Search

• Hydrological digital twins (HDT) are essential tools to understand the hydrological cycle especially in a changing climate. • We present, test, and validate an open-source framework to implement HDT in the large anthropized alpine Adige catchment. • The HDT is calibrated/validated on a variety of hydrological processes and data sources providing satisfactory performances. Understanding and simulating the hydrological cycle, especially in a context of climate change, is crucial for quantitative water risk assessment and basin management. The hydrological cycle is complex as it is a combination of non-linear natural processes and anthropogenic influences that alter landforms and water flows. Human-induced changes of relevance, including changes in land uses, construction of dams and artificial reservoirs, and diversion of the river course, lead to changes in water flows throughout the basin. These should be explicitly accounted for a realistic representation of the anthropogenically altered hydrological cycle. Such a realistic representation of the hydrological cycle is a necessary input for the water risk assessment in a particular region. In this paper, we present a hydrological digital twin (HDT) model of a large anthropized alpine basin: the Adige basin located in the northeast of Italy. Most catchments model often overlook land-uses changes over time and forget to model reservoir operation and their influence over time on water flow. Yet, for example, the Adige basin has>30 reservoirs affecting the water flow. We therefore use the GEOframe modeling framework to demonstrate the ability to create a hydrological twin model accounting for these anthropogenic changes. Specifically, we model each component of the water cycle over 39 years (1980–2018) at daily timescale through calibration of the Adige HDT with a multi-site approach using discharge data of 33 stations, based on a high-resolution (1 km) temperature and precipitation dataset and a calculated crop potential evapotranspiration (PETc) dataset, which accounts for human-induced change of the land cover over time. The modeling system also includes the simulation of artificial reservoirs and dams by the dynamically zoned target release (DZTR) reservoir model. The Adige HDT is assessed/validated/compared through a variety of hydrological processes (i.e., river and reservoir discharges, PETc and actual evapotranspiration, snow, and soil moisture) and data sources (i.e., observations and remote sensing data). Overall, the HDT reproduces well the measured discharge in space and time with a Kling Gupta Efficiency (KGE) above 0.7 (0.8) for 30 (23) of the 33 gauge-stations. For 7 artificial reservoirs with available data, the reservoir turbinated discharges are successfully reproduced with an average KGE of 0.92. A comparison between modeled and MODIS remote sensing snow data showed an average error of < 10% across the entire basin; the model also presented a good spatio-temporal agreement both with GLEAMS potential (and actual evapotranspiration) with an average KGE of 0.63 (0.60) and a high-level of correlation (0.5 on average) with the ASCAT satellite retrieved soil moisture. The findings of this paper demonstrate the potential of the open-source, component-based, GEOframe system to build a HDT, to provide a reliable and long term (39 years) estimation of all the water cycle components in a complex anthropized river basin at high spatial resolution. Spatially detailed HDT models results of this type can be used to inform basin-wise adaptation policy decisions and better water management practices in a time of changing climate. [ABSTRACT FROM AUTHOR]

• The Stacking-Hybrid model is proposed by combining process-driven and data-driven models, considering multiple sources of uncertainty. • A series of comparison models are applied to verify the effectiveness of the proposed model in terms of deterministic, interval and probabilistic forecasts. • The Stacking-hybrid model has promising applications in real-time flood forecasting at an hourly scale. Accurate and reliable incoming flood forecasting is an important prerequisite for flood warning, flood risk analysis and reservoir flood control operation. This paper proposes a hybrid model for real-time flood forecasting that couples process-driven hydrological models (HMs) with data-driven models (DDMs). The generic hybrid model framework adds DDMs as the post-processing procedure for residual correction to the original results of HMs, and considers multiple uncertainties in input data, parameter and model structure simultaneously. The performance of the hybrid model is evaluated comprehensively in terms of deterministic forecast accuracy, interval forecast reliability, and the reliability and sharpness of probabilistic forecast. Taking the multireservoir system at the east Pi River as a study case, the results indicate that: (1) Compared to the benchmark model (ensemble XAJ model), the hybrid model with additional residual analysis show a significant improvement. The average continuous ranked probability score (CRPS) metric values calculated by the Stacking-Hybrid model improved by 71.5 %, 67.0 % and 38.1 % in the three data samples. Furthermore, the adaptability of the Stacking-Hybrid model for residual correction during short-duration intense rainfall events has been validated, with the relative error of the peak discharge improved to within ± 10 %. (2) The Stacking-Hybrid model, which also takes into account structure uncertainty, is able to better exploit the combined advantages and improve the stability of the model performance compared to those that only apply a single DDM. (3) When the number of iterations within the BOA reaches 300, the parameter optimization process is capable to search for the hyperparameters that bring out the best performance of the DDMs. (4) When the ensemble size reaches 200, the uncertainty of HM parameters can be fully defined, and the consumed computational resources can be controlled within an acceptable bound while ensuring stable model performance. Overall, the hybrid model that takes into account multiple sources of uncertainty generates both interval and probabilistic forecast in addition to deterministic forecast, which can provide richer risk information for subsequent flood warning and reservoir operation, making the flood prevention decisions more reliable and scientific. [ABSTRACT FROM AUTHOR]

Graphical abstract Highlights • The paper presents a comprehensive review and application of ELM model. • The review comprises all the implementations of the hydrological processes. • The reviewed studies evaluated and assessed with wide-ranging of aspects. • The enhanced ELM model performed remarkably on the river flow forecasting. • Several milestone propositions are anticipated for future research directions. Abstract Despite the massive diversity in the modeling requirements for practical hydrological applications, there remains a need to develop more reliable and intelligent expert systems used for real-time prediction purposes. The challenge in meeting the standards of an expert system is primarily due to the influence and behavior of hydrological processes that is driven by natural fluctuations over the physical scale, and the resulting variance in the underlying model input datasets. River flow forecasting is an imperative task for water resources operation and management, water demand assessments, irrigation and agriculture, early flood warning and hydropower generations. This paper aims to investigate the viability of the enhanced version of extreme learning machine (EELM) model in river flow forecasting applied in a tropical environment. Herein, we apply the complete orthogonal decomposition (COD) learning tool to tune the output-hidden layer of the ELM model's internal neuronal system, instead of the conventional multi-resolution tool (e.g., singular value decomposition). To demonstrate the application of EELM model, the Kelantan River, located in the Malaysian peninsular, selected as a case study. For a comparison of the EELM model, and further model evaluation, two distinct data-intelligent models are developed (i.e., the classical ELM and the support vector regression, SVR model). An exhaustive list of diagnostic indicators are used to evaluate the EELM model in respect to the benchmark algorithms, namely, SVR and ELM. The model performance indicators exhibit superior results for the EELM model relative to ELM and SVR models. In addition, the EELM model is presented as a more accurate, alternative predictive tool for modelling the tropical river flow patterns and its underlying characteristic perturbations in the physical space. Several statistical metrics defined as the coefficient of determination (r), Nash-Sutcliffe efficiency (E ns), Willmott's Index (WI), root-mean-square error (RMSE) and mean absolute error (MAE) are computed to assess the model's effectiveness. In quantitative terms, superiority of EELM over ELM and SVR models was exhibited by E ns = 0.7995, 0.7434 and 0.665, r = 0.894, 0.869 and 0.818 and WI = 0.9380, 0.9180 and 0.8921, respectively. Whereas, EELM model attained lower (RMSE and MAE) values by approximately (11.61–22.53%) and (8.26–8.72%) relative to ELM and SVR models, respectively. The obtained results reveal that the EELM model is a robust expert model and can be embraced practically in real-life water resources management and river sustainability decisions. As a complementary component of this paper, we also review state-of-art research works where scholars have embraced extensive implementation of the ELM model in water resource engineering problems. A comprehensive evaluation is carried out to recognize the current limitations, and also to propose potential opportunities of applying improved variants of the ELM model presented as a future research direction. [ABSTRACT FROM AUTHOR]

Highlights • Consideration of both regional development patterns and site-scale regulations. • Future urbanization projections generated through ANN machine learning model. • Floodplain increases up to 12.5% observed under projected urbanization scenarios. Abstract In the United States, fluvial flood risk is managed by the National Flood Insurance Program (NFIP), which delineates areas that have a 1% chance of flooding each year (i.e. the 100-year floodplain). Development within and adjacent to riverine floodplains exacerbates flood losses by increasing peak discharge and runoff volume, shortening the time to peak, and altering the extent of the floodplain. This paper improves understanding of how future development can alter the 100-year floodplain in a watershed in Houston, TX through land use projection, hydrologic and hydraulic modeling, and implements a novel method for considering the impact of both regional trends in development and site-scale development policies on runoff response. The methodology presented in this paper integrates future development scenarios from a machine learning land use projection model with distributed hydrologic modeling and coupled 1D/2D unsteady hydraulic modeling to produce future floodplain estimates. Current site-scale detention requirements are represented within the hydrologic model to evaluate the regional effectiveness of these policies under future development conditions in 2050. Results indicate that the 100-year floodplain can expand by up to 12.5% across the watershed as a result of projected development in 2050 using current stormwater mitigation policies, and the number of parcels within the floodplain can increase by up to 18.8%. This study demonstrates how incremental land use changes can significantly alter the reality of flood risk in urbanizing watersheds, and how existing land use policies may be insufficient to mitigate impacts from future development. [ABSTRACT FROM AUTHOR]

Multiple barriers constrain the widespread application of participatory methods in water management, including the more technical focus of most water agencies, additional cost and time requirements for stakeholder involvement, as well as institutional structures that impede collaborative management. This paper presents a stepwise methodological framework that addresses the challenges of context-sensitive initiation, design and institutionalization of participatory modeling processes. The methodological framework consists of five successive stages: (1) problem framing and stakeholder analysis, (2) process design, (3) individual modeling, (4) group model building, and (5) institutionalized participatory modeling. The Management and Transition Framework is used for problem diagnosis (Stage One), context-sensitive process design (Stage Two) and analysis of requirements for the institutionalization of participatory water management (Stage Five). Conceptual modeling is used to initiate participatory modeling processes (Stage Three) and ensure a high compatibility with quantitative modeling approaches (Stage Four). This paper describes the proposed participatory model building (PMB) framework and provides a case study of its application in Québec, Canada. The results of the Québec study demonstrate the applicability of the PMB framework for initiating and designing participatory model building processes and analyzing barriers towards institutionalization. [ABSTRACT FROM AUTHOR]

• Integrating an agent-based model into a large-scale distributed hydrologic model. • Evaluating the co-evolution of the CNHS at multiple spatial scales. • Demonstrating the trade-off among FEW sectors under climate change. Managing water resources to meet increasing energy and food demands while maintaining environmental sustainability under climate change is a major challenge, especially when this nexus occurred in a coupled natural–human system (CNHS), where heterogeneous human activities affect the natural hydrologic cycle and vice versa. The relevant research has been limited by the lack of models that can effectively integrate human dynamics and hydrologic conditions with spatial details to examine co-evolutionary systems. To address this challenge, this paper develops a modeling framework that integrates an agent-based model (ABM; human behavior model) into a large-scale, process-based distributed hydrologic model to simulate human decisions endogenously in the hydrologic cycle. We then apply the Decision Scaling approach, an ex-post scenario analysis method, with our integrated model to study the bidirectional feedback of the CNHS under future changing climate conditions. With the Columbia River Basin (CRB) selected as the case study area, the calibration results show that the integrated model can simultaneously capture the historical irrigated water consumption and streamflow dynamics. Modeling results show that the trade-off between irrigated water consumption, hydropower generation, and streamflow will become more pronounced under hotter and wetter climate conditions at both the entire basin and regional (states and provinces) levels. Special attention should be given to "temperature thresholds" of different regions when the trade-off pattern started. The trade-off results can potentially inform the Columbia River Treaty renegotiation and provide insights for long-term water management policies. [ABSTRACT FROM AUTHOR]

[Display omitted] • Hydrological model predictions are crucial for risk-based water management decisions. • Errors in hydrological predictions are typically neglected for risk-based decision-making. • New framework shows neglecting errors over-estimates yield, under-estimates risks. • Outcomes underscore the need to account for hydrological errors in decision-making. Risk-based decision making for water resource systems often relies on streamflow predictions from hydrological models. These predictions are integral for estimating the frequency of high consequence extreme events, such as floods and droughts. However, streamflow predictions are known to have errors due to various factors such as incomplete hydrological understanding, parameter misspecification, and uncertain data. Despite these errors being well known, they are frequently neglected when undertaking risk-based decision-making. This paper demonstrates that neglecting hydrological errors can impact on drought risk estimation for high stakes decisions with potentially severe consequences for water resource system performance. A generic framework is introduced to evaluate the impact of hydrological errors for a wide range of water resource system properties. This framework is applied in two Australian case study catchments, where we use a stochastic rainfall model, the GR4J hydrological model, a residual error model, and a simplified reservoir storage model to estimate water resource performance metrics (risk and yield). The results underscore the impact of neglecting hydrological errors on decision-making. In one case study catchment, the yield was over-estimated by ∼15%–55%, resulting in the (actual) risk of running out of water being ∼2–30 times larger than reservoir design. The magnitude of these errors in water resource performance metrics is striking, especially considering that the streamflow predictions appear reasonable based on typical performance metrics (e.g., NSE of ∼0.7). The errors in performance metrics stem from the complex propagation of hydrological errors through the water resource system modelling chain. By accounting for critically important hydrological errors we can mitigate highly erroneous risk estimates and improve decision-making related to water resource management. [ABSTRACT FROM AUTHOR]

[Display omitted] • Five discharge datasets have median correlation coefficients > 0.6 with the observations. • Overestimation of monthly mean flow is more common than underestimation. • GRFR showed the best estimate at around 40% of the locations. • None of the models provides good estimates for rivers and streams in arid regions. High-resolution global flow discharge datasets are critical to understand the terrestrial water cycle and river ecosystem services. Recently, a number of global hydrological models have been developed to capture global river flow dynamics at locally relevant scales. However, the relative accuracy of these simulated flow discharges is largely unknown. In this paper, using the largest observed streamflow datasets from > 26,000 gauging stations, we evaluated the performance of monthly and sub-monthly flow variation (1980–2014) generated from five major global hydrological models, Catchment-based Macro-scale Floodplain model (CaMa-Flood), Global Flood Awareness System (GloFAS-reanalysis), Global Reach-level Flood Reanalysis (GRFR), PCRaster GLOBal Water Balance model (PCR-GLOBWB), and Water Global Assessment and Prognosis (WaterGAP). Significant spatial heterogeneity in model performance was identified. We found that estimates by all five discharge datasets have a relatively high correlation with the observations (medians > 0.6). But none of them showed adequate correlations between simulated and observed flows for systems in Southwest or Central U.S.A., South Africa, or South Australia. Across all models, overestimation of monthly mean flow was more common than underestimation and high flows at a sub-monthly scale were more accurately characterized than low flows. Performance of all models was poor in Africa and Oceania. No single model performed universally best at all gauging stations, but GRFR showed the best estimate at around 40 % of the locations. None of the models provided satisfying estimates for rivers and streams in arid regions. We also found a significant negative effect of flow regulation on the predictive accuracy of all models. Our study highlights the importance of thorough evaluation of global discharge datasets before their local and regional application and provides guidance for future enhancement of global hydrological models. [ABSTRACT FROM AUTHOR]

• Due to skilful forcing, hindcasts of runoff and discharge have much skill in north and in southeast South America. • Due to initial conditions, hydrological hindcasts have much skill in the drier regions of the continent. • Spatial patterns of theoretical (compared to reference simulation) and actual skill (compared to observed discharge) are similar. The first aim of the present paper is the determination of the magnitude, annual variation and spatial distribution of skill in seasonal hindcasts of runoff and discharge in the entire continent of South America. We evaluated 35 years of hindcasts generated with the Variable Infiltration Capacity (VIC) hydrological model forced with SEAS5 hindcasts. Initial conditions of terrestrial water and so-called pseudo-observations were computed with a reference (i.e. historic) simulation. Skill was determined with monthly temporal resolution for the entire annual cycle and mostly using the pseudo-observations for verification. The second aim of the paper is the explanation of skill in terms of its sources, namely meteorological forcing and the initial conditions. Therefore, two sets of restricted hindcasts, which isolate the sources of skill, were analysed. The SEAS5 precipitation hindcasts exhibit significant skill even at the longest lead times (7 months). Beyond the first lead month we found significant skill in 13–43 % of the grid cells, depending on target and lead month. Levels of skill are higher in the full hindcasts (significant skill in 31–89 % of the grid cells). At the continental scale more of the skill is caused by the forcing than by the initial conditions. The runoff hindcasts are skilful in large parts of the continent. In a 1000 km wide band along the north coast of the continent and in southeast South America, most of the skill is due to the forcing. A typical feature of these regions is an increase of skill with lead time during specific parts of the year, which is against the common tendency. In Argentina and north Chile most of the skill in the runoff hindcasts can be attributed to the initial conditions of soil moisture. Verification with real observations of discharge broadly confirmed the skill pattern obtained with pseudo-observations. [ABSTRACT FROM AUTHOR]

• A spatial random forest-based downscaling and merging method (SRF-DM) is proposed. • SRF-DM performs better than the original satellite-based precipitation products. • SRF-DM is more accurate than the classical machine learning-based methods. • Spatial autocorrelation is the most important variable. • SRF-DM has a good performance even when the gauge density is low. Publicly available satellite precipitation products (SPPs) are essential inputs for hydrological models. However, the existing SPPs suffer from coarse resolutions and large uncertainties, which greatly hinder their widespread applications. Thus, to fully utilize the unique characteristics of each individual SPP, a spatial random forest (SRF)-based downscaling and merging (SRF-DM) method is proposed to fuse multiple SPPs and gauge observations in this paper. The proposed method incorporates spatial autocorrelation information between precipitation observations through a covariate. This covariate is estimated using ordinary kriging interpolation on spatial neighbors that have a high correlation with the target location. The proposed method was used to generate 1-km daily precipitation data using the datasets from five satellite precipitation products (IMERG, GSMaP, CHIRPS, MORPH, and PERSIANN) and gauge observations in the period from January 1, 2015 to December 31, 2019 over Sichuan province, China. The performance of SRF-DM was compared with the original SPPs, XGBoost-based downscaling and merging (XGB-DM), two double-layer machine learning (ML) methods, namely RF-RF and RF-artificial neural network (RF-ANN), and three single ML methods, namely RF, ANN, and RF-based Merging Procedure (RF-MEP). The experimental results demonstrate the following findings: (i) SRF-DM is more accurate than the original SPPs across different temporal scales (i.e., daily, monthly, and seasonal) in terms of statistical and categorical accuracy measures, (ii) SRF-DM outperforms the six ML-based methods in estimating high precipitation and capturing spatial precipitation patterns, (iii) the spatial autocorrelation between neighboring precipitation data is the most significant factor in both downscaling and merging, and (iv) the performance of SRF-DM is influenced by gauge density, but it still yields good results even with a limited number of gauges. On the whole, SRF-DM is considered an effective tool for generating daily precipitation products with high accuracy and resolution, which is primarily attributed to its satisfactory performance in regions with diverse precipitation patterns and topography. [ABSTRACT FROM AUTHOR]

• U-Net and its improved models were used for radar echo extrapolation and precipitation nowcasting. • The nowcasted precipitation was used as input to the HEC-HMS hydrological model for flood forecasting. • Nested-Unet and U-Net 3Plus were helpful to improve the accuracy of precipitation nowcasting and flood forecasting. Accurate and timely short-term precipitation nowcasting is important for achieving reliable flood forecasting. The data-driven approaches have performed well in radar echo extrapolation to nowcast precipitation. In this paper, U-Net and its improved models, including SmaAt-Unet, Nested-Unet, and U-Net 3Plus were applied to perform radar echo extrapolation and precipitation nowcasting for 0.5 h, 1 h, and 2 h lead times of typical rainfall processes. The nowcasted precipitation was used as input to the HEC-HMS hydrological model for flood forecasting to compare the effect of different structural improvements to U-Net on the accuracy of flood forecasting. The results demonstrated that Nested-Unet and U-Net 3Plus aided in enhancing the accuracy of the extrapolation of moderate intensity radar echoes. With fewer discrepancies and better correlation with measured rainfall, the U-Net and U-Net 3Plus precipitation nowcasting results also produced improved flood forecasting outcome. The precipitation nowcasting and flood forecasting for SmaAt-Unet were slightly worse than other models; the relative errors of both flood peak and depth for Nested-Unet at 0.5 h lead time were less than 20 %, showing a good performance. Moreover, in a separate control experiment, the accuracy of the echo extrapolation was significantly decreased when convolutional block attention module (CBAM) was added to each model. However, all models have better extrapolation accuracy than the basic ConvLSTM. In general, Nested-Unet and U-Net 3Plus were helpful to improve the accuracy of precipitation nowcasting and flood forecasting, and the forecasted flood with 0.5 h and 1 h lead times could match the actual flood processes, but the peak discharge from nowcasting with 2 h lead time were severely underestimated, while the peak occurrence time could be forecasted correctly. These conclusions and attempts can provide effective guidelines for regional precipitation nowcasting and flood forecasting. [ABSTRACT FROM AUTHOR]

[Display omitted] • Ten multi-model combination methods (MMCMs) are created from 29 models in 50 basins. • MMCMs improve some performance indicators, especially the Granger-Ramanathan method. • MMCMs do not improve performance in reproducing distributions of hydrological signatures. • Application with series of targeted signatures does not improve MMCM performance. • MMCM performance is improved by selecting more robust candidate models for ensembles. Multi-model combination (averaging) methods (MMCMs) are used to improve the accuracy of hydrological (precipitation-runoff) outputs in simulation or forecasting/prediction modes. In this paper, we examined if the application of MMCMs can improve model performance in reproducing distributions of hydrological signatures, such as annual maxima or minima of varying durations. To this end, ten MMCMs were applied to 29 bucket-type models to simulate runoff in 50 high-latitude catchments. The MMCMs were evaluated by comparing the resulting simulated flows to the reference (i.e., best-performing) individual model, considering various commonly used performance indicators, as well as model performance in reproducing the distributions of signatures. Additionally, we analysed whether (1) the selection of the candidate models, or (2) targeting specific signatures, such as annual maxima or minima, can improve performance of the model combinations. The results suggest that the application of MMCMs can improve accuracy of runoff simulations in terms of traditional performance indicators, but fails to improve performance in reproducing the distributions of signatures. Neither excluding poor-performing models nor applying the MMCMs with the targeted signatures, improves this aspect of model performance. These findings clearly reveal the need for further research aiming at enhancing model performance in reproducing the distributions of hydrological signatures, which is essential for climate-change impact studies. [ABSTRACT FROM AUTHOR]

Models like the Rangeland Hydrology and Erosion Model (RHEM) are useful for estimating soil erosion, however, they rely on input parameters that are sometimes difficult or expensive to measure. Specifically, RHEM requires information about foliar and ground cover fractions that generally must be measured in situ, which makes it difficult to use models like RHEM to produce erosion or soil risk maps for areas exceeding the size of a hillslope such as a large watershed. We previously developed a deep learning emulator of RHEM that has low computational expense and can, in principle, be run over large areas (e.g., over the continental US). In this paper, we develop a deep learning model to estimate the RHEM ground cover inputs from remote sensing time series, reducing the need for extensive field surveys to produce erosion maps. We achieve a prediction accuracy on hillslope runoff of R 2 ≈ 0. 9 , and on soil loss and sediment yield of R 2 ≈ 0. 4 at 66,643 field locations within the US. We demonstrate how this approach can be used for mapping by developing runoff, soil loss, and sediment yield maps over a 1356 km2 region of interest in Nebraska. • We developed a deep learning model estimating the ground cover inputs of the RHEM. • Litter and Shrubs are the most and Bio Crusts is the least accurate estimations. • We achieved an R 2 ≈ 0. 9 for runoff and ≈ 0. 4 for soil loss by the estimated covers. [ABSTRACT FROM AUTHOR]

• Guidelines for the use of snow data combined with runoff to calibrate models. • Improvement of the snow simulations without impairing runoff performance. • The improvement in the model consistency is not accompanied by lower equifinality. • Using snow data does not improve model robustness to climate variability. This paper compares different calibration strategies for using snow data combined with streamflow records to constrain model optimisation in mountain catchments. In particular, it assesses to what extent the use of snow observations makes it possible to improve the consistency, identifiability and robustness of the calibrated parameters. To answer this question, a semi-distributed snow and ice model was used on top of a rainfall-runoff model using the SAFRAN meteorological reanalysis as input dataset on several catchments in the Alps and Pyrenees. Model calibration and control were based on streamflow observations, remotely-sensed snow-covered areas and in-situ snow water equivalent (SWE) measurements. The snow and rainfall-runoff parameters were calibrated sequentially and simultaneously against objective functions integrating different combinations of runoff, snow cover and SWE criteria. Statistical assessment of model performances in independent evaluation periods showed that sequential calibration of the snow parameters gives too much weight to snow compared to runoff. Instead, incorporating snow data in the simultaneous calibration of the parameters improves snow simulations without impairing runoff performance. This can be achieved by limiting the weight of the snow criteria to 25% in the objective function. Although local SWE measurements were found to be more useful than satellite observations for identifying more consistent parameters, it is advisable to include both in the calibration process through a compromise objective function. However, the improved model consistency was not accompanied by a significant reduction in equifinality and optimisation times. On the other hand, improving internal consistency made it possible to reduce the interdependence between the parameters of the snow model and those of the rainfall-runoff model. This also made it possible to identify the least sensitive snow parameters in order to fix them at general values without impairing model performance while reducing equifinality with a more parsimonious model. Finally, there was no evidence that using snow observations in the calibration process improves model robustness with respect to climate variability. [ABSTRACT FROM AUTHOR]

• Developing a gated recurrent unit neural network for streamflow prediction. • Proposing a method for testing and benchmarking deep learning models. • Investigating deep learning model performances with different inputs. • Studying deep learning model performances with various sizes of training data. Machine learning (ML) algorithms have produced remarkable advances in streamflow prediction, exceeding the performance of calibrated conceptual and physics-based hydrological models that have been developed over many decades. ML algorithms seem to overcome the issue of errors known to be present in rainfall and streamflow estimates that have hindered the performance of hydrological models for decades. In this paper, we propose a methodology for testing and benchmarking ML algorithms using artificial data generated by physically-based hydrological models. Our approach makes it possible to design controlled numerical experiments that can improve our understanding of this new generation of black-box models. We conducted a diagnostics study to demonstrate our methodology in which we attempted to determine if ML algorithms can identify a function relating streamflow and rainfall. This exercise combined the implementation of the distributed hillslope-link hydrological model (HLM) on a 4,385 km2 basin driven by precipitation fields created using the stochastic storm transposition (SST) framework, and an advanced deep learning algorithm based on gated recurrent unit (GRU)-Attention neural networks. The data generated allowed us to create prediction scenarios that are equivalent to the hindcast and real-time forecast problems. We proposed a set of scale-independent performance metrics to evaluate the results of our experiment and found that the GRU can correctly identify a predictive function for all analyzed locations in the river network. We concluded that under the circumstances tested in this study, deep learning can identify the transformation function when trained in Hindcast Mode, making it a powerful tool to determine the streamflow response of a basin to predetermined rainfall scenarios. However, it fails to significantly outperform the predictions of temporal persistence when tested in Forecast Mode. [ABSTRACT FROM AUTHOR]

Summary Paired catchment approach probably provides the most robust method to detect the effects of land-use change on catchments’ flow characteristics. This approach is limited by the availability of two neighbor catchments with and without land-use change under similar climate conditions. This paper uses a hydrological model to detect the hydrological change caused by urbanization. This study describes (1) use a statistical method to evaluate change detection relative to variation of land use change, (2) simulation of non-urban condition for the urban catchment with an alternative approach, to this aim stream flow series of urban catchments have been reconstructed from the period that urbanization had not taken place yet, and (3) the model validation with observed data. This paper intends to compare the flow changes detected by two different approaches: a regional statistical approach (the paired-catchment approach) and a conceptual modeling approach (the residual approach) on the particular case of urbanized catchments. To investigate the sensitivity of the results to the settings of both approaches, the comparison is made on a relatively large number of 43 catchments located in the United States, with relatively large gradients in terms of geomorphology and hydroclimatic characteristics. Results show that the two approaches are generally in relative good agreement in terms of detection and quantification of changes for the three flow characteristics analyzed (mean annual flow, high and low flow characteristics). Besides, it is found that the impact of urbanization on the catchment’s hydrologic response is difficult to generalize: the proportion of nonsignificant trends, significantly increasing decreasing trends are on the same order of magnitude, even if an increase in urban areas generally has a greater impact on mean flows and high flows than on low flows. [ABSTRACT FROM AUTHOR]

• Defining the risk of multi-reservoir flood control operation considering model structure and hydrological uncertainties. • Establishing a risk analysis model based on the Bayesian network theory. • Quantitatively analyzing the risk under double uncertainties and its spatial characteristics. In the real-time flood control operation of a multi-reservoir system, the multi-reservoir real-time flood control hybrid operation (MRFCHO) model with reduced structure is an effective way to improve the efficiency, but the reduction of the model structure may cause greater flood risk. This paper focuses on this scientific problem and evaluates the reliability of the MRFCHO model in real-time flood control operation through risk analysis. Firstly, the flood risk is defined. Then, the Bayesian network risk analysis model is established through four parts: Bayesian network structure construction, training samples generation, parameter learning, and inference. This paper takes the flood control system in upstream of the Lutaizi point in the Huaihe River basin as a case study. The flood risk of the MRFCHO model caused by the model structure reduction is quantitatively described, the spatial distribution and propagation characteristics of flood risk are evaluated, and the flood risk caused by the double uncertainties of the model structure and flood forecasting is analyzed. Besides, the factors affecting the risk of the public flood control point are diagnosed. The results show that (1) the model structure reduction does not significantly increase the flood risk at the public flood control point, and there is no obvious propagation law of flood risk in space; (2) compared with only considering the model structure reduction, considering the double uncertainties of flood forecasting and model structure reduction has a 65% probability of increasing the risk; (3) Runoff from the Runheji point in the upper reaches of main stream and the lateral inflow of Lutaizi point are the main reasons for the risk of Lutaizi point. [ABSTRACT FROM AUTHOR]

• Defining the risk of multi-reservoir flood control operation considering model structure and hydrological uncertainties. • Establishing a risk analysis model based on the Bayesian network theory. • Quantitatively analyzing the risk under double uncertainties and its spatial characteristics. In the real-time flood control operation of a multi-reservoir system, the multi-reservoir real-time flood control hybrid operation (MRFCHO) model with reduced structure is an effective way to improve the efficiency, but the reduction of the model structure may cause greater flood risk. This paper focuses on this scientific problem and evaluates the reliability of the MRFCHO model in real-time flood control operation through risk analysis. Firstly, the flood risk is defined. Then, the Bayesian network risk analysis model is established through four parts: Bayesian network structure construction, training samples generation, parameter learning, and inference. This paper takes the flood control system in upstream of the Lutaizi point in the Huaihe River basin as a case study. The flood risk of the MRFCHO model caused by the model structure reduction is quantitatively described, the spatial distribution and propagation characteristics of flood risk are evaluated, and the flood risk caused by the double uncertainties of the model structure and flood forecasting is analyzed. Besides, the factors affecting the risk of the public flood control point are diagnosed. The results show that (1) the model structure reduction does not significantly increase the flood risk at the public flood control point, and there is no obvious propagation law of flood risk in space; (2) compared with only considering the model structure reduction, considering the double uncertainties of flood forecasting and model structure reduction has a 65% probability of increasing the risk; (3) Runoff from the Runheji point in the upper reaches of main stream and the lateral inflow of Lutaizi point are the main reasons for the risk of Lutaizi point. [ABSTRACT FROM AUTHOR]

• A novel framework for assessing Bayesian Causality performance. • An advance in Bayesian causal modelling process for predictive assessment. • An optimal factors combination for Bayesian Causality Learning-Training Processes. • External and internal driving factors for Bayesian Causality performance. • Applicability to different time scales of normalizable data for analysis-forecast. A strong innovative tendency is nowadays emerging that largely comprises new hydrological modelling approaches, based on Causal Reasoning through Probabilistic Graphical Modelling (PGM), because its ability to support probabilistic reasoning from data with uncertainty. These novel modelling frameworks are quite diverse and disperse not only in terms of techniques but also regarding its aims. It seems necessary to find a general and robust methodology for assessing its performance. This paper aims to provide a novel general methodology for assessing the performance of PGM based on Bayesian Causality for modelling and analysing the riverś runoff behaviour. For it, a structured four-step approach is developed and showed throughout the paper. The proposed methodology begins with the identification of the two main factors that condition the Bayesian Causal (BC) Modelling: the number of synthetic series (data amount) and the number of intervals for probability distributions for training and learning processes. The developed analysis comprises the definition of three levels for the first factor and seven levels for the second one, as well as the design of an innovative stability framework that assesses the level of BC Modelling performance. Furthermore, it has been necessary to create-define two novel indexes, named "Similarity and Stability Indexes" from 21 scenarios arising from the combination of the levels of the both factors. The optimal combination of factors is identified through a bi-objectives recursive approach based on previous indexes. Main results drawn successfully show a high relationship between the level of modelling performance, measured in terms of stability, and the river runoff temporal behaviour, measured in terms of temporal dependence. This research may help water managers and engineers to develop more rigorous and robust hydrological causal modelling implementations. [ABSTRACT FROM AUTHOR]

Summary Growing water extractions combined with emerging demands for environment protection increase competition for scarce water resources worldwide, especially in arid and semiarid regions. In those regions, climate change is projected to exacerbate water scarcity and increase the recurrence and intensity of droughts. These circumstances call for methodologies that can support the design of sustainable water management. This paper presents a hydro-economic model that links a reduced form hydrological component, with economic and environmental components. The model is applied to an arid and semiarid basin in Southeastern Spain to analyze the effects of droughts and to assess alternative adaptation policies. Results indicate that drought events have large impacts on social welfare, with the main adjustments sustained by irrigation and the environment. The water market policy seems to be a suitable option to overcome the negative economic effects of droughts, although the environmental effects may weaken its advantages for society. The environmental water market policy, where water is acquired for the environment, is an appealing policy to reap the private benefits of markets while protecting ecosystems. The current water management approach in Spain, based on stakeholders’ cooperation, achieves almost the same economic outcomes and better environmental outcomes compared to a pure water market. These findings call for a reconsideration of the current management in arid and semiarid basins around the world. The paper illustrates the potential of hydro-economic modeling for integrating the multiple dimensions of water resources, becoming a valuable tool in the advancement of sustainable water management policies. [ABSTRACT FROM AUTHOR]

Summary Retrospective forecasts of precipitation, temperature, and streamflow were generated with the Hydrologic Ensemble Forecast Service (HEFS) of the U.S. National Weather Service (NWS) for a 20-year period between 1979 and 1999. The hindcasts were produced for two basins in each of four River Forecast Centers (RFCs), namely the Arkansas–Red Basin RFC, the Colorado Basin RFC, the California–Nevada RFC and the Middle Atlantic RFC. In a companion paper, temperature and precipitation hindcasts were produced with the Meteorological Ensemble Forecast Processor (MEFP) and verified against observed temperature and precipitation, respectively. Inputs to the MEFP comprised raw precipitation and temperature forecasts from the frozen (circa 1997) version of the NWS Global Forecast System (MEFP-GFS) and a conditional or “resampled” climatology (MEFP-CLIM). For this paper, streamflow hindcasts were produced with the Community Hydrologic Prediction System and were bias-corrected with the Ensemble Post-processor (EnsPost). In order to separate the meteorological and hydrologic uncertainties, the raw streamflow forecasts were verified against simulated streamflows, as well as observed flows. Also, when verifying the bias-corrected streamflow forecasts, the total skill was decomposed into contributions from the MEFP-GFS and the EnsPost. In general, the streamflow forecasts are substantially more skillful when using the MEFP-GFS together with the EnsPost than using the MEFP with resampled climatology alone. However, both the raw and bias-corrected streamflow forecasts have lower biases, stronger correlations and are more skillful in CB- and CN-RFCs than AB- and MA-RFCs. In addition, there are strong variations in forecast quality with streamflow amount, forecast lead time, season and aggregation period. The relative importance of the meteorological and hydrologic uncertainties also varies between basins and is modulated by the same controls on forecast quality. For example, the MEFP-GFS accounts for the majority of skill in the CNRFC basins. This is associated with the greater predictability of large storms in the North Coast Ranges during the winter months. In CBRFC, much of the skill in the streamflow forecasts originates from the hydrologic modeling and the EnsPost, particularly during the snowmelt period. In AB- and MA-RFCs, the contributions from the MEFP and the EnsPost are more variable. This paper summarizes the verification results, describes the expected performance and limitations of the HEFS for short- to medium-range streamflow forecasting, and provides recommendations for future research. [ABSTRACT FROM AUTHOR]

Summary The selection and identification of a suitable hydrological model structure is a more challenging task than fitting parameters of a fixed model structure to reproduce a measured hydrograph. The suitable model structure is highly dependent on various criteria, i.e. the modeling objective, the characteristics and the scale of the system under investigation and the available data. Flexible environments for model building are available, but need to be assisted by proper diagnostic tools for model structure selection. This paper introduces a qualitative method for model component sensitivity analysis. Traditionally, model sensitivity is evaluated for model parameters. In this paper, the concept is translated into an evaluation of model structure sensitivity. Similarly to the one-factor-at-a-time (OAT) methods for parameter sensitivity, this method varies the model structure components one at a time and evaluates the change in sensitivity towards the output variables. As such, the effect of model component variations can be evaluated towards different objective functions or output variables. The methodology is presented for a simple lumped hydrological model environment, introducing different possible model building variations. By comparing the effect of changes in model structure for different model objectives, model selection can be better evaluated. Based on the presented component sensitivity analysis of a case study, some suggestions with regard to model selection are formulated for the system under study: (1) a non-linear storage component is recommended, since it ensures more sensitive (identifiable) parameters for this component and less parameter interaction; (2) interflow is mainly important for the low flow criteria; (3) excess infiltration process is most influencing when focussing on the lower flows; (4) a more simple routing component is advisable; and (5) baseflow parameters have in general low sensitivity values, except for the low flow criteria. [ABSTRACT FROM AUTHOR]

Summary The effects of climate change on water resources have been studied extensively throughout the world through the use of hydrologic models coupled with General Circulation Model (GCM) output or climate sensitivity scenarios. This paper examines the effects of hydrologic model parameterization uncertainty or equifinality, where multiple unique hydrologic model parameter sets can result in adequate calibration metrics, on hydrologic projections from downscaled GCMs for three snowmelt-dependent watersheds (upper reaches of the Clearwater, Gunnison, and Sacramento River watersheds) in the western United States. The hydrologic model used in this study is the Soil and Water Assessment Tool (SWAT) and is calibrated for discharge at the watershed outlet in each watershed. Despite achieving similar calibration metrics, a majority of hydrologic projections of average annual streamflow during the 2080s were statistically different, with differences in magnitude and direction (increase or decrease) compared to historical annual streamflows. At the average monthly time-scale, a majority of the hydrologic projections varied in peak streamflow timing, peak streamflow magnitude, summer streamflows, as well as overall increases or decreases compared to the historical monthly streamflows. Snowmelt projections from the SWAT model also widely varied, both in depth and snowmelt peak timing, for all watersheds. Since a large portion of the runoff-producing regions in the western United States is snowmelt-dependent, this has large implications for the prediction of the amount and timing of streamflow in the coming century. This paper shows that hydrologic model parameterizations that give similar adequate calibration metrics can lead to statistically significant differences in hydrologic projections under climate change. Therefore, researchers and water resource managers should account for this uncertainty by assembling ensemble projections from both multiple parameter sets and GCMs. [ABSTRACT FROM AUTHOR]

• A framework to evaluate stochastic rainfall models with a quantitative scale. • 'Good' modelled rainfall does not guarantee 'good' modelled streamflow. • Streamflow-based evaluation is more holistic than rainfall-based evaluation. • Component-wise evaluation can misrepresent ultimate model performance. Stochastic rainfall models are important tools for evaluating hydrological risks such as flooding and drought because of their ability to randomly generate alternative plausible climatic timeseries. The stochastic generation of climatic timeseries is not an end in itself, since they are typically applied to a catchment to determine the performance of water-related infrastructure systems, such as reservoirs or flood-control measures. This methodology typically involves a train of models to determine the end-of-system impact, yet the evaluation of stochastically generated rainfall timeseries is usually a stand-alone procedure focused on metrics directly related to the stochastic generator. This paper demonstrates discrepancies in this approach by evaluating two, daily-timestep, stochastic rainfall models in terms of rainfall metrics and their subsequently generated flow metrics after rainfall-runoff transformation. The two models are a Markov-based model and a latent-variable model, where each model is calibrated and evaluated showing 'overall good' performance. Stochastically generated timeseries, alongside observed rainfall timeseries are inputted to a calibrated catchment model (GR4J) to derive daily flow timeseries. Whereas the rainfall metrics typically showed 'good' performance, streamflow-based metrics are not necessarily 'good'. The procedure is repeated for 277 stations from Australia and 106 stations from the United States of America. Depending on the strictness of the flow-based comparison and region analysed, using the Markov-based model 12–26% of sites were classified as 'poor' performing, and 1%-9% of sites were classified as 'poor' using the latent-variable model. The results demonstrate that catchment-based performance of flow metrics is more holistic since it magnifies features of the rainfall not otherwise visible to rainfall-based evaluation. [ABSTRACT FROM AUTHOR]

• A framework for characterizing inundation depth (ID) during compound events is presented. • Relative importance of surge and inland flows in controlling flood ID is quantified. • A new topographic index is introduced to capture the variability of flood ID from river to coast. • With the index areas are identified where compound events are more prominent. Storm surge and river runoff can result in compound flooding in coastal areas. The impact of these events can be more significant than that generated by each component individually. This paper describes a framework for characterizing flooding during compound events and quantifying the relative importance of surge and river flows in controlling inundation depth (ID). For the analysis, we considered 1051 simulations of historical flood events covering about 40 years in Connecticut (CT), USA, and several simulated storms associated with synthetic climate scenarios. We simulated river discharge time series for each event using a physically based distributed hydrological model and retrieved the storm surge from tidal stations. These time series are used as upstream and downstream boundary conditions in 2D hydrodynamic simulation. We focused the analysis on seven locations along the coast of Connecticut covered by LiDAR-derived 1 m DEM. To capture the variability of inundation characteristics over the full-scale gradient from river to coast, we introduced a new topographic index, D-Index, that distinguishes topologic variabilities. Results from this study highlight that there is a correlation of ID to different drivers in distinct categories of the D-Index, which can correctly label the main source of hazard, either coast, river, or compound. We identified thresholds of standard deviation (σ) of the D-Index to identify areas where ID strongly correlates with either flow peak and volume, surge peak, or a combination of both. The study also shows that it is possible to use the results obtained from the 1 m analysis to generalize the findings for the entire CT coast to derive zones dominated by surge, river flow, or the compound effect of the two. This study can help coastal communities better understand the risk of the compound impacts of coastal storms, which is essential for increasing climate resilience. [ABSTRACT FROM AUTHOR]

• Two novel filters for state estimation with parameter uncertainty are proposed. • The CCKF and ACKF are tested by updating the states of a hydrological model. • Both filters can improve state accuracy by including parameter uncertainty. • The ACKF can suffer from diverging problems in real-world applications. • The CCKF can be preferred for operational forecasting due to its robustness. Filtering methods are popular for improving the forecast performances of hydrological models. Parameter uncertainty can result in poor state estimates; hence, augmented state-parameter estimation methods have gained much attention in the hydrologic literature. However, researchers have not treated the robustness of augmented methods in much detail. Another powerful approach to include the parameter uncertainty in the state estimation process is the consider Kalman filter (consider KF), which accounts for the parameter uncertainty without updating the parameters. The consider KF has received little to no attention in the hydrologic context. This paper investigates the usefulness of the consider KF by updating the states of a hydrological model (without updating the parameters), with the augmented state-parameter estimation method as the benchmark. Two nonlinear extensions for the augmented state-parameter estimation method (augmented cubature Kalman filter, ACKF) and consider Kalman filter (consider cubature Kalman filter, CCKF) are developed since hydrological models are mostly nonlinear. Both filters are tested with a lumped hydrological model. The results of a state update synthetic case show that both filters can improve the accuracy of the updated states by including the parameter uncertainty, while the ACKF may deteriorate the original performance when given perfect parameters. The results of a reforecast experiment show that both filters can improve the forecast performance by updating the model states based on the observed discharge, while the ACKF may suffer from potential divergence. Either the ACKF or CCKF can be considered for state estimation in the presence of parameter uncertainty in hydrological forecasting, with the CCKF being the more robust option. [ABSTRACT FROM AUTHOR]

• Three models were used to cover model structure uncertainty for each catchment. • AET is larger in the karst-dominated than the adjacent nonkarst-dominated catchment. • AET of karst-dominated catchment is more sensitive to the change in climate. • The karst-dominated catchment will experience much drier hydrological projections. Karst systems are well-known for their special features, such as epikarst zone and quick flow, compared to other hydrological systems. However, few studies explored the difference in hydrological responses to climate change between karst and nonkarst catchments. In this paper, two adjacent catchments dominated by karst and nonkarst in Southwest China (subtropical monsoon climate) were selected to investigate their different hydrological behaviors under future climates. To account for potential model structure uncertainties, three different karst hydrological models and three traditional catchment models were used to simulate these two catchments. The results demonstrate that different models yield comparable streamflow simulations within each catchment, allowing for the focus to be on the impact of climate change and catchment hydrological characteristics. The karst-dominated catchment is modeled to have larger actual evapotranspiration (AET) than the nonkarst-dominated catchment under similar climate conditions, with the AET of the former being more sensitive to changes in precipitation and potential evapotranspiration. These differences are primarily due to the presence of the epikarst in the unsaturated zone of the karst system, which can store more infiltration rainwater for evapotranspiration. Both catchments are expected to experience a wetter projection in the wet season and a drier projection in the dry season in the future. However, the karst-dominated catchment will face a drier climate with a significant increase in precipitation in just two months (July and August). Additionally, because of a higher sensitivity of monthly AET to climate change, the karst-dominated catchment will experience a greater monthly discharge reduction in the first two wet months (April and May), a relatively lower discharge increase in other wet months, and a lower monthly discharge in dry months, while the monthly discharge of the nonkarst-dominated catchment increases in the wet season and decrease in the dry season. Furthermore, the karst-dominated catchment will have a much lower annual discharge than the nonkarst-dominated catchment, particularly during extremely dry years in the future period due to its larger AET consumption. These all suggest that the karst-dominated catchment will be more vulnerable to droughts. Overall, due to the drier climate and the different hydrological backgrounds of the karst-dominated catchment, it is expected to have much drier hydrological conditions in the future compared to the nonkarst-dominated catchment. [ABSTRACT FROM AUTHOR]

Summary: The Office of Hydrologic Development (OHD) of the National Oceanic and Atmospheric Administration’s (NOAA) National Weather Service (NWS) conducted the second phase of the Distributed Model Intercomparison Project (DMIP 2). After DMIP 1, the NWS recognized the need for additional science experiments to guide its research-to-operations path towards advanced hydrologic models for river and water resources forecasting. This was accentuated by the need to develop a broader spectrum of water resources forecasting products (such as soil moisture) in addition to the more traditional river, flash flood, and water supply forecasts. As it did for DMIP 1, the NWS sought the input and contributions from the hydrologic research community. DMIP 1 showed that using operational precipitation data, some distributed models could indeed perform as well as lumped models in several basins and better than lumped models for one basin. However, in general, the improvements were more limited than anticipated by the scientific community. Models combining so-called conceptual rainfall-runoff mechanisms with physically-based routing schemes achieved the best overall performance. Clear gains were achieved through calibration of model parameters, with the average performance of calibrated models being better than uncalibrated models. DMIP 1 experiments were hampered by temporally-inconsistent precipitation data and few runoff events in the verification period for some basins. Greater uncertainty in modeling small basins was noted, pointing to the need for additional tests of nested basins of various sizes. DMIP 2 experiments in the Oklahoma (OK) region were more comprehensive than in DMIP 1, and were designed to improve our understanding beyond what was learned in DMIP 1. Many more stream gauges were located, allowing for more rigorous testing of simulations at interior points. These included two new gauged interior basins that had drainage areas smaller than the smallest in DMIP 1. Soil moisture and routing experiments were added to further assess if distributed models could accurately model basin-interior processes. A longer period of higher quality precipitation data was available, and facilitated a test to note the impacts of data quality on model calibration. Moreover, the DMIP 2 calibration and verification periods contained more runoff events for analysis. Two lumped models were used to define a robust benchmark for evaluating the improvement of distributed models compared to lumped models. Fourteen groups participated in DMIP 2 using a total of sixteen models. Ten of these models were not in DMIP 1. This paper presents the motivation for DMIP 2 Oklahoma experiments, discusses the major project elements, and describes the data and models used. In addition, the paper introduces the findings, which are covered in a companion results paper (). Lastly, the paper summarizes the DMIP 1 and 2 experiments with commentary from the NWS perspective. Future papers will cover the DMIP 2 experiments in the western USA mountainous basins. [ABSTRACT FROM AUTHOR]

Summary: Catchment hydrological models are widely used in climate change impact studies. The considered climate scenarios often involve increases in rainfall intensities. These increases might go beyond the range of historical events used in model calibration and validation. The hydrological impact results of climate change consequently are under these circumstances highly uncertain. This paper presents a method to test the validity of hydrological models for that type of impact analysis. The method is based on the evaluation of peak flow increases due to different levels of rainfall increases. The method is demonstrated, based upon the results of three conceptual rainfall–runoff models (NAM, PDM, VHM) for three catchments in Belgium. It is shown how the performance of the model in predicting impacts of rainfall increases can be integrated in the calibration process and how this method can increase the reliability of climate change impact results on peak flows. The paper addresses the importance of the relation between the soil moisture content and the overland flow coefficient, as it controls the generation of peak flows in the considered rainfall–runoff models. [ABSTRACT FROM AUTHOR]

• Improving M2S streamflow forecasting for rainfall-dominated basins. • Hydrologic EnKF data assimilation utilizing streamflow observations. • EnKF is more influential in improving correlation coefficient. Among different sources of uncertainty in hydrologic modeling (i.e., model structure, parameter estimation, input data, etc.), consecutive error reduction of model initial conditions can prevent a model from drifting away from reality and consequently improving model estimates. Most approaches that evaluated the correction of initial conditions through data assimilation (DA) have focused on improving hydrologic model simulations (i.e., under observed forcings) rather than evaluating the model performance in a forecasting context. This paper investigates the utility of Ensemble Kalman Filter (EnKF) data assimilation in which available observed streamflow is exploited to update state variables of a conceptual water balance model for forecasting monthly streamflow over 340 rainfall-dominated river basins across the contiguous United States (CONUS). Our results demonstrate that after EnKF application, streamflow simulation skill improves in terms of both Relative Root Mean Square Error (R-RMSE) and correlation coefficient (CC) for almost 90% of the selected river basins. Evaluating the model performance under different flow conditions shows that EnKF has stronger positive effect on monthly low flow predictions comparing to monthly high flows particularly during the summer season. The utility of EnKF is also assessed in the context of 1-month ahead streamflow forecasting. Due to the updated model initial conditions, streamflow forecasts are improved throughout the year even though the skill in hydrologic forecasts is predominantly dependent on the accuracy of precipitation forecasts. [ABSTRACT FROM AUTHOR]

• Literature-based and alternative equations for quick-flow runoff are tested. • A new empirical equation was found to outperform previously used equations. • Monthly maps of quick-flow runoff are produced and evaluated for the CONUS. • Patterns of increased quick-flow runoff in winter-spring months are observed. The quantitative estimation of the quick-flow runoff component of streamflow is required for many hydrologic applications. Estimation at the monthly timescale and national spatial scale would be particularly useful for national water availability modeling. This paper reviews a sample of commonly used equations for quick-flow runoff, including several currently in use in continental-scale models. The review shows the wide range of equation forms or heuristics currently in use to predict quick-flow runoff, the limited spatial scale over which these equations are often developed or calibrated, and the scarcity of well-tested equations available for quick-flow runoff at the monthly timescale. Data were gathered from a set of 1301 gaged watersheds across the United States to test a range of equations from the literature, along with several alternative equations, to assess and compare their performance in predicting quick-flow runoff at the monthly timescale. The highest-performing equation was selected for application to monthly maps of explanatory variables to produce monthly quick-flow runoff water budget contribution maps. This equation is a regression against precipitation, soil saturated hydraulic conductivity, surficial geology type, and slope data. Its application indicates that average quick-flow runoff across the conterminous United States in the winter exceeds that in the summer by up to a factor of three. The monthly maps were explored and evaluated for the timespan of 2000–2015. The comparison of equation forms and produced monthly maps will be useful for a variety of hydrologic modeling and monitoring applications. [ABSTRACT FROM AUTHOR]

• A technique is proposed for the simulation of gridded daily precipitation. • Proposed method is applicable to regions with limited precipitation information. • A conditional gridded hourly temperature generator is developed. • Generators have the capacity to produce temporally coherent fields of data. • Coherent generation is essential for hydrologic modeling applications. In this study, synchronized gridded precipitation and temperature generators were developed and evaluated in terms of their ability to reconstruct the spatial structure and other statistical properties of high-resolution CaPA (Canadian Precipitation Analysis) precipitation and GEM (Global Environmental Multiscale Model) temperature products. The precipitation generator is an extension to the gridded multi-site occurrence-amount model. While the multi-site framework is not efficiently applicable to regions with limited precipitation information and sparse observation network, this paper provides the full mathematical description of estimating the model based on high-resolution gridded products. The temperature model, on the other hand, introduces the first gridded hourly generator in the literature in which cyclic temperature components are conditioned using wet and dry day status information. This was achieved using harmonic functions establishing the connection between the simulated daily precipitation occurrence time-series and the moments of the generated temperature data. One of the key features of the proposed gridded generators is the capacity to produce temporally coherent fields of precipitation and temperature; an essential prerequisite for hydrologic modeling applications. The generators were evaluated in terms of their ability to reproduce the standard, interannual, and extreme-value characteristics of observed gridded products. The gridded precipitation model can preserve key statistical and spatial properties of the gridded CaPA observations to an acceptable level, while similar to the original multi-site framework, it falls short to capture long dry spells and the interannual variability of monthly total precipitation. The temperature model, on the other hand, preserves the majority of key features of the gridded GEM temperature observations, including the first two moments of the hourly data and the cumulative frequency of maximum and minimum temperature thresholds. The only notable defect identified is the departure of the interannual variability of the generated maximum and minimum temperatures from those observed. [ABSTRACT FROM AUTHOR]

• A new framework is developed for detailed flash flood risk mapping in ungauged basin. • Performance of the developed model is assessed with −0.004 m of BIAS for water level. • Flood inundation parameters and socioeconomic features are considered in the model. Flash floods are among the most devastating natural hazards but the analyses of flash floods are hindered by the lack of hydro-meteorological data. In particular, data issues are critical in ungauged basins. This paper presents an investigation of flash flood modeling using coupled hydrological and hydrodynamic models. Specifically, the TOPography-based hydrological MODEL (TOPMODEL) is employed to obtain the flood hydrograph and the coupled 1D-2D hydrodynamic model MIKEFLOOD is used to simulate floodplain inundation. The methods are tested in the Pajiang River, a mountainous basin in South China. The results show that the TOPMODEL effectively simulates the flood hydrograph and the Nash–Sutcliffe coefficients of the flow rate are larger than 0.80 for both calibration and validation. Also, MIKEFLOOD effectively simulates the flood inundation extent and flooding depth and the BIAS for simulated water level is −0.004 m. Further, using socioeconomic and historical flood disaster datasets, areas with high flash flood risk are identified. A flash flood risk assessment shows a clear trend of change in the flood hazard and the spatial heterogeneity of the flood vulnerability is significant. A sensitivity analysis of the flood vulnerability to the socioeconomic indicators indicates that cultivated land is the most sensitive indicators, and the population and the construction areas have similar influences. Overall, the coupled hydrological and hydrodynamic models prove effective and can be used in practical applications for flash flood risk mapping. [ABSTRACT FROM AUTHOR]

Highlights • Hydrological signatures about dynamics derived from distributed water level and soil moisture. • Information relevant to document impact of land use, rainfall type, soil moisture. • Information useful to diagnose simulated processes in distributed hydrological models. • Information useful to prioritize future model improvements or parameters estimation. • Qualitative information not sufficient to improve simulated water volume. Abstract This paper emphasizes the importance of integrating outlet discharge and observed internal variables in the evaluation of distributed hydrological models outputs. It proposes a general methodology for a diagnostic evaluation of a complex distributed hydrological model, based on discharge data at the outlet and additional distributed information such as water level and surface soil moisture data. The proposed methodology is illustrated using the PUMMA model in the Mercier sub-catchment (6.6 km2). Model parameters are specified according to field data and a previous study performed in a neighbouring catchment (Jankowfsky et al., 2014), without calibration. The distributed water level and soil moisture network of sensors were useful in the model evaluation process. Thus, model parameters are specified either using in situ information or results from previous studies. A stepwise approach is used for model evaluation. It includes standard water balance assessment as well as comparison of observed and simulated outlet discharge, whether on annual or event timescales. Soil moisture sensors are used to assess the ability of the model to simulate seasonal water storage dynamics based on a normalized index. The water level sensors network is used on two timescales: on a seasonal timescale, sensors network is used to assess the model's ability to simulate intermittency; whereas on event timescales, sensors network is used in determining the model's ability to reproduce observed reaction as well as response times. Event timescales do also focus on the correlation between hydrological response and either rainfall event or antecedent soil moisture variables. Results show that the non-calibrated model is quite effective at capturing water flow and soil water-storage dynamics, but it fails to reproduce observed runoff volume during events. There is strong indication of a deficiency in the characterization of catchment storage and upstream flowpath description. The soil water content and a network of water level sensors provide interesting information about soil moisture and river flow dynamics. They however fail to provide quantitative information about catchment storage. This study opens interesting perspectives for the evaluation of distributed hydrological models using hydrological signatures. Furthermore, it highlights the requirement of quantitative as well as qualitative signatures for improving such models. [ABSTRACT FROM AUTHOR]

Highlights • Comparison of two hydrological modeling studies at the local scale in the Everest region. • At high altitudes, snow-pack is the main driver of the discrepancy between the two models. • The time step impacts the snow-pack simulation, in particular during the summer season. • Estimations of soil processes are associated with high uncertainties due to models structure. Abstract This paper compares the hydrological responses at the local scale of two models using different degrees of refinement to represent physical processes in sparsely instrumented mountainous Himalayan catchments. This work presents the novelty of applying, at a small spatio-temporal scale and under the same forcing conditions, a fully distributed surface scheme based on mass and energy balance equations (ISBA surface scheme), and a semi-distributed calibrated model (J2000 hydrological model). A new conceptual module coupled to the ISBA surface scheme for flow routing is presented. Two small catchments located in mid- and high- mountain environments were chosen to represent the very different climatic and physiographic characteristics of the Central Himalayas in the Everest region of eastern Nepal. The results show that both models globally represent the dynamic of the processes for evaporation, quick runoff and discharge in a similar way. The differences in the model structures and results mainly concern the snow processes and the soil processes. In particular for the high-mountain catchment, the snow-pack simulation is shown to be the main driver of the discrepancy between the two models. The sub-daily variations of snow processes are shown to significantly influence the estimation of the snow-melt contribution to discharge. [ABSTRACT FROM AUTHOR]

Highlights • 3DNet-Catch provides balanced model complexity and adaptability to local conditions. • 3DNet-Catch includes the interception, snow, soil, and runoff and channel routing. • A comprehensive evaluation of model parameterisation and performance is conducted. • Most model parameters are physically meaningful, well-identifiable and uncorrelated. • Consistently good performance despite input data from sparse meteorological network. Abstract Hydrologic models are important for effective water resources management. They vary in complexity from parsimonious, spatially lumped, to physically-based, fully distributed models, which are generally expected to outperform the former. Wide applications of complex models are limited due to high data and computational demands. Therefore, a new approach based on well-balanced model complexity is needed to obtain reasonable simulation results with low data requirements. This paper presents a novel 3DNet-Catch hydrologic model, developed to represent key processes in sloped catchments under a temperate climate with modest data requirements. 3DNet-Catch includes runoff simulations within computational units by employing the interception, snow and soil routines, as well as runoff and channel routing. The soil routine, which is the key model feature, combines the SCS-CN method, an analytically integrated nonlinear outflow equation and the Brooks-Corey relation for unsaturated conductivity in an innovative manner. To advance runoff routing in 3DNet-Catch, an approach for analytical integration of the linear and nonlinear outflow equations is implemented. Most model parameters are physically meaningful, thus facilitating model calibration. The model structure can be adjusted according to soil and groundwater flow data, and it can include hydraulic structures, thereby providing adaptability to local conditions. A comprehensive hydrologic evaluation framework is established and conducted to examine whether 3DNet-Catch is adequately parameterised and can accurately reproduce catchment hydrologic response. The model parameterisation is evaluated by sensitivity, identifiability and correlation analyses. Model efficiency is quantified in terms of performance measures, hydrologic signatures and plausibility of the simulated hydrological processes. The results show high sensitivity of the hydrologic variables and performance measures to the model parameters, particularly to those of the soil routine. The parameters are uncorrelated and generally well identifiable. The model performs equally well in the calibration and evaluation periods. High efficiency in the hydrological signatures related to the soil routine indicates its robustness. The results, therefore, suggest that 3DNet-Catch is a comprehensively parameterised, versatile hydrologic model. It realistically reproduces observed hydrographs with modest data requirements, thus being appropriate for both engineering applications and investigative catchment dynamics studies. [ABSTRACT FROM AUTHOR]

Summary: The objective of this paper is to present the development and application of a grid based hydrologic model using an object oriented framework within geographic information systems (GIS). The proposed model is called Storage Released based Distributed Hydrologic Model (STORE DHM). Development of STORE DHM is a part of a broader objective to develop a modular hydrologic modeling system within GIS. In this paper, the conceptual framework of STORE DHM including its application to three watersheds in Indiana is presented. The model is tested in the study watersheds by first calibrating it against observed flow hydrograph for a single event, and then verifying it for three additional storm events using both point (gauged) and continuous (radar) rainfall data. Results show that SOTRE DHM is able to predict runoff hydrographs for different types of events in terms of storm duration, peak flow magnitude and time-to-peak. The average Nash–Sutcliffe coefficient is greater than 0.8 for all runoff hydrograph during validation, and the errors in peak flow and time-to-peak predictions are within 5% and 15%, respectively. In addition, STORE DHM output is compared with outputs from two hydrologic models including Hydrologic Engineering Center’s Hydrologic Modeling System (HEC-HMS) and time variant Spatially Distributed Direct Hydrograph travel time method (SDDH). Results from these comparisons show that STORE DHM outperforms both HEC-HMS and SDDH in terms of overall hydrograph shape and flow magnitude. Although the results from application of STORE DHM are encouraging, the work presented in this paper is just an initial step towards developing a comprehensive tightly coupled GIS modeling framework for hydrology. [ABSTRACT FROM AUTHOR]

Summary: This paper analyses the relationship between the Saturated Area Ratio (SAR) and the Relative Water Content (RWC) within a hydrological catchment. The SAR–RWC relationship is fundamental in lumped modelling in correctly representing the dominant physical processes. According to the simulations provided by a distributed model (TOPKAPI) on several catchments, the SAR–RWC relationship is affected by hysteresis, which is also supported by observations and numerical simulations by other researchers. Due to the strong linearity, this relationship is difficult to derive analytically. In this paper we propose a physical interpretation of the hysteresis and present a methodology to model it. By incorporating this mechanism in a lumped model, we demonstrate improvement in discharge reproduction as evaluated by comparative streamflow simulations. [Copyright &y& Elsevier]

Summary: This paper presents a detailed analysis of the September 8–9, 2002 flash flood event in the Gard region (southern France) using two distributed hydrological models: CVN built within the LIQUID® hydrological platform and MARINE. The models differ in terms of spatial discretization, infiltration and water redistribution representation, and river flow transfer. MARINE can also account for subsurface lateral flow. Both models are set up using the same available information, namely a DEM and a pedology map. They are forced with high resolution radar rainfall data over a set of 18 sub-catchments ranging from 2.5 to 99km2 and are run without calibration. To begin with, models simulations are assessed against post field estimates of the time of peak and the maximum peak discharge showing a fair agreement for both models. The results are then discussed in terms of flow dynamics, runoff coefficients and soil saturation dynamics. The contribution of the subsurface lateral flow is also quantified using the MARINE model. This analysis highlights that rainfall remains the first controlling factor of flash flood dynamics. High rainfall peak intensities are very influential of the maximum peak discharge for both models, but especially for the CVN model which has a simplified overland flow transfer. The river bed roughness also influences the peak intensity and time. Soil spatial representation is shown to have a significant role on runoff coefficients and on the spatial variability of saturation dynamics. Simulated soil saturation is found to be strongly related with soil depth and initial storage deficit maps, due to a full saturation of most of the area at the end of the event. When activated, the signature of subsurface lateral flow is also visible in the spatial patterns of soil saturation with higher values concentrating along the river network. However, the data currently available do not allow the assessment of both patterns. The paper concludes with a set of recommendations for enhancing field observations in order to progress in process understanding and gather a larger set of data to improve the realism of distributed models. [Copyright &y& Elsevier]

Abstract: The importance of hydrological uncertainty analysis has been emphasized in recent years and there is an urgent need to incorporate uncertainty estimation into water resources assessment procedures used in the southern Africa region. The region is characterized by a paucity of accurate data and limited human resources, but the need for informed development decisions is critical to social and economic development. One of the main sources of uncertainty is related to the estimation of the parameters of hydrological models. This paper proposes a framework for establishing parameter values, exploring parameter inter-dependencies and setting parameter uncertainty bounds for a monthly time-step rainfall-runoff model (Pitman model) that is widely used in the region. The method is based on well-documented principles of sensitivity and uncertainty analysis, but recognizes the limitations that exist within the region (data scarcity and accuracy, model user attitudes, etc.). Four example applications taken from different climate and physiographic regions of South Africa illustrate that the methods are appropriate for generating behavioural stream flow simulations which include parameter uncertainty. The parameters that dominate the model response and their degree of uncertainty vary between regions. Some of the results suggest that the uncertainty bounds will be too wide for effective water resources decision making. Further work is required to reduce some of the subjectivity in the methods and to investigate other approaches for constraining the uncertainty. The paper recognizes that probability estimates of uncertainty and methods to include input climate data uncertainties need to be incorporated into the framework in the future. [Copyright &y& Elsevier]

Summary: Artificial neural network ensemble (ANN ensemble) prediction is a technique in which the outputs of a set of separately trained ANNs are combined to form one unified prediction. ANN ensemble models are developed in this paper to improve the results of single artificial neural network (single ANN) for the estimation of the ice thickness in a number of selected Canadian lakes during the early winter ice growth period. An effective ensemble consists of a set of ANNs that may not be highly performing when they are used separately, but have their prediction errors greatly reduced once combined. This paper evaluates the effectiveness of a number of ensemble techniques including randomization, bagging and boosting for creating members of an ensemble, then averaging and stacking techniques for combining ensemble members. The experiments show that, in the context of estimation of lake ice thickness, boosting is much better than randomization, and sometimes better than bagging. Stacking was found to be more competitive than averaging. Overall, ANN ensemble models for the estimation of ice thickness proved to be more accurate than single ANN models, especially when boosting is used for combining ensemble members and when stacking is used to combine the outputs from individual members. ANN ensembles achieve the best generalization performance when the ensemble size is increased to around 20. [Copyright &y& Elsevier]

Summary: Quantification of uncertainty of hydrological models has attracted much attention in hydrologic research in recent years. Many methods for quantification of uncertainty have been reported in the literature, of which GLUE and formal Bayesian method are the two most popular methods. There have been many discussions in the literature concerning differences between these two methods in theory (mathematics) and results, and this paper focuses on the computational efficiency and differences in their results, but not on philosophies and mathematical rigor that both methods rely on. By assessing parameter and modeling uncertainty of a simple conceptual water balance model (WASMOD) with the use of GLUE and formal Bayesian method, the paper evaluates differences in the results of the two methods and discusses the reasons for these differences. The main findings of the study are that: (1) the parameter posterior distributions generated by the Bayesian method are slightly less scattered than those by the GLUE method; (2) using a higher threshold value (>0.8) GLUE results in very similar estimates of parameter and model uncertainty as does the Bayesian method; and (3) GLUE is sensitive to the threshold value used to select behavioral parameter sets and lower threshold values resulting in a wider uncertainty interval of the posterior distribution of parameters, and a wider confidence interval of model uncertainty. More study is needed to generalize the findings of the present study. [Copyright &y& Elsevier]

Summary: This work examines the possibility of using a surface hydrology model in a karst environment. In these kinds of environments, deterministic models such as SWAT, and the majority of surface hydrology models, are unable to compute the added contribution due to subsurface circulation. This can result in a substantial under-estimation of the river flows. Therefore such models are generally not applied by researchers to complex watersheds. Often the simulation of spring discharges is in itself very complex, making it even more difficult to simulate river discharge in karst environments, where the flow contribution coming from karst conduits cannot be directly measured. The present paper investigates the advantages of applying Wavelet Analysis (WA) to this field. The statistical method of WA has been recently applied in hydrogeology for interpreting spring discharges, and in hydrology to describe the variability of streamflows, thus providing insights into their dynamical link with the dominant modes of climate variability. However it has not yet been directly used to support the understanding of river flow components in karst environments. What is more, the present work proposes using WA not merely to assist in the analysis of experimental series, but also as a tool for interpreting the data series generated by the modelling process. In this paper we shall describe how WA can potentially be applied to help define the nature and behaviour of the karst contribution to river flows, thereby improving the future performance of surface hydrological modelling. [Copyright &y& Elsevier]

• A trivariate statistical framework is proposed for flood frequency analysis. • It uses a combination of a storm generator and a distributed hydrological model. • The trivariate methodology considers the main flood-generating mechanisms. • Initial soil moisture and spatio-temporal variability of storms are explicit. • This methodology can be replicated thanks its processes-based orientation. This paper proposes a trivariate methodology for flood frequency estimation. It combines the flood peak, storm magnitude, and initial soil moisture condition (ISMC) as the main flood-related statistical variables to be considered. The semi-arid Mediterranean "Rambla del Poyo" catchment has been used as a representative case study where the influence of the spatio-temporal variability of the storms and the ISMC on floods can lead to differences of up to two orders of magnitude in quantiles when the most commonly used methods are applied. In order to incorporate the main flood-generating mechanisms, the integrated use of a multidimensional storm generator with distributed hydrological modelling is proposed. Flood quantiles are then estimated by combining the maximum flows with the storm magnitude and ISMC in a trivariate probability distribution function through the application of Bayes' theorem and Lagrange's Mean Value theorem. Although the methodology proposed in this paper has been applied and tested in only one case study, it can be extended to other case studies due to its process-based orientation. [ABSTRACT FROM AUTHOR]

The objective of the authors’ work in the area of distributed modeling is to determine the manner with which rainfall input and model parameter uncertainty shapes the character of the flow simulation and prediction uncertainty of distributed hydrologic models. Toward this end and as a tool for the investigation, a distributed model, HRCDHM, has been formulated and tested as part of the NOAA Distributed Model Intercomparison Project (DMIP). This paper examines hourly flow simulations from HRCDHM applied with operational data obtained for the DMIP study watersheds. HRCDHM is a catchment-based, distributed input, distributed parameter hydrologic model. The hydrologic processes of infiltration/percolation, evapotranspiration, surface and subsurface flow (includes leakage to deep groundwater) are modeled along the vertical direction on a subcatchment basis in a manner similar to the Sacramento Soil Moisture Accounting model, and kinematic channel routing carries the flow through the network of subcatchments to the watershed outlet, providing capability for spatially distributed flow simulations. Subcatchment physical properties are derived from various digital terrain and land-characteristics databases through GIS processing and they are used to derive spatially distributed model parameter values. The NWS operational WSR-88D hourly radar rainfall estimates (Stage III product with pixel scale of approximately 4 km) constitute the rainfall forcing and a combination of model-derived and observed hourly surface meteorological data are used to produce the potential evapotranspiration forcing. HRCDHM was applied to and was calibrated for five watersheds for the period May 1993 through June 2000. Validation was done with data not used during the calibration period. This application shows that: (a) the HRCDHM, when forced with hourly data, is able to reproduce well the observed hourly streamflow at the outlet of each study watershed; and (b) beyond these outlet locations, HRCDHM is able to reproduce adequately the hourly flows at several interior locations. A companion paper [J. Hydrol. (2004)], in this issue details the use of the model for the characterization of simulation uncertainty within a Monte Carlo framework. [Copyright &y& Elsevier]

• Simplified hydrologic models of Urban Drainage Systems reliably predict runoff. • Simplified hydrologic model parameters can be assigned from cartography. • Simplified hydrologic model reliably predicts Combined Sewer Overflows. The management of an Urban Drainage System (UDS) is a complex task, as it requires extensive knowledge about precipitation regime, hydrological features of the catchment, hydraulic characteristics of the drainage network, and information about the water use by the served inhabitants. Complex semi-distributed hydrological and physically based hydraulic models are nowadays available to summarise such information and run simulations. However, in many cases, the uncertainty of the available hydrological information hampers the use of complex models. Hence, simple models with few parameters and small computational effort may be preferable, especially for UDS management problems requiring the execution of many simulations. This paper proposes a convenient approach to define effective lumped Simplified Models (SMs) of UDSs, the parameters of which can be estimated directly from cartographic information. For several case studies of UDS with different morphological and topological characteristics, SMs were built, capable of reproducing the hydrographs provided by available semi-distributed Detailed Models (DMs), assumed as benchmark in absence of measured hydrographs. To this aim, the SWMM simulation software was used, and the SM lumped parameters were calibrated by maximising the goodness of fit between the hydrograph of the DM and of the SM. The results show that SMs satisfactorily predict the hydrographs for all the case studies, and that robust relationships between the calibrated parameters and morphological and topological characteristics of the UDS can be established. This suggests that SMs can be used by decision makers for preliminary design, planning studies and management problems of UDSs, as their parameters can be soundly estimated from cartographic information. An example of application of SMs to Combined Sewer Overflow prediction is also presented. [ABSTRACT FROM AUTHOR]

Highlights • The predictive drought capacity of a seasonal forecast system was assessed using contingency tables. • An aggregate index for assessing predictive drought capacity was proposed. • The proposed methodology is useful for assessing the skill of climate models to predict droughts. • The dynamical System4 coupled model can be used for forecasting drought events. Abstract Droughts cause significant socio-economic and environmental impacts, so it has become an extremely important element in decision-making within water resource systems. For this reason, the research in this field has increased considerably over the last few decades. In order to be capable of making early decisions and reducing drought impacts, it is necessary to predict the occurrence of such events months or even years in advance. In this sense, various methods have been used to predict the occurrence of droughts. At present, seasonal forecast data can be used to forecast meteorological, hydrological, agricultural and operational droughts. However, the seasonal forecast data of these dynamical ocean-atmosphere coupled models must be analyzed in an exhaustive way, since it is known that these models may not adequately represent the climatic variability at river basin scale. Hence, this paper presents a new methodology for assessing the skill of a climate forecasting system in order to predict the occurrence of droughts by using contingency tables. The indices obtained from the contingency tables are necessary to perform the analysis of the predictive ability of the model in a semi-distributed way. All this taking into account the intensity of droughts using different scenarios based on the threshold below which it is considered to be in drought. Finally, a single value is obtained to determine the predictive ability of the forecasting model for the entire basin. The proposed methodology is applied to the Júcar river basin in Spain. It has been found that the analyzed forecast model shows better results than those obtained using an autoregressive model. Further work is needed to enhance climate forecasting from the perspective of water resources management, however, it should be mentioned that this type of data could be used for drought forecasting, allowing possible mitigation measures. [ABSTRACT FROM AUTHOR]

Hydrological and non-point source pollution (H/NPS) predictions in ungagged basins have become the key problem for watershed studies, especially for those large-scale catchments. However, few studies have explored the comprehensive impacts of rainfall data scarcity on H/NPS predictions. This study focused on: 1) the effects of rainfall spatial scarcity (by removing 11%–67% of stations based on their locations) on the H/NPS results; and 2) the impacts of rainfall temporal scarcity (10%–60% data scarcity in time series); and 3) the development of a new evaluation method that incorporates information entropy. A case study was undertaken using the Soil and Water Assessment Tool (SWAT) in a typical watershed in China. The results of this study highlighted the importance of critical-site rainfall stations that often showed greater influences and cross-tributary impacts on the H/NPS simulations. Higher missing rates above a certain threshold as well as missing locations during the wet periods resulted in poorer simulation results. Compared to traditional indicators, information entropy could serve as a good substitute because it reflects the distribution of spatial variability and the development of temporal heterogeneity. This paper reports important implications for the application of Distributed Hydrological Models and Semi-distributed Hydrological Models, as well as for the optimal design of rainfall gauges among large basins. [ABSTRACT FROM AUTHOR]

Snow is an important component of the hydrological cycle. As a major part of the cryosphere, snow cover also represents a valuable terrestrial water resource. In the context of climate change, the dynamics of snow cover play a crucial role in rebalancing the global energy and water budgets. Remote sensing, hydrological modeling and in situ observations are three techniques frequently utilized for snow cover investigations. However, the uncertainties caused by systematic errors, scale gaps, and complicated snow physics, among other factors, limit the usability of these three approaches in snow studies. In this paper, an overview of the advantages, limitations and recent progress of the three methods is presented, and more effective ways to estimate snow cover properties are evaluated. The possibility of improving remotely sensed snow information using ground-based observations is discussed. As a rapidly growing source of volunteered geographic information (VGI), web-based geotagged photos have great potential to provide ground truth data for remotely sensed products and hydrological models and thus contribute to procedures for cloud removal, correction, validation, forcing and assimilation. Finally, this review proposes a synergistic framework for the future of snow cover research. This framework highlights the cross-scale integration of in situ and remotely sensed snow measurements and the assimilation of improved remote sensing data into hydrological models. [ABSTRACT FROM AUTHOR]

Data assimilation techniques such as the Ensemble Kalman Filter (EnKF) are often applied to hydrological models with minimal state volume/capacity constraints enforced during ensemble generation. Flux constraints are rarely, if ever, applied. Consequently, model states can be adjusted beyond physically reasonable limits, compromising the integrity of model output. In this paper, we investigate the effect of constraining the EnKF on forecast performance. A “free run” in which no assimilation is applied is compared to a completely unconstrained EnKF implementation, a ‘typical’ hydrological implementation (in which mass constraints are enforced to ensure non-negativity and capacity thresholds of model states are not exceeded), and then to a more tightly constrained implementation where flux as well as mass constraints are imposed to force the rate of water movement to/from ensemble states to be within physically consistent boundaries. A three year period (2008–2010) was selected from the available data record (1976–2010). This was specifically chosen as it had no significant data gaps and represented well the range of flows observed in the longer dataset. Over this period, the standard implementation of the EnKF (no constraints) contained eight hydrological events where (multiple) physically inconsistent state adjustments were made. All were selected for analysis. Mass constraints alone did little to improve forecast performance; in fact, several were significantly degraded compared to the free run. In contrast, the combined use of mass and flux constraints significantly improved forecast performance in six events relative to all other implementations, while the remaining two events showed no significant difference in performance. Placing flux as well as mass constraints on the data assimilation framework encourages physically consistent state estimation and results in more accurate and reliable forward predictions of streamflow for robust decision-making. We also experiment with the observation error, which has a profound effect on filter performance. We note an interesting tension exists between specifying an error which reflects known uncertainties and errors in the measurement versus an error that allows “optimal” filter updating. [ABSTRACT FROM AUTHOR]