Research papers
Disaggregated monthly hydrological models can outperform daily models in providing daily flow statistics and extrapolate well to a drying climate

https://doi.org/10.1016/j.jhydrol.2021.126471Get rights and content

Highlights

  • New disaggregation scheme for rainfall-runoff models based on soil moisture.

  • Assessed ability of disaggregated monthly models to match daily flow statistics.

  • Disaggregation largely performed better than daily models, even during drought.

  • Disaggregation performance especially suited to eco-hydrological investigations.

  • Allows complex models and assessments to be conceptualised in a simpler way.

Abstract

Daily timescale hydrological information is important for many purposes such as flood estimation, predicting the consequences of catchment management and meeting the needs of freshwater ecology. In hydrological assessments, daily timestep modelling is typically used because of the availability of daily data and many of the processes governing impacts occur at this timescale. However, daily models can suffer from poor performance in certain contexts (e.g. drier climates), and their computational requirements can make it difficult to efficiently explore many sources of uncertainty in some situations, such as understanding the impacts of climate change in larger water resource systems. Here, we test an alternate approach based on monthly modelling with a post-processing step involving monthly-to-daily disaggregation using historic flow patterns conditioned on soil moisture estimates. We apply our approach to 214 catchments across Australia representing a wide range of climate and hydrological conditions, and assess outcomes for multiple objectives spanning water supply, flood magnitude and freshwater ecological outcomes, and validate performance over an extreme multi-year drought with substantially different rainfall and streamflow characteristics. Our results show that for many metrics including sustained low flows, annual flow maxima, and high and low flow spells, the results based on monthly hydrologic modelling with daily disaggregation are generally better than those based on daily hydrological modelling. This was especially true for ecologically relevant flow metrics. In addition, the disaggregation approach fared better than the daily model when extrapolating to the multi-year dry period. Our approach also has the potential to greatly reduce the effort required to explore uncertainty in large river systems.

Introduction

Obtaining daily timestep hydrological information is key for many water resource management purposes. This can include understanding hydrologic regimes, flood frequency analysis, providing the input for ecological, social, or economic models, and predicting the consequences of change – such as infrastructure development or a drying climate. A key challenge in understanding and managing hydrologic changes to river systems is how best to model impacts given limitations of existing modelling capability (Saft et al., 2016) and the significant uncertainties involved in predicting climatic changes (Hawkins and Sutton, 2011) and hydrologic responses (Andréassian et al., 2001, Knoben et al., 2020, McMillan et al., 2012)

The current approach to meet this challenge is to model impacts at a daily time-step to explicitly represent rainfall-runoff processes, river operational decisions and responses (such as ecological outcomes that respond to daily flow triggers). This is particularly the case for rainfall-runoff modelling which is dominated by daily timestep models (Knoben et al., 2019), specifically conceptual rainfall-runoff models calibrated to historic records. This dominance is understandable given that certain hydrometeorological data (e.g. precipitation) are commonly available at a daily timestep, and rarely finer. Thus, the daily timestep is usually the first choice for certain modelling purposes, such as exploration of dominant hydrological processes (Khatami et al., 2019) or to replicate historical river operations in models that support water allocation (Perera et al., 2005). However, the costs and benefits of daily conceptual models must be carefully weighed giving due consideration to the following challenges.

Firstly, it is well known that daily rainfall-runoff models are subject to significant trade-offs in performance (Euser et al., 2013, Gharari et al., 2013). For example, models calibrated to replicate floods simulate low-flow periods poorly, and vice versa (Pushpalatha et al., 2012). Similarly, models calibrated in wet periods typically do not perform well in dry periods (Coron et al., 2012), and also are particularly vulnerable to overestimation of flows during extended droughts (Saft et al., 2016). This can be problematic for ecological purposes, where multiple aspects of flow regimes such as flood timing and peak flow combine to support species needs (Poff et al., 1997). Often a lack of favorable trade-off solutions (Fowler et al., 2016) forces modelers to choose which aspect of performance to emphasize at the expense of the others (Li et al., 2012). Because of these trade-offs from daily calibration, we hypothesize that modelling with a monthly timestep paired with a disaggregation that directly uses observed flow patterns, may provide more favorable outcomes across a range of hydrologic metrics related to flow volumes, timing and sequencing. In this way, monthly calibration can focus more on volumetric accuracy whilst extra information is provided by incorporating observed daily flow patterns.

Second, from a practical perspective, more computational effort and model complexity is required to apply daily hydrological models especially when they are combined in integrated water resource models. This is particularly apparent for larger rivers or regional assessments where model inputs cannot be easily represented by area-averaged information and additional complexity is required to accommodate spatially correlated behavior at current and lagged time steps. Such integrated models may also have several sub-components that are heavily parameterized such as water demand or ecological response models, and their structure or parameterization may contribute significantly to output uncertainty (Arnold et al., 2009). Climate uncertainty in future conditions has contributed to the development of scenario-neutral or decision scaling approaches (Brown et al., 2012, Brown and Wilby, 2012), although these methods are not solely suited to climate change analysis. Unfortunately, their computational requirements can quickly become demanding, particularly when assessing complex systems with multiple objectives (Whateley et al., 2016). To overcome this, many modelers use simple statistical meta-models that are parameterized based on a few simulations of the more detailed system model (Poff et al., 2016, Turner et al., 2014, Whateley et al., 2016). However, these simple methods may obscure system dynamics and introduce additional uncertainty into modelling outputs, again, especially for ecological purposes where rates of species decline can depend on initial conditions (Bond et al., 2018, Wang et al., 2018). In modelling the hydrological impacts of climate change, many studies use simplified scaling approaches that provide daily outputs but do not simulate changes to the frequency distribution of daily rainfall (Anandhi et al., 2011). Although there are tradeoffs in computational effort and model ‘realism,’ there is a place for simpler simulation approaches as they more easily allow consideration of epistemic uncertainties (Helgeson et al., 2020).

In this paper, we tackle these challenges by investigating whether alternative approaches to daily modelling of hydrological inputs can provide equivalent information for multiple hydrological metrics and purposes for less computational effort. We developed a method that combines monthly hydrological modelling and a disaggregation approach to provide daily timestep flows as an alternative to direct daily hydrological modelling using conceptual rainfall-runoff models calibrated on historic records. Our emphasis on model performance is on characterizing a streamflow regime rather than the simulation of catchment response to a specific climatic sequence. In other words, “history matching” to a daily observed record is not a requirement, and this allows disaggregation approaches to be applied in a way that replicates daily statistics relevant to the full flow regime. While other studies have looked at the reproducibility of daily flow statistics using synthetic flow models, including disaggregation (You et al., 2014), this approach has not been compared to conceptual daily hydrological models and evaluated for a wide range of hydroclimatic conditions at large scale, nor tested when extrapolated to hydroclimatic periods outside of the range of calibration data.

Section snippets

Methods

We compare the more traditional approach of using daily conceptual rainfall-runoff modelling to generate daily streamflow with an alternative approach that uses a monthly rainfall-runoff model and then disaggregates to daily streamflow based on sampling informed by catchment wetness. The approach is firstly applied to two unregulated systems of differing catchment conditions in the Goulburn River in Victoria, Australia. It is then also tested using a range of unregulated catchments across

Case study for Goulburn River tributaries

Selected model calibration statistics for the daily and monthly rainfall runoff models are given in Table 2. Note that the statistics are calculated on monthly values for the WAPABA model and daily values for the Sacramento model, hence are not directly comparable for a given catchment. PBIAS (or the percentage of model bias) is calculated as proposed by Moriasi et al., (2007). Nonetheless, both the Sacramento and WAPABA models show good performance in replicating historic flows across wet and

Strengths of disaggregation method

The calibration strategy used in hydrological modelling will affect the ability of models to meet certain purposes, particularly for drying climates (Fowler et al., 2018). A key strength of the disaggregation approach is that it avoids some of the trade-offs that are normally required in calibrating daily rainfall-runoff models. For example, when high flows are important a calibration strategy might focus on Nash-Sutcliffe efficiency or mean-squared error as an objective function that weights

Conclusion

It is clear from the detailed case study and wider exploration on catchments across Australia that the disaggregation approach is not only a viable alternative to daily modelling, but that it outperforms the daily model for a large range of purposes. For example, water resource management is assisted by better replication of typical runoff yields and variability. Assessing flood risk is aided through improved simulation of high flows and annual maxima. Ecological modelling and management

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

This study was funded by the Australian Research Council (ARC Linkage Project LP170100598), Australian Commonwealth Government under a Research Training Program Scholarship, and several partner agencies including the Department of Environment, Land, Water and Planning; the Victorian Environmental Water Holder and the Bureau of Meteorology. Avril Horne was funded by Australian Research Council DECRA DE180100550.

References (88)

  • L. Zhang et al.

    Water balance modeling over variable time scales based on the Budyko framework - Model development and testing

    J. Hydrol.

    (2008)
  • A. Acharya et al.

    Simple method for streamflow disaggregation

    J. Hydrol. Eng.

    (2014)
  • A. Anandhi et al.

    Examination of change factor methodologies for climate change impact assessment

    Water Resour. Res.

    (2011)
  • K.E. Anderson et al.

    Instream flow needs in streams and rivers: the importance of understanding ecological dynamics

    Front. Ecol. Environ.

    (2006)
  • S. Arnold et al.

    Uncertainty in parameterisation and model structure affect simulation results in coupled ecohydrological models

    Hydrol. Earth Syst. Sci.

    (2009)
  • G. Blöschl et al.

    Scale issues in hydrological modelling: a review

    Hydrol. Process.

    (1995)
  • G. Blöschl et al.

    Twenty-three unsolved problems in hydrology (UPH)–a community perspective

    Hydrol. Sci. J.

    (2019)
  • N.R. Bond et al.

    Assessment of environmental flow scenarios using state-and-transition models

    Freshw. Biol.

    (2018)
  • C. Brown et al.

    An alternate approach to assessing climate risks

    Eos Trans. Am. Geophys. Union

    (2012)
  • C. Brown et al.

    Decision scaling: linking bottom-up vulnerability analysis with climate projections in the water sector

    Water Resour. Res.

    (2012)
  • Burnash, R., Ferral, R., 1973. A generalized streamflow simulation system, NOAA Technical Report. US Department of...
  • L. Coron et al.

    Crash testing hydrological models in contrasted climate conditions: an experiment on 216 Australian catchments

    Water Resour. Res.

    (2012)
  • R.G. Death et al.

    Resetting the river template: the potential for climate-related extreme floods to transform river geomorphology and ecology

    Freshw. Biol.

    (2015)
  • T. Euser et al.

    A framework to assess the realism of model structures using hydrological signatures

    Hydrol. Earth Syst. Sci.

    (2013)
  • Fowler, K., Acharya, S.C., Addor, N., Chou, C., Peel, M., 2020. CAMELS-AUS v1: Hydrometeorological time series and...
  • H.J. Fowler et al.

    Anthropogenic intensification of short-duration rainfall extremes

    Nat. Rev. Earth Environ.

    (2021)
  • K.J.A. Fowler et al.

    Simulating runoff under changing climatic conditions: revisiting an apparent deficiency of conceptual rainfall-runoff models

    Water Resour. Res.

    (2016)
  • K. Fowler et al.

    Improved rainfall-runoff calibration for drying climate: choice of objective function

    Water Resour. Res.

    (2018)
  • D.C. Frank et al.

    Water-use efficiency and transpiration across European forests during the Anthropocene

    Nat. Clim. Chang.

    (2015)
  • A.J.E. Gallant et al.

    An experimental streamflow reconstruction for the River Murray, Australia, 1783–1988

    Water Resour. Res.

    (2011)
  • J. Gergis et al.

    On the long-term context of the 1997–2009 “Big Dry” in South-Eastern Australia: insights from a 206-year multi-proxy rainfall reconstruction

    Clim. Change

    (2012)
  • S. Gharari et al.

    An approach to identify time consistent model parameters: sub-period calibration

    Hydrol. Earth Syst. Sci.

    (2013)
  • J.C. Grygier et al.

    Condensed disaggregation procedures and conservation corrections for stochastic hydrology

    Water Resour. Res.

    (1988)
  • J.O. Haerter et al.

    Climate model bias correction and the role of timescales

    Hydrol. Earth Syst. Sci.

    (2011)
  • T. Hallouin et al.

    Calibration of hydrological models for ecologically relevant streamflow predictions: a trade-off between fitting well to data and estimating consistent parameter sets?

    Hydrol. Earth Syst. Sci.

    (2020)
  • N. Hansen et al.

    Evaluating the CMA evolution strategy on multimodal test functions

    Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics)

    (2004)
  • E.d. Hawkins et al.

    The potential to narrow uncertainty in projections of regional precipitation change

    Clim. Dyn.

    (2011)
  • C. Helgeson et al.

    Why simpler computer simulation models can be epistemically better for informing decisions

    Philos. Sci.

    (2021)
  • Horne, A.C., Nathan, R., Poff, N.L., Bond, N.R., Webb, J.A., Wang, J., John, A., 2019. Modeling Flow-Ecology Responses...
  • IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment...
  • John, A., Nathan, R., Horne, A., Stewardson, M., Angus Webb, J., 2020. How to incorporate climate change into modelling...
  • O.D. Jones et al.

    Modelling the effects of fire and rainfall regimes on extreme erosion events in forested landscapes

    Stoch. Environ. Res. Risk Assess.

    (2014)
  • D.A. Jones et al.

    High-quality spatial climate data-sets for Australia

    Aust. Meteorol. Oceanogr. J.

    (2009)
  • S. Khatami et al.

    Equifinality and flux mapping: a new approach to model evaluation and process representation under uncertainty

    Water Resour. Res.

    (2019)
  • Cited by (0)

    View full text