scispace - formally typeset
Open accessJournal ArticleDOI: 10.5194/HESS-25-1069-2021

Behind the scenes of streamflow model performance

02 Mar 2021-Hydrology and Earth System Sciences (Copernicus GmbH)-Vol. 25, Iss: 2, pp 1069-1095
Abstract: . Streamflow is often the only variable used to constrain hydrological models. In a previous international comparison study, eight research groups followed an identical protocol to calibrate a total of twelve hydrological models using observed streamflow of catchments within the Meuse basin. In the current study, we hypothesize that these twelve process-based models with similar streamflow performance have similar representations of internal states and fluxes. We test our hypothesis by comparing internal states and fluxes between models and we assess their plausibility using remotely-sensed products of evaporation, snow cover, soil moisture and total storage anomalies. Our results indicate that models with similar streamflow performance represent internal states and fluxes differently. Substantial dissimilarities between models are found for annual and seasonal evaporation and interception rates, the number of days per year with water stored as snow, the mean annual maximum snow storage and the size of the root-zone storage capacity. Relatively small root-zone storage capacities for several models lead to drying-out of the root-zone storage and significant reduction of evaporative fluxes each summer, which is not suggested by remotely-sensed estimates of evaporation and root-zone soil moisture. These differences in internal process representation imply that these models cannot all simultaneously be close to reality. Using remotely-sensed products, we could evaluate the plausibility of model representations only to some extent, as many of these internal variables remain unknown, highlighting the need for experimental research. We also encourage modelers to rely on multi-model and multi-parameter studies to reveal to decision-makers the uncertainties inherent to the heterogeneity of catchments and the lack of evaluation data.

... read more

Topics: Streamflow (53%), Snow (51%)
Citations
  More

9 results found


Open access
01 Apr 2015-
Abstract: Reproducibility and repeatability of experiments are the fundamental prerequisites that allow researchers to validate results and share hydrological knowledge, experi- ence and expertise in the light of global water management problems. Virtual laboratories offer new opportunities to en- able these prerequisites since they allow experimenters to share data, tools and pre-defined experimental procedures (i.e. protocols). Here we present the outcomes of a first col- laborative numerical experiment undertaken by five differ- ent international research groups in a virtual laboratory to address the key issues of reproducibility and repeatability. Moving from the definition of accurate and detailed experi- mental protocols, a rainfall-runoff model was independently applied to 15 European catchments by the research groups and model results were collectively examined through a web- based discussion. We found that a detailed modelling proto- col was crucial to ensure the comparability and reproducibil- ity of the proposed experiment across groups. Our results suggest that sharing comprehensive and precise protocols and running the experiments within a controlled environment (e.g. virtual laboratory) is as fundamental as sharing data and tools for ensuring experiment repeatability and reproducibil-

... read more

53 Citations


Open accessJournal ArticleDOI: 10.1016/J.ADVWATRES.2020.103667
Abstract: Twelve actual evaporation datasets are evaluated for their ability to improve the performance of the fully distributed mesoscale Hydrologic Model (mHM) The datasets consist of satellite-based diagnostic models (MOD16A2, SSEBop, ALEXI, CMRSET, SEBS), satellite-based prognostic models (GLEAM v32a, GLEAM v33a, GLEAM v32b, GLEAM v33b), and reanalysis (ERA5, MERRA-2, JRA-55) Four distinct multivariate calibration strategies (basin-average, pixel-wise, spatial bias-accounting and spatial bias-insensitive) using actual evaporation and streamflow are implemented, resulting in 48 scenarios whose results are compared with a benchmark model calibrated solely with streamflow data A process-diagnostic approach is adopted to evaluate the model responses with in-situ data of streamflow and independent remotely sensed data of soil moisture from ESA-CCI and terrestrial water storage from GRACE The method is implemented in the Volta River basin, which is a data scarce region in West Africa, for the period from 2003 to 2012 Results show that the evaporation datasets have a good potential for improving model calibration, but this is dependent on the calibration strategy All the multivariate calibration strategies outperform the streamflow-only calibration The highest improvement in the overall model performance is obtained with the spatial bias-accounting strategy (+29%), followed by the spatial bias-insensitive strategy (+26%) and the pixel-wise strategy (+24%), while the basin-average strategy (+20%) gives the lowest improvement On average, using evaporation data in addition to streamflow for model calibration decreases the model performance for streamflow (-7%), which is counterbalance by the increase in the performance of the terrestrial water storage (+11%), temporal dynamics of soil moisture (+6%) and spatial patterns of soil moisture (+89%) In general, the top three best performing evaporation datasets are MERRA-2, GLEAM v33a and SSEBop, while the bottom three datasets are MOD16A2, SEBS and ERA5 However, performances of the evaporation products diverge according to model responses and across climatic zones These findings open up avenues for improving process representation of hydrological models and advancing the spatiotemporal prediction of floods and droughts under climate and land use changes

... read more

Topics: Hydrological modelling (52%), Streamflow (50%)

17 Citations


Open access
01 Apr 2018-
Abstract: The calibration of hydrological models without streamflow observations is problematic, and the simultaneous, combined use of remotely sensed products for this purpose has not been exhaustively tested thus far. Our hypothesis is that the combined use of products can (1) reduce the parameter search space and (2) improve the representation of internal model dynamics and hydrological signatures. Five different conceptual hydrological models were applied to 27 catchments across Europe. A parameter selection process, similar to a likelihood weighting procedure, was applied for 1,023 possible combinations of 10 different data sources, ranging from using 1 to all 10 of these products. Distances between the two empirical distributions of model performance metrics with and without using a specific product were determined to assess the added value of a specific product. In a similar way, the performance of the models to reproduce 27 hydrological signatures was evaluated relative to the unconstrained model. Significant reductions in the parameter space were obtained when combinations included Advanced Microwave Scanning Radiometer - Earth Observing System and Advanced Scatterometer soil moisture, Gravity Recovery and Climate Experiment total water storage anomalies, and, in snow-dominated catchments, the Moderate Resolution Imaging Spectroradiometer snow cover products. The evaporation products of Land Surface Analysis - Satellite Application Facility and MOD16 were less effective for deriving meaningful, well-constrained posterior parameter distributions. The hydrological signature analysis indicated that most models profited from constraining with an increasing number of data sources. Concluding, constraining models with multiple data sources simultaneously was shown to be valuable for at least four of the five hydrological models to determine model parameters in absence of streamflow.

... read more

7 Citations


Open accessPosted ContentDOI: 10.5194/HESS-25-5839-2021
Yang Yang1, Ting Fong May Chui1Institutions (1)
Abstract: . Sustainable urban drainage systems (SuDS) are decentralized stormwater management practices that mimic natural drainage processes. The hydrological processes of SuDS are often modeled using process-based models. However, it can require considerable effort to set up these models. This study thus proposes a machine learning (ML) method to directly learn the statistical correlations between the hydrological responses of SuDS and the forcing variables at sub-hourly timescales from observation data. The proposed methods are applied to two SuDS catchments with different sizes, SuDS practice types, and data availabilities in the USA for discharge prediction. The resulting models have high prediction accuracies (Nash–Sutcliffe efficiency, NSE, >0.70 ). ML explanation methods are then employed to derive the basis of each ML prediction, based on which the hydrological processes being modeled are then inferred. The physical realism of the inferred hydrological processes is then compared to that would be expected based on the domain-specific knowledge of the system being modeled. The inferred processes of some models, however, are found to be physically implausible. For instance, negative contributions of rainfall to runoff have been identified in some models. This study further empirically shows that an ML model's ability to provide accurate predictions can be uncorrelated with its ability to offer plausible explanations to the physical processes being modeled. Finally, this study provides a high-level overview of the practices of inferring physical processes from the ML modeling results and shows both conceptually and empirically that large uncertainty exists in every step of the inference processes. In summary, this study shows that ML methods are a useful tool for predicting the hydrological responses of SuDS catchments, and the hydrological processes inferred from modeling results should be interpreted cautiously due to the existence of large uncertainty in the inference processes.

... read more

7 Citations


Open accessJournal ArticleDOI: 10.5194/HESS-25-5287-2021
Abstract: . Distributed hydrological models rely on hydrography data such as flow direction, river length, slope and width. For large-scale applications, many of these models still rely on a few flow direction datasets, which are often manually derived. We propose the Iterative Hydrography Upscaling (IHU) method to upscale high-resolution flow direction data to the typically coarser resolutions of distributed hydrological models. The IHU aims to preserve the upstream–downstream relationship of river structure, including basin boundaries, river meanders and confluences, in the D8 format, which is commonly used to describe river networks in models. Additionally, it derives representative sub-grid river length and slope parameters, which are required for resolution-independent model results. We derived the multi-resolution MERIT Hydro IHU dataset at resolutions of 30 arcsec ( ∼ 1 km), 5 arcmin ( ∼ 10 km) and 15 arcmin ( ∼ 30 km) by applying IHU to the recently published 3 arcsec MERIT Hydro data. Results indicate improved accuracy of IHU at all resolutions studied compared to other often-applied upscaling methods. Furthermore, we show that MERIT Hydro IHU minimizes the errors made in the timing and magnitude of simulated peak discharge throughout the Rhine basin compared to simulations at the native data resolutions. As the method is open source and fully automated, it can be applied to other high-resolution hydrography datasets to increase the accuracy and enhance the uptake of new datasets in distributed hydrological models in the future.

... read more

4 Citations


References
  More

122 results found


Open access
01 Jan 1979-
Abstract: A hydrological forecasting model is presented that attempts to combine the important distributed effects of channel network topology and dynamic contributing areas with the advantages of simple lumped parameter basin models. Quick response flow is predicted from a storage/contributing area relationship derived analytically from the topographic structure of a unit within a basin. Average soil water response is represented by a constant leakage infiltration store and an exponential subsurface water store. A simple non-linear routing procedure related to the link frequency distribution of the channel network completes the model and allows distinct basin sub-units, such as headwater and sideslope areas to be modelled separately. The model parameters are physically based in the sense that they may be determined directly by measurement and the model may be used at ungauged sites. Procedures for applying the model and tests with data from the Crimple Beck basin are described. Using only measured and estimated parameter values, without optimization, the model makes satisfactory predictions of basin response. The modular form of the model structure should allow application over a range of small and medium sized basins while retaining the possibility of including more complex model components when suitable data are available.

... read more

5,714 Citations


Abstract: In an introductory review it is reemphasized that the large-scale parameterization of the surface fluxes of sensible and latent heat is properly expressed in terms of energetic considerations over land while formulas of the bulk aerodynamic type are most suitahle over the sea. A general framework is suggested. Data from a number of saturated land sites and open water sites in the absence of advection suggest a widely applicable formula for the relationship between sensible and latent heat fluxes. For drying land surfaces, we assume that the evaporation rate is given by the same formula for evaporation multiplied by a factor. This factor is found to remain at unity while an amount of water, varying from one site to another, is evaporated. Following this a linear decrease sets in, reducing the evaporation rate to zero after a further 5 cm of evaporation, the same at several sites examined.

... read more

Topics: Latent heat (61%), Potential evaporation (60%), Evaporation (58%) ... read more

5,380 Citations


Open accessJournal ArticleDOI: 10.1080/02626667909491834
Keith Beven, Mike Kirkby1Institutions (1)
Abstract: A hydrological forecasting model is presented that attempts to combine the important distributed effects of channel network topology and dynamic contributing areas with the advantages of simple lum...

... read more

4,163 Citations


Journal ArticleDOI: 10.1002/HYP.3360060305
Keith Beven1, Andrew Binley1Institutions (1)
Abstract: This paper describes a methodology for calibration and uncertainty estimation of distributed models based on generalized likelihood measures. The GLUE procedure works with multiple sets of parameter values and allows that, within the limitations of a given model structure and errors in boundary conditions and field observations, different sets of values may be equally likely as simulators of a catchment. Procedures for incorporating different types of observations into the calibration; Bayesian updating of likelihood values and evaluating the value of additional observations to the calibration process are described. The procedure is computationally intensive but has been implemented on a local parallel processing computer.

... read more

Topics: Calibration (statistics) (59%), Swat-CUP (57%), GLUE (51%) ... read more

3,857 Citations


Journal ArticleDOI: 10.13031/2013.26773
Abstract: MEASURED lysimeter evapotranspiration of Alta fescue grass (a cool season grass) is taken as an index of reference crop evapotranspiration (ETo). An equation is presented that estimates ETo from measured values of daily or mean values of maximum and minimum temperature. This equation is compared with various other methods for estimating ETo. The equation was developed using eight years of daily lysimeter data from Davis, California and used to estimate values of ETo for other locations. Comparisons with other methods with measured cool season grass evapotranspiration at Aspendale, Australia; Lompoc, California; and Seabrook, New Jersey; with lysimeter data from Damin, Haiti; and with the modified Penman for various locations in Bangladesh indicated that the method usually does not require local calibration and that the estimated values are probably as reliable and useable as those from the other estimating methods used for comparison. Considering the scarcity of complete and reliable climatic data for estimating crop water requirements in developing countries, this proposed method can do much to improve irrigation planning design and scheduling in the developing countries.

... read more

Topics: Blaney–Criddle equation (62%), Lysimeter (61%), Crop coefficient (59%) ... read more

2,768 Citations