Root Mean Square Error Research Articles

Abstract 3566: OncoPSM as a tool for cost-effectiveness analysis in oncology trial using Kaplan Meier curve data

Abstract Introduction: Cost-effectiveness analysis (CEA) is a key tool for evaluating the economic viability of new treatments compared to current standards of care. However, many CEA methodologies are not specifically designed to address the intricate cost structures in oncology trials. Methods: We employed a public image recognition tool to extract data from Kaplan-Meier (KM) curves published in clinical trials. Using an iterative algorithm based on the iKM method, we reconstructed individual patient data (IPD). A partitioned survival models (PSM) were then fitted to the IPD data. Under this model, we estimated the probability of each survival state for each model-cycle and combine these probabilities with utility values to calculate both the effect per model-cycle and the incremental effect for the experimental group. For cost analysis, we performed a treatment-cycle-specific cost evaluation, accounting for cost uncertainty through a gamma distribution. By setting model-cycle equals to the treatment-cycle, we could calculate the state-weighted cost, applied a discount rate, and determined the incremental cost for the experimental group. The Incremental Cost-Effectiveness Ratio (ICER) was then calculated based on the incremental effect and incremental cost. Results: The OncoPSM application, available at (https://zhaofeilong.shinyapps.io/app_oncopsa/), was validated with real-world data from the CHOICE-01 trial. OncoSPM accurately reconstructed IPD from KM curve data, with root mean square error (RMSE) values below 0.004 across all curves. Log-rank p-values for the experimental and control groups (PFS: <0.0001; OS: 0.0104) closely aligned with those reported in the original publication (PFS: <0.0001; OS: 0.0108). Hazard ratios (HR) from the reconstructed IPD data (PFS: 0.504, 95%CI: 0.4-0.63, OS: 0.731, 95%CI: 0.57-0.93) were consistent with the original study (PFS: 0.49, 95%CI: 0.39-0.61, OS: 0.73, 95%CI: 0.57-0.93). The log-logistic model provided the best fit for both PFS and OS curves, based on the Akaike Information Criterion (AIC). Extrapolating the survival curve to a 10-year horizon, we created the PSM, deriving the average state probabilities per treatment-cycle to calculate state-weighted costs. The incremental cost for the experimental group was ¥42, 291, with an incremental quality-adjusted life year (QALY) of 0.35, yielding an ICER of 122, 319 ¥/QALY, which is well below the willingness-to-pay (WTP) threshold of 268, 200 ¥/QALY. Uncertainty analysis showed a 99.6% probability that the experimental group was cost-effective at the WTP threshold. Conclusions: OncoPSM offers an accurate and user-friendly treatment-cycle-based cost analysis tool that addresses the complexities of costs in oncology research. By integrating and visualizing the entire CEA process, OncoPSM supports decision-makers in making well-informed choices based on both statistical and intuitive insights. Citation Format: Baohui Han, Feilong Zhao, Ding Zhang. OncoPSM as a tool for cost-effectiveness analysis in oncology trial using Kaplan Meier curve data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3566.

Computational Fluid Dynamics (CFD) Optimization in Smart Factories: AI-Based Predictive Modelling

In the era of Industry 4.0, optimizing fluid flow systems in smart factories is essential to improve energy efficiency and operational stability. Traditional Computational Fluid Dynamics (CFD) simulations provide accurate fluid flow analysis but require extensive computational resources and long processing times, making real-time applications challenging. To address this limitation, this study aims to develop an AI-based predictive model for CFD simulations, utilizing Convolutional Neural Networks (CNN) and Extreme Gradient Boosting (XGBoost) to accelerate the estimation of fluid flow characteristics in industrial environments. The research methodology involves generating CFD simulation datasets, preprocessing data, and training AI models to predict key fluid parameters such as pressure, velocity, and temperature. The evaluation results show that CNN achieves a Mean Squared Error (MSE) of 0.0025 and a Root Mean Squared Error (RMSE) of 0.05, outperforming XGBoost, which records an MSE of 0.0030 and an RMSE of 0.055. Moreover, CNN predicts fluid dynamics in just 15.2 seconds, while XGBoost achieves results in 10.5 seconds, compared to the 1200.5 seconds required by traditional CFD simulations. These findings highlight the potential of AI in reducing computation time by over 98%, making real-time fluid flow analysis feasible in industrial settings. This study contributes to the advancement of AI-integrated CFD modeling, demonstrating that AI can significantly enhance the efficiency of fluid dynamics analysis without compromising accuracy. Future research should focus on expanding AI models to handle more complex flow conditions and integrating AI with smart factory design tools for real-time optimization

Research on the Z-direction thermal error modelling method based on an optimized CNN-BiLSTM-Attention

ABSTRACT High-speed drilling and tapping centres are widely employed in the 3C industry. However, the heat sources within their spindle systems and Z-axis feed systems are concentrated, leading to thermal errors on the Z-axis that negatively impact machining accuracy. Given the limited adaptability of traditional thermal error modelling methods in complex environments, this paper presents a high-precision model for identifying Z-axis thermal errors to enable real-time compensation across varying environmental temperatures and spindle speeds. This study identifies key temperature-sensitive points through finite element analysis and introduces an R-CBLA thermal error identification model that integrates the RIME optimization algorithm (RIME), temporal convolutional networks (CNNs), bidirectional long short-term memory networks (BiLSTM), and attention mechanisms. Using a specific high-speed drilling and tapping centre as the research subject, the authors validated the model through no-load experiments and assessed its performance against seven comparative models. The experimental results demonstrate that the root mean square errors (RMSEs) of the R-CBLA model are 2.8419 and 3.7308, respectively, achieving a prediction accuracy of 98.8%; this significantly enhances both the accuracy and stability of thermal error predictions, thereby providing a novel approach for thermal error compensation in CNC machine tools.

A case study on long short-term memory-based modeling for predicting chlorine and free ammonia levels at water treatment plant testing sites

ABSTRACT This work explores the accurate prediction of water quality, a critical aspect of water environment protection. Leveraging advanced algorithms and statistical models, including regression analyses, moving average, and artificial neural networks (ANN), we analyze multivariate time-series datasets from tap water testing sites. Our approach involves model fitting and evaluation based on the Akaike information criterion (AIC) and root mean square error (RMSE). The work introduces the long short-term memory (LSTM) neural network and the seasonal autoregressive integrated moving average (SARIMA) model, detailing their architectures and implementations. Results highlight the effectiveness of a multivariate LSTM model in forecasting total chlorine concentrations, complemented by the SARIMA model. Evaluation metrics reveal comparable predictive accuracy, with a slight advantage observed for the mixed method combining both models. This work underscores the significance of integrating technological and statistical methodologies for enhanced water quality predictions, contributing to proactive water management strategies.

Hybrid ANN-PSO Based MPPT Optimization for Enhanced Solar Panel Efficiency

In some cases of Solar Power Generation System (PLTS) optimization, AI algorithms can be used to solve complex problems such as efficiency problems. In this research, a hybrid approach that combines Artificial Neural Network (ANN) and Particle Swarm Optimization (PSO) algorithms is used to optimize the Maximum Power Point Tracking (MPPT) system for solar panels. The hybrid technique seeks to maximize power output by precisely determining the ideal voltage and current points, which will increase the efficiency of solar panels. This study includes the measurement of parameters such as current (I), voltage (V), and power (W) in the MPPT system. The research shows that the hybrid ANN-PSO approach performs better than the traditional ANN method, producing mean squared error (MSE) and root mean squared error (RMSE) values that are lower. Moreover, research results show that the hybrid system maintains a load efficiency of approximately 51% in real-world measurements and about 67% in simulation data, indicating better performance and implementation ease.

A Multi-Scale Fusion Convolutional Network for Time-Series Silicon Prediction in Blast Furnaces

In steel production, the blast furnace is a critical element. In this process, precisely controlling the temperature of the molten iron is indispensable for attaining efficient operations and high-grade products. This temperature is often indirectly reflected by the silicon content in the hot metal. However, due to the dynamic nature and inherent delays of the ironmaking process, real-time prediction of silicon content remains a significant challenge, and traditional methods often suffer from insufficient prediction accuracy. This study presents a novel Multi-Scale Fusion Convolutional Neural Network (MSF-CNN) to accurately predict the silicon content during the blast furnace smelting process, addressing the limitations of existing data-driven approaches. The proposed MSF-CNN model extracts temporal features at two distinct scales. The first scale utilizes a Convolutional Block Attention Module, which captures local temporal dependencies by focusing on the most relevant features across adjacent time steps. The second scale employs a Multi-Head Self-Attention mechanism to model long-term temporal dependencies, overcoming the inherent delay issues in the blast furnace process. By combining these two scales, the model effectively captures both short-term and long-term temporal dependencies, thereby enhancing prediction accuracy and real-time applicability. Validation using real blast furnace data demonstrates that MSF-CNN outperforms recurrent neural network models such as Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU). Compared with LSTM and the GRU, MSF-CNN reduces the Root Mean Square Error (RMSE) by approximately 22% and 21%, respectively, and improves the Hit Rate (HR) by over 3.5% and 4%, highlighting its superiority in capturing complex temporal dependencies. These results indicate that the MSF-CNN adapts better to the blast furnace’s dynamic variations and inherent delays, achieving significant improvements in prediction precision and robustness compared to state-of-the-art recurrent models.

Winter Oilseed Rape LAI Inversion via Multi-Source UAV Fusion: A Three-Dimensional Texture and Machine Learning Approach

Leaf area index (LAI) serves as a critical indicator for evaluating crop growth and guiding field management practices. While spectral information (vegetation indices and texture features) extracted from multispectral sensors mounted on unmanned aerial vehicles (UAVs) holds promise for LAI estimation, the limitations of single-texture features necessitate further exploration. Therefore, this study conducted field experiments over two consecutive years (2021–2022) to collect winter oilseed rape LAI ground truth data and corresponding UAV multispectral imagery. Vegetation indices were constructed, and canopy texture features were extracted. Subsequently, a correlation matrix method was employed to establish novel randomized combinations of three-dimensional texture indices. By analyzing the correlations between these parameters and winter oilseed rape LAI, variables with significant correlations (p < 0.05) were selected as model inputs. These variables were then partitioned into distinct combinations and input into three machine learning models—Support Vector Machine (SVM), Backpropagation Neural Network (BPNN), and Extreme Gradient Boosting (XGBoost)—to estimate winter oilseed rape LAI. The results demonstrated that the majority of vegetation indices and texture features exhibited significant correlations with LAI (p < 0.05). All randomized texture index combinations also showed strong correlations with LAI (p < 0.05). Notably, the three-dimensional texture index NDTTI exhibited the highest correlation with LAI (R = 0.725), derived from the spatial combination of DIS5, VAR5, and VAR3. Integrating vegetation indices, texture features, and three-dimensional texture indices as inputs into the XGBoost model yielded the highest estimation accuracy. The validation set achieved a determination coefficient (R2) of 0.882, a root mean square error (RMSE) of 0.204 cm2cm−2, and a mean relative error (MRE) of 6.498%. This study provides an effective methodology for UAV-based multispectral monitoring of winter oilseed rape LAI and offers scientific and technical support for precision agriculture management practices.

Autoencoder-Based Hyperspectral Unmixing with Simultaneous Number-of-Endmembers Estimation

Hyperspectral unmixing plays a fundamental role in mining meaningful information from hyperspectral data. It promotes advancements in various scientific, environmental, and industrial applications by extracting meaningful information from hyperspectral data. However, it is still hindered by several challenges, including accurately identifying the number of endmembers in a hyperspectral image, extracting the endmembers, and estimating their abundance fractions. This research addresses these challenges by employing a convolutional-neural-network-based autoencoder that leverages both the spatial and spectral information present in the hyperspectral image. Additionally, a self-learning module utilizing a fuzzy clustering algorithm is designed to determine the number of endmembers. A novel approach is also introduced that estimates the abundances of the endmembers from the autoencoder and the clustering output. Real datasets and relevant performance metrics were used to validate and evaluate the performance of the proposed method. The results demonstrate that our approach outperforms related methods, achieving improvements of 47% in Spectral Angle Distance (SAD) and 42% in root-mean-square error (RMSE).

Compressive strength prediction of ultra lightweight cement using data-driven modeling techniques

The advancement of predictive modeling techniques using machine learning (ML) has emerged as a pivotal component in optimizing the performance of materials in the oil and gas industry. This paper presents an in-depth analysis of employing four distinct ML algorithms—Artificial Neural Networks (ANN), Gradient Boosting Regressor (GBR), Random Forest (RF), and Support Vector Regression (SVR)—to accurately estimate the Unconfined Compressive Strength (UCS) of lightweight cement formulations. This study tested 516 different cement samples, and the models were trained and validated using a portion of the dataset. The ANN and the GBR ensemble approach produced highly accurate and consistent results when tested on unseen data, supported by Root Mean Squared Error (RMSE) scores of 140 psi and 130 psi and Mean Absolute Percentage Error (MAPE) of approximately 10% and 9%, respectively. All models had computed coefficient of determination (R2) values exceeding 0.9 for all dataset splits, indicating good model fits. By leveraging a dataset comprising various cement samples enhanced with strength-improving additives, the study explores the effectiveness of each algorithm in predicting UCS under different wellbore conditions. Through meticulous hyperparameter tuning and validation processes, the research assesses the performance metrics of each algorithm, thereby providing valuable insights into their applicability for real-time cementing operations in the oil and gas sector. The findings illustrate the potential of these advanced modeling techniques to enhance decision-making processes, reduce costs, and improve the reliability of cementing practices in challenging environments.

A robust pressure drop prediction model in vertical multiphase flow: a machine learning approach

Predicting pressure drop in multiphase flow is crucial for optimizing tubing size, wellhead pressure (WHP), completion design, cost management, and other production-phase objectives. While various correlations and models exist to calculate pressure drops, their accuracy diminishes significantly when applied to outlier datasets beyond their intended parameter ranges. Not only that, as reported by many studies, most of the empirical correlations and mechanistic models have an error, which is quite high and intolerable. This study introduces a novel Adaptive Neuro-Fuzzy Inference System (ANFIS) model for accurately predicting pressure drops in vertical wells carrying multiphase fluids. A comprehensive dataset of 335 experimental records was compiled from diverse sources to encompass a wide range of parameters, ensuring the robustness of the proposed model. Key input parameters include WHP, oil rate, water rate, gas rate, inner diameter, surface temperature, and flow length. The ANFIS model was rigorously evaluated using different methods, such as cross plot, error distribution, Kruskal-Wallis (KW) test, error boxplot and violin graphs, Confidence Interval (CI), and statistical error metrics. The latter includes Average Absolute Percentage Error (AAPE), Root Mean Square Error (RMSE), and the coefficient of determination (R²) to prove the ANFIS model’s performance. The proposed ANFIS model was compared with the most commonly used published models. Results demonstrate that the ANFIS model outperforms existing methods, achieving an AAPE of 2.92%, an RMSE of 1.9638%, and an R² of 0.9645. KW test, error boxplot, violin graphs, and CI results indicated that the best model for predicting pressure drops is the proposed ANFIS model. These findings establish the ANFIS model as a reliable and superior tool for predicting pressure drops in vertical wells with multiphase flow, offering significant advancements in design accuracy and operational efficiency.

Deep Mining on the Formation Cycle Features for Concurrent SOH Estimation and RUL Prognostication in Lithium-Ion Batteries

Lithium-ion batteries (LIBs) are widely utilized in consumer electronics, electric vehicles, and large-scale energy storage systems due to their high energy density and long lifespan. Accurately estimating the state of health (SOH) and predicting the remaining useful life (RUL) of cells is crucial to ensuring their safety and preventing potential risks. Existing state estimation methodologies primarily rely on electrical signal measurements, which predominantly capture electrochemical reaction dynamics but lack sufficient integration of thermomechanical process data critical to holistic system characterization. In this study, relevant thermal and mechanical features collected during the formation process are extracted and incorporated as additional data sources for battery state estimation. By integrating diverse datasets with advanced algorithms and models, we perform correlation analyses of parameters such as capacity, voltage, temperature, pressure, and strain, enabling precise SOH estimation and RUL prediction. Reliable predictions are achieved by considering the interaction mechanisms involved in the formation process from a mechanistic perspective. Full lifecycle data of batteries, gathered under varying pressures during formation, are used to predict RUL using convolutional neural networks (CNN) and Gaussian process regression (GPR). Models that integrate all formation-related data yielded the lowest root mean square error (RMSE) of 2.928% for capacity estimation and 16 cycles for RUL prediction, highlighting the significant role of surface-level physical features in improving accuracy. This research underscores the importance of formation features in battery state estimation and demonstrates the effectiveness of deep learning in performing thorough analyses, thereby guiding the optimization of battery management systems.

Evaluating the impact of waste marble on the compressive strength of traditional concrete using machine learning

Waste marble, an industrial byproduct generated from marble cutting and polishing processes, can be effectively utilized as a partial replacement in concrete mixtures. Incorporating waste marble in concrete not only addresses environmental concerns related to marble waste disposal but also contributes to the sustainability of construction materials. Using machine learning (ML) to predict the impact of waste marble on the compressive strength of traditional concrete offers several advantages over repeated laboratory experiments. ML offers a powerful alternative to costly and time-consuming laboratory experiments, enabling faster and more sustainable exploration of the potential of waste marble in improving concrete’s compressive strength. This research has focused on evaluating the impact of waste marble on the compressive strength of traditional concrete using machine learning (ML). Advanced ML techniques such as the Group Methods Data Handling Neural Network (GMDH-NN), Support Vector Regression (SVR), K-Nearest Neighbors (kNN) and Adaptive Boosting (AdaBoost) have been applied in this research work. The GMDH-NN model was created using GMDH Shell 3.0 software, while AdaBoost, SVR and kNN models were created using “Orange Data Mining” software version 3.36. Error indices such as the sum of squared error (SSE), mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), and Error (%), and performance metrics such as Accuracy % and the R2 between predicted and calculated compressive strength parameters were used to evaluate the overall behavior of the models. Finally, the Hoffman sensitivity analysis procedure was applied to determine the individual relative impact of the input variables on the output. At the end of the processes, a total of 1135 waste marble concrete entries were collected containing constituents such as the cement density (C), waste marble (WM), fine aggregate (FAg), coarse aggregate (CAg), water (W), superplasticizer (PL) and curing age (Age) used as input variables of the waste marble concrete model. The records were divided into training set (900 records = 80%) and validation set (235 records = 20%) following standard partitioning pattern reported in the literature. The kNN and AdaBoost, with SSE of 1408.5 MPa2 and 1397 MPa2 respectively and a tie Accuracy of 95.5% and R2 of 0.985 showed the best models suggesting excellent model performance while GMDH-NN showed the worst. Conversely, RF balances accuracy and model complexity, making it a practical alternative to kNN and AdaBoost. And lastly, Age, Coarse Aggregates, Water, and Plasticizer play the most significant roles in determining the compressive strength, while Cement, Waste Marble, and Fine Aggregates have comparatively smaller impacts. However, considering the standard proportion required for waste marble powder to replace cement, it showed a remarkable influence on the behavior of the concrete thus a recommended potential for its used as replacement for cement.

Future precipitation projections and model evaluation in the Hengduan Mountains based on CMIP6.

Future precipitation projections and model evaluation in the Hengduan Mountains based on CMIP6.

Uncertainty-aware approach for multiple imputation using conventional and machine learning models: a real-world data study

Missing data poses a significant challenge in clinical real-world studies, often arising from unplanned data collection, misplacement, patient loss to follow-up, and other factors. While multiple imputation by chained equations (MICE) is a widely used method, its sequential nature introduces uncertainty, potentially impacting the prediction model performance. We proposed and evaluated three uncertainty-aware functions (i.e., uncertainty sampling (US), probability of improvement (PI), and expected improvement (EI)) integrated with linear regression (LinearReg), decision tree (DT), random forest (RF), and extreme gradient boosting (XGBoost) using three large datasets: chronic kidney disease (CKD, n = 31,043), hypertension cohort from Ramathibodi Hospital (HT-RAMA, n = 140,047) and Khon Kaen University Hospital (HT-KKU, n = 108,942) with high missing rates. In the CKD cohort, uncertainty-aware models significantly improved performance (evaluated by root mean squared error (RMSE) and mean absolute error (MAE)) over standard MICE, except for XGBoost. LinearReg-EI performed best (RMSE 0.12, MAE 0.36), followed by RF-EI (RMSE 0.22, MAE 0.34), and DT-EI (RMSE 0.21, MAE 0.38). In HT-RAMA, LinearReg-US performed best (RMSE 0.24, MAE 8.15), outperforming RF-US (RMSE 0.92, MAE 8.58) and DT-PI (RMSE 0.96, MAE 8.74). Similarly, in HT-KKU, LinearReg-US performed best (RMSE 0.98, MAE 12.00), followed by RF-PI (RMSE 1.93, MAE 12.90) and DT-US (RMSE 2.10, MAE 12.63). Uncertainty-aware models produced imputed distributions closely resembling the original data, unlike standard MICE. Our findings suggest that incorporating uncertainty functions can improve MICE, particularly for LinearReg, RF and DT. Further research is warranted to validate these findings across diverse clinical settings and model types.

Predicting Life Expectancy using Machine Learning Techniques

Life expectancy is a key measure of a country's overall health, socioeconomic development, and quality of life. The main objective of the study is to identify key factors influencing life expectancy using ‘Cleaned-Life-Exp’ standardized data from the World Health Organization (WHO) and to compare the performances of life expectancy prediction using various machine learning algorithms. The key influencing features on life expectancy are selected using Boruta and Regularized Random Forest (RRF) algorithms. Eight machine learning models such as Linear Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Gradient Boosting (GB), XGBoost (XGB), and Neural Network (NN) are evaluated for predictive performance of life expectancy. Evaluation metrics such as the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) are applied to evaluate the performance of the models. Boruta and Regularized Random Forest (RRF) algorithms identified the same 20 significant predictors, including Income Composition of Resources, HIV/AIDS, Adult Mortality, and Schooling, as the most influential features. Among the eight machine learning models evaluated, Random Forest achieves the highest performance (R2 = 0.969, RMSE = 0.179, MAE = 0.116), highlighting the superiority of ensemble methods. Support Vector Machine (SVM) performs well, while Decision Tree and KNN show moderate performance. Linear Regression and Neural Networks have the lowest predictive performances. This study will help to provide a better predictive framework using machine learning models, which can guide policymakers in improving life expectancy prediction. IJSS, Vol. 25(1), March, 2025, pp 55-70

TEMLI: A High-Resolution Air Temperature Estimation Using Machine Learning and Land Surface Data Across Morocco

Abstract Air temperature (Tair) is an important parameter to understand climate dynamics and support decision-making across various sectors. However, accurately estimating Tair at high spatial resolution remains challenging, particularly in data-sparse regions. This study addresses the research gap that arises from the scarcity of observed temperature data by developing a robust machine learning-based estimation model, TEMLI (Temperature Estimation with ML and Land Input). The approach leverages satellite-derived inputs such as Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), albedo, and wind speed. The objective is to improve the accuracy of Tair estimation by capturing the nonlinear relationships between these variables. Among the models tested within the TEMLI framework, the Multilayer Perceptron (MLP) demonstrated superior performance, achieving a Root Mean Square Error (RMSE) of less than 1.6 °C. Notably, the model excelled in estimating extreme temperatures, especially in high-altitude regions. Trained using Moroccan weather station data, the model’s generalizability across diverse climatic conditions was also validated. This study concludes that integrating satellite-derived data with machine learning offers a scalable and reliable approach for Tair estimation, particularly in regions with limited in situ observations. The findings have significant implications for advancing climate modeling, monitoring extreme events, and informing decision-making in various climate-sensitive sectors. Graphical Abstract This visual summary offers a concise and accessible overview of the study’s main objectives, methodology, and findings. The TEMLI (Temperature Estimation with Machine Learning and Land Inputs) framework aims to improve air temperature (Tair) estimation at high spatial resolution (1 km) using satellite-derived inputs. The left panel introduces the research context, emphasizing the need for accurate temperature datasets in regions with sparse observations. The central panel outlines the methodological flowchart, detailing the various inputs (LST, NDVI, NDWI, albedo, wind speed), the machine learning models tested, and the selection of the optimal model (MLP), and the evaluation metrics and process. The right panel highlights key results, demonstrating TEMLI’s improved performance over ERA5 products with RMSE < 1.6 °C and R² > 0.9, particularly in high-altitude areas. It also summarizes practical applications, such as enhanced hydrological modeling, improved agricultural climate monitoring, and better temperature-based climate adaptation strategies.

Modeling an All-Copper Redox Flow Battery for Microgrid Applications: Impact of Current and Flow Rate on Capacity Fading and Deposition

The copper redox flow battery (CuRFB) stands out as a promising hybrid redox flow battery technology, offering significant advantages in electrolyte stability. Within the CuBER H-2020 project framework, this study addresses critical phenomena such as electrodeposition at the negative electrode during charging and copper crossover through the membrane, which influence capacity fading. A comprehensive two-dimensional physicochemical model of the CuRFB cell was developed using COMSOL Multiphysics, providing insights into the distribution of electroactive materials over time. The model was validated against experimental cycling data, demonstrating a Root Mean Square Error (RMSE) of 0.0212 in voltage estimation. Least-squares parameter estimation, utilizing Bound Optimization by Quadratic Approximation, was conducted to determine active material diffusivities and electron transfer coefficients. The results indicate that higher current densities and lower flow rates lead to increased copper deposition near the inlet, significantly impacting the battery’s State of Health (SoH). These findings highlight the importance of considering fluid dynamics and ion concentration distribution to improve battery performance and longevity. The study’s insights are crucial for optimizing and scaling up CuRFB operations, guiding potential cell-scale-up strategies into stack-level configurations.

Trend in global ocean heat content into different depth layers from 1940 to 2050

Abstract Oceans are crucial to Earth’s climate regulation and Earth’s energy, water, and carbon cycles, significantly affecting human society. Ocean heat content (OHC) is a critical indicator of global warming, as the ocean absorbs over 90% of excess heat trapped by greenhouse gases. This heat storage delays atmospheric warming but drives stronger storms, shifts ocean currents, and intensifies marine heatwaves. Understanding changes in OHC and monitoring them at the surface and various depth layers are becoming more and more crucial as global warming picks up speed. This study models and forecasts global monthly OHC time-series data within the 0–700 m, 0–2000 m, 0–6000 m, 700–2000 m, and 2000–6000 m depth layers until 2050 using the seasonal autoregressive integrated moving average (SARIMA) model. Initially, the global monthly OHC data set for all depth layers was analyzed from 1940 to 2023. Monthly and annual forecasts are then created until 2050. Results reveal that the SARIMA models exhibit coefficient of determination (R2) values greater than 0.9972 and root mean square error (RMSE) values less than 5.3 Zetajoules (ZJ), and that they also accurately predict the trends and variability of the OHC data. By 2050, the global OHC anomalies relative to the 2006–2015 average in the 0-700, 0-2000, 0-6000, 700–2000, and 2000–6000 m depth layers are projected to reach 373 ZJ, 477 ZJ, 564 ZJ, 100 ZJ, and 95 ZJ, respectively.

Integration of Machine Learning and VIRS for Soil Potassium Content Prediction

ABSTRACT At present, the combination of machine learning and visible and infrared spectroscopy (VIRS) stands out as one of the most promising methods for predicting soil potassium content. In this study, a total of 100 soil samples were collected in the vicinity of Xunsi River, Hubei University of Technology. Spectral information and corresponding soil potassium content were obtained for these 100 soil samples. Subsequently, the original spectra of the soil samples were denoised using the locally weighted regression (loess) method. Feature wavelength selection was performed using ∆Gini and Successive Projections Algorithm (SPA) for dimensionality reduction. The model was then optimized by adjusting its parameters. Finally, random forest (RF), back propagation (BP), Genetic algorithm-back propagation (GA-BP) and particle swarm optimization-back propagation (PSO-BP) prediction models were established to predict soil potassium content. The results show that the RF model exhibits superior prediction performance compared to other algorithms, with a coefficient of determination (R2) of 0.851 and a root mean square error (RMSE) of 0.503. The RF model accurately predicts soil potassium content and validates the stability of the VIRS method for rapid estimation. Among the BP model and its optimization variants, the PSO-BP-Levenberg-Marquardt model demonstrates the best predictive capability, yielding an R2 of 0.791 and an RMSE of 0.621. This combination algorithm holds potential as a valuable reference for the selection of optimization algorithms and training functions for the BP model within the research domain.

Convolutional neural network-based deep learning for landslide susceptibility mapping in the Bakhtegan watershed

Landslides pose a significant threat to infrastructure, ecosystems, and human safety, necessitating accurate and efficient susceptibility assessment methods. Traditional models often struggle to capture the complex spatial dependencies and interactions between geological and environmental factors. To address this gap, this study employs a deep learning approach, utilizing a convolutional neural network (CNN) for high-precision landslide susceptibility mapping in the Bakhtegan watershed, southwestern Iran. A comprehensive landslide inventory was compiled using 235 documented landslide locations, validated through remote sensing and field surveys. An equal number of non-landslide locations were systematically selected to ensure balanced model training. Fifteen key conditioning factors—including topographical, geological, hydrological, and climatological variables—were incorporated into the model. While traditional statistical methods often fail to extract spatial hierarchies, the CNN model effectively processes multi-dimensional geospatial data, learning intricate patterns influencing slope instability. The CNN model outperformed other classification approaches, achieving an accuracy of 95.76% and a precision of 95.11%. Additionally, error metrics confirmed its reliability, with a mean absolute error (MAE) of 0.11864, mean squared error (MSE) of 0.18796, and root mean squared error (RMSE) of 0.18632. The results indicate that the northern and northeastern regions of the Bakhtegan watershed are highly susceptible to landslides, highlighting areas where proactive mitigation strategies are crucial. This study demonstrates that deep learning, particularly CNNs, offers a powerful and scalable solution for landslide susceptibility assessment. The findings provide valuable insights for urban planners, engineers, and policymakers to implement effective risk reduction strategies and enhance resilience in landslide-prone regions.