Taylor Diagram Research Articles

Prediction of groundwater level fluctuations using artificial intelligence-based models and GMS

Groundwater level fluctuations are one of the main components of the hydrogeological cycle and one of the required variables for many water resources operation models. The numerical models can estimate groundwater level (GWL) based on extensive statistics and information and using complex equations in any area. But one of the most important challenges in analyzing and predicting groundwater depletion in water management is the lack of reliable and complete data. For this reason, the use of artificial intelligence models with high predictive accuracy and due to the need for less data is inevitable. In recent years, the use of different numerical models has been noticed as an efficient solution. These models are able to estimate groundwater levels in any region based on extensive statistics and information and also various field experiments such as pumping tests, geophysics, soil and land use maps, topography and slope data, different boundary conditions and complex equations. In the current research, first, by using available statistics, information and maps, the groundwater level fluctuations of the Sonqor plain are simulated by the GMS model, and the accuracy of the model is evaluated in two stages of calibration and validation. Then, due to the need for much less data volume in artificial intelligence-based methods, the GA-ANN and ICA-ANN hybrid methods and the ELM and ORELM models are utilized. The results display that the output of the ORELM model has the best fit with observed data with a correlation coefficient equal to 0.96, and it also has the best and closest scatter points around the 45 degrees line, and in this sense, it is considered as the most accurate model. To ensure the correct selection of the best model, the Taylor diagram is also used. The results demonstrate that the closest point to the reference point is related to the ORELM method. Therefore, to predict the groundwater level in the whole plain, instead of using the complex GMS model with a very large volume of data and also the very time-consuming process of calibration and verification, the ORELM model can be used with confidence. This approach greatly helps researchers to predict groundwater level variations in dry and wet years using artificial intelligence with high accuracy instead of numerical models with complex and time-consuming structures.

Prediction of the Dynamic Viscosity of N-Alcohol by Three Intelligent Models (ANN, LSSVM, and ANFIS) in Operation Conditions

Viscosity is an essential property in chemical engineering for different applications. If predicted accurately, the viscosity has a significant effect on chemical applications. In this paper, the capability of the three intelligence models, artificial neural network (ANN), least squares support vector machine (LSSVM), and adaptive neuro-fuzzy inference system (ANFIS), were evaluated to model the dynamic viscosity of n-alcohol at the different operational conditions. The models were improved based on a 237 data set collected from reliable articles. The used databank contains temperature (T), pressure (P), and carbon number of n-alcohols (n-C) were chosen as the input of models. The result of these models was studied by Statistical parameters such as Mean of the squares errors (MSE), Root mean of the squares errors (RMSE), Maximum absolute error (MAAE %), Mean absolute error (MEAE %), correlation coefficient and graphical technique like Taylor diagram and William plot. The proposed models are known to appropriately estimate the viscosity of n-alcohol at the different operational conditions. It was found that the ANN with R2= 0.999, MSE=0.000017, MAAE %= 1.6, MEAE %=0.32 for Test, and R2= 0.999, MSE=0.0000094, MAAE %= 0.83, MEAE %=0.23 for Train exhibited a high performance than LSSVM and ANFIS for predicting dynamic viscosity of n-alcohol at the operational conditions.

Application of augmented bat algorithm with artificial neural network in forecasting river inflow in Malaysia

Hydrologists rely extensively on anticipating river streamflow (SF) to monitor and regulate flood management and water demand for people. Only a few simulation systems, where previous techniques failed to anticipate SF data quickly, let alone cost-effectively, and took a long time to execute. The bat algorithm (BA), a meta-heuristic approach, was used in this study to optimize the weights and biases of the artificial neural network (ANN) model. The proposed hybrid work was validated in five different study areas in Malaysia. The statistical tests analysis of the preliminary results revealed that hybrid BA-ANN was superior to forecasting the SF at all five selected study areas, with average RMSE values of 0.103 m3/s for training and 0.143 m3/s for testing as compared to ANN standalone training and testing yielding 0.091 m3/s and 0.116 m3/s, respectively. This finding signifies that the implementation of BA into the ANN model resulted in a 20% improvement. In addition, with an R2 score of 0.951, the proposed model showed a better correlation than the 0.937 value of R2 of standard ANN. Nonetheless, while the proposed work outperformed the conventional ANN, the Taylor diagram, violin plot, relative error, and scatter plot findings confirmed the disparities in the proposed work’s performance throughout the research regions. The findings of these evaluations highlighted that the adaptability of the proposed works would need detailed investigation because its performance differed from case to case.

Future Changes in Temperature and Precipitation over Northeastern Brazil by CMIP6 Model

Global warming is causing an intensification of extreme climate events with significant changes in frequency, duration, and intensity over many regions. Understanding the current and future influence of this warming in northeastern Brazil (NEB) is important due to the region’s greater vulnerability to natural disasters, as historical records show. In this paper, characteristics of climate change projections (precipitation and air temperature) over NEB are analyzed using 15 models of Coupled Model Intercomparison Project Phase 6 (CMIP6) under four Shared Socioeconomic Pathways (SSPs: SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) scenarios. By using the Taylor diagram, we observed that the HadGEM3-GC31-MM model simulates the seasonal behavior of climate variables more efficiently. Projections for NEB indicate an irreversible increase in average air temperature of at least 1 °C throughout the 21st century, with a reduction of up to 30% in annual rainfall, as present in scenarios of regional rivalry (SSP3-7.0) and high emissions (SSP5-8.5). This means that a higher concentration of greenhouse gases (GHG) will increase air temperature, evaporation, and evapotranspiration, reducing rainfall and increasing drought events. The results obtained in this work are essential for the elaboration of effective strategies for adapting to and mitigating climate change for the NEB.

Satellite-based estimation of daily suspended sediment load using hybrid intelligent models

ABSTRACT This study uses a combination of support vector regression models, particle swarm optimization, and grey wolf optimization algorithms to predict suspended sediment load. For this purpose, The Satellite Precipitation of Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) and Global Land Data Assimilation System (GLDAS) soil moisture products are utilized as the predictors. The prediction models are evaluated based on various visual and quantitative indicators. The Taylor and radar diagrams confirm that the support vector regression-particle swarm optimization best agrees with the observed values. Moreover, the obtained quantitative indices show that the support vector regression-particle swarm optimization model offers better performance than other models used in the present study. The values of the best indices are: Pearson correlation coefficient of 0.997, relative root mean square error of 13.17, percentage bias of 4.05, and Nash-Sutcliffe efficiency of 0.995.

Droughts in Germany: performance of regional climate models in reproducing observed characteristics

Abstract. Droughts are among the most relevant natural disasters related to climate change. We evaluated different regional climate model outputs and their ability to reproduce observed drought indices in Germany and its near surroundings between 1980–2009. Both outputs of an ensemble of six EURO-CORDEX models of 12.5 km grid resolution and outputs from a high-resolution (5 km) Weather Research and Forecasting (WRF) run were employed. The latter model was especially tailored for the study region regarding the physics configuration. We investigated drought-related variables and derived the 3-month standardized precipitation evapotranspiration index (SPEI-3) to account for meteorological droughts. Based on that, we analyzed correlations, the 2003 event, trends and drought characteristics (frequency, duration and severity) and compared the results to E-OBS. Methods used include Taylor diagrams, the Mann–Kendall trend test and the spatial efficiency (SPAEF) metric to account for spatial agreement of patterns. Averaged over the domain, meteorological droughts were found to occur approximately 16 times in the study period with an average duration of 3.1 months and average severity of 1.47 SPEI units. WRF's resolution and setup were shown to be less important for the reproduction of the single drought event and overall drought characteristics. Depending on the specific goals of drought analyses, computation resources could therefore be saved, since a coarser resolution can provide similar results. Benefits of WRF were found in the correlation analysis. The greatest benefits were identified in the trend analysis: only WRF was able to reproduce the observed negative SPEI trends to a fairly high spatial accuracy, while the other regional climate models (RCMs) completely failed in this regard. This was mainly due to the WRF model settings, highlighting the importance of appropriate model configuration tailored to the target region. Our findings are especially relevant in the context of climate change studies, where the appropriate reproduction of trends is of high importance.

Modeling and Evaluation of the Permeate Flux in Forward Osmosis Process with Machine Learning

Predicting the permeate flux is critical for evaluating and optimizing the performance of the forward osmosis (FO) process. However, the solution diffusion models have poor applicability in accessing the FO process. Recently, the data-driven eXtreme Gradient Boosting (XGBoost) algorithm has been proven to be effective in processing structure data in engineering problems and has not been utilized to assess the FO process. Herein, a combination of the XGBoost model with a genetic algorithm (GA) was first proposed to predict the permeate flux, highlighting its superiority in the FO process through comparison of the support vector regression (SVR) model, the artificial neural network (ANN), and the multiple linear regression (MLR). Moreover, the performance of these models was optimized by tuning hyperparameters with a genetic algorithm (GA) and compared via Taylor Diagram. Among these machine learning (ML) models, the GA-based XGBoost model is superior to the other three models in terms of mean square error (MSE, 2.7326) and coefficient of determination (R2, 0.9721) on the test data, and its prediction power was compared to that of the solution diffusion (SD) model in the literature. Finally, further insight into the feature importance that affects the permeate flux in the FO process was examined by utilizing the SHapley Additive exPlanations (SHAP) to estimate the contribution value of various variables. The results demonstrated that the XGBoost model could predict the permeate flux in the FO system with high accuracy and good generalization ability for the given data set and even on the unseen data. Furthermore, the findings of the SHAP method show that the osmotic pressure difference, the osmotic pressure difference of draw solution and FS solution, the crossflow velocity of the feed solution and draw solution, and the water permeability coefficient have a significant impact on water flux.

Projection of extreme precipitation in the Minjiang River Basin, Southeast China

AbstractExtreme precipitation events, which have intensified with global warming, will have a serious influence on society. Based on the latest generation of CMIP6 climate models and high-resolution grid observation data, the quantile mapping method and Taylor diagrams were used to correct and evaluate the simulation values. Six extreme precipitation indices for the near (2021–2040), middle (2051–2070), and far (2081–2100) periods of the 21st century were analyzed under shared socioeconomic pathway scenarios (SSP2-4.5 and SSP5-8.5). The results show that most of the extreme precipitation indices will increase over the Minjiang River Basin in the future, and both extreme precipitation and persistent drought will increase at the same time, which is more likely to cause extreme drought and flood disasters. For seasonal variation, the total amount and intensity of extreme precipitation will increase fastest in autumn, while the frequency of extreme precipitation will increase most in summer. Multimodel median changes show a decrease in the return period of RX1day (the maximum 1-day precipitation). By the end of the 21st century, under SSP5-8.5, the extreme events expected once every 100 years under the current climate are expected to occur approximately every 18.6 years.

A machine learning-based analysis for predicting fragility curve parameters of buildings

Fragility curves are one of the substantial means required for seismic risk assessment of buildings in the framework of performance-based earthquake engineering (PBEE). Obtaining fragility curves, however, needs an extensive analytical analysis which makes it time-consuming and sometimes inaccurate due to errors. In this study, hence, machine learning (ML)-based models are proposed for predicting fragility parameters of structures namely dispersion, β, and μ (log (PGA), based on a reduced set of information avoiding cumbersome computation. Firstly, a brief review on the analytical models proposed for deriving fragility curves is provided as well as ML-based models developed so far. Then, to achieve the research objective, a comprehensive database including 214 datasets from peer-reviewed international publications is collected. It is then divided into training (85%) and testing (15%) sub-datasets for the purpose of training the models and assessing the results, respectively. The most effective parameters on the target outputs are defined as input variables including construction material, building plan area, building height, damage state, buildings' period and soil classification. β and μ were estimated utilizing various ML-based techniques namely nonlinear regression, decision tree, random forest, K-Nearest Neighbours (KNN) and Artificial Neural Network (ANN). The actual values and the values predicted by the proposed models are compared. Moreover, the models' accuracy is assessed through performance metrics and Taylor diagram. The results demonstrated the high ability of the models for learning the relationship between inputs and outputs. According to the accuracy assessment, Decision tree was the most accurate model for predicting both β and μ. A sensitivity study was also conducted by changing input variables, and estimation equations are provided accordingly. • A critical review on the conventional and machine learning-based methods proposed for obtaining fragility curves of buildings is presented. • A database including 214 fragility curves' parameters is collected. • Different machine learning-based techniques are developed and their accuracy is compared and discussed. • Sensitivity analysis of parameters affecting fragility curves is performed.

Evaluation of CORDEX‐CORE regional climate models in simulating rainfall variability in Rwanda

AbstractIn this study, three regional climate models (RCMs), CCLM5‐0‐15, RegCM4‐7 and REMO2015, from CORDEX‐CORE (AFR‐22) are evaluated in their ability to reproduce rainfall variability in Rwanda for the period 1981–2005. They are driven by three different global climate models (GCMs), namely MPI‐M‐MPI‐ESM‐LR, NCC‐NorESM1‐M and MOHC‐HadGEM2‐ES, and the European Centre for Medium‐Range Weather Forecasts Reanalysis (ECMWF‐ERAINT). Simulated rainfall is evaluated against observations from Rwanda Meteorology Agency to assess models' performance. A set of metrics are used to quantify discrepancies of models' simulations from observations. A possible association of El Niño–Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD) to rainfall over Rwanda is investigated. It is found that in general, all RCMs, their ensemble and multimodel ensemble means reproduce satisfactorily the spatial distribution of the mean seasonal rainfall (MSR), the mean rainfall annual cycle, and the interannual variability of the MSR for both March–April–May (MAM) and October–November–December (OND). However, significant biases in individual RCMs are observed with varying magnitude of bias in space. Observed MSR indicates a positive trend of 0.045 and 0.058 mm·day·year−1, respectively, for MAM and OND at 0.05 significance level, but almost all models indicate no significant trend (at 0.05 significance level). The seasonal correlations between observed rainfall anomalies and sea surface temperature (SST) anomalies indices across the tropical Pacific (Niño1+2 and Niño3.4) and Indian Oceans associated, respectively, with ENSO and IOD, although relatively weak, are reproduced by the three RCMs driven by ECMWF‐ERAINT and the multimodel ensemble means of ECMWF‐ERAINT and MPI‐M‐MPI‐ESM‐LR. Analysis of the Taylor diagram indicates that CCLM5‐0‐15_MPI‐M‐MPI‐ESM‐LR and the multimodel ensemble mean of MPI‐M‐MPI‐ESM‐LR outperform individual models. Overall, the evaluation finds reasonable model skill in representing seasonal rainfall climatology and variability, suggesting the potential use of CORDEX‐CORE (AFR‐22) RCMs for the assessment of future climate projections in Rwanda.

A comparison of carbon monoxide retrievals between the MOPITT satellite and Canadian high-Arctic ground-based NDACC and TCCON FTIR measurements

Abstract. Measurements of Pollution In The Troposphere (MOPITT) is an instrument on NASA's Terra satellite that has measured tropospheric carbon monoxide (CO) from early 2000 to the present day. Validation of data from satellite instruments like MOPITT is often conducted using ground-based measurements to ensure the continued accuracy of the space-based instrument's measurements and its scientific results. Previous MOPITT validation studies generally found a larger bias in the MOPITT data poleward of 60∘ N. In this study, we use data from 2006 to 2019 from the Bruker IFS 125HR Fourier Transform Infrared spectrometer (FTIR) located at the Polar Environment Atmospheric Research Laboratory (PEARL) in Eureka, Nunavut, Canada, to validate the MOPITT version 8 (V8) retrievals. These comparisons utilize mid- and near-infrared FTIR measurements made as part of the Network for the Detection for Atmospheric Composition Change (NDACC) and the Total Carbon Column Observing Network (TCCON), respectively. All MOPITT version 8 retrievals within a radius of 110 km (1∘) from the PEARL Ridge Laboratory and within a 24 h time interval are used in this validation study. MOPITT retrieval products include those from the near-infrared (NIR) channel, the thermal infrared (TIR) channel, and a joint product from the thermal and near-infrared (TIR–NIR) channels. Each channel's detector has 4 pixels. We calculated the MOPITT pixel-to-pixel biases for each pixel, which were found to vary based on the season and surface type (land or water). The systematic bias for pixel 1 over land is larger than that for other pixels, which can reach up to 20 ppb. We use a small-region approximation method to find filtering criteria. We then apply the filters to the MOPITT dataset to minimize the MOPITT pixel bias and the number of outliers in the dataset. The sensitivity of each MOPITT pixel and each product is examined over the Canadian high Arctic. We then follow the methodologies recommended by NDACC and TCCON for the comparison between the FTIR and satellite total column retrievals. MOPITT averaging kernels are used to weight the NDACC and TCCON retrievals and take into account the different vertical sensitivities between the satellite and PEARL FTIR measurements. We use a modified Taylor diagram to present the comparison results from each pixel for each product over land and water with NDACC and TCCON measurements. Our results show overall consistency between MOPITT and the NDACC and TCCON measurements. When compared to the FTIR, the NIR MOPITT retrievals have a positive bias of 3 %–10 % depending on the pixel. The bias values are negative for the TIR product, with values between −5 % and 0 %. The joint TIR–NIR products show differences of −4 % to 7 %. The drift in MOPITT biases (in units of % yr−1) relative to NDACC and TCCON varies by MOPITT data product. In the NIR, drifts vs. TCCON are smaller than those vs. NDACC; however, this scenario is reversed for the MOPITT TIR and joint TIR–NIR products. Overall, this study aims to provide detailed validation for MOPITT version 8 measurements in the Canadian high Arctic.

Enhancing Sustainability of Corroded RC Structures: Estimating Steel-to-Concrete Bond Strength with ANN and SVM Algorithms

The bond strength between concrete and corroded steel reinforcement bar is one of the main responsible factors that affect the ultimate load-carrying capacity of reinforced concrete (RC) structures. Therefore, the prediction of accurate bond strength has become an important parameter for the safety measurements of RC structures. However, the analytical models are not enough to estimate the bond strength, as they are built using various assumptions and limited datasets. The machine learning (ML) techniques named artificial neural network (ANN) and support vector machine (SVM) have been used to estimate the bond strength between concrete and corroded steel reinforcement bar. The considered input parameters in this research are the surface area of the specimen, concrete cover, type of reinforcement bars, yield strength of reinforcement bars, concrete compressive strength, diameter of reinforcement bars, bond length, water/cement ratio, and corrosion level of reinforcement bars. These parameters were used to build the ANN and SVM models. The reliability of the developed ANN and SVM models have been compared with twenty analytical models. Moreover, the analyzed results revealed that the precision and efficiency of the ANN and SVM models are higher compared with the analytical models. The radar plot and Taylor diagrams have also been utilized to show the graphical representation of the best-fitted model. The proposed ANN model has the best precision and reliability compared with the SVM model, with a correlation coefficient of 0.99, mean absolute error of 1.091 MPa, and root mean square error of 1.495 MPa. Researchers and designers can apply the developed ANN model to precisely estimate the steel-to-concrete bond strength.

Prediction of evaporation from dam reservoirs under climate change using soft computing techniques.

This study aimed to predict evaporation from dam reservoirs using artificial intelligence considering climate change. Mahabad Dam, near Lake Urmia, in northwestern Iran, is used to investigate the proposed approach. There are three parts to the study presented herein. In the first part, two machine learning models, namely group method of data handling (GMDH) and least squares support vector regression (LS-SVR), were used to model the inflow to the dam reservoir. Temperature, precipitation, and inflow during the previous month from 1990 to 2017 were used as input data. In the second part, the evaporation from the dam reservoir was modeled using the adaptive neuro-fuzzy inference system (ANFIS) and optimized ANFIS using Harris hawks optimization (HHO) and the arithmetic optimization algorithm (AOA) optimization algorithms. The input parameters in this part were temperature, precipitation, inflow to the dam reservoir, along with evaporation from the dam reservoir in the previous month. In the third part, precipitation and temperature were predicted using the fifth report of IPCC based on RCP2.6, RCP4.5, and RCP8.5 scenarios for the period 2020-2040. Out of 28 models presented in the fifth report, EC-ERATH and FIO-ESM had the greatest similarity with observational data of temperature and precipitation, respectively. The results of scatter plots and Taylor's diagram showed the higher performance of LS-SVR (root mean square error (RMSE), mean absolute percentage error (MAPE), and Nash-Sutcliffe efficiency (NSE) of 8.65, 4.69, and 0.96) compared to GMDH (RMSE, MAPE, and NSE of 11.65, 7.81, and 0.93) in modeling the inflow. Moreover, both hybrid modes (AOA-ANFIS and HHO-ANFIS) improved the performance of ANFIS in modeling the evaporation from the dam reservoir. The RMSE, MAPE, and NSE values for ANFIS were 0.56, 0.52, and 0.89, respectively, while these values for the AOA-ANFIS (RMSE, MAPE, and NSE of 0.31, 0.24, and 0.93) and HHO-ANFIS (RMSE, MAPE, and NSE of 0.20, 0.30, and 0.96) were improved remarkably. The impact of climate change reduced the inflow to the dam reservoir by about 0.45, 0.80, and 1.7 MCM in RCP2.6, RCP4.5, and RCP8.5 scenarios, respectively. Also, the effect of climate change caused the evaporation from the dam reservoir to increase by about 0.2, 0.9, and 1 MCM in RCP2.6, RCP4.5, and RCP8.5 scenarios, respectively. The findings of this study show that the correct management of dam reservoirs needs to consider the potential effects of climate change in the future. Moreover, the hybrid machine learning models used in this study can be used to predict the amount of evaporation in other reservoirs.

Determination of the size of rock fragments using RVM, GPR, and MPMR

For predicting the size of rock fragments during drilling and blasting operations, this article uses GPR, RVM, and MPMR. The current analysis makes use of a blast data set generated in a prior investigation. In this study, a portion of the blast data was utilized to train a model to determine the mean particle size arising from blast fragmentation for each of the similarity groups generated. The particle size was modeled as a function of seven different variables. The dataset contains information about the bench height and drilled burden ratio (H / B), spacing to burden ratio (S / B), burden to hole diameter ratio (B / D), stemming to burden ratio (T / B), powder factor (Pf ), modulus of elasticity (E), and in-situ block size (XB) are the input and output is X50. By comparing forecasts with actual mean particle size values and predictions based on one of the most widely used fragmentation estimation techniques in the blasted literature, the capacity of the generated models may be established. The statistical parameters, actual vs predicted curve, Taylor diagram, error bar, and developed discrepancy ratio are used to analysis the performance of models. A comparative study has been carried out between the developed RVM, GPR, and MPMR. The results show the developed models have the capability for prediction of X50. From these comparisons, the MPMR has the highest value with a high degree of precision and robustness in the size of rock fragments X50.

Vulnerability assessment of drought in India: Insights from meteorological, hydrological, agricultural and socio-economic perspectives

Drought is a complicated and diverse natural calamity that causes huge global economic losses. Drought, particularly in dominant agricultural areas, is becoming more severe, posing a threat to global food security. Drought stress has a wide range of effects on a region, and it is largely dependent upon climate variability, groundwater availability and soil moisture condition. Droughts are influenced by a wide range of factors, including climatic factors like rainfall and temperature and economic factors like population density, irrigated land, and so on. In India, the recurrence of drought is largely owed to sole physical and climatic susceptibilities. As a result, this research study makes an intensive effort to assess drought vulnerability assessment in India considering “meteorological, hydrological, agricultural and socio-economic drought” groups. The statistical method of “analytical hierarchy process (AHP)” and machine learning algorithms of “random forest (RF)” and “Bayesian additive regression trees (BART)” were used for respective modelling purposes. The modelling outcomes were assessed using “receiver operating characteristics area under the curve (ROC-AUC)”, “mean square error (MSE)”, “root mean square error (RMSE)”, and the Taylor diagram. The evaluation measure indicates that BART is the most optimal (AUC = 0.901, r = 0.931) followed by RF (AUC = 0.872, r = 0.901). The findings suggest that the method used to determine drought vulnerability in the region is successful, which will aid planners in developing drought mitigation measures.

Stacking- and voting-based ensemble deep learning models (SEDL and VEDL) and active learning (AL) for mapping land subsidence.

This contribution presents a novel methodology based on the feature selection, ensemble deep learning (EDL) models, and active learning (AL) approach for prediction of land subsidence (LS) hazard and rate, and its uncertainty in an area involving two important plains - the Minab and Shamil-Nian plains - in the Hormozgan province, southern Iran. The important features controlling LS hazard were identified by ridge regression. Then, two EDL models were constructed by stacking (SEDL) and voting (VEDL) five dense deep learning (DL) models (model 1 to model 5) for mapping LS hazard. Thereafter, the predictive model performance was assessed by a precision-recall curve and Kolmogorov-Smirnov (KS) plot. A partial dependence plot (PDP), individual conditional expectation plots (ICEP), game theory, and a sensitivity analysis were used for the interpretability of the predictive DL model. According to SEDL - a model with higher accuracy - 34% (1624 km2), 14.7% (698 km2), and 19.2% (912 km2) of the total area were classified as being of very low, low, and moderate hazards, whereas 17.7% (845 km2) and 14.4% (683 km2) of area were classified as being of high and very high hazards, respectively. Based on all interpretability techniques, aquifer loss or groundwater drawdown is the most important feature controlling LS hazard, and it having the greatest impact on the SEDL model output. Based on a Taylor diagram and R2 as model performance assessment indicators, SEDL-AL (with R2 > 95% for training and test datasets) performed better than SEDL for quantify LS rate, the rate of LS ranging between 0 and 48.1cm. The highest rate of LS occurred in the Minab plain - an area located downstream of the Minab Esteghlal dam. SEDL-AL was used to quantify the uncertainty associated with the LS rate. The observed values fell within predictions provided by SEDL-AL, which indicates a high accuracy of our predictive model. Overall, our newly developed modeling techniques are helpful tools for the spatial mapping of LS susceptibility and rate, and its uncertainty.

Performance assessment of gridded snow products in the Upper Euphrates

Snow observations are important in many mountain areas to quantify the water stored in snowpacks and to predicting runoff during the melting period. In this study we compare the performance of five different regional-scale gridded snow products to reproduce snow water equivalent (SWE) in the Upper Euphrates region (Karasu Basin, 10,275 km2), with observations from automatic weather stations in the catchment through Taylor diagrams. The products compared are the ERA5, ERA5-Land, MERRA-2, snow data from a dynamical downscaling of ERA-5 (period 2000-2018) and SWE generated from microwave satellite data (SWE-E(H13) period 2013-2015 product of the EUMETSAT H SAF project). The H13 product presented deficiencies in terms of not being able to reproduce the spatial and temporal variability of the snowpack. ERA-5 and, in particular, ERA-Land products, at 30 and 9 km grid size, respectively, showed good performance in reproducing snow evolution compared to four available observation sites. MERRA2 at 50 km resolution showed lower skills compared to the above-mentioned products. Resulting snow data from WRF at 10 km resolution did not show any improvement with respect to the global datasets. The impossibility of testing different configurations due to the lack of observations to compare and the computational constraints to test different parametrizations may be the reasons to explain the low performance although they remain speculative. All the gridded datasets showed good performance in reproducing snow duration over the basin, compared to remotely sensed data. Results highlight ERA-Land dataset as a very promising tool for regional snow studies in mountainous regions with limited observations, in a cost-effective way.

Hybridized Adaptive Neuro-Fuzzy Inference System with Metaheuristic Algorithms for Modeling Monthly Pan Evaporation

Precise estimation of pan evaporation is necessary to manage available water resources. In this study, the capability of three hybridized models for modeling monthly pan evaporation (Epan) at three stations in the Dongting lake basin, China, were investigated. Each model consisted of an adaptive neuro-fuzzy inference system (ANFIS) integrated with a metaheuristic optimization algorithm; i.e., particle swarm optimization (PSO), whale optimization algorithm (WOA), and Harris hawks optimization (HHO). The modeling data were acquired for the period between 1962 and 2001 (480 months) and were grouped into several combinations and incorporated into the hybridized models. The performance of the models was assessed using the root mean square error (RMSE), mean absolute error (MAE), Nash–Sutcliffe Efficiency (NSE), coefficient of determination (R2), Taylor diagram, and Violin plot. The results showed that maximum temperature was the most influential variable for evaporation estimation compared to the other input variables. The effect of periodicity input was investigated, demonstrating the efficacy of this variable in improving the models’ predictive accuracy. Among the models developed, the ANFIS-HHO and ANFIS-WOA models outperformed the other models, predicting Epan in the study stations with different combinations of input variables. Between these two models, ANFIS-WOA performed better than ANFIS-HHO. The results also proved the capability of the models when they were used for the prediction of Epan when given a study station using the data obtained for another station. Our study can provide insights into the development of predictive hybrid models when the analysis is conducted in data-scare regions.

Evaluation of the forecast skill of North American Multi‐Model Ensemble for monthly and seasonal precipitation forecasts over Iran

AbstractNorth American Multi‐Model Ensemble (NMME) precipitation forecast skill over Iran is evaluated using Taylor diagrams and ranked probability skill scores (RPSS) as determined over a 29‐year test period (1991–2019). The forecast skill for both monthly (October through June for lead‐times of 0.5–3.5 months) and seasonal (October–December [OND], January–March [JFM], and April–June [AMJ] for lead‐times of 1.5–3.5 months) timescales is evaluated using six NMME models as well as multi‐model ensemble means (MMM). The latest versions of these models for forecasting Iran's precipitation have not been evaluated thus far. The Global Precipitation Climatology Center (GPCC) version 2020 dataset is used to verify the models. Among individual NMME models, Geophysical Fluid Dynamics Laboratory‐Seamless System for Prediction and Earth System Research (GFDL‐SPEAR) has generally the highest forecast skill. Both Taylor diagrams and RPSS of most of the models have indicated that the highest forecast skill is found for the month of November such that the Pearson correlation for both SPEAR and MMM is statistically significant for all lead‐times. For both monthly and seasonal timescales, the temporal Pearson correlation (TPC) between the observed and forecasts of the MMM is higher than the TPC of the individual models. The spatial Pearson correlation (SPC) and normalized centred root mean square error (NCRMSE) of the SPEAR is close to MMM, but the normalized standard deviation (NSD) of the SPEAR is closer to one compared to the MMM for months from November to March and two seasons (OND and JFM seasons). The MMM precipitation forecasts are underestimated over the northern regions and Zagros mountains for JFM and OND seasons for both 1.5‐ and 2.5‐month lead‐times. The degree to which the forecast skill of MMM is dependent on the El Niño–Southern Oscillation (ENSO) connections with precipitation over Iran is examined. Significant Spearman correlations between simultaneous observed Niño3.4 index and Iran precipitation are found for OND, but not for JFM and AMJ. The MMM reproduces the observed ENSO teleconnections to the tropical Pacific in OND, consistent with forecast skill in that season. However, the MMM also produces forecast skill in JFM and AMJ when the ENSO influence is marginal, showing that ENSO is not the only source of skill in the models.

Investigating the Bond Strength of FRP Laminates with Concrete Using LIGHT GBM and SHAPASH Analysis.

The corrosion of steel reinforcement necessitates regular maintenance and repair of a variety of reinforced concrete structures. Retrofitting of beams, joints, columns, and slabs frequently involves the use of fiber-reinforced polymer (FRP) laminates. In order to develop simple prediction models for calculating the interfacial bond strength (IBS) of FRP laminates on a concrete prism containing grooves, this research evaluated the nonlinear capabilities of three ensemble methods-namely, random forest (RF) regression, extreme gradient boosting (XGBoost), and Light Gradient Boosting Machine (LIGHT GBM) models-based on machine learning (ML). In the present study, the IBS was the desired variable, while the model comprised five input parameters: elastic modulus x thickness of FRP (EfTf), width of FRP plate (bf), concrete compressive strength (fc'), width of groove (bg), and depth of groove (hg). The optimal parameters for each ensemble model were selected based on trial-and-error methods. The aforementioned models were trained on 70% of the entire dataset, while the remaining data (i.e., 30%) were used for the validation of the developed models. The evaluation was conducted on the basis of reliable accuracy indices. The minimum value of correlation of determination (R2 = 0.82) was observed for the testing data of the RF regression model. In contrast, the highest (R2 = 0.942) was obtained for LIGHT GBM for the training data. Overall, the three models showed robust performance in terms of correlation and error evaluation; however, the trend of accuracy was obtained as follows: LIGHT GBM > XGBoost > RF regression. Owing to the superior performance of LIGHT GBM, it may be considered a reliable ML prediction technique for computing the bond strength of FRP laminates and concrete prisms. The performance of the models was further supplemented by comparing the slopes of regression lines between the observed and predicted values, along with error analysis (i.e., mean absolute error (MAE), and root-mean-square error (RMSE)), predicted-to-experimental ratio, and Taylor diagrams. Moreover, the SHAPASH analysis revealed that the elastic modulus x thickness of FRP and width of FRP plate are the factors most responsible for IBS in FRP.