Taylor Diagram Research Articles

Monthly Streamflow Prediction Using ANN, KNN and ANFIS models: Example of Gediz River Basin

Stream flow forecasting is very important in many aspects such as water supply, irrigation, building water infrastructures, and taking precautions against floods. The ability to forecast future streamflow helps us anticipate and plan for upcoming flooding, decreasing property destruction, preventing deaths and managing water in the best way possible. Different hydrological models have been developed for predicting streamflow and they have different characteristics, driven by the research area and available data. İn this study, three types of Artificial Intelligence models; K-Nearest Neighbor (KNN), Artificial Neural Network (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) have been used to study the Gediz River Basin which is located in the Aegean region of western Turkey. The results varied due to the complication of the data and different parts of the study area as well as the structure of the models, over all, looking at Regression coefficient (R2), Root Mean Square Error (RMSE) and Wilcoxon (WT) values, ANFIS is more accurate compared to ANN and KNN models. Conversely, according to Taylor diagram, KNN is more accurate compared to ANN and ANFIS.

Optimization of rice husk ash concrete design towards economic and environmental assessment

Rice husk ash (RHA) is one of the main agricultural wastes that holds great potential for utilization in concrete production. Previous studies have mainly focused on its technical feasibility through experimental investigations, neglecting the crucial aspects of environmental and economic viability in practical applications. To address this gap and promote the efficient utilization of RHA in concrete, this study employs three advanced machine learning algorithms to develop predictive models for compressive strength. These models aim to design RHA Concrete (RHAC) with the minimum carbon footprint and financial cost. Among the algorithms used, the ant lion algorithm and eXtreme Gradient Boosting (ALO-XGB) model exhibit superior performance, demonstrating high accuracy across four indicators and showcasing their effectiveness in a Taylor diagram. The optimized model is further interpreted using the Shapley Additive Profile (SHAP). Additionally, multi-objective optimization is conducted using the ALO-XGB, to maximize the compressive strength while controlling material cost and CO2 emissions. Our results provide the optimal mix ratios for RHAC within different strength ranges and offer insights into the influence of RHA content on mechanical, economic, and environmental factors. Specifically, the RHAC designed in this study can reach a compressive strength of 40–60 MPa while delivering significant environmental benefits. By replacing traditional concrete in housing and bridge structures, RHAC can effectively reduce the carbon footprint by 12.2–17.5 kg CO2 eq./m3 in Italy. However, it is important to acknowledge that the current high cost of producing this sustainable concrete necessitates government incentives and subsidies to alleviate financial burdens on companies and facilitate market expansion.

Application of boosted tree algorithm with new data preprocessing techniques in the forecasting one day ahead streamflow values in the Tigris basin, Türkiye

Accurate streamflow forecasting is very useful in water resources management, design of hydraulic structures, and almost all issues related to the use of water and water resources, especially in arid regions that have increased in recent years. Since water is the source of all life and the most important basic element for humanity to continue its life, streamflow prediction studies increase its importance daily. This research combined the boosted tree (BT) model with robust empirical mode decomposition, empirical mode decomposition, complete ensemble empirical mode decomposition with adaptive noise, empirical wavelet transforms and variational mode decomposition for predicting daily average streamflow data. While historical streamflow data was input in the model's setup, one-day lead-time streamflow data was used as the target. 70% of the data is reserved for training and the rest for testing. 5-fold cross-validation technique was used to solve the over-fitting problem. The coefficient of determination, mean squared error, Nash-Sutcliffe efficiency and percent bias statistical criteria and Taylor diagrams, polar plot, scattering diagram, and violin plot were used to determine the algorithm's success. At the end of the study, it was found that the most successful streamflow predictions were made with the variational mode decomposition-based BT hybrid approach.

Vegetation dynamics and its driving factors in agro-climatic zones of Upper Bhima Sub-basin, Western India

Climate change is a key threat to biodiversity and ecosystems. The impacts of climate change on vegetation must be evaluated to control the sustainability of the ecosystem. Precipitation and temperature are the most important climatic parameters affecting vegetation growth. It is important to analyze the spatial and temporal variations of vegetation under changing climatic parameters to envisage the response to the regional ecosystem. In this study, variation of the Normalized Difference Vegetation Index over five different agro-climatic zones in the Upper Bhima Sub-basin from 2003 to 2013 was analyzed to find the effect of climatic and hydrological variables on vegetation dynamics. The correlation analysis has been performed by using the Pearson correlation method and Geographical Detector Model. Taylor diagrams are generated to highlight the variation in the correlation values of NDVI with other variables. Results show that precipitation and temperature are key parameters for the growth of the vegetation, but Western Ghat Zone shows contradictory results. To identify the reasons behind contradictory results in the Western Ghat Zone, hydrological parameters such as soil moisture, terrestrial water storage and groundwater levels are analyzed. NDVI shows a positive correlation with precipitation in Water Scarcity Zone, and Assured Rainfall Zone with correlation values 0.37, and 0.462 respectively, but, in the Western Ghat Zone, NDVI has a negative correlation (− 0.38) though more than average precipitation occurs in that region. The result of the Geographical Detector Model reveals that the terrestrial water storage and precipitation are dominant factors affecting vegetation in WGZ and WSZ, respectively.

Prioritizing the selection of CMIP6 model ensemble members for downscaling projections of CONUS temperature and precipitation

Given the mismatch between the large volume of data archived for the sixth phase of the Coupled Model Intercomparison Project (CMIP6) and limited personnel and computational resources for downscaling, only a small fraction of the CMIP6 archive can be downscaled. In this work, we develop an approach to robustly sample projected hydroclimate states in CMIP6 for downscaling to test whether the selection of a single initial condition (IC) ensemble member from each CMIP6 model is sufficient to span the range of modeled hydroclimate over the conterminous United States (CONUS) and CONUS sub-regions. We calculate the pattern-centered root mean square difference of IC ensemble member anomalies relative to each model’s historical climatology for shared socioeconomic pathway (SSP) projections over 30-year time periods and compare the ratio of inter-model to intra-model variability for this metric. Regardless of SSP, inter-model variability is generally much greater than intra-model variability at the scales of the CONUS as a whole, as well as for most CONUS sub-regions. However for some variables and scenarios, inter- and intra-model variability are similar at sub-CONUS scales, indicating that selecting a single IC ensemble member per model may be sufficient to sample the range of projected hydroclimate states in the 21st Century across CONUS, but for specific regions and variables, more careful selection of ensemble members may be necessary. Regionally-resolved Taylor diagrams identify where more IC ensemble member downscaling efforts should be focused if resources are available to do so. Our results suggest that, with parsimonious sampling, the requisite computational expense of downscaling temperature and precipitation fields over the CONUS for subsequent CMIP activities may increase only marginally despite the great increase in data volumes with each successive CMIP phase.

Predicting carbonation depth of concrete using a hybrid ensemble model

This study aims to develop a robust and accurate machine learning-based model to improve the prediction accuracy of carbonation depth in complex concrete structures. Two hybrid ensemble methods, the inverse variance method and the artificial neural network-based ensemble method, were proposed to integrate multiple algorithms. The models were trained on a dataset of 532 data points with 6 input variables. Performance evaluation metrics and the Taylor diagram were used to compare the models. The results showed that the hybrid ensemble models outperformed the single models, with the inverse variance-based model achieving the highest performance (R = 0.975, RMSE = 2.978 mm). Additionally, the contribution analysis indicated that carbonation time, CO2 concentration, and the amount of binder were the most influential factors. This study provides an efficient prediction model for carbonation depth and valuable insights for carbonation-resistant design in concrete structures.

The Potential Relationship Between Income Inequality and Environmental Quality in Iran: The Panel Quantile Regression Approach

This paper addresses the relationship between income inequality and environmental quality in the agriculture sector as the most related sector to the environment. In this context, we used a panel data set for 28 provinces of Iran during 2003-2017 and implemented panel quantile regression. To choose the best econometric specification the Taylor diagram is used. To explore whether the effect on income inequality on CO2 emission is different for rich and poor provinces, we consider the GINI × GDP interaction term in the model. Results confirmed that EKC in the agriculture sector of Iran, and a major and significant effect of energy consumption on CO2 emission. Findings indicated a threshold per capita income ($ 17.60 thousand) from which the effect of income inequality on carbon emissions changes. Based on these results, Marginal Properties to Emit (MPE) is more significant for poor people in low-income provinces than rich ones. Therefore, the government should adopt appropriate policies to increase the income of farmers so that the policy of income distribution, along with social justice, also improves the quality of the environment.

Performance evaluation of CMIP6 HighResMIP models in simulating precipitation over Madagascar

AbstractThe present study evaluates the performance of high‐resolution global climate models derived from Coupled Model Intercomparison Project Phase 6 (CMIP6 HighResMIP), in simulating rainfall characteristics over Madagascar on an annual and seasonal scales for the period 1981–2014. The models and their ensemble mean are assessed based on two observational datasets sourced from Climate Hazards Group Infrared Precipitation with Station data version 2 (CHIRPS v2.0) data and the European Centre for Medium‐Range Weather Forecasts (ECMWF) reanalysis fifth generation‐Land dataset (ERA5) as the references throughout the diverse analyses. A Taylor diagram, accompanied by the Taylor skill score (TSS), is used for the annual and seasonal model‐rankings and the overall performance of the models. The best‐performing models are EC‐Earth3P‐HR, ECMWF‐IFS‐HR, ECMWF‐IFS‐LR and HadGEM3‐GC31‐MM. The least‐recommended models with remarkable biases are BCC‐CSM2‐HR, CAMS‐CSM1‐0, FGOALS‐f3‐H, MPI‐ESM1‐2‐HR and MPI‐ESM1‐2‐XR. It is worth mentioning that FGOALS‐f3‐H tends to overestimate rainfall in most analyses, while MPI‐ESM1‐2‐HR and MPI‐ESM1‐2‐XR underestimate it. The findings of this study are of great importance to climatologists and present an opportunity for further investigation of underlying processes responsible for the observed wet/dry biases in order to improve the forecast skills in the models over the study area.

Evaluating the Performance of CMIP6 GCMs to Simulate Precipitation and Temperature Over the Vietnamese Mekong Delta

This study evaluates the performance of simulated precipitation and maximum and minimum temperatures in the historical runs of the Climate Model Intercomparison Project Phase 6 (CMIP6) for the Vietnamese Mekong Delta (VMD). The precipitation, as well as maximum and minimum temperatures outputs from 16 general circulation models (GCMs), were compared with observations from 12 stations for the period 1980–2014, using a set of statistical metrics, namely, normalised root mean square error (NRMSE), percentage of bias (PBIAS), Nash–Sutcliffe efficiency (NSE), coefficient of determination (R2), and volumetric efficiency (VE). Finally, ranking (total score - TS) was carried out and the probability distribution function (PDF) and Taylor diagram were used to confirm rankings. The results show that different statistical indicators reveal variation ranking order of the 16 GCMs. Based on RS ranking, it is indicated that each simulation GCM performed differently under the different metrics and no single model performed best for all metrics. The top five highest ranked GCMs based on TS were HadGEM3-GC31-LL, ACCESS-CM2, CanESM5, NESM3 and CanESM5-CanOE for precipitation; and CNRM-CM6-1, CNRM-ESM2-1, GFDL-ESM4, NESM3 and INM-CM5-0 for the maximum; and CNRM-CM6-1, CNRM-ESM2-1, GFDL-ESM4, NESM3 and INM-CM5-0 for minimum temperatures, respectively. We also observed an underestimation of precipitation and an overestimation of temperature over the study area. The TS method demonstrates efficiency to aggregate the multi-model ensemble GCMs based on different statistical indicators which were sometimes contradictory. The findings from this study provide useful guidance in the selection of GCMs for climate change applications in the VMD.

LGBM-based modeling scenarios to compressive strength of recycled aggregate concrete with SHAP analysis

Concrete production contributes significantly to global greenhouse gas emissions, and its manufacture requires substantial natural resources. These concerns can be partly mitigated by recycling construction demolition waste as aggregates to produce Recycled Aggregate Concrete (RAC). RAC has gained momentum due to its lower environmental impact, production costs, and increased sustainability. The aim of this study was to advance the reasonable use of recycled aggregate in concrete and achieve optimal mixture ratio design. Four advanced machine learning algorithms, Support Vector Machine (SVR), Light Gradient Boosting Machine (LGBM), Random Forest (RF), and Multi-Layer Perceptron (MLP), were employed, and the novel optimization algorithms, biogeography-based optimization (BBO), Multi-Verse Optimizer (MVO) and Gravitational Search Algorithm (GSA), were integrated to to predict the compressive strength of RAC. Six potential influential factors for RAC strength were considered in the models. The study employed four evaluation metrics, Taylor diagrams and Regression Error Characteristic plots to compare model performance. The result shows LGBM-based hybrid model outperformed other methods, demonstrating high accuracy in predicting compressive strength. The Shapley Additive Explanation (SHAP) results emphasize the importance of understanding the interactions between the various factors and their effects on the mechanical properties of the RAC. The findings can inform the development of more sustainable and environmentally friendly building materials.

Experimental validation of Orcaflex-based numerical models for the PEWEC device

To design a Wave Energy Converter mooring system that ensures maintainability, resistance, and low costs, without affecting productivity, it is necessary to foster reliable numerical models. In the present paper, the validation of Orcaflex© simulations against the experiment results of the Pendulum Wave Energy Converter (PEWEC) 1:25 scale model, conducted at the University of Naples Federico II, is presented. The experiments consist of free-decay and static pull-out tests to assess the inertial properties of the model and mooring system; tests in operative and extreme regular and irregular waves to fully characterize the mooring system and the device dynamics. The same wave records measured in the towing tank have been used in numerical simulations. PEWEC motion results from numerical simulations and experiments are given in terms of Response Amplitude Operators (RAOs) for regular waves, as statistical values in tabular form, and in Taylor's diagrams for irregular waves. The most probable fairleads tensions for the load dimensioning have been obtained from the Generalized Extreme Value (GEV) distribution of experimental and numerical data. The obtained differences show that the numerical model accurately predicts the experimental data and correctly estimates the dimensioning load of the mooring lines.

Quantification of river total phosphorus using integrative artificial intelligence models

Total phosphorus (T-P) refers to the concentration of phosphorus in water and is one of the important parameters for eutrophication in lakes and rivers. In current research, neuroscience dependent (i.e., singular and double-platform synthetic) approaches were employed to predict the river T-P concentration. Singular techniques were developed utilizing machine learning based (support vector machines (SVMs), stochastic gradient boosting (SGB)), and deep learning based multilayer perceptron (DMLP) models. Besides, double-platform synthetic techniques were developed by integrating a prior data-processing (Discrete-wavelet) algorithm with the singular techniques. Six input scenarios conditional on different water quantity and quality parameters acquired from two stations (Hwangji and Toilchun), South Korea, were employed for appraising singular and double-platform synthetic techniques. The various promoted models were evaluated using four statistical standards viz., mean absolute error (MAE), root mean square error (RMSE), scatter index (SI), and Pearson correlation coefficient (CCp), and four visual references viz., scatter diagram, box-and-whisker plot, violin plot, and Taylor diagram. It can be indicated from the outcomes that the double-platform synthetic techniques did not always lead to more accurate predictions than the singular techniques. Further, results also supplied that the SGB with the 6th input (MAE = 0.012 mg/L, RMSE = 0.014 mg/L, and CCp = 0.650 for Hwangji; MAE = 0.011 mg/L, RMSE = 0.017 mg/L, and CCp = 0.963 for Toilchun) scenario model demonstrated the best accuracy for predicting river T-P concentration by the singular techniques for both stations, whereas the discrete-wavelet SVMs with the 4th input (MAE = 0.007 mg/L, RMSE = 0.011 mg/L, and CCp = 0.765 for Hwangji) and discrete-wavelet SGB with the 5th input (MAE = 0.012 mg/L, RMSE = 0.017 mg/L, and CCp = 0.953 for Toilchun) scenario models provided the best predictive accuracy among double-platform synthetic techniques, respectively.

Assessing the skill of gridded satellite and reanalysis precipitation products over in East and Southern Africa

Validation of gridded precipitation products (GPP) increases the users’ confidence and highlights possible improvements in the algorithms to handle complex rain-forming processes. We evaluated the skill of three GGPs (CHIRPS-v2, CHELSA, and TerraClimate) in estimating the rain gauge observations and compared the precipitation trends derived from these products across the East and Southern Africa (ESA) region. We used Taylor diagrams and Kling-Gupta Efficiency (KGE) to assess the accuracy. A modified Mann-Kendal test and a Sen’s slope estimator were utilized to determine the trends’ significance and magnitude, respectively. The three GPPs had varied performance over temporal and altitudinal ranges. The skill of the three GPPs, at a monthly scale, was generally high but showed lower performance at elevations over 1500 masl, especially during the October-November-December (OND) season. The three GPPs performed equally well between the 1001 – 1500 masl elevation range. CHELSA-v2.1 was most accurate at 0-500 masl but had the lowest skill in both 501 – 1000 and above 1500 masl elevations, which caused over-estimation of the annual and seasonal precipitation trends over mountainous terrain and large inland water bodies. The quantified precipitation trends revealed high spatial-temporal variability. Generally, the skill and precipitation trends derived from CHIRPS-v2 and TC data showed substantial convergence except in Tanzania. Our results emphasize the importance of validating climate datasets to avoid error propagation in different models and applications. Moreover, we demonstrate that new or higher-resolution precipitation data are not always accurate since an algorithm update can introduce artifacts or biases.

Integration of machine learning models and metaheuristic algorithms for predicting compressive strength of waste granite powder concrete

The utilization of Granite Powder (GP), a by-product of granite processing, has arisen great attention due to environmental challenge posed by its casual disposal. To promote reasonable use of GP in concrete and obtain optimal ratio design of Granite Powder Concrete (GPC), this study explored four advanced machine learning models (Support Vector Machine (SVM), eXtreme Gradient Boosting (XGB), Random Forest (RF), and Multi-Layer Perceptron (MLP)) integrated with two novel metaheuristic algorithms (Salp Swarm Algorithms (SSA) and Ant Lion Optimizer (ALO)) to predict the compressive strength of GPC. The models consider seven effect factors on the compressive strength: cement, water, fine aggregate, granite powder, superplasticizer, coarse aggregates, and curing age. The comprehensive score ranking system was utilized four evaluation metrics to assess models’ performance: Coefficient of Determination (R2), Variance Accounted For (VAF), Mean Absolute Error (MAE), and Mean Square Error (MSE). In addition, Taylor diagrams and Regression Error Characteristic (REC) curves were used to compare the performance of the models. The results showed that the XGB-based hybrid model outperformed the other methods, with an optimal model of ALO-XGB achieving a high level of accuracy in predicting compressive strength. Furthermore, the study highlights the importance of understanding the interplay between various factors and their effect on the mechanical properties of GPC, which can inform the development of more sustainable and environmentally friendly building materials.

Simulation of extreme rainfall events in Uruguay: Role of initialization and large scale dynamics

We investigate the role of initial conditions in the simulation of extreme rainfall events in summer (December–January-February) in Uruguay using RegCM4.7. Although extreme events are defined for the south of the country, considering the dimensions of these kind of events, the validation is performed in a larger region. Firstly, the quality of the observational database is analyzed and filtered. Then, RegCM4.7 is run in two different modes. The first mode is weather like (WL), where the model is initialized with reanalysis 24 h before the extreme event date. The second one is climate mode (CM), that simulates 5 months (starting 1st October) and their initial conditions are realistic only for the first day. Then, simulated rainfall were compared to the observations with Taylor diagrams, Spearman correlations and accumulated values. The difference in the ability of both modes is associated to the role of initial conditions. Overall, half of the extreme events were simulated with good skill in WL while CM was able to skillfully simulate 28% of the events. This difference is attributed to the better simulation of atmospheric circulation in WL, especially in low levels. However, in some particular cases, relevant atmospheric variables were more realistic in CM and its rainfall simulations did also. Finally, we found that in CM, most of the model's skill comes from the large scale circulation anomalies associated with ENSO teleconnections in subtropical South America.

Prediction of Future Urban Rainfall and Waterlogging Scenarios Based on CMIP6: A Case Study of Beijing Urban Area

Extreme weather events will become more frequent and severe as a result of climate change, necessitating an immediate need for cities to adapt to future climate change. Therefore, the prediction of future precipitation and waterlogging is of utmost importance. Using Beijing as an example, the simulation capability of different models was evaluated, and the optimal model for the study area was screened using Taylor diagrams and interannual variability scores, along with actual monthly precipitation data from Chinese weather stations from 1994 to 2014 and historical monthly precipitation data from 10 coupled models from Coupled Model Intercomparison Project Phase 6 (CMIP6). The SWMM model was then used to simulate future rainfall and waterlogging scenarios for the study area using precipitation forecast data for 2020–2050 from the best model to investigate the impact of climate change on future rainfall and waterlogging in urban areas. CMIP6 brings together the most recent simulation data from major climate models on a global scale, providing a broader and more diverse range of model results and thereby making future predictions more accurate and dependable, and its findings provide a theoretical foundation for the emergency management of and scientific responses to urban flooding events. The following major conclusions were reached: 1. The best-performing models are EC-Earth3, GFDL-ESM4, and MPI- ESM1-2-HR. EC-Earth3 is a modular Earth system model developed collaboratively by a European consortium. MPI-ESM1-2 is a climate precipitation prediction model developed in Germany and promoted for global application, whereas the GFDL-ESM4 model was developed in the United States and is currently employed for global climate precipitation simulations. 2. Under future climate circumstances, the total annual precipitation in the example region simulated by all three models increases by a maximum of 40%. 3. Under future climatic conditions, urban surface runoff and nodal overflow in the study area will be more significant. The node overflow will become more severe with the increase in climate scenario oppression, and the potential overflow nodes will account for 1.5%, 2.7%, and 2.9% of the total number of nodes under the SSP1–2.6, SSP2–4.5, and SSP5–8.5 scenarios, respectively. 4. In the future, the effectiveness of stormwater drainage systems may diminish. To increase climate change resilience, the impacts of climate change should be considered when planning the scope of stormwater optimization and the integrated improvement of gray–green–blue facilities.

ANFIS-based soft computing models for forecasting effective drought index over an arid region of India

Abstract Drought is a natural hazard that is characterized by a low amount of precipitation in a region. In order to evaluate the drought-related issues that cause chaos for human well-being, drought indices have become increasingly important. In this study, the monthly precipitation data from 1964 to 2013 (about 50 years) of the Jodhpur district in the drought-prone Rajasthan state of India was used to derive the effective drought index (EDI). The machine learning models hybridized with evolutionary optimizers such as the genetic algorithm adaptive neuro-fuzzy inference system (GA-ANFIS) and particle swarm optimization ANFIS (PSO-ANFIS) were used in addition to the generalized regression neural network (GRNN) to predict the EDI index. Using the partial autocorrelation function (PACF), models for forecasting the monthly EDI were constructed with 2-, 3- and 5-input combinations to evaluate their outcomes based on various performance indices. The results of the different combination models were compared. With reference to 2-input and 3-input combination models, both GA-ANFIS and PSO-ANFIS show better performance results with R2 = 0.75, while among the models with 5-input combination, GA-ANFIS depicts better performance results compared to other models with R2 = 0.78. The results are presented suitably with the aid of scatter plots, Taylor's diagram and violin plots. Overall, the GA-ANFIS and PSO-ANFIS models outperformed the GRNN model.

A novel integrated approach of RUNge Kutta optimizer and ANN for estimating compressive strength of self-compacting concrete

In this study machine learning models are used that forecast the compressive strength of self-compacting concrete (SCC) depending on percentage replacement of cement by supplementary cementitious material. In order to predict the mechanical property of SCC, a hybrid artificial neural network (ANN) along with metaheuristic optimization techniques and two traditional models were employed which comprised of 300 datasets. Several input factors are employed in the modelling process, including cement, the water-binder ratio, coarse aggregate, fine aggregate, fly ash (FA) and the superplasticizer. ANN hybridised with RUNge Kutta optimizer (RUN) was chosen as one of the top predictive models based on coefficient of determination ((R2 = 0.933 (in training) and R2 = 0.9203 (in testing))) findings and model validation. According to the rank analysis, the ANN-RUN surpassed other models with a score of 128. Permutation feature importance, statistical analysis, and comparisons amongst regression models are used to determine the models' efficacy. The best input parameters were selected based on person correlation and density plot. Also, the use of Taylor diagrams and accuracy matrices was also examined for the visual comprehension of the data which has shown similar results as discussed above. The results show that the suggested machine learning models attained consistent accuracy and can be used as a robust technique to predict the compressive strength of SCC.

Modeling of H2S solubility in ionic liquids: comparison of white-box machine learning, deep learning and ensemble learning approaches

In the context of gas processing and carbon sequestration, an adequate understanding of the solubility of acid gases in ionic liquids (ILs) under various thermodynamic circumstances is crucial. A poisonous, combustible, and acidic gas that can cause environmental damage is hydrogen sulfide (H2S). ILs are good choices for appropriate solvents in gas separation procedures. In this work, a variety of machine learning techniques, such as white-box machine learning, deep learning, and ensemble learning, were established to determine the solubility of H2S in ILs. The white-box models are group method of data handling (GMDH) and genetic programming (GP), the deep learning approach is deep belief network (DBN) and extreme gradient boosting (XGBoost) was selected as an ensemble approach. The models were established utilizing an extensive database with 1516 data points on the H2S solubility in 37 ILs throughout an extensive pressure and temperature range. Seven input variables, including temperature (T), pressure (P), two critical variables such as temperature (Tc) and pressure (Pc), acentric factor (ω), boiling temperature (Tb), and molecular weight (Mw), were used in these models; the output was the solubility of H2S. The findings show that the XGBoost model, with statistical parameters such as an average absolute percent relative error (AAPRE) of 1.14%, root mean square error (RMSE) of 0.002, standard deviation (SD) of 0.01, and a determination coefficient (R2) of 0.99, provides more precise calculations for H2S solubility in ILs. The sensitivity assessment demonstrated that temperature and pressure had the highest negative and highest positive affect on the H2S solubility in ILs, respectively. The Taylor diagram, cumulative frequency plot, cross-plot, and error bar all demonstrated the high effectiveness, accuracy, and reality of the XGBoost approach for predicting the H2S solubility in various ILs. The leverage analysis shows that the majority of the data points are experimentally reliable and just a small number of data points are found beyond the application domain of the XGBoost paradigm. Beyond these statistical results, some chemical structure effects were evaluated. First, it was shown that the lengthening of the cation alkyl chain enhances the H2S solubility in ILs. As another chemical structure effect, it was shown that higher fluorine content in anion leads to higher solubility in ILs. These phenomena were confirmed by experimental data and the model results. Connecting solubility data to the chemical structure of ILs, the results of this study can further assist to find appropriate ILs for specialized processes (based on the process conditions) as solvents for H2S.

Performance comparison of IHACRES, random forest and copula-based models in rainfall-runoff simulation

In this study, two models of Random Forest and copula-based simulation were used to evaluate the accuracy and efficiency of the IHACRES rainfall-runoff model in simulating the daily discharge of the Siminehroud River in the south of Lake Urmia basin, Iran. A trivariate copula-based model was created using discharge, rainfall and temperature data on a daily scale in the period 1992–2018. Vine family models and their conditional densities were used to implement the copula-based model. By calibrating the IHACRES model and also selecting the tree sequence in accordance with the data, rainfall-runoff simulations were performed in the study area. The accuracy and efficiency of the studied models were evaluated using RMSE and NSE criteria, and also violin plot and Taylor diagram. The results of comparing the error rate of rainfall-runoff simulation in the study area showed that the vine-based model reduces the RMSE statistics by about 14.5 and 16.5%, respectively, compared to the IHACRES and Random Forest models. According to the presented diagrams, the efficiency and certainty of IHACRES and copula-based simulation models are acceptable. While the Random Forest model does not have acceptable accuracy and efficiency in the study area. The copula-based simulation model has a good performance due to the unique tree sequence as well as involving the marginal distributions fitted to the data. Although the copula-based simulation model has increased the efficiency of the model in simulating the daily discharge by about 5% compared to the IHACRES model, it is not significant compared to the mathematical complexity of the copula-based model.