Elsevier

Journal of Hydrology

Volume 599, August 2021, 126394
Journal of Hydrology

Research papers
Physically consistent conceptual rainfall–runoff model for urbanized catchments

https://doi.org/10.1016/j.jhydrol.2021.126394Get rights and content

Highlights

  • We modified an hourly conceptual model to account for urban features.

  • We attempted to explicitly account for rapid runoff from impervious surfaces.

  • Modifications improved model performances over a sample of 273 urbanized catchments.

  • Newly added parameters were highly correlated with mean catchment imperviousness.

Abstract

Hydrological models should be tested and evaluated for a wide variety of levels of urbanization before they are used to predict the impact of urbanization on catchment behavior. In this study, we illustrate a top–down approach of modifying step by step an hourly conceptual model structure (GR4H) to account for urbanization features. Modifying the original model structure included accounting explicitly for runoff from impervious surfaces by bypassing the soil moisture reservoir and varying the partitioning between quick flow and slow flow. These adaptations were chosen based on the reported specificities of urbanized catchments, namely, decreasing infiltration, increasing runoff, and fast runoff dynamics. Using a split-sample test, the relevance of each modification with regard to the reproduction of catchment response (i.e., observed streamflow) was assessed for a large sample of 273 urbanized catchments, located in France and the United States, for which mean total impervious area (TIA) varied between 0.05 and 0.59. Six continuous and three event-based criteria were used, and two statistical tests were applied to assess the significance of improvements. Results showed the following: (i) Tested modifications improved the ability of the model to reproduce the catchment response, especially high flows and observed streamflow amid dry conditions. (ii) Event-based evaluation using more than 45,000 events showed an improvement in predicting the event peak flow and event runoff volume, whereas no significant improvements were obtained in predicting the timing of peak flow. (iii) Newly added parameters were moderately to highly correlated with TIA, especially the calibrated proportion of impervious surfaces, which is promising as a hydrological validation of estimated urbanization measures from land cover. The tested modifications improved both the representation of urbanization processes and the reproduction of the observed streamflow, yielding a simple and credible model for predicting the impact of future urbanization scenarios on catchment response.

Introduction

There is strong evidence that urbanization modifies the hydrological behavior of catchments (Braud et al., 2013, Diem et al., 2018, Fletcher et al., 2013, Leopold, 1968, McGrane, 2016, Miller and Hess, 2017). Nonetheless, predicting and quantifying the impact of urbanization on the rainfall–runoff relationship at the catchment scale is still a challenge (Oudin et al., 2018, Redfern et al., 2016). To this end, two types of approaches are generally applied (Braud et al., 2013, Salavati et al., 2016): a statistical approach and a modeling approach. The statistical approach seeks to either (i) identify temporal trends in the behavior of a catchment provided that long periods of hydroclimatic data are available across the urbanization period over the catchment area (Haase, 2009), or (ii) compare the behavior of an urbanized catchment with a non-urbanized one that has similar climatic and geomorphological characteristics (e.g., slope, elevation, lithology), as in paired-catchment experiments (Bonneau et al., 2018, Prosdocimi et al., 2015). These approaches are useful to detect and quantify past changes, but they do not provide physically sound links between hydrological processes and urbanization, which are necessary to reliably predict how the catchment behavior would be altered under future urbanization scenarios. In this respect, the modeling approach is advantageous because it backs the statistical method by providing a synthesis of the catchment behavior via the model parameters. This helps not only to detect change in catchment behavior (Pathiraja et al., 2018, Saadi et al., 2020a), but also to create hydrological scenarios that correspond to urbanization scenarios (De Niel et al., 2020, McIntyre et al., 2014, Niehoff et al., 2002, Sanzana et al., 2019).

There is a spectrum of modeling tools along which a compromise is made between detailed representation of the spatial variability of hydrological processes and model simplicity (Hrachowitz and Clark, 2017, McIntyre et al., 2014, Salvadore et al., 2015). Distributed models that use small-scale equations to represent the main hydrological processes are applied to account for the highly heterogeneous nature of hydrological processes in urban areas, which is accompanied by rapid dynamics of runoff generation on impervious surfaces (Cristiano et al., 2017, Ogden et al., 2011, Salvadore et al., 2015). Their application is also advocated because they explicitly represent catchment properties, which enables a direct assessment of the impact of urbanization by changing model parameters and scaling up the impact (Beven, 2002, Bronstert et al., 2002). This generally leads to heavily parametrized model structures (e.g., Cuo et al., 2008, Jankowfsky et al., 2014, Jia et al., 2001, Sanzana et al., 2019, Stavropulos-Laffaille et al., 2018), which impedes testing their ability to reproduce the rainfall–runoff relationship for many catchments with a mix of rural and urban areas. Also, constraining these models requires a large volume of data, available only for a handful of monitored catchments (Petrucci and Bonhomme, 2014, Rodriguez et al., 2003).

At the other side of the spectrum, lumped, conceptual models that rely on relatively simple parametrization of water fluxes to describe the catchment-scale manifestation of small-scale heterogeneities are more easy to implement and less data-demanding. Thus, they offer the possibility of testing their robustness under different climate and land-cover situations, as has been shown by numerous model testing experiments using large samples of catchments (Gupta et al., 2014). Perrin et al. (2001) compared 19 lumped daily models across 429 catchments located in France and the United States to discuss the issue of model complexity. Le Moine et al. (2007) tested different model modifications to account for daily inter-catchment groundwater flows on 1040 catchments located in France. At the hourly time step, van Esse et al. (2013) compared the fixed and the flexible conceptual modeling approaches using 237 catchments located in France. Recently, Ficchì et al. (2019) improved the consistency of the fluxes of a conceptual model across multiple sub-daily time steps using a set of 240 French catchments. These and many other studies focusing on model development, regionalization, or evaluation, as reviewed by Gupta et al. (2014), were mostly concerned with non-urbanized catchments, a fact that is mirrored by the existing large samples of catchment-scale hydroclimatic data that did not necessarily focus on (or even excluded) highly urbanized cases (Addor et al., 2020).

Despite their relative simplicity, most of existing applications of conceptual models to urbanized catchments were limited to only few places at each time (Salvadore et al., 2015), highlighting the need for intensive testing of conceptual models using many urbanized catchments. Huang et al. (2008) quantified the impact of increasing imperviousness in the Wu-Tu catchment, Taiwan, on peak flow recurrence using a combination of Nash model and the Curve Number (CN) method. Dotto et al. (2011) conducted a sensitivity analysis of MUSIC and KAREN models on five urban Australian catchments with different imperviousness levels. Recently, De Niel et al. (2020) developed a methodology based on the NAM model to incorporate and project the impact of rapid urbanization on the hydrological behavior of two Belgian catchments, and Fidal and Kjeldsen (2020a) improved the conceptual model URMOD to account for soil moisture across 28 urbanized catchments in United Kingdom. Nonetheless, direct projection of the impact of urbanization on catchment behavior using conceptual tools is still undermined by the lack of explicit links between their structures and landscape properties. One could cite the CN method as an exception, but its use for continuous applications is impeded by inconsistencies in its formulation (Michel et al., 2005), and its ability to predict the impact of land-use changes (including urbanization) on catchment response is generally unverified (Ogden et al., 2017).

A promising way of developing robustly tested models while at the same time enabling an explicit link between model structure and urbanization features lies in modifying already existing conceptual models to incorporate some physical properties of catchments (Euser et al., 2015, Gharari et al., 2014, Hrachowitz et al., 2014, Kirchner, 2006). This can be guided by learning about the behavioral specificities of urbanized catchments using hydrological signatures (Gupta et al., 2008, McMillan et al., 2011, Saadi et al., 2020a). An example was shown by Kjeldsen et al. (2013), who modified a non-urban model structure to account for urbanization features. They explicitly linked the model parameters (up to four) to the proportion of urban cover in the catchment, but their test concerned only seven catchments located in the United Kingdom. Another study by Hamel and Fletcher (2014) illustrated an improvement of the representation of low-flow components (interflow and groundwater flow) by gradually adding reservoirs to distinguish the contributions of the different parts of an urbanized catchment (including the impervious part of stream area and riparian zone). Their model modifications helped them test different stormwater management strategies to analyze their role in mitigating the impact of catchment imperviousness on baseflow. They defined their model modifications by analyzing the hydrological signatures of the McMahons Creek catchment in Australia. Both studies illustrate a top–down approach of improving the representation of urban hydrological processes in conceptual models, although their application was constrained to small sets of catchments. Thus, an attempt is required to balance the simplicity of conceptual models with robust model assessment using large samples of catchments, in order to detect their weaknesses and increase their credibility (Andréassian et al., 2009, Gupta et al., 2014, Klemeš, 1986), especially in the case of catchments with changing landscape.

Our research motivation stems from the fact that currently available models for urbanized catchments lack intensive testing on large number of cases. Among the 43 model applications in urbanized catchments reported in the exhaustive literature review by Salvadore et al. (2015), none used more than 6 catchments to develop/test their modeling tool. We argue that using a large sample of urbanized catchments is a necessity to (i) advance our general understanding of how catchments with a mix of non-urban (or pervious) and urban (or impervious) covers behave hydrologically, by building on what we already know from non-urbanized ones, and thus (ii) make a credible extrapolation of catchment behavior under future urban planning schemes. Simple conceptual models developed for non-urbanized catchments have been proved to be flexible enough to adapt their model parameters from rural to urbanized contexts since the hydrological processes are not that different (Fletcher et al., 2013, Redfern et al., 2016, Saadi et al., 2020a). Adapting their structures to urbanized environments should fulfill the following requirements: (i) an equal or improved ability (relative to the original conceptual model) to simulate the response of urbanized catchments (Fidal and Kjeldsen, 2020b), and (ii) more explicit links between the model parameters/structure and the urban characteristics of the catchments in order to enhance the physical consistency of the conceptual models with respect to landscape specificities (Gharari et al., 2014, Hrachowitz et al., 2014).

To fill this gap, we used a large sample of 273 urbanized catchments located in France and the United States, for which the mean total impervious area (TIA) ranged between 0.05 and 0.59. We conducted a step-by-step modification of the hourly non-urban GR4H model (Ficchì et al., 2019). Each modification was evaluated using a set of six continuous and three event-based evaluation criteria, and two statistical tests were applied to evaluate the statistical significance of each improvement with regard to the original non-urban model structure (Fidal and Kjeldsen, 2020b). By the modifications, we aimed at improving the physical consistency of the original conceptual structure by (i) testing modifications that were in agreement with the behavioral specificities of urbanized catchments, and (ii) comparing the added model parameters with observable urban characteristics.

This paper is organized as follows: In Section 2.1, we present the catchment set. Section 2.2 details the tested modifications of model structure to account for urbanization, and Section 2.3 describes the framework of their calibration and evaluation. Results are presented in Section 3, followed by a discussion and some perspectives in Section 4.

Section snippets

Catchment sample

We selected a large sample of 273 urbanized catchments, located in the United States (US) and France, based on four criteria (Saadi et al., 2019):

  • (1)

    Availability of at least 8 years of hourly streamflow, precipitation, and daily temperature between 1997 and 2017. We considered that a minimum of 4 years are required for model calibration (Merz et al., 2009, Perrin et al., 2007).

  • (2)

    Limited snow influence, as snow melting was not addressed in the tested model.

  • (3)

    Limited impact of large artificial

Calibration performances and distributions of calibrated parameters

Calibrating more parameters resulted in improved calibration performances. Moreover, accounting for TIA resulted in better calibration performances for the majority of catchments, even with no additional calibrated parameters (MU4H vs. GR4H), as shown by the distributions of RKGESQ in Fig. 4. The median reduction of error ranged between 2% (RKGESQ = 0.01) and 11% (RKGESQ = 0.06), and improvements were observed for the majority of cases. Calibrating only X5 yielded higher calibration

Improvement of model performances

The attempted modifications aimed not only to enhance the link between model structure and urbanization features but also to improve the simulation of observed streamflow over the 273 urbanized catchments. Fig. 7, Fig. 8 show relatively limited but significant improvements in evaluation scores, which might not be convincing enough. First, a comparison with existing studies can guide the interpretation. Le Moine et al. (2007) compared a number of strategies to account for groundwater–surface

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

The authors thank the Handling Editor and the four anonymous reviewers for their constructive comments and suggestions which helped improve the manuscript.

References (113)

  • J. Fidal et al.

    Accounting for soil moisture in rainfall-runoff modelling of urban areas

    J. Hydrol.

    (2020)
  • T.D. Fletcher et al.

    Understanding, management and modelling of urban hydrology and its consequences for receiving waters: A state of the art

    Adv. Water Resour.

    (2013)
  • C. Furusho et al.

    Adapting the coupled hydrological model ISBA-TOPMODEL to the long-term hydrological cycles of suburban rivers: Evaluation and sensitivity analysis

    J. Hydrol.

    (2013)
  • H.V. Gupta et al.

    Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling

    J. Hydrol.

    (2009)
  • D. Haase

    Effects of urbanisation on the water balance – A long-term trajectory

    Environ. Impact Assess. Rev.

    (2009)
  • C. Homer et al.

    Completion of the 2011 national land cover database for the conterminous United States – Representing a decade of land cover change information

    Photogramm. Eng. Remote Sens.

    (2015)
  • S.-Y. Huang et al.

    Identifying peak-imperviousness-recurrence relationships on a growing-impervious watershed, Taiwan

    Journal of Hydrology

    (2008)
  • S. Jankowfsky et al.

    Assessing anthropogenic influence on the hydrology of small peri-urban catchments: Development of the object-oriented PUMMA model by integrating urban and rural hydrological models

    J. Hydrol.

    (2014)
  • Y. Mei et al.

    A hydrograph separation method based on information from rainfall and runoff records

    J. Hydrol.

    (2015)
  • R. Merz et al.

    Regionalisation of catchment model parameters

    J. Hydrol.

    (2004)
  • J.D. Miller et al.

    Urbanisation impacts on storm runoff along a rural-urban gradient

    J. Hydrol.

    (2017)
  • J.E. Nash et al.

    River flow forecasting through conceptual models part I — A discussion of principles

    J. Hydrol.

    (1970)
  • D. Niehoff et al.

    Land-use impacts on storm-runoff generation: Scenarios of land-use change and simulation of hydrological response in a meso-scale catchment in SW-Germany

    J. Hydrol Adv. Flood Res.

    (2002)
  • L. Oudin et al.

    Which potential evapotranspiration input for a lumped rainfall–runoff model?

    J. Hydrol.

    (2005)
  • L. Oudin et al.

    Hydrological impacts of urbanization at the catchment scale

    J. Hydrol.

    (2018)
  • C. Perrin et al.

    Improvement of a parsimonious model for streamflow simulation

    J. Hydrol.

    (2003)
  • C. Perrin et al.

    Does a large number of parameters enhance model performance? Comparative assessment of common catchment model structures on 429 catchments

    J. Hydrol.

    (2001)
  • G. Petrucci et al.

    The dilemma of spatial representation for urban hydrology semi-distributed modelling: Trade-offs among complexity, calibration and geographical data

    J. Hydrol.

    (2014)
  • F. Rodriguez et al.

    Surface runoff in urban catchments: Morphological identification of unit hydrographs from urban databanks

    J. Hydrol.

    (2003)
  • B. Salavati et al.

    Modeling approaches to detect land-use changes: Urbanization analyzed on a set of 43 US catchments

    J. Hydrol.

    (2016)
  • E. Salvadore et al.

    Hydrological modelling of urbanized catchments: A review and future directions

    J. Hydrol.

    (2015)
  • N. Addor et al.

    Large-sample hydrology: recent progress, guidelines for new datasets and grand challenges

    Hydrol. Sci. J.

    (2020)
  • R.G. Allen et al.

    Crop evapotranspiration – Guidelines for computing crop water requirements

    FAO Irrig. Drain. Pap.

    (1998)
  • W.M. Alley et al.

    Effective impervious area in urban runoff modeling

    J. Hydraul. Eng.

    (1983)
  • V. Andréassian et al.

    All that glitters is not gold: the case of calibrating hydrological models

    Hydrol. Process.

    (2012)
  • V. Andréassian et al.

    HESS Opinions “Crash tests for a standardized evaluation of hydrological models”

    Hydrol. Earth Syst. Sci.

    (2009)
  • C.L. Arnold et al.

    Impervious surface coverage: The emergence of a key environmental indicator

    J. Am. Plann. Assoc.

    (1996)
  • H.E. Beck et al.

    Present and future Köppen-Geiger climate classification maps at 1-km resolution

    Sci. Data

    (2018)
  • K. Beven

    Towards an alternative blueprint for a physically based digitally simulated hydrologic response modelling system

    Hydrol. Process.

    (2002)
  • T. Blume et al.

    Rainfall—runoff response, event-based runoff coefficients and hydrograph separation

    Hydrol. Sci. J.

    (2007)
  • D.B. Booth et al.

    Urbanization of aquatic systems: Degradation thresholds, stormwater detection, and the limits of mitigation

    JAWRA J. Am. Water Resour. Assoc.

    (1997)
  • M.J. Boyd et al.

    Pervious and impervious runoff in urban catchments

    Hydrol. Sci. J.

    (1993)
  • E. Brabec et al.

    Impervious surfaces and water quality: A review of current literature and its implications for watershed planning

    J. Plan. Lit.

    (2002)
  • A. Bronstert et al.

    Effects of climate and land-use change on storm runoff generation: present knowledge and modelling capabilities

    Hydrol. Process.

    (2002)
  • Cicco, L.A.D., Lorenz, D., Hirsch, R.M., Watkins, W. 2018. dataRetrieval: R packages for discovering and retrieving...
  • C.J. Clopper et al.

    The use of confidence or fiducial limits illustrated in the case of the binomial

    Biometrika

    (1934)
  • W. Collischonn et al.

    Defining parameters for Eckhardt’s digital baseflow filter

    Hydrol. Process.

    (2013)
  • L. Congedo et al.

    Copernicus high-resolution layers for land cover classification in Italy

    J. Maps

    (2016)
  • R.E. Criss et al.

    Do Nash values have value? Discussion and alternate proposals

    Hydrol. Process.

    (2008)
  • E. Cristiano et al.

    Spatial and temporal variability of rainfall and their effects on hydrological response in urban areas – A review

    Hydrol. Earth Syst. Sci.

    (2017)
  • Cited by (13)

    • Satellite-derived spatiotemporal data on imperviousness for improved hydrological modelling of urbanised catchments

      2022, Journal of Hydrology
      Citation Excerpt :

      However, spatially variable climate and landuse (impervious surface) data over an urban catchment has not been represented to a distributed conceptual rainfall-runoff models to examine any possible improvement. Second, most of conceptual rainfall-runoff models could not consider explicitly landuse characteristics (e.g., spatiotemporal landuse maps) in hydrological relationships used for streamflow simulation (Chiew et al., 2008; Saadi et al., 2021). A few attempts have been made to incorporate landuse characteristics into conceptual rainfall-runoff models.

    View all citing articles on Scopus
    View full text