Articles | Volume 25, issue 6
Research article 03 Jun 2021
Research article | 03 Jun 2021
Evaluation of random forests for short-term daily streamflow forecasting in rainfall- and snowmelt-driven watersheds
Leo Triet Pham et al.
No articles found.
Feng Ma, Lifeng Luo, Aizhong Ye, and Qingyun Duan
Hydrol. Earth Syst. Sci., 22, 5697–5709,Short summary
Predicting meteorological droughts more than 2 months in advance became difficult due to low predictability, leading to weak skill for hydrological droughts in wet seasons. Hydrological drought forecasts showed skills up to 3–6 lead months due to the memory of initial hydrologic conditions in dry seasons. Human activities have increased hydrological predictability during wet seasons in the MHRB. This fills gaps in understanding drought and predictability predictions in endorheic and arid basins.
Related subject area
Subject: Catchment hydrology | Techniques and Approaches: Modelling approachesPerformance of automated methods for flash flood inundation mapping: a comparison of a digital terrain model (DTM) filling and two hydrodynamic methodsA novel method for cold-region streamflow hydrograph separation using GRACE satellite observationsA Bayesian approach to understanding the key factors influencing temporal variability in stream water quality – a case study in the Great Barrier Reef catchmentsProjected changes in Rhine River flood seasonality under global warmingTechnical note: Diagnostic efficiency – specific evaluation of model performanceHow catchment characteristics influence hydrological pathways and travel times in a boreal landscapeRainfall–runoff prediction at multiple timescales with a single Long Short-Term Memory networkUser-oriented hydrological indices for early warning systems with validation using post-event surveys: flood case studies in the Central Apennine DistrictSpace–time variability in soil moisture droughts in the Himalayan regionA multi-sourced assessment of the spatiotemporal dynamics of soil moisture in the MARINE flash flood modelClimate change impacts model parameter sensitivity – implications for calibration strategy and model diagnostic evaluationThe value of water isotope data on improving process understanding in a glacierized catchment on the Tibetan PlateauImpact of karst areas on runoff generation, lateral flow and interbasin groundwater flow at the storm-event timescaleTriple oxygen isotope systematics of evaporation and mixing processes in a dynamic desert lake systemIrrigation, damming, and streamflow fluctuations of the Yellow RiverBehind the scenes of streamflow model performanceLearning from satellite observations: increased understanding of catchment processes through stepwise model improvementDiagnosis toward predicting mean annual runoff in ungauged basinsThe era of infiltrationA time-varying parameter estimation approach using split-sample calibration based on dynamic programmingTiming and magnitude of future annual runoff extremes in contrasting Alpine catchments in AustriaA history of TOPMODELProgressive water deficits during multiyear droughts in basins with long hydrological memory in ChileA comparison of catchment travel times and storage deduced from deuterium and tritium tracers using StorAge Selection functionsThe role and value of distributed precipitation data in hydrological modelsFlood spatial coherence, triggers, and performance in hydrological simulations: large-sample evaluation of four streamflow-calibrated modelsFlexible vector-based spatial configurations in land modelsTwo-stage variational mode decomposition and support vector regression for streamflow forecastingPredicting probabilities of streamflow intermittency across a temperate mesoscale catchmentA new fractal-theory-based criterion for hydrological model calibrationImportance of the informative content in the study area when regionalising rainfall-runoff model parameters: the role of nested catchments and gauging station densityMachine Learning Deciphers CO2 Sequestration and Subsurface Flowpaths from Stream ChemistryWhich rainfall score is more informative about the performance in river discharge simulation? A comprehensive assessment on 1318 basins over EuropeAssimilation of Soil Moisture and Ocean Salinity (SMOS) brightness temperature into a large-scale distributed conceptual hydrological model to improve soil moisture predictions: the Murray–Darling basin in Australia as a test caseFrequency and magnitude variability of Yalu River flooding: numerical analyses for the last 1000 yearsAssessing the degree of detail of temperature-based snow routines for runoff modelling in mountainous areas in central EuropeAdaptive clustering: reducing the computational costs of distributed (hydrological) modelling by exploiting time-variable similarity among model elementsClimate elasticity of evapotranspiration shifts the water balance of Mediterranean climates during multi-year droughtsFuture streamflow regime changes in the United States: assessment using functional classificationRisks and opportunities for a Swiss hydroelectricity company in a changing climateSurvival of the Qaidam mega-lake system under mid-Pliocene climates and its restoration under future climatesHydrological evaluation of open-access precipitation data using SWAT at multiple temporal and spatial scalesUnderstanding coastal wetland conditions and futures by closing their hydrologic balance: the case of the Gialova lagoon, GreeceWhy does a conceptual hydrological model fail to correctly predict discharge changes in response to climate change?Sensitivity of meteorological-forcing resolution on hydrologic variablesUsing altimetry observations combined with GRACE to select parameter sets of a hydrological model in a data-scarce regionAssessing the impact of seasonal-rainfall anomalies on catchment-scale water balance componentsCrossing hydrological and geochemical modeling to understand the spatiotemporal variability of water chemistry in a headwater catchment (Strengbach, France)Future shift in winter streamflow modulated by the internal variability of climate in southern OntarioOn the shape of forward transit time distributions in low-order catchments
Nabil Hocini, Olivier Payrastre, François Bourgin, Eric Gaume, Philippe Davy, Dimitri Lague, Lea Poinsignon, and Frederic Pons
Hydrol. Earth Syst. Sci., 25, 2979–2995,Short summary
Efficient flood mapping methods are needed for large-scale, comprehensive identification of flash flood inundation hazards caused by small upstream rivers. An evaluation of three automated mapping approaches of increasing complexity, i.e., a digital terrain model (DTM) filling and two 1D–2D hydrodynamic approaches, is presented based on three major flash floods in southeastern France. The results illustrate some limits of the DTM filling method and the value of using a 2D hydrodynamic approach.
Shusen Wang, Junhua Li, and Hazen A. J. Russell
Hydrol. Earth Syst. Sci., 25, 2649–2662,Short summary
Separating river flow into baseflow and surface runoff provides useful information for hydrology and climate studies, but traditional methods have critical limitations in the lack of physics, identifying snowmelt runoff and watershed size. This study developed a novel model using the GRACE satellite observations to address these limitations. It also includes estimates for watershed hydraulic conductivity and drainable water storage, which help assess aquifer properties and water resources.
Shuci Liu, Dongryeol Ryu, J. Angus Webb, Anna Lintern, Danlu Guo, David Waters, and Andrew W. Western
Hydrol. Earth Syst. Sci., 25, 2663–2683,Short summary
Riverine water quality can change markedly at one particular location. This study developed predictive models to represent the temporal variation in stream water quality across the Great Barrier Reef catchments, Australia. The model structures were informed by a data-driven approach, which is useful for identifying important factors determining temporal changes in water quality and, in turn, providing critical information for developing management strategies.
Erwin Rottler, Axel Bronstert, Gerd Bürger, and Oldrich Rakovec
Hydrol. Earth Syst. Sci., 25, 2353–2371,Short summary
The mesoscale hydrological model (mHM) forced with an ensemble of climate projection scenarios was used to assess potential future changes in flood seasonality in the Rhine River basin. Results indicate that future changes in flood characteristics are controlled by increases in precipitation sums and diminishing snowpacks. The decreases in snowmelt can counterbalance increasing precipitation, resulting in only small and transient changes in streamflow maxima.
Robin Schwemmle, Dominic Demand, and Markus Weiler
Hydrol. Earth Syst. Sci., 25, 2187–2198,Short summary
A better understanding of the reasons why model performance is unsatisfying represents a crucial part for meaningful model evaluation. We propose the novel diagnostic efficiency (DE) measure and diagnostic polar plots. The proposed evaluation approach provides a diagnostic tool for model developers and model users and facilitates interpretation of model performance.
Elin Jutebring Sterte, Fredrik Lidman, Emma Lindborg, Ylva Sjöberg, and Hjalmar Laudon
Hydrol. Earth Syst. Sci., 25, 2133–2158,Short summary
A numerical model was used to estimate annual and seasonal mean travel times across 14 long-term nested monitored catchments in the boreal region. The estimated travel times and young water fractions were consistent with observed variations of base cation concentration and stable water isotopes, δ18O. Soil type was the most important factor regulating the variation in mean travel times among sub-catchments, while the areal coverage of mires increased the young water fraction.
Martin Gauch, Frederik Kratzert, Daniel Klotz, Grey Nearing, Jimmy Lin, and Sepp Hochreiter
Hydrol. Earth Syst. Sci., 25, 2045–2062,Short summary
We present multi-timescale Short-Term Memory (MTS-LSTM), a machine learning approach that predicts discharge at multiple timescales within one model. MTS-LSTM is significantly more accurate than the US National Water Model and computationally more efficient than an individual LSTM model per timescale. Further, MTS-LSTM can process different input variables at different timescales, which is important as the lead time of meteorological forecasts often depends on their temporal resolution.
Annalina Lombardi, Valentina Colaiuda, Marco Verdecchia, and Barbara Tomassetti
Hydrol. Earth Syst. Sci., 25, 1969–1992,Short summary
The paper presents a modelling approach for the assessment of extremes in the hydrological cycle at a multi-catchment scale. It describes two new hydrological stress indices, innovative instruments that could be used by Civil Protection operators, for flood mapping in early warning systems. The main advantage in using the proposed indices is the possibility of displaying hydrological-stress information over any geographical domain.
Santosh Nepal, Saurav Pradhananga, Narayan Kumar Shrestha, Sven Kralisch, Jayandra P. Shrestha, and Manfred Fink
Hydrol. Earth Syst. Sci., 25, 1761–1783,Short summary
This paper examines soil moisture drought in the central Himalayan region by applying a process-based hydrological model. Our results suggest that both the occurrence and severity of droughts have increased over the last 3 decades, especially in the winter and pre-monsoon seasons. The insights provided into the frequency, spatial coverage, and severity of the drought conditions can provide valuable inputs towards improved management of water resources and greater agricultural productivity.
Judith Eeckman, Hélène Roux, Audrey Douinot, Bertrand Bonan, and Clément Albergel
Hydrol. Earth Syst. Sci., 25, 1425–1446,Short summary
The risk of flash flood is of growing importance for populations, particularly in the Mediterranean area in the context of a changing climate. The representation of soil processes in models is a key factor for flash flood simulation. The importance of the various methods for soil moisture estimation are highlighted in this work. Local measurements from the field as well as data derived from satellite imagery can be used to assess the performance of model outputs.
Lieke Anna Melsen and Björn Guse
Hydrol. Earth Syst. Sci., 25, 1307–1332,Short summary
Certain hydrological processes become more or less relevant when the climate changes. This should also be visible in the models that are used for long-term predictions of river flow as a consequence of climate change. We investigated this using three different models. The change in relevance should be reflected in how the parameters of the models are determined. In the different models, different processes become more relevant in the future: they disagree with each other.
Yi Nan, Lide Tian, Zhihua He, Fuqiang Tian, and Lili Shao
Hydrol. Earth Syst. Sci. Discuss.,
Revised manuscript accepted for HESSShort summary
This study integrated a water isotope module into the hydrological model THREW. The isotope-aided model was subsequently applied for process understanding in the glacierized watershed of Karuxung River on the Tibetan Plateau. The model was used to quantify the contribution of runoff component, and estimate the water travel time in the catchment. Model uncertainties were significantly constrained by using additional isotopic data, improving the process understanding in the catchment.
Martin Le Mesnil, Roger Moussa, Jean-Baptiste Charlier, and Yvan Caballero
Hydrol. Earth Syst. Sci., 25, 1259–1282,Short summary
We present an innovative approach consisting of the statistical analysis and comparison of 15 hydrological descriptors, characterizing catchment response to rainfall events. The distribution of these descriptors is analysed according to the occurrence of karst areas inside 108 catchments. It shows that karst impacts on storm events mainly result in river losses and that interbasin groundwater flows can represent a significant part of the catchment water budget ah the event timescale.
Claudia Voigt, Daniel Herwartz, Cristina Dorador, and Michael Staubwasser
Hydrol. Earth Syst. Sci., 25, 1211–1228,Short summary
Evaporation trends in the stable isotope composition (18O/16O, 17O/16O, 2H/1H) of throughflow ponds in a hydrologically complex and seasonally dynamic lake system can be reliably predicted by the classic Craig–Gordon isotope evaporation model. We demonstrate that the novel 17O-excess parameter is capable of resolving different types of evaporation with and without recharge and of identifying mixing processes that cannot be resolved using the classic δ2H–δ18O system alone.
Zun Yin, Catherine Ottlé, Philippe Ciais, Feng Zhou, Xuhui Wang, Polcher Jan, Patrice Dumas, Shushi Peng, Laurent Li, Xudong Zhou, Yan Bo, Yi Xi, and Shilong Piao
Hydrol. Earth Syst. Sci., 25, 1133–1150,Short summary
We improved the irrigation module in a land surface model ORCHIDEE and developed a dam operation model with the aim to investigate how irrigation and dams affect the streamflow fluctuations of the Yellow River. Results show that irrigation mainly reduces the annual river flow. The dam operation, however, mainly affects streamflow variation. By considering two generic operation rules, flood control and base flow guarantee, our dam model can sustainably improve the simulation accuracy.
Laurène J. E. Bouaziz, Fabrizio Fenicia, Guillaume Thirel, Tanja de Boer-Euser, Joost Buitink, Claudia C. Brauer, Jan De Niel, Benjamin J. Dewals, Gilles Drogue, Benjamin Grelier, Lieke A. Melsen, Sotirios Moustakas, Jiri Nossent, Fernando Pereira, Eric Sprokkereef, Jasper Stam, Albrecht H. Weerts, Patrick Willems, Hubert H. G. Savenije, and Markus Hrachowitz
Hydrol. Earth Syst. Sci., 25, 1069–1095,Short summary
We quantify the differences in internal states and fluxes of 12 process-based models with similar streamflow performance and assess their plausibility using remotely sensed estimates of evaporation, snow cover, soil moisture and total storage anomalies. The dissimilarities in internal process representation imply that these models cannot all simultaneously be close to reality. Therefore, we invite modelers to evaluate their models using multiple variables and to rely on multi-model studies.
Petra Hulsman, Hubert H. G. Savenije, and Markus Hrachowitz
Hydrol. Earth Syst. Sci., 25, 957–982,Short summary
Satellite observations have increasingly been used for model calibration, while model structural developments largely rely on discharge data. For large river basins, this often results in poor representations of system internal processes. This study explores the combined use of satellite-based evaporation and total water storage data for model structural improvement and spatial–temporal model calibration for a large, semi-arid and data-scarce river system.
Yuan Gao, Lili Yao, Ni-Bin Chang, and Dingbao Wang
Hydrol. Earth Syst. Sci., 25, 945–956,Short summary
Mean annual runoff prediction is of great interest but still poses a challenge in ungauged basins. The purpose of this study is to diagnose the data requirement for predicting mean annual runoff in ungauged basins based on a water balance model, in which the effects of climate variability are explicitly represented. The performance of predicting mean annual runoff can be improved by employing better estimation of soil water storage capacity including the effects of soil, topography, and bedrock.
Hydrol. Earth Syst. Sci., 25, 851–866,Short summary
Inspired by a quotation from Howard Cook in 1946, this paper traces the evolution of the infiltration theory of runoff from the work of Robert Horton and LeRoy Sherman in the 1930s to the early digital computer models of the 1970s and 1980s. Reconsideration of the perceptual model for many catchments, partly as a result of the greater appreciation of the contribution of subsurface flows to the hydrograph indicated by tracer studies, suggests a reconsideration of hydrological nomenclature.
Xiaojing Zhang and Pan Liu
Hydrol. Earth Syst. Sci., 25, 711–733,Short summary
Rainfall–runoff models are useful tools for streamflow simulation. However, efforts are needed to investigate how their parameters vary in response to climate changes and human activities. Thus, this study proposes a new method for estimating time-varying parameters, by considering both simulation accuracy and parameter continuity. The results show the proposed method is effective for identifying temporal variations of parameters and can simultaneously provide good streamflow simulation.
Sarah Hanus, Markus Hrachowitz, Harry Zekollari, Gerrit Schoups, Miren Vizcaino, and Roland Kaitna
Hydrol. Earth Syst. Sci. Discuss.,
Revised manuscript accepted for HESSShort summary
This study investigates the effects of climate change on runoff patterns in six alpine catchments in Austria at the end of the 21st century. Our results indicate a substantial shift to earlier occurrences in annual maximum and minimum flows in high-elevation catchments. Magnitudes of annual extremes are projected to increase under a moderate emission scenario in all catchments. Changes are generally more pronounced for high-elevation catchments.
Keith J. Beven, Mike J. Kirkby, Jim E. Freer, and Rob Lamb
Hydrol. Earth Syst. Sci., 25, 527–549,Short summary
The theory that forms the basis of TOPMODEL was first outlined by Mike Kirkby some 45 years ago. This paper recalls some of the early developments: the rejection of the first journal paper, the early days of digital terrain analysis, model calibration and validation, the various criticisms of the simplifying assumptions, and the relaxation of those assumptions in the dynamic forms of TOPMODEL, and it considers what we might do now with the benefit of hindsight.
Camila Alvarez-Garreton, Juan Pablo Boisier, René Garreaud, Jan Seibert, and Marc Vis
Hydrol. Earth Syst. Sci., 25, 429–446,Short summary
The megadrought experienced in Chile (2010–2020) has led to larger than expected water deficits. By analysing 106 basins with snow-/rainfall regimes, we relate such intensification with the hydrological memory of the basins, explained by snow and groundwater. Snow-dominated basins have larger memory and thus accumulate the effect of persistent precipitation deficits more strongly than pluvial basins. This notably affects central Chile, a water-limited region where most of the population lives.
Nicolas Björn Rodriguez, Laurent Pfister, Erwin Zehe, and Julian Klaus
Hydrol. Earth Syst. Sci., 25, 401–428,Short summary
Different parts of water have often been used as tracers to determine the age of water in streams. The stable tracers, such as deuterium, are thought to be unable to reveal old water compared to the radioactive tracer called tritium. We used both tracers, measured in precipitation and in a stream in Luxembourg, to show that this is not necessarily true. It is, in fact, advantageous to use the two tracers together, and we recommend systematically using tritium in future studies.
Ralf Loritz, Markus Hrachowitz, Malte Neuper, and Erwin Zehe
Hydrol. Earth Syst. Sci., 25, 147–167,Short summary
This study investigates the role and value of distributed rainfall in the runoff generation of a mesoscale catchment. We compare the performance of different hydrological models at different periods and show that a distributed model driven by distributed rainfall yields improved performances only during certain periods. We then step beyond this finding and develop a spatially adaptive model that is capable of dynamically adjusting its spatial model structure in time.
Manuela I. Brunner, Lieke A. Melsen, Andrew W. Wood, Oldrich Rakovec, Naoki Mizukami, Wouter J. M. Knoben, and Martyn P. Clark
Hydrol. Earth Syst. Sci., 25, 105–119,Short summary
Assessments of current, local, and regional flood hazards and their future changes often involve the use of hydrologic models. A reliable model ideally reproduces both local flood characteristics and regional aspects of flooding. In this paper we investigate how such characteristics are represented by hydrologic models. Our results show that both the modeling of local and regional flood characteristics are challenging, especially under changing climate conditions.
Shervan Gharari, Martyn P. Clark, Naoki Mizukami, Wouter J. M. Knoben, Jefferson S. Wong, and Alain Pietroniro
Hydrol. Earth Syst. Sci., 24, 5953–5971,Short summary
This work explores the trade-off between the accuracy of the representation of geospatial data, such as land cover, soil type, and elevation zones, in a land (surface) model and its performance in the context of modeling. We used a vector-based setup instead of the commonly used grid-based setup to identify this trade-off. We also assessed the often neglected parameter uncertainty and its impact on the land model simulations.
Ganggang Zuo, Jungang Luo, Ni Wang, Yani Lian, and Xinxin He
Hydrol. Earth Syst. Sci., 24, 5491–5518,Short summary
A two-stage variational mode decomposition and support vector regression is designed to reduce the influence of boundary effects without removing or correcting boundary-affected decompositions. The proposed model significantly reduces the boundary effect consequences, saves modeling time and computation resources, barely overfits the calibration samples, and forecasts monthly runoff reasonably well compared to the benchmark models.
Nils Hinrich Kaplan, Theresa Blume, and Markus Weiler
Hydrol. Earth Syst. Sci., 24, 5453–5472,Short summary
In recent decades the demand for detailed information of spatial and temporal dynamics of the stream network has grown in the fields of eco-hydrology and extreme flow prediction. We use temporal streamflow intermittency data obtained at various sites using innovative sensing technology as well as spatial predictors to predict and map probabilities of streamflow intermittency. This approach has the potential to provide intermittency maps for hydrological modelling and management practices.
Zhixu Bai, Yao Wu, Di Ma, and Yue-Ping Xu
Hydrol. Earth Syst. Sci. Discuss.,
Revised manuscript accepted for HESSShort summary
We think that fractal dimension of streamflow series can be used to improve the calibration of hydrological models. Thus, we designed the E-RD (efficiency-ratio of fractal dimension) strategy and examined its usability in the calibration of lumped model. The results reveal that, from most aspects, introducing RD into model calibration makes the simulation of streamflow components more reasonable. Besides, in calibration, only little decrease of E occurs when pursuing better RD.
Mattia Neri, Juraj Parajka, and Elena Toth
Hydrol. Earth Syst. Sci., 24, 5149–5171,Short summary
One of the most informative ways to gain information on ungauged river sections is through the implementation of a rainfall-runoff model, exploiting the information collected in gauged catchments in the study area. This study analyses how the performances of different model regionalisation approaches are influenced by the informative content of the available regional data set, in order to identify the methods that are more suitable for the data availability in the region.
Andrew R. Shaughnessy, Xin Gu, Tao Wen, and Susan L. Brantley
Hydrol. Earth Syst. Sci. Discuss.,
Revised manuscript accepted for HESSShort summary
It is often difficult to determine the sources of solutes in streams and how much each source contributes. We developed a new method of unmixing stream chemistry via machine learning. We found that sulfate in three watersheds are related to groundwater flowpaths. Our results emphasize that acid rain reduces a watershed’s capacity to remove CO2 from the atmosphere, a key geological control on climate. Our method will help scientists unmix stream chemistry in watersheds where sources are unknown.
Stefania Camici, Christian Massari, Luca Ciabatta, Ivan Marchesini, and Luca Brocca
Hydrol. Earth Syst. Sci., 24, 4869–4885,Short summary
The paper performs the most comprehensive European-scale evaluation to date of satellite rainfall products for river flow prediction. In doing so, how errors transfer from satellite-based rainfall products into flood simulation is investigated in depth and, for the first time, quantitative guidelines on the use of these products for hydrological applications are provided. This result can represent a keystone in the use of satellite rainfall products, especially in data-scarce regions.
Renaud Hostache, Dominik Rains, Kaniska Mallick, Marco Chini, Ramona Pelich, Hans Lievens, Fabrizio Fenicia, Giovanni Corato, Niko E. C. Verhoest, and Patrick Matgen
Hydrol. Earth Syst. Sci., 24, 4793–4812,Short summary
Our objective is to investigate how satellite microwave sensors, particularly Soil Moisture and Ocean Salinity (SMOS), may help to reduce errors and uncertainties in soil moisture simulations with a large-scale conceptual hydro-meteorological model. We assimilated a long time series of SMOS observations into a hydro-meteorological model and showed that this helps to improve model predictions. This work therefore contributes to the development of faster and more accurate drought prediction tools.
Hui Sheng, Xiaomei Xu, Jian Hua Gao, Albert J. Kettner, Yong Shi, Chengfeng Xue, Ya Ping Wang, and Shu Gao
Hydrol. Earth Syst. Sci., 24, 4743–4761,Short summary
This paper investigates the variability of past flooding by applying a numerical model coupled with historical records of regional climate and anthropogenic activity under the deficiency of observations. We conclude that trends in flooding frequency were predominantly modulated by the intensity and frequency of extreme rainfall events, which highlights the need for the implementation of effective river engineering measures to counteract increasing flood risks as a result of the future.
Marc Girons Lopez, Marc J. P. Vis, Michal Jenicek, Nena Griessinger, and Jan Seibert
Hydrol. Earth Syst. Sci., 24, 4441–4461,Short summary
Snow processes are crucial for runoff in mountainous areas, but their complexity makes water management difficult. Temperature models are widely used as they are simple and do not require much data, but not much thought is usually given to which model to use, which may lead to bad predictions. We studied the impact of many model alternatives and found that a more complex model does not necessarily perform better. Finding which processes are most important in each area is a much better strategy.
Uwe Ehret, Rik van Pruijssen, Marina Bortoli, Ralf Loritz, Elnaz Azmi, and Erwin Zehe
Hydrol. Earth Syst. Sci., 24, 4389–4411,Short summary
In this paper we propose adaptive clustering as a new method for reducing the computational efforts of distributed modelling. It consists of identifying similar-acting model elements during the runtime, clustering them, running the model for just a few representatives per cluster, and mapping their results to the remaining model elements in the cluster. With the example of a hydrological model, we show that this saves considerable computation time, while largely maintaining the output quality.
Francesco Avanzi, Joseph Rungee, Tessa Maurer, Roger Bales, Qin Ma, Steven Glaser, and Martha Conklin
Hydrol. Earth Syst. Sci., 24, 4317–4337,Short summary
Multi-year droughts in Mediterranean climates often see a lower fraction of precipitation allocated to runoff compared to non-drought years. By comparing observed water-balance components with simulations by a hydrologic model (PRMS), we reinterpret these shifts as a hysteretic response of the water budget to climate elasticity of evapotranspiration. Our results point to a general improvement in hydrologic predictions across drought and recovery cycles by including this mechanism.
Manuela I. Brunner, Lieke A. Melsen, Andrew J. Newman, Andrew W. Wood, and Martyn P. Clark
Hydrol. Earth Syst. Sci., 24, 3951–3966,Short summary
Streamflow seasonality is changing and expected to further change under the influence of climate change. We here assess how annual streamflow hydrographs will change in future by using a newly developed classification scheme. Our comparison of future with current annual hydrograph classes shows that robust changes are expected only for currently melt-influenced regions in the Rocky Mountains. These upstream changes may require the adaptation of management strategies in downstream regions.
Kirsti Hakala, Nans Addor, Thibault Gobbe, Johann Ruffieux, and Jan Seibert
Hydrol. Earth Syst. Sci., 24, 3815–3833,Short summary
Under a changing climate, reliable information on future hydrological conditions is necessary to inform water resource management. Here, we collaborated with a hydropower company that selected streamflow and energy demand indices. Using these indices, we identified stakeholder needs and used this to tailor the production of our climate change impact projections. We show that opportunities and risks for a hydropower company depend on a range of factors beyond those covered by traditional studies.
Hydrol. Earth Syst. Sci., 24, 3835–3850,Short summary
During the Pliocene, the Qaidam Basin on the Tibetan Plateau contained a mega-lake system. During the Pleistocene, it disappeared almost completely. Today, hyperarid climates prevail in the low-altitude parts of the basin. This study reveals that today's mean water balance of the Qaidam Basin is nearly zero and is positive during warmer, less dry years. The results explain how the mega-lake system could survive for a long time in the past and could eventually be restored in the future.
Jianzhuang Pang, Huilan Zhang, Quanxi Xu, Yujie Wang, Yunqi Wang, Ouyang Zhang, and Jiaxin Hao
Hydrol. Earth Syst. Sci., 24, 3603–3626,Short summary
As frequently used precipitation products, Gauge, CPC, and CHIRPS presented different behaviors in describing precipitation on different spatial and temporal scales, yet these dissimilarities could be concealed in hydrological modeling by parameter calibration and validation. Parameter adjustment in hydrologic modeling, however, would yield different water balance components and thus alter hydrologic mechanisms, demonstrating the complexity in physically describing natural hydrologic processes.
Stefano Manzoni, Giorgos Maneas, Anna Scaini, Basil E. Psiloglou, Georgia Destouni, and Steve W. Lyon
Hydrol. Earth Syst. Sci., 24, 3557–3571,Short summary
A modeling tool is developed to assess the vulnerability of coastal wetlands to climatic and water management changes. Applied to the case study of the Gialova lagoon (Greece), this tool highlights the reliance of the lagoon functionality on scarce freshwater sources already under high demand from agriculture. Climatic changes will likely increase lagoon salinity, despite efforts to improve water management.
Doris Duethmann, Günter Blöschl, and Juraj Parajka
Hydrol. Earth Syst. Sci., 24, 3493–3511,Short summary
We investigate why a conceptual hydrological model failed to correctly predict observed discharge changes in response to increasing precipitation and air temperature in 156 Austrian catchments. Simulations indicate that poor model performance is related to two problems, namely a model structure that neglects changes in vegetation dynamics and inhomogeneities in precipitation data caused by changes in stations density with time. Other hypotheses did not improve simulated discharge changes.
Fadji Z. Maina, Erica R. Siirila-Woodburn, and Pouya Vahmani
Hydrol. Earth Syst. Sci., 24, 3451–3474,Short summary
Projecting the changes in water resources under a no-analog future climate requires integrated hydrologic models. However, these models are plagued by several sources of uncertainty. A hydrologic model was forced with various resolutions of meteorological forcing (0.5 to 40.5 km) to assess its sensitivity to these inputs. We show that most hydrologic variables reveal biases that are seasonally and spatially dependent, which can have serious implications for calibration and water management.
Petra Hulsman, Hessel C. Winsemius, Claire I. Michailovsky, Hubert H. G. Savenije, and Markus Hrachowitz
Hydrol. Earth Syst. Sci., 24, 3331–3359,Short summary
In the absence of discharge data in ungauged basins, remotely sensed river water level data, i.e. altimetry, may provide valuable information to calibrate hydrological models. This study illustrated that for large rivers in data-scarce regions, river altimetry data from multiple locations combined with GRACE data have the potential to fill this gap when combined with estimates of the river geometry, thereby allowing a step towards more reliable hydrological modelling in data-scarce regions.
Paolo Nasta, Carolina Allocca, Roberto Deidda, and Nunzio Romano
Hydrol. Earth Syst. Sci., 24, 3211–3227,Short summary
Rainfall seasonal anomalies in a Mediterranean climate are assessed by using two distinct approaches: a static approach based on the standardized precipitation index and a dynamic approach that identifies the rainy season by considering rainfall magnitude, timing, and duration. The impact of rainfall seasonality on catchment-scale water balance components is evaluated through scenario-based simulations of the Soil Water Assessment Tool in the upper Alento River catchment in southern Italy.
Julien Ackerer, Benjamin Jeannot, Frederick Delay, Sylvain Weill, Yann Lucas, Bertrand Fritz, Daniel Viville, and François Chabaux
Hydrol. Earth Syst. Sci., 24, 3111–3133,
Olivier Champagne, M. Altaf Arain, Martin Leduc, Paulin Coulibaly, and Shawn McKenzie
Hydrol. Earth Syst. Sci., 24, 3077–3096,Short summary
Using 50 members of one regional climate model and a processed-based hydrological model applied in four river basins in southern Ontario, this work focused on the winter streamflow projection uncertainties for the first half of 21st century. The results show a January–February increase of streamflow for the 50 projections due to early snowmelt and a rainfall increase. The streamflow projections are also modulated by the change of pressure patterns advecting different air masses over the region.
Ingo Heidbüchel, Jie Yang, Andreas Musolff, Peter Troch, Ty Ferré, and Jan H. Fleckenstein
Hydrol. Earth Syst. Sci., 24, 2895–2920,Short summary
With the help of a 3-D computer model we examined how long the water of different rain events stays inside small catchments before it is discharged and how the nature of this discharge is controlled by different catchment and climate properties. We found that one can only predict the discharge dynamics when taking into account a combination of catchment and climate properties (i.e., there was not one single most important predictor). Our results can help to manage water pollution events.
Adamowski, J. F.: Development of a short-term river flood forecasting method for snowmelt driven floods based on wavelet and cross-wavelet analysis, J. Hydrol., 353, 247–266, 2008. a
Altman, D. G. and Bland, J. M.: Statistics notes Variables and parameters, Brit. Med. J., 318, 1667, 1999. a
Aubert, D., Loumagne, C., and Oudin, L.: Sequential assimilation of soil moisture and streamflow data in a conceptual rainfall–runoff model, J. Hydrol., 280, 145–161, 2003. a
Boyle, D. P., Gupta, H. V., and Sorooshian, S.: Toward improved calibration of hydrologic models: Combining the strengths of manual and automatic methods, Water Resour. Res., 36, 3663–3674, 2000. a
Calle, M. L. and Urrea, V.: Letter to the editor: stability of random forest importance measures, Brief. Bioinform., 12, 86–89, 2010. a
Cayan, D. R., Redmond, K. T., and Riddle, L. G.: ENSO and hydrologic extremes in the western United States, J. Climate, 12, 2881–2893, 1999. a
Chen, X. and Ishwaran, H.: Random forests for genomic data analysis, Genomics, 99, 323–329, 2012. a
Coulston, J. W., Blinn, C. E., Thomas, V. A., and Wynne, R. H.: Approximating prediction uncertainty for random forest regression models, Photogramm. Eng. Rem. S., 82, 189–197, 2016. a
Dawson, C. W., Abrahart, R. J., Shamseldin, A. Y., and Wilby, R. L.: Flood estimation at ungauged sites using artificial neural networks, J. Hydrol., 319, 391–409, 2006. a
Dibike, Y. B. and Solomatine, D. P.: River flow forecasting using artificial neural networks, Phys. Chem. Earth Pt. B, 26, 1–7, 2001. a
Dingman, S. L.: Physical hydrology, Waveland Press, Long Grove, Illinois, 104–106, 2015. a
Elsner, M. M., Cuo, L., Voisin, N., Deems, J. S., Hamlet, A. F., Vano, J. A., Mickelson, K. E., Lee, S.-Y., and Lettenmaier, D. P.: Implications of 21st century climate change for the hydrology of Washington State, Climatic Change, 102, 225–260, 2010. a
Graham, C. B., Barnard, H. R., Kavanagh, K. L., and McNamara, J. P.: Catchment scale controls the temporal connection of transpiration and diel fluctuations in streamflow, Hydrol. Process., 27, 2541–2556, 2013. a
Gregorutti, B., Michel, B., and Saint-Pierre, P.: Correlation and variable importance in random forests, Stat. Comput., 27, 659–678, 2017. a
Gupta, H. V., Sorooshian, S., and Yapo, P. O.: Status of automatic calibration for hydrologic models: Comparison with multilevel expert calibration, J. Hydrol. Eng., 4, 135–143, 1999. a
Gupta, H. V., Kling, H., Yilmaz, K. K., and Martinez, G. F.: Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling, J. Hydrol., 377, 80–91, 2009. a
Hwang, S. H., Ham, D. H., and Kim, J. H.: A new measure for assessing the efficiency of hydrological data-driven forecasting models, Hydrolog. Sci. J., 57, 1257–1274, 2012. a
Ishwaran, H. and Lu, M.: Standard errors and confidence intervals for variable importance in random forest regression, classification, and survival, Stat. Med., 38, 558–582, 2019. a
James, G., Witten, D., Hastie, T., and Tibshirani, R.: An introduction to statistical learning,Springer, New York, 113, 246–247, 2013. a
Johnstone, J. A.: A quasi-biennial signal in western US hydroclimate and its global teleconnections, Clim. Dynam., 36, 663–680, 2011. a
Karran, D. J., Morin, E., and Adamowski, J.: Multi-step streamflow forecasting using data-driven non-linear methods in contrasting climate regimes, J. Hydroinform., 16, 671–689, 2013. a
Knowles, N., Dettinger, M. D., and Cayan, D. R.: Trends in snowfall versus rainfall in the western United States, J. Climate, 19, 4545–4559, 2006. a
Knowles, N., Dettinger, M., and Cayan, D.: Trends in snowfall versus rainfall for the western united states, 1949–2001, prepared for California energy commission public interest energy research program, Sacramento, California, 2007. a
Kuhn, M. et al.: Building predictive models in R using the caret package, J. Stat. Softw., 28, 1–26, 2008. a
Li, D., Wrzesien, M. L., Durand, M., Adam, J., and Lettenmaier, D. P.: How much runoff originates as snow in the western United States, and how will that change in the future?, Geophys. Res. Lett., 44, 6163–6172, 2017. a
Li, X., Sha, J., and Wang, Z.-L.: Comparison of daily streamflow forecasts using extreme learning machines and the random forest method, Hydrolog. Sci. J., 64, 1857–1866, 2019. a
Louppe, G., Wehenkel, L., Sutera, A., and Geurts, P.: Understanding variable importances in forests of randomized trees, in: Advances in neural information processing systems, 26, 431–439, 2013. a
Lundquist, J. D., Dettinger, M. D., Stewart, I. T., and Cayan, D. R.: Variability and trends in spring runoff in the western United States, Climate warming in western North America: evidence and environmental effects, University of Utah Press, Salt Lake City, Utah, USA, in: Climate Warming in Western North America: Evidence and Environmental Effects, 63–76, 2009. a
Mantua, N., Tohver, I., and Hamlet, A. F.: Impacts of Climate Change on Key Aspects of Freshwater Salmon Habitat in Washington State, The Washington Climate Change Impacts Assessment: Evaluating Washington's Future in a Changing Climate, University of Washington Climate Impacts Group, Seattle, WA, https://doi.org/10.7915/CIG6QZ23J, 2009. a, b
Mass, C.: The weather of the Pacific Northwest, University of Washington Press, Seattle, Washington, 34–35, 2015. a
Mentch, L. and Hooker, G.: Quantifying uncertainty in random forests via confidence intervals and hypothesis tests, J. Mach. Learn. Res., 17, 841–881, 2016. a
Mittermaier, M. P.: The potential impact of using persistence as a reference forecast on perceived forecast skill, Weather Forecast., 23, 1022–1031, 2008. a
Mote, P. W., Li, S., Lettenmaier, D. P., Xiao, M., and Engel, R.: Dramatic declines in snowpack in the western US, NPJ Climate and Atmospheric Science, 1, 1–6, 2018. a
Nicodemus, K. K.: Letter to the editor: On the stability and ranking of predictors from random forest variable importance measures, Brief. Bioinform., 12, 369–373, 2011. a
Pal, M.: Random forest classifier for remote sensing classification, Int. J. Remote Sens., 26, 217–222, 2005. a
Pan, M., Sheffield, J., Wood, E. F., Mitchell, K. E., Houser, P. R., Schaake, J. C., Robock, A., Lohmann, D., Cosgrove, B., Duan, Q., and Luo, L.: Snow process modeling in the North American Land Data Assimilation System (NLDAS): 2. Evaluation of model simulated snow water equivalent, J. Geophys. Res.-Atmos., 108, 8850, https://doi.org/10.1029/2003JD003994, 2003. a, b
Papacharalampous, G. A. and Tyralis, H.: Evaluation of random forests and Prophet for daily streamflow forecasting, Advances in Geosciences, 45, 201–208, 2018. a
Payne, J. T., Wood, A. W., Hamlet, A. F., Palmer, R. N., and Lettenmaier, D. P.: Mitigating the effects of climate change on the water resources of the Columbia River basin, Climatic Change, 62, 233–256, 2004. a
Pham, L. T.: Random Forest Streamflow Forecast (2020), GitHub, available at: https://github.com/leopham95/RandomForestStreamflowForecast, last access: 15 June 2020. a
Ralph, F., Dettinger, M., White, A., Reynolds, D., Cayan, D., Schneider, T., Cifelli, R., Redmond, K., Anderson, M., Gherke, F., and Jones, J.: A vision for future observations for western US extreme precipitation and flooding, Journal of Contemporary Water Research & Education, 153, 16–32, 2014. a, b
Regonda, S. K., Rajagopalan, B., Clark, M., and Pitlick, J.: Seasonal cycle shifts in hydroclimatology over the western United States, J. Climate, 18, 372–384, 2005. a
Safeeq, M., Mauger, G. S., Grant, G. E., Arismendi, I., Hamlet, A. F., and Lee, S.-Y.: Comparing large-scale hydrological model predictions with observed streamflow in the Pacific Northwest: effects of climate and groundwater, J. Hydrometeorol., 15, 2501–2521, 2014. a
Salathé Jr, E. P., Hamlet, A. F., Mass, C. F., Lee, S.-Y., Stumbaugh, M., and Steed, R.: Estimates of twenty-first-century flood risk in the Pacific Northwest based on regional climate model simulations, J. Hydrometeorol., 15, 1881–1899, 2014. a
Seibold, H., Bernau, C., Boulesteix, A.-L., and De Bin, R.: On the choice and influence of the number of boosting steps for high-dimensional linear Cox-models, Comput. Stat., 33, 1195–1215, 2018. a
Shortridge, J. E., Guikema, S. D., and Zaitchik, B. F.: Machine learning methods for empirical streamflow simulation: a comparison of model accuracy, interpretability, and uncertainty in seasonal watersheds, Hydrol. Earth Syst. Sci., 20, 2611–2628, https://doi.org/10.5194/hess-20-2611-2016, 2016. a, b
Tohver, I. M., Hamlet, A. F., and Lee, S.-Y.: Impacts of 21st-century climate change on hydrologic extremes in the Pacific Northwest region of North America, J. Am. Water Resour. As., 50, 1461–1476, 2014. a
Tyralis, H., Papacharalampous, G., and Langousis, A.: A brief review of random forests for water scientists and practitioners and their recent history in water resources, Water, 11, p. 910, 2019. a
U.S. Geological Survey: U.S. Geological Survey, 2019, National Hydrography Dataset (ver. USGS National Hydrography Dataset Best Resolution (NHD) for Hydrologic Unit (HU) 4 – 2001), available at: https://www.usgs.gov/core-science-systems/ngp/national-hydrography/access-national-hydrography-products (last access: 6 June 2020), 2020. a
Van Rijn, J. N. and Hutter, F.: Hyperparameter importance across datasets, in: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2367–2376, 2018. a
Vano, J. A., Nijssen, B., and Lettenmaier, D. P.: Seasonal hydrologic responses to climate change in the Pacific Northwest, Water Resour. Res., 51, 1959–1976, 2015. a
Wager, S., Hastie, T., and Efron, B.: Confidence intervals for random forests: The jackknife and the infinitesimal jackknife, J. Mach. Learn. Research, 15, 1625–1651, 2014. a
Wilcoxon, F., Katti, S., and Wilcox, R. A.: Critical values and probability levels for the Wilcoxon rank sum test and the Wilcoxon signed rank test, Selected tables in mathematical statistics, 1, 171–259, 1970. a
Zheng, X., Wang, Q., Zhou, L., Sun, Q., and Li, Q.: Predictive Contributions of Snowmelt and Rainfall to Streamflow Variations in the Western United States, Adv. Meteorol., 2018, p. 14, 2018. a
Model evaluation metrics suggest that RF performs better in snowmelt-driven watersheds. The largest improvements in forecasts compared to benchmark models are found among rainfall-driven watersheds. RF performance deteriorates with increases in catchment slope and soil sandiness. We note disagreement between two popular measures of RF variable importance and recommend jointly considering these measures with the physical processes under study.
Model evaluation metrics suggest that RF performs better in snowmelt-driven watersheds. The...