in supply chains, energy systems, chip design and manufacturing, and sustainable food systems. By combining machine learning and mathematical optimization, AI4OPT strives to develop AI‐assisted optimization systems that bring orders of magnitude improvements in efficiency, perform accurate uncertainty quantification, and address challenges in resiliency and sustainability. AI4OPT also applies its “teaching the teachers” philosophy to provide longitudinal educational pathways in AI for engineering.

Innovative Neuroimaging Biomarker Distinction of Major Depressive Disorder and Bipolar Disorder through Structural Connectome Analysis and Machine Learning Models.

Major depressive disorder (MDD) and bipolar disorder (BD) share clinical features, which complicates their differentiation in clinical settings. This study proposes an innovative approach that integrates structural connectome analysis with machine learning models to discern individuals with MDD from individuals with BD. High-resolution MRI images were obtained from individuals diagnosed with MDD or BD and from HCs. Structural connectomes were constructed to represent the complex interplay of brain regions using advanced graph theory techniques. Machine learning models were employed to discern unique connectivity patterns associated with MDD and BD. At the global level, both BD and MDD patients exhibited increased small-worldness compared to the HC group. At the nodal level, patients with BD and MDD showed common differences in nodal parameters primarily in the right amygdala and the right parahippocampal gyrus when compared with HCs. Distinctive differences were found mainly in prefrontal regions for BD, whereas MDD was characterized by abnormalities in the left thalamus and default mode network. Additionally, the BD group demonstrated altered nodal parameters predominantly in the fronto-limbic network when compared with the MDD group. Moreover, the application of machine learning models utilizing structural brain parameters demonstrated an impressive 90.3% accuracy in distinguishing individuals with BD from individuals with MDD. These findings demonstrate that combined structural connectome and machine learning enhance diagnostic accuracy and may contribute valuable insights to the understanding of the distinctive neurobiological signatures of these psychiatric disorders.

Machine learning based lightning parameterizations for CONUS

Abstract The diversity in the lightning parameterizations for numerical weather and climate models causes a considerable uncertainty in lightning prediction. In this study, we take a data-driven approach to address the lightning parameterization problem, by combining machine learning (ML) techniques with the rich lightning observations from World-Wide Lightning Location Network. Three ML algorithms are trained over the Contiguous United States (CONUS) to predict lightning stroke density in a 1° box based on the information about the atmospheric variables in the same grid (local), or over the entire CONUS (non-local). The performance of the ML-based lightning schemes is examined and compared with that of a simple, conventional lightning parameterization scheme of Romps et al. (2014). We find that all ML-based lightning schemes exhibit a performance that are superior to that of the conventional scheme in the regions and in the seasons with climatologically higher lightning stroke density. To the west of Rocky Mountains, the non-local ML lightning scheme achieves the best overall performance, with lightning stroke density predictions being 70% more accurate than the conventional scheme. Our results suggest that the ML-based approaches have the potential to improve the representation of lightning and other types of extreme weather events in the weather and climate models.

A 2020 forest age map for China with 30 m resolution

Abstract. A high-resolution, spatially explicit forest age map is essential for quantifying forest carbon stocks and carbon sequestration potential. Prior attempts to estimate forest age on a national scale in China have been limited by sparse resolution and incomplete coverage of forest ecosystems, attributed to complex species composition, extensive forest areas, insufficient field measurements, and inadequate methods. To address these challenges, we developed a framework that combines machine learning algorithms (MLAs) and remote sensing time series analysis for estimating the age of China's forests. Initially, we identify and develop the optimal MLAs for forest age estimation across various vegetation divisions based on forest height, climate, terrain, soil, and forest-age field measurements, utilizing these MLAs to ascertain forest age information. Subsequently, we apply the LandTrendr time series analysis to detect forest disturbances from 1985 to 2020, with the time since the last disturbance serving as a proxy for forest age. Ultimately, the forest age data derived from LandTrendr are integrated with the result of MLAs to produce the 2020 forest age map of China. Validation against independent field plots yielded an R2 ranging from 0.51 to 0.63. On a national scale, the average forest age is 56.1 years (standard deviation of 32.7 years). The Qinghai–Tibet Plateau alpine vegetation zone possesses the oldest forest with an average of 138.0 years, whereas the forest in the warm temperate deciduous-broadleaf forest vegetation zone averages only 28.5 years. This 30 m-resolution forest age map offers crucial insights for comprehensively understanding the ecological benefits of China's forests and to sustainably manage China's forest resources. The map is available at https://doi.org/10.5281/zenodo.8354262 (Cheng et al., 2023a).

Structure Discovery in Atomic Force Microscopy Imaging of Ice.

The interaction of water with surfaces is crucially important in a wide range of natural and technological settings. In particular, at low temperatures, unveiling the atomistic structure of adsorbed water clusters would provide valuable data for understanding the ice nucleation process. Using high-resolution atomic force microscopy (AFM) and scanning tunneling microscopy, several studies have demonstrated the presence of water pentamers, hexamers, and heptamers (and of their combinations) on a variety of metallic surfaces, as well as the initial stages of 2D ice growth on an insulating surface. However, in all of these cases, the observed structures were completely flat, providing a relatively straightforward path to interpretation. Here, we present high-resolution AFM measurements of several water clusters on Au(111) and Cu(111), whose understanding presents significant challenges due to both their highly 3D configuration and their large size. For each of them, we use a combination of machine learning, atomistic modeling with neural network potentials, and statistical sampling to propose an underlying atomic structure, finally comparing its AFM simulated images to the experimental ones. These results provide insights into the early phases of ice formation, which is a ubiquitous phenomenon ranging from biology to astrophysics.

Predicting soil cone index and assessing suitability for wind and solar farm development in using machine learning techniques

This study proposes a novel approach that combines machine learning models to predict soil compaction using the soil cone index values. The methodology incorporates support vector regression (SVR) to gather input data on key soil parameters, and the output data from SVR are used as inputs for additional machine learning techniques such as Gradient Boosting, Decision Tree, Artificial Neural Networks, and Adaptive Neuro-Fuzzy Inference System. Evaluation of Artificial Intelligent techniques shows that the XGBoost model outperforms others, exhibiting high accuracy and reliability with low mean square error and high correlation coefficient. The effectiveness of the XGBoost model has implications for soil management, agricultural productivity, and land suitability evaluations, particularly for renewable energy projects. By integrating advanced AI techniques, stakeholders can make informed decisions about land use planning, sustainable farming practices, and the feasibility of renewable energy installations. Overall, this research contributes to soil science by demonstrating the potential of AI techniques, specifically the XGBoost model, in accurately predicting soil compaction and supporting optimal soil management practices.

Combining Machine Learning and Edge Computing: Opportunities, Challenges, Platforms, Frameworks, and Use Cases

In recent years, we have been observing the rapid growth and adoption of IoT-based systems, enhancing multiple areas of our lives. Concurrently, the utilization of machine learning techniques has surged, often for similar use cases as those seen in IoT systems. In this survey, we aim to focus on the combination of machine learning and the edge computing paradigm. The presented research commences with the topic of edge computing, its benefits, such as reduced data transmission, improved scalability, and reduced latency, as well as the challenges associated with this computing paradigm, like energy consumption, constrained devices, security, and device fleet management. It then presents the motivations behind the combination of machine learning and edge computing, such as the availability of more powerful edge devices, improving data privacy, reducing latency, or lowering reliance on centralized services. Then, it describes several edge computing platforms, with a focus on their capability to enable edge intelligence workflows. It also reviews the currently available edge intelligence frameworks and libraries, such as TensorFlow Lite or PyTorch Mobile. Afterward, the paper focuses on the existing use cases for edge intelligence in areas like industrial applications, healthcare applications, smart cities, environmental monitoring, or autonomous vehicles.

Determination of Novel SARS-CoV-2 Inhibitors by Combination of Machine Learning and Molecular Modeling Methods.

Within the scope of the project, this study aimed to find novel inhibitors by combining computational methods. In order to design inhibitors, it was aimed to produce molecules similar to the RdRp inhibitor drug Favipiravir by using the deep learning method. For this purpose, a Trained Neural Network (TNN) was used to produce 75 molecules similar to Favipiravir by using Simplified Molecular Input Line Entry System (SMILES) representations. The binding properties of molecules to Viral RNA-dependent RNA polymerase (RdRp) were studied by using molecular docking studies. To confirm the accuracy of this method, compounds were also tested against 3CL protease (3CLpro), which is another important enzyme for the progression of SARS-CoV-2. Compounds having better binding energies and RMSD values than favipiravir were searched with similarity analysis on the ChEMBL drug database in order to find similar structures with RdRp and 3CLpro inhibitory activities. A similarity search found new 200 potential RdRp and 3CLpro inhibitors structurally similar to produced molecules, and these compounds were again evaluated for their receptor interactions with molecular docking studies. Compounds showed better interaction with RdRp protease than 3CLpro. This result presented that artificial intelligence correctly produced structures similar to favipiravir that act more specifically as RdRp inhibitors. In addition, Lipinski's rules were applied to the molecules that showed the best interaction with RdRp, and 7 compounds were determined to be potential drug candidates. Among these compounds, a Molecular Dynamic simulation study was applied for ChEMBL ID:1193133 to better understand the existence and duration of the compound in the receptor site. The results confirmed that the ChEMBL ID:1193133 compound showed good Root Mean Square Deviation (RMSD), Root Mean Square Fluctuation (RMSF), hydrogen bonding, and remaining time in the active site; therefore, it was considered that it could be active against the virus. This compound was also tested for antiviral activity, and it was determined that it did not delay viral infection, although it was cytotoxic between 5mg/mL-1.25mg/mL concentrations. However, if other compounds could be tested, it might provide a chance to obtain activity, and compounds should also be tested against the enzymes as well as the other types of viruses.

Vulnerable voices: using topic modeling to analyze newspaper coverage of climate change in 26 non-Annex I countries (2010–2020)

News media influence how climate change is represented, understood, and discussed in the public sphere. To date, media and climate change research has primarily focused on Annex I countries, or treated non-Annex I countries as a homogenous bloc, despite the global nature of climate change and its geographically uneven impacts. This study uses a mixed-method approach, combining machine learning (topic modeling), econometrics, and qualitative analyses, to investigate newspaper coverage of climate change in 26 non-Annex I countries. We compiled a dataset of 95 216 news articles (dated between 2010 and 2020 from 50 sources) in 26 lower-middle and upper-middle income non-Annex I countries. In line with previous research results, we find that most common topics represented are international governance of climate change, the economics of energy transitions, and the impacts of climate change. Advancing current research understanding, we also demonstrate heterogeneity in coverage between non-Annex I countries and discover that a country’s vulnerability to climate change is positively associated with the diversity of topics (based on an article-level entropy index) portrayed by its domestic news media outlets.

Parameterization of the Ångström–Prescott formula based on machine learning benefit estimation of reference crop evapotranspiration with missing solar radiation data

AbstractAccurately estimated reference evapotranspiration (ET0) is essential to regional water management. The FAO recommends coupling the Penman–Monteith (P‐M) model with the Ångström–Prescott (A‐P) formula as the standard method for ET0 estimation with missing Rs measurements. However, its application is usually restricted by the two fundamental coefficients (a and b) of the A‐P formula. This paper proposes a new method for estimating ET0 with missing Rs by combining machine learning with physical‐based P‐M models (PM‐ET0). The benchmark values of the A‐P coefficients were first determined at the daily, monthly, and yearly scales, and further evaluated in Rs and ET0 estimates at 80 national Rs measuring stations. Then, three empirical models and four machine‐learning methods were evaluated in estimating the A‐P coefficients. Machine learning methods were also used to estimate ET0 (ML‐ET0) to compare with the PM‐ET0. Finally, the optimal estimation method was used to estimate the A‐P coefficients for the 839 regular weather stations for ET0 estimation without Rs measurement for China. The results demonstrated a descending trend for coefficient a from northwest to southeast China, with larger values in cold seasons. However, coefficient b showed the opposite distribution as the coefficient a. The FAO has recommended a larger a but a smaller b for southeast China, which produced the region's largest Rs and ET0 estimation errors. Additionally, the A‐P coefficients calibrated at the daily scale obtained the best estimation accuracy for both Rs and ET0, and slightly outperformed the monthly and yearly coefficients without significant difference in most cases. The machine learning methods outperformed the empirical methods for estimating the A‐P coefficients, especially for the sites with extreme values. Further, ML‐ET0 outperformed the PM‐ET0 with yearly A‐P coefficients but underperformed those with daily and monthly ones. This study indicates an exciting potential for combining machine learning with physical models for estimating ET0. However, we found that using the A‐P coefficients with finer time scales is unnecessary to deal with the missing Rs measurements.

Determining protein conformational ensembles by combining machine learning and small-angle scattering

Determining protein conformational ensembles by combining machine learning and small-angle scattering