Votre recherche

Abstract This paper focuses on evaluating the uncertainty of three common regionalization methods for predicting continuous streamflow in ungauged basins. A set of 268 basins covering 1.6 million km 2 in the province of Quebec was used to test the regionalization strategies. The multiple linear regression, spatial proximity, and physical similarity approaches were evaluated on the catchments using a leave‐one‐out cross‐validation scheme. The lumped conceptual HSAMI hydrological model was used throughout the study. A bootstrapping method was chosen to further estimate uncertainty due to parameter set selection for each of the parameter set/regionalization method pairs. Results show that parameter set selection can play an important role in regionalization method performance depending on the regionalization methods (and their variants) used and that equifinality does not contribute significantly to the overall uncertainty witnessed throughout the regionalization methods applications. Regression methods fail to consistently assign behavioral parameter sets to the pseudoungauged basins (i.e., the ones left out). Spatial proximity and physical similarity score better, the latter being the best. It is also shown that combining either physical similarity or spatial proximity with the multiple linear regression method can lead to an even more successful prediction rate. However, even the best methods were shown to be unreliable to an extent, as successful prediction rates never surpass 75%. Finally, this paper shows that the selection of catchment descriptors is crucial to the regionalization strategies' performance and that for the HSAMI model, the optimal number of donor catchments for transferred parameter sets lies between four and seven. , Key Points Uncertainty can be limited in regionalization Physical similarity method is best, followed by spatial proximity Regression‐augmented methods can yield better performance

Abstract. Efficient adaptation strategies to climate change require the estimation of future impacts and the uncertainty surrounding this estimation. Over- or underestimating future uncertainty may lead to maladaptation. Hydrological impact studies typically use a top-down approach in which multiple climate models are used to assess the uncertainty related to the climate model structure and climate sensitivity. Despite ongoing debate, impact modelers have typically embraced the concept of “model democracy”, in which each climate model is considered equally fit. The newer Coupled Model Intercomparison Project Phase 6 (CMIP6) simulations, with several models showing a climate sensitivity larger than that of Phase 5 (CMIP5) and larger than the likely range based on past climate information and understanding of planetary physics, have reignited the model democracy debate. Some have suggested that “hot” models be removed from impact studies to avoid skewing impact results toward unlikely futures. Indeed, the inclusion of these models in impact studies carries a significant risk of overestimating the impact of climate change. This large-sample study looks at the impact of removing hot models on the projections of future streamflow over 3107 North American catchments. More precisely, the variability in future projections of mean, high, and low flows is evaluated using an ensemble of 19 CMIP6 general circulation models (GCMs), 5 of which are deemed hot based on their global equilibrium climate sensitivity (ECS). The results show that the reduced ensemble of 14 climate models provides streamflow projections with reduced future variability for Canada, Alaska, the Southeast US, and along the Pacific coast. Elsewhere, the reduced ensemble has either no impact or results in increased variability in future streamflow, indicating that global outlier climate models do not necessarily provide regional outlier projections of future impacts. These results emphasize the delicate nature of climate model selection, especially based on global fitness metrics that may not be appropriate for local and regional assessments.

Abstract Streamflow sensitivity to different hydrologic processes varies in both space and time. This sensitivity is traditionally evaluated for the parameters specific to a given hydrologic model simulating streamflow. In this study, we apply a novel analysis over more than 3000 basins across North America considering a blended hydrologic model structure, which includes not only parametric, but also structural uncertainties. This enables seamless quantification of model process sensitivities and parameter sensitivities across a continuous set of models. It also leads to high-level conclusions about the importance of water cycle components on streamflow predictions, such as quickflow being the most sensitive process for streamflow simulations across the North American continent. The results of the 3000 basins are used to derive an approximation of sensitivities based on physiographic and climatologic data without the need to perform expensive sensitivity analyses. Detailed spatio-temporal inputs and results are shared through an interactive website.

Abstract In spring 2011, an unprecedented flood hit the complex eastern United States (U.S.)–Canada transboundary Lake Champlain–Richelieu River (LCRR) Basin, destructing properties and inducing negative impacts on agriculture and fish habitats. The damages, covered by the Governments of Canada and the U.S., were estimated to C$90M. This natural disaster motivated the study of mitigation measures to prevent such disasters from reoccurring. When evaluating flood risks, long‐term evolving climate change should be taken into account to adopt mitigation measures that will remain relevant in the future. To assess the impacts of climate change on flood risks of the LCRR basin, three bias‐corrected multi‐resolution ensembles of climate projections for two greenhouse gas concentration scenarios were used to force a state‐of‐the‐art, high‐resolution, distributed hydrological model. The analysis of the hydrological simulations indicates that the 20‐year return period flood (corresponding to a medium flood) should decrease between 8% and 35% for the end of the 21st Century (2070–2099) time horizon and for the high‐emission scenario representative concentration pathway (RCP) 8.5. The reduction in flood risks is explained by a decrease in snow accumulation and an increase in evapotranspiration expected with the future warming of the region. Nevertheless, due to the large climate inter‐annual variability, short‐term flood probabilities should remain similar to those experienced in the recent past.

Abstract. This study investigates the ability of long short-term memory (LSTM) neural networks to perform streamflow prediction at ungauged basins. A set of state-of-the-art, hydrological model-dependent regionalization methods are applied to 148 catchments in northeast North America and compared to an LSTM model that uses the exact same available data as the hydrological models. While conceptual model-based methods attempt to derive parameterizations at ungauged sites from other similar or nearby catchments, the LSTM model uses all available data in the region to maximize the information content and increase its robustness. Furthermore, by design, the LSTM does not require explicit definition of hydrological processes and derives its own structure from the provided data. The LSTM networks were able to clearly outperform the hydrological models in a leave-one-out cross-validation regionalization setting on most catchments in the study area, with the LSTM model outperforming the hydrological models in 93 % to 97 % of catchments depending on the hydrological model. Furthermore, for up to 78 % of the catchments, the LSTM model was able to predict streamflow more accurately on pseudo-ungauged catchments than hydrological models calibrated on the target data, showing that the LSTM model's structure was better suited to convert the meteorological data and geophysical descriptors into streamflow than the hydrological models even calibrated to those sites in these cases. Furthermore, the LSTM model robustness was tested by varying its hyperparameters, and still outperformed hydrological models in regionalization in almost all cases. Overall, LSTM networks have the potential to change the regionalization research landscape by providing clear improvement pathways over traditional methods in the field of streamflow prediction in ungauged catchments.

Abstract Ensemble forecasting applied to the field of hydrology is currently an established area of research embracing a broad spectrum of operational situations. This work catalogs the various pathways of ensemble streamflow forecasting based on an exhaustive review of more than 700 studies over the last 40 years. We focus on the advanced state of the art in the model‐based (dynamical) ensemble forecasting approaches. Ensemble streamflow prediction systems are categorized into three leading classes: statistics‐based streamflow prediction systems, climatology‐based ensemble streamflow prediction systems and numerical weather prediction‐based hydrological ensemble prediction systems. For each ensemble approach, technical information, as well as details about its strengths and weaknesses, are provided based on a critical review of the studies listed. Through this literature review, the performance and uncertainty associated with the ensemble forecasting systems are underlined from both operational and scientific viewpoints. Finally, the remaining key challenges and prospective future research directions are presented, notably through hybrid dynamical ‐ statistical learning approaches, which obviously present new challenges to be overcome in order to allow the successful employment of ensemble streamflow forecasting systems in the next decades. Targeting students, researchers and practitioners, this review provides a detailed perspective on the major features of an increasingly important area of hydrological forecasting. , Key Points This work summarizes the 40 years of research in the generation of streamflow forecasts based on an exhaustive review of studies Ensemble prediction systems are categorized into three classes: statistics‐based, climatology‐based and numerical weather prediction‐based hydrological ensemble prediction systems For each ensemble forecasting system, thorough technical information is provided

Abstract. Model intercomparison studies are carried out to test and compare the simulated outputs of various model setups over the same study domain. The Great Lakes region is such a domain of high public interest as it not only resembles a challenging region to model with its transboundary location, strong lake effects, and regions of strong human impact but is also one of the most densely populated areas in the USA and Canada. This study brought together a wide range of researchers setting up their models of choice in a highly standardized experimental setup using the same geophysical datasets, forcings, common routing product, and locations of performance evaluation across the 1×106 km2 study domain. The study comprises 13 models covering a wide range of model types from machine-learning-based, basin-wise, subbasin-based, and gridded models that are either locally or globally calibrated or calibrated for one of each of the six predefined regions of the watershed. Unlike most hydrologically focused model intercomparisons, this study not only compares models regarding their capability to simulate streamflow (Q) but also evaluates the quality of simulated actual evapotranspiration (AET), surface soil moisture (SSM), and snow water equivalent (SWE). The latter three outputs are compared against gridded reference datasets. The comparisons are performed in two ways – either by aggregating model outputs and the reference to basin level or by regridding all model outputs to the reference grid and comparing the model simulations at each grid-cell. The main results of this study are as follows: The comparison of models regarding streamflow reveals the superior quality of the machine-learning-based model in the performance of all experiments; even for the most challenging spatiotemporal validation, the machine learning (ML) model outperforms any other physically based model. While the locally calibrated models lead to good performance in calibration and temporal validation (even outperforming several regionally calibrated models), they lose performance when they are transferred to locations that the model has not been calibrated on. This is likely to be improved with more advanced strategies to transfer these models in space. The regionally calibrated models – while losing less performance in spatial and spatiotemporal validation than locally calibrated models – exhibit low performances in highly regulated and urban areas and agricultural regions in the USA. Comparisons of additional model outputs (AET, SSM, and SWE) against gridded reference datasets show that aggregating model outputs and the reference dataset to the basin scale can lead to different conclusions than a comparison at the native grid scale. The latter is deemed preferable, especially for variables with large spatial variability such as SWE. A multi-objective-based analysis of the model performances across all variables (Q, AET, SSM, and SWE) reveals overall well-performing locally calibrated models (i.e., HYMOD2-lumped) and regionally calibrated models (i.e., MESH-SVS-Raven and GEM-Hydro-Watroute) due to varying reasons. The machine-learning-based model was not included here as it is not set up to simulate AET, SSM, and SWE. All basin-aggregated model outputs and observations for the model variables evaluated in this study are available on an interactive website that enables users to visualize results and download the data and model outputs.