Kling-Gupta Efficiency Research Articles

Modeling sediment concentrations and loads for two small agricultural watersheds in Prince-Edward-Island (Canada): present conditions and a future scenario

The degradation of soils and its detrimental consequences on aquatic environments is an important research topic in agricultural regions such as Prince Edward Island (PEI, Canada). Enhanced information related to suspended sediments in watercourses can serve as an effective decision-making tool in agricultural land management. This study aims to compare flow, suspended sediment concentrations (SSC), and loads using the Soil and Water Assessment Tool (SWAT) in two watersheds in Prince Edward Island (PEI). The final investigations will focus on the potential variations in hydrological and sedimentary values in the future using a relatively pessimistic climate change scenario. Finally, the projected sediment concentrations and loads will be analyzed, considering their potential impacts on ecosystems. Water level and turbidity were recorded using two water level loggers and two optical backscatter sensors (OBS) deployed in the Tuplin Creek and Spring Valley watersheds. These instruments continuously recorded suspended sediments and flow data from June 2021 to September 2022. The data were used to manually calibrate the hydrological and suspended sediment models. The understanding of sediment loads and the benefits of proposed changes to agricultural practices can be tested with the SWAT model, as it incorporates a land use index that varies spatially and temporally. Calibration and validation of both the hydrological and sediment models were satisfactory, with Kling-Gupta Efficiency coefficients varying between 0.51 and 0.73 and Nash-Sutcliffe coefficients varying between 0.61 and 0.73 respectively, indicating successful simulation of both variables in an agricultural context in spite of relatively short calibration and validation periods. Under the selected climate change scenario (RCP 8.5), daily flows and suspended sediment concentrations were simulated until 2,100, showing a slight increase in the average suspended sediment concentration (CSS). For Tuplin Creek, extremely high sediment peaks (>1,500 mg/L) could become significantly more frequent, potentially causing more frequent and severe ecosystem disturbances according to the simulations.

Reconstructing monthly 0.25° terrestrial evapotranspiration data in a remote arid region using Bayesian-driven ensemble learning method

This study presents a Bayesian-driven ensemble learning framework to minimize errors in evapotranspiration (ET) data assimilation, thereby generating a continuous-scale ET estimation with improved accuracy for the data-scarce arid Central Asia (CA). Five state-of-the-art gridded ET products were extensively tested and utilized to reconstruct new long-term ET data at grid scale, including two Global Land Data Assimilation System (GLDAS)-Land Surface Models (LSM) (Catchment and NOAH), two reanalysis products (European Centre for Medium-Range Weather Forecasts Reanalysis 5 (ERA5) and Modern-Era Retrospective Analysis for Research and Applications – ver. 2 (MERRA2)), along with the Global Land Evaporation Amsterdam (GLEAM) satellite ET product ver. 3.6b. Taylor Skill Score was employed to evaluate the input gridded ET datasets, using in situ measurements across four arid biomes — cropland, shrubland, farmland, and grassland —as benchmarks. Bayes-Markov Chain Monte Carlo (Bayes-MCMC)-derived logarithmically transformed predictors from the posterior ensemble anomalies were standardized to resolve the mismatches in the cumulative distribution functions of ET data relating to the in situ observations. We found that the constrained residual errors were more profound between 45th and 95th percentiles for the rescaled total ET anomalies across all the input ET datasets at 95 % confidence interval. Moreover, the weighted ensemble estimates derived from the Bayes-MCMC were instrumental in correcting ET anomalies, leading to the robust reconstruction of a more reliable ensemble mean ET data for Central Asia (CAETens) at monthly temporal and 0.25° spatial resolutions, spanning from 2003 to 2020. Overall, CAETens exhibited superior performance, achieving the highest averaged Kling-Gupta Efficiency of 0.75 and registering the lowest Root Mean Square Error of 6.85 mm/m compared to the standalone gridded ET products. Comparisons of spatial distributions of fine-scale features, and evaluations using standard metrics revealed that all ET products varied significantly (p < 0.05) by biome types. Remarkably, only CAETens exhibited a reasonable low positive bias and pronounced ability to resolve the complex spatiotemporal patterns associated with the input gridded ET products. The study bridges the gap in continuous ET monitoring for sustainable water resources management in CA.

Large-Scale Hydrological Models and Transboundary River Basins

Large-scale hydrological modeling is an emerging approach in river hydrology, especially in regions with limited available data. This research focuses on evaluating the performance of two well-known large-scale hydrological models, namely E-HYPE and LISFLOOD, for the five transboundary rivers of Greece. For this purpose, discharge time series at the rivers’ outlets from both models are compared with observed datasets wherever possible. The comparison is conducted using well-established statistical measures, namely, coefficient of determination, Percent Bias, Nash–Sutcliffe Efficiency, Root-Mean-Square Error, and Kling–Gupta Efficiency. Subsequently, the hydrological models’ time series are bias corrected through scaling factor, linear regression, delta change, and quantile mapping methods, respectively. The outputs are then re-evaluated against observations using the same statistical measures. The results demonstrate that neither of the large-scale hydrological models consistently outperformed the other, as one model performed better in some of the basins while the other excelled in the remaining cases. The bias-correction process identifies linear regression and quantile mapping as the most suitable methods for the case study basins. Additionally, the research assesses the influence of upstream waters on the rivers’ water budget. The research highlights the significance of large-scale models in transboundary hydrology, presents a methodological approach for their applicability in any river basin on a global scale, and underscores the usefulness of the outputs in cooperative management of international waters.

Integrated modeling of snow-covered areas and hydrological processes in the Larji Sub-Basin, Himachal Pradesh, India, using SRM and SWAT models

Abstract Alpine snow is crucial for the water cycle, as the runoff and environment of arid and semi-arid regions rely entirely on these glaciers. Mountain-fed rivers provides the water for domestic, agricultural irrigation, hydroelectric power, and other uses. Due to climate variability, river catchment flows may shift, causing floods and droughts that will aggravate the economy. This study estimated the SCA using Google Earth Engine (GEE) for the Larji River basin, situated in Himachal Pradesh, from the year 2001 to 2021. The snow cover area (SCA) varies from 0.7–22% in the basin. The present study highlights the effectiveness of cloud computing for assessing trends in snow-cover regions in the basin. Moderate Resolution Imaging Spectroradiometer (MODIS) snow product (MOD10A1 V6 Snow Cover Daily Global 500m product) is utilized for SCA calculation. This study simulates runoff from the Larji river, using Snow-melt Runoff Model (SRM) and Soil Water Assessment Tool (SWAT) for years (2019-2021). Furthermore, the Coefficient of determination (R2) Coefficient of corelation (r), Nash-Sutcliffe Efficiency (NSE) and Kling-Gupta Efficiency (KGE) were used to evaluate the efficacy of models. This study will help the policymakers and stakeholders to carry out the water resource management practices in the study area.

Persistent neural calibration for discharges modelling in drought-stressed catchments

Cross-sector coordination between all water uses and the environmental flows is an essential target for achieving sustainable management in any hydrographic region. In this framework, a novel neural approach was developed and implemented, by the software ANNPI 1.0, to characterise and infer of the discharges regime in a specific basin using only a few attributes as independent variables. The calibration procedure is controlled by the Persistence Index (PI), which is function of a determined estimation lead-time, to facilitate the dynamic character of these simulations. A model validation was carried out in the Lower Guadiana Transboundary Basin, in the Southwest Iberian Peninsula, characterised by moderate and severe drought cyclical events. The best neural approaches included as input variable, between others, the Standardized Precipitation Index at a twelve-month scale SPI(12) that is indicator of hydrological drought, obtaining results statistically very good with determination coefficients higher to 0.77, Nash-Sutcliffe Efficiency coefficients higher to 0.75, Kling-Gupta Efficiency coefficients higher to 0.87 and Persistence Indexes higher to 0.60 in three of the four reservoirs analysed. These accuracy measures showed the ability of the software ANNPI 1.0 to reduce the naïve effect in the forecasting of streamflows time series and could therefore facilitate the development of decision-support systems to make reliable reservoir water balance simulations which will allow to assess future water availability to ensure the main ecosystem services.

Calibration using R-programming and parallel processing at the HUC12 subbasin scale in the Mid-Atlantic region: Development of national SWAT hydrologic calibration

The first phase of a national scale Soil and Water Assessment Tool (SWAT) model calibration effort at the HUC12 (Hydrologic Unit Code 12) watershed scale was demonstrated over the Mid-Atlantic Region (R02), consisting of 3036 HUC12 subbasins. An R-programming based tool was developed for streamflow calibration including parallel processing for SWAT-CUP (SWAT- Calibration and Uncertainty Programs) to streamline the computational burden of calibration. Successful calibration of streamflow for 415 gages (KGE ≥0.5, Kling-Gupta efficiency; PBIAS ≤15%, Percent Bias) out of 553 selected monitoring gages was achieved in this study, yielding calibration parameter values for 2106 HUC12 subbasins. Additionally, 67 more gages were calibrated with relaxed PBIAS criteria of 25%, yielding calibration parameter values for an additional 150 HUC12 subbasins. This first phase of calibration across R02 increases the reliability, uniformity, and replicability of SWAT-related hydrological studies. Moreover, the study presents a comprehensive approach for efficiently optimizing large-scale multi-site calibration.

Reconstruction of missing streamflow series in human-regulated catchments using a data integration LSTM model

Study regionYellow River Basin in China, where streamflow dynamics were significantly impacted by human activities. Study focusWe introduced a deep learning-based method, i.e., Data Integration (DI) with Long Short-Term Memory (LSTM), which leverages Global Flood Awareness System (GloFAS) streamflow data. Multiscale (Catchment, River) attributes were incorporated into the DI LSTM to represent human disturbances on land surface. We employed this method to reconstruct daily streamflow series in 60 human-regulated catchments across the Yellow River Basin, and identified the sensitivity of the DI LSTM model to the multiscale attributes. New hydrological Insights for the RegionOur findings revealed that the DI LSTM model achieved favourable performance in streamflow estimation, with the highest Kling-Gupta efficiency (KGE) reaching up to 0.9, outperforming the Regular LSTM model, which was forced by meteorological variables. Multiscale attributes can enhance the DI model performance, particularly in large catchments with significant human activities. A two-step validation demonstrated the high accuracy of the reconstructed streamflow data across the Yellow River Basin, as the KGEs for streamflow estimation in 40 catchments are over 0.6. In summary, the DI LSTM model shows great potential for reconstructing streamflow in human-regulated catchments in arid regions. The reconstructed daily streamflow data contribute valuable insights for monitoring changing hydrological conditions, especially in regions lacking extensive streamflow monitoring networks.

Assessing the impact of missing data on water quality index estimation: a machine learning approach

Despite the regulations and controls implemented worldwide by governments and institutions to ensure the availability and quality of water resources, many water sources remain susceptible to contamination. This contamination poses significant risks to human health and can lead to substantial economic losses. One of the challenges in this context is the presence of missing or incomplete data, which can arise from various factors such as the methodology used or the expertise of personnel involved in sample collection and analysis. The existence of such data gaps hampers the accurate analysis that can be conducted. To address this issue and estimate a water quality index from the available samples, it is crucial to handle missing information appropriately to avoid biased calculations. This study focuses on the application of machine learning methods for imputing missing data in water samples. Furthermore, it quantifies the performance of different models based on the distribution of the obtained data. By applying 10 distinct methods to a sample of water quality data, the most effective approaches, namely Bayesian Ridge, Gradient Boosting, Ridge, Support Vector Machine, and Theil-Sen regressors, were identified. The selection of these models was based on the evaluation of two estimation error metrics: average percent bias (PBIAS) and Kling-Gupta Efficiency statistic (KGEss). The respective metric values for the aforementioned methods are as follows: ⟨PBIAS⟩0.5=14.665,19.555,14.300,15.380,15.920\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\langle \\hbox {PBIAS}\\rangle _{0.5}=14.665, 19.555, 14.300, 15.380, 15.920$$\\end{document} and ⟨KGEss⟩0.5=0.670,0.585,0.655,0.620,0.595\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\langle \\hbox {KGEss}\\rangle _{0.5}=0.670, 0.585, 0.655, 0.620, 0.595$$\\end{document}. The results obtained from these models have been utilized to establish unbiased relationships among physical, chemical, and biological parameters based on the information retrieved through the applied imputation methods.

Fusion-based approach for hydrometeorological drought modeling: a regional investigation for Iran.

The objective of this study was to model a new drought index called the Fusion-based Hydrological Meteorological Drought Index (FHMDI) to simultaneously monitor hydrological and meteorological drought. Aiming to estimate drought more accurately, local measurements were classified into various clusters using the AGNES clustering algorithm. Four single artificial intelligence (SAI) models-namely, Gaussian Process Regression (GPR), Ensemble, Feedforward Neural Networks (FNN), and Support Vector Regression (SVR)-were developed for each cluster. To promote the results of single of products and models, four fusion-based approaches, namely, Wavelet-Based (WB), Weighted Majority Voting (WMV), Extended Kalman Filter (EKF), and Entropy Weight (EW) methods, were used to estimate FHMDI in different time scales, precipitation, and runoff. The performance of single and combined products and models was assessed through statistical error metrics, such as Kling-Gupta efficiency (KGE), Mean Bias Error (MBE), and Normalized Root Mean Square Error (NRMSE). The performance of the proposed methodology was tested over 24 main river basins in Iran. The validation results of the FHMDI (the compliance of the index with the pre-existing drought index) revealed that it accurately identified drought conditions. The results indicated that individual products performed well in some river basins, while fusion-based models improved dataset accuracy more compared to local measurements. The WMV with the highest accuracy (lowest NRMSE) had a good performance in 60% of the cases compared to all other products and fusion-based models. WMV also showed higher efficiency in 100% of the cases than all other fusion-based and SAI models for simultaneous hydrological and meteorological drought estimation. In light of these findings, we recommend the use of fusion-based approach to improve drought modeling.

Advancing objective functions in hydrological modelling: Integrating knowable moments for improved simulation accuracy

Runoff prediction is a crucial aspect of water resource management and risk mitigation. Despite hydrological modelling plays a vital role in accurately representing catchment behaviour, calibration still poses a significant challenge. This study explores the use of knowable moments (KMoments), a category of high-order statistical moments, as part of the core of objective functions in hydrological model calibration. Traditional objective functions, such as Nash-Sutcliffe Efficiency (NSE) and Kling-Gupta Efficiency (KGE), often make assumptions about data distribution and are sensitive to outliers. KMoments offer a promising alternative by enabling reliable estimation and effective description of high-order statistics from typical hydrological samples and therefore, reducing uncertainty in their estimation and computation of the objective functions in question. Three daily lumped hydrological models (GR4J, VIC, and HYMOD) were employed to test the performance of different calibration strategies using KMoments-based objective functions and compare them with conventional approaches. The hydrological consistency of the simulations was also assessed through 27 hydrological signatures. Our findings highlight the advantages of using KMoments, including improved performance metrics and enhanced hydrological signature reproduction. The findings contribute to advancing hydrological modelling techniques and provide valuable insights for researchers and practitioners seeking to enhance simulation accuracy and reliability.

A comparative analysis of missing data imputation techniques on sedimentation data

Sediment data pertains to various hydrological variables with complex sediment hydrodynamics such as sedimentation rates which are often incompletely presented. Thus, the availability of sedimentation data is of utmost necessity for data accessibility. A comparative analysis on the missing fine sediment data imputation performance was made based on four different techniques, namely the k-Nearest Neighbourhood (k-NN), Support Vector Regression (SVR), Multiple Regression (MR), and Artificial Neural Network (ANN), under the single imputation (SI) and multiple imputation (MI) regimes. Across different missing data proportions (10%-50%), the ANN demonstrated optimal results with consistent performance metrics recorded over both SI and MI regimes. For the highest missing data proportion (50%), the ANN presented the best imputation performance with a reported root mean squared error (RMSE) 0.000882, mean absolute error (MAE) 0.000595, coefficient of determination (R2) 71%, and Kling-Gupta Efficiency (KGE) 72%. The imputation performance ranking is as follows: ANN, SVR, MR, and k-NN.

Novel hybrid intelligence predictive model based on successive variational mode decomposition algorithm for monthly runoff series

A high-accuracy estimation of the runoff has always been an extremely relevant and challenging subject in hydrology science.Therefore, in the current research, a novel hybrid decomposition-integration-optimization based model is developed to enhance the estimation precision of the runoff. The suggested predictive model is a combination of successive variational mode decomposition (SVMD) technique and Multi-Layer Perceptron neural network (MLP) model integrated with particle swarm optimization (PSO) meta-heuristic algorithm (i.e., hybrid SVMD-MLP-PSO model). To test its performance, the mean monthly runoff data recorded from Sep 1986-Aug 2017 in Dez River basin (MRDRm), southwest of Iran, are used. The performance of the recommended model is also matched with other different hybrid and single models including MLP-PSO, SVMD-MLP, and MLP as the benchmark model. In all models, the sequence-to-one regression module of forecasting (i.e., without using meteorological parameters recorded in the study region) is utilized. In the SVMD based hybrid models, the optimal value of compactness of mode (α) for the original MRDRm time series is achieved at 100. Then, the PACF(partial autocorrelation function) diagram related to the lag length from each decomposed intrinsic mode function (IMF) sub-signals sequence generated is operated to select the ideal input variables. Performance evaluation metrics prove that the hybrid SVMD-MLP-PSO model under the best predictor and meta-parameters, outperformed with an R2 of 0.89, modified 2012 version of Kling-Gupta efficiency (KGEʹ) of 0.83, volumetric efficiency (VE) of 0.91, Nash–Sutcliffe efficiency (NSE) of 0.88, and RMSE of 13.91 m3/s. Comparatively, the standalone MLP as the benchmark model results in an R2 of 0.24, VE of 0.33, KGEʹ of 0.2, NSE of 0.29, and RMSE of 153.39 m3/s.

Multimodel ensemble estimation of Landsat-like global terrestrial latent heat flux using a generalized deep CNN-LSTM integration algorithm

Accurate estimates of high-spatial-resolution global terrestrial latent heat flux (LE) from Landsat data are crucial for many basic and applied research. Yet current Landsat-derived LE products were developed using single algorithm with large uncertainties and discrepancies. Here we proposed a convolutional neural network-long short-term memory (CNN-LSTM)-based integrated LE (CNN-LSTM-ILE) framework that integrates five Landsat-derived physical LE algorithms, topography-related variables (elevation, slope and aspect) and eddy covariance (EC) observations to estimate 30-m global terrestrial LE. CNN-LSTM-ILE not only conserves good performance of LE estimation from pure deep learning (DL) algorithm, but partially inherits physical mechanism of the individual physical algorithms for improving the generalization of the integration algorithms for extreme cases. CNN-LSTM is an algorithm that combines two deep learning structures (CNN and LSTM) to effectively utilize the spatial and temporal information contained in the forcing inputs. The data were collected from 190 globally distributed EC observations from 2000 to 2015 and were provided by FLUXNET. The cross-validation results indicated that the CNN-LSTM integration algorithm improved the LE estimates by reducing the root mean square error (RMSE) of 5–8 W/m2 and increasing Kling-Gupta efficiency (KGE) of 0.05–0.16 when compared with the individual LE algorithms and the results of three other machine learning integration algorithms (multiple linear regression, MLR; random forest, RF; and deep neural networks, DNN). The CNN-LSTM integration algorithm had highest KGE (0.81) and R2 (0.66) compared to ground-measured and was applied to generate the Landsat-like regional and global terrestrial LE. An innovation of our strategy is that the CNN-LSTM-ILE model integrates pixel proximity effects and daily LE variations to enhance the accuracy of 16-day LE estimations. This approach can produce a more reliable Landsat-like global terrestrial LE product to improve the representativeness of heterogeneous regions for monitoring hydrological variables.

On the application of symbolic regression in the energy sector: Estimation of combined cycle power plant electrical power output using genetic programming algorithm

This paper focuses on the estimation of electrical power output (Pe) in a combined cycle power plant (CCPP) using ambient temperature (AT), vacuum in the condenser (V), ambient pressure (AP), and relative humidity (RH). The study stresses accurate estimation for better CCPP performance and energy efficiency through responsive control to changing conditions. The novelty lies in applying genetic programming (GP) on a publicly available dataset to generate Symbolic Expressions (SEs) for high-accuracy Pe. To address the challenge of numerous GP hyperparameters, a random hyperparameter values search method (RHVS) is introduced to find optimal combinations, resulting in SEs with higher accuracy. SEs are created with varying input variables, and their performance is evaluated using multiple metrics (coefficient of determination (R2), mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE), Kling–Gupta Efficiency (KGE), and Bland–Altman (B-A) analysis). A key innovation involves combining the best SEs through an Averaging ensemble (AE), leading to a robust estimation accuracy. Notably, the AE YVE−2 achieves the highest (Pe) accuracy, including R2=0.9368, MAE=3.3378, MSE=18.4800, RMSE=4.2985, MAPE=0.7354%, and KGE=0.9479. The investigation highlights AT as the most influential variable, underscoring the importance of choosing inputs aligned with physical processes. This paper’s outlined procedure, combining GP, hyperparameter optimization, and ensemble techniques, offers an efficient method for estimating Pe in CCPP. It promises simplicity and effectiveness in real-world applications. B-A analysis proves valuable for SE selection, enhancing the proposed methodology.

On optimization of calibrations of a distributed hydrological model with spatially distributed information on snow

Abstract. In northern cold-temperate countries, a large portion of annual streamflow is produced by spring snowmelt, which often triggers floods. It is important to have spatial information about snow variables such as snow water equivalent (SWE), which can be incorporated into hydrological models, making them more efficient tools for improved decision-making. The present research implements a unique spatial pattern metric in a multi-objective framework for calibration of hydrological models and attempts to determine whether raw SNODAS (SNOw Data Assimilation System) data can be utilized for hydrological model calibration. The spatial efficiency (SPAEF) metric is explored for spatially calibrating SWE. Different calibration experiments are performed combining Nash–Sutcliffe efficiency (NSE) for streamflow and root-mean-square error (RMSE) and SPAEF for SWE, using the Dynamically Dimensioned Search (DDS) and Pareto Archived Dynamically Dimensioned Search multi-objective optimization (PADDS) algorithms. Results of the study demonstrate that multi-objective calibration outperforms sequential calibration in terms of model performance (SWE and discharge simulations). Traditional model calibration involving only streamflow produced slightly higher NSE values; however, the spatial distribution of SWE could not be adequately maintained. This study indicates that utilizing SPAEF for spatial calibration of snow parameters improved streamflow prediction compared to the conventional practice of using RMSE for calibration. SPAEF is further implied to be a more effective metric than RMSE for both sequential and multi-objective calibration. During validation, the calibration experiment incorporating multi-objective SPAEF exhibits enhanced performance in terms of NSE and Kling–Gupta efficiency (KGE) compared to calibration experiment solely based on NSE. This observation supports the notion that incorporating SPAEF computed on raw SNODAS data within the calibration framework results in a more robust hydrological model. The novelty of this study is the implementation of SPAEF with respect to spatially distributed SWE for calibrating a distributed hydrological model.

Assessing Objective Functions in Streamflow Prediction Model Training Based on the Naïve Method

Reliable streamflow forecasting is a determining factor for water resource planning and flood control. To better understand the strengths and weaknesses of newly proposed methods in streamflow forecasting and facilitate comparisons of different research results, we test a simple, universal, and efficient benchmark method, namely, the naïve method, for short-term streamflow prediction. Using the naïve method, we assess the streamflow forecasting performance of the long short-term memory models trained with different objective functions, including mean squared error (MSE), root mean squared error (RMSE), Nash–Sutcliffe efficiency (NSE), Kling–Gupta efficiency (KGE), and mean absolute error (MAE). The experiments over 273 watersheds show that the naïve method attains good forecasting performance (NSE &gt; 0.5) in 88%, 65%, and 52% of watersheds at lead times of 1 day, 2 days, and 3 days, respectively. Through benchmarking by the naïve method, we find that the LSTM models trained with squared-error-based objective functions, i.e., MSE, RMSE, NSE, and KGE, perform poorly in low flow forecasting. This is because they are more influenced by training samples with high flows than by those with low flows during the model training process. For comprehensive short-term streamflow modeling without special demand orientation, we recommend the application of MAE instead of a squared-error-based metric as the objective function. In addition, it is also feasible to perform logarithmic transformation on the streamflow data. This work underscores the critical importance of appropriately selecting the objective functions for model training/calibration, shedding light on how to effectively evaluate the performance of streamflow forecast models.

Assessing the growing threat of heat stress in the North Africa and Arabian Peninsula region connected to climate change

Climate change exacerbates extreme heat events in the North Africa and Arabian Peninsula (NAAP) region, posing a significant threat to human health and well-being. This study proposes a novel approach for reliable mapping of population exposure in NAAP to different UTCI levels under future climate change scenarios by improved prediction of UTCI using machine learning (ML) methods. Daily mean UTCI was calculated using hourly ERA-HEAT data from 1979 to 2022. Sobol's global sensitivity analysis was used to identify the most influential meteorological variables affecting UTCI. Three ML models were trained to predict monthly UTCI based on these variables. The best-performing model was used to generate historical and future projections of UTCI for four CMIP6 climate models under Paris Agreement goal scenarios. Finally, the projected change in population exposure to mean UTCI was assessed by comparing historical and future projections. The results showed a significant influence of air temperature, wind speed, surface pressure, solar radiation, and humidity on UTCI in NAAP. Estimation of UTCI using these variables showed a higher performance of random forest (RF) (Kling-Gupta Efficiency = 0.84) compared to other models. The projection of UTCI from CMIP6 climatic model simulations using the RF model revealed a decrease in UTCI for the 1.5 °C temperature rise scenario by 0.25–0.75 °C, while an increase by 0.25–0.75 °C for the 2.0 °C temperature rise scenario in most of NAAP. The study revealed that even a modest increase of 0.5 °C in global warming could lead to a 20-day increase in the number of days with moderate thermal stress in Egypt and Sudan and a significant increase in the number of days with strong thermal stress in the Arabian Peninsula. The population exposure to heat stress would increase significantly even under the most ambitious Paris Agreement goals. These findings highlight the urgent need for adaptation and mitigation strategies to protect public health and well-being in the NAAP region from escalating risks.

Suspended sediment load modeling using Hydro-Climate variables and Machine learning

Machine learning (ML) models have the potential to improve the accuracy of Suspended Sediment Load (SSL) predictions. However, their ability to incorporate climate data to enhance SSL predictions remains underexplored. This study investigates the efficacy of ML models in modeling sediment transport using hydro-climate variables as input data. Boosted decision tree models, including Xgboost, LightGBM, and Catboost, were employed to estimate SSL using hydro-climate variables such as streamflow and precipitation, as well as maximum and minimum temperatures. The models were trained, optimized, and evaluated using either climate/streamflow data or a combination of both, encompassing 47 gauges across the contiguous United States (CONUS) between 1950 and 2022. The results demonstrate that streamflow data alone is insufficient for accurate SSL estimation in all rivers. Consequently, integrating climate data and streamflow data proved effective in predicting SSL at the daily scale across diverse climate zones with commendable accuracy. The models achieved an average Nash-Sutcliffe Efficiency (NSE) of > 0.66, Kling-Gupta Efficiency (KGE) of > 0.67, and Willmott Index (WI) of > 0.83 during the unseen (test) phase. These findings highlight the potential of ML models, particularly when combining climate data with streamflow, to enhance SSL predictions and improve our understanding of sediment transport dynamics.

Multi-objective assessment of hydrological model performances using Nash–Sutcliffe and Kling–Gupta efficiencies on a worldwide large sample of watersheds

We introduce a new diagnosis tool that is well suited to analyzing simulation results over large samples of watersheds. It consists of a modification of the classical Taylor diagram to simultaneously visualize several error components (based on bias, standard deviation or squared errors) that are commonly used in efficiency criteria (such as the Nash–Sutcliffe efficiency (NSE) or the Kling–Gupta efficiency (KGE)) to evaluate hydrological model performance. We propose a methodological framework that explicitly links the graphical and numerical evaluation approaches, and show how they can be usefully combined to visually interpret numerical experiments conducted on large datasets. The approach is illustrated using results obtained by testing two rainfall-runoff models on a sample of 2050 watersheds from 8 countries and calibrated with two alternative objective functions (NSE and KGE). The assessment tool clearly highlights well-documented problems related to the use of the NSE for the calibration of rainfall-runoff models, which arise due to interactions between the ratio of simulated to observed standard deviations and the correlation coefficient. We also illustrate the negative impacts of classical mathematical transformations (square root) applied to streamflow when employing NSE and KGE as metrics for model calibration.

A Transfer Learning Approach Based on Radar Rainfall for River Water-Level Prediction

River water-level prediction is crucial for mitigating flood damage caused by torrential rainfall. In this paper, we attempt to predict river water levels using a deep learning model based on radar rainfall data instead of data from upstream hydrological stations. A prediction model incorporating a two-dimensional convolutional neural network (2D-CNN) and long short-term memory (LSTM) is constructed to exploit geographical and temporal features of radar rainfall data, and a transfer learning method using a newly defined flow–distance matrix is presented. The results of our evaluation of the Oyodo River basin in Japan show that the presented transfer learning model using radar rainfall instead of upstream measurements has a good prediction accuracy in the case of torrential rain, with a Nash–Sutcliffe efficiency (NSE) value of 0.86 and a Kling–Gupta efficiency (KGE) of 0.83 for 6-h-ahead forecast for the top-four peak water-level height cases, which is comparable to the conventional model using upstream measurements (NSE = 0.84 and KGE = 0.83). It is also confirmed that the transfer learning model maintains its performance even when the amount of training data for the prediction site is reduced; values of NSE = 0.82 and KGE = 0.82 were achieved when reducing the training torrential-rain-period data from 12 to 3 periods (with 105 periods of data from other rivers for transfer learning). The results demonstrate that radar rainfall data and a few torrential rain measurements at the prediction location potentially enable us to predict river water levels even if hydrological stations have not been installed at the prediction location.