Elsevier

Journal of Hydrology

Volume 376, Issues 3–4, 15 October 2009, Pages 362-377
Journal of Hydrology

PROMET – Large scale distributed hydrological modelling to study the impact of climate change on the water flows of mountain watersheds

https://doi.org/10.1016/j.jhydrol.2009.07.046Get rights and content

Summary

Climate change will change availability, quality and allocation of regional water resources. Appropriate modelling tools should therefore be available to realistically describe reactions of watersheds to climate change and to identify efficient and effective adaptation strategies on the regional scale. The paper presents the hydrologic model PROMET (Processes of Radiation, Mass and Energy Transfer), which was developed within the GLOWA-Danube project as part of the decision support system DANUBIA. PROMET covers the coupled water and energy fluxes of large-scale (A  100,000 km2) watersheds. It is fully spatially distributed, raster-based with raster-elements of 1 km2 area, runs on an hourly time step, strictly conserves mass and energy and is not calibrated using measured discharges. Details on the model concept and the individual model components are given.

An application case of PROMET is given for the mountainous Upper-Danube watershed in Central Europe (A = 77,000 km2). The water resources are intensively utilized for hydropower, agriculture, industry and tourism. The water flows are significantly influenced by man-made structures like reservoirs and water diversions. A 33-years model run covering the period from 1971 to 2003 using the existing meteorological station network as input is used to validate the performance of PROMET against measured stream flow data. Three aspects of the model performance were validated with good to very good results: the annual variation of the water balance of the whole watershed and selected sub-watersheds, the daily runoff for the whole period at selected gauges and the annual flood peaks and low flows (minimum 7-days average).

PROMET is used to investigate the impact of climate change on the water cycle of the Upper Danube. A stochastic climate generator is fed with two scenarios of climate development until 2060. One assumes no future temperature change, the other uses the temperature trends of the IPCC-A1B climate change scenario. PROMET is run with both climate data sets. No change in low-flow is detected when no temperature change is assumed. The IPCC-A1B climate scenario results in marked decreases of low-flow at the outlet of the watershed.

Introduction

Climate change will change availability, quality and allocation of water resources on the regional scale and will force to adapt to changing future boundary conditions. Appropriate modelling tools should therefore be available to realistically describe reactions of watersheds to climate change and to identify efficient and effective adaptation strategies on the regional scale.

Regional climate models will be able to cover areas of watersheds of the order of 100,000 km2 with adequate spatial detail and an improved description of the atmospheric process within the next few years (Jacob et al., 2007). Is the hydrological modelling community prepared to make full and adequate use of these regional climate models? A mutual cooperation of hydrological and atmospheric models opens opportunities to more deeply study the impacts of climate change on the regional water cycle on the land surface in order to determine decision alternatives to adapt to the changing boundary conditions. For a recent review on land surface schemes presently used in regional climate models, which usually include land surface hydrology at different levels of complexity, see Pitman (2003). A review of hydrologic models by Singh and Woolhiser (2002) also clearly pointed out the need for a more regional to global view in future hydrologic models. Provisions should therefore be taken to allow hydrologic models to fully couple with climate models on the regional scale. Two-way coupling with regional climate models poses constraints on the architecture of hydrologic models. They should close the energy and mass balances, should be valid even under changing climate conditions and should cover large-scale watersheds. The main challenge, though, one faces when extending the catchment area to regional scales is that both complexity and heterogeneity of the watersheds generally increases considerably. With increasing watershed area different characteristic hydrologic features, like snow dominated mountain hydrology, forest hydrology at mountain foothills and agricultural hydrology in the lowlands as well as different rainfall regimes, land use patterns and geological settings simultaneously influence the overall reaction of the watershed and have to be treated in a consistent way within one model framework.

The hydrologic cycle on the land surface is complex. The processes involved couple the energy and water cycle and incorporate a large range of compartments of and physical interactions within the Earth system. On the regional and local scale the hydrologic cycle is increasingly influenced by humans, who make decisions on land use and runoff conditions including the construction of hydraulic structures. This highly complex system of interactions cannot be treated adequately with black- or grey-box input–output relations especially when future states of the system should be simulated under changing boundary conditions.

Spatial heterogeneity of the land surface on all considered scales adds an additional dimension of complexity. The factors dominating heterogeneity though depend on the size of the watershed. Large watersheds stretch over different climatic zones. For small watersheds the hydrologic regime can be assumed constant. Large-scale watersheds as defined in the context of this paper are influenced by different hydrologic regimes within one climate zone. This especially applies to mountainous watersheds and their lowlands. An excellent overview of the challenges of modelling mountainous watersheds is given by Klemes (1990).

Numerous hydrologic model studies exist on different levels of complexity which investigate water fluxes in small to large-scale watersheds. They range from attempts to fully physically describe all water flows in small catchments within, e.g. the multi-purpose model SHE (Bathurst and O’Connell, 1992) to simplified, empirical approaches with no physical foundation like the unit-hydrograph approach for flood forecasting (Sherman, 1932). Most models concentrate on the simulation of river runoff with an hourly to daily temporal resolution. Hydrologic model approaches can further be divided into lumped and spatially distributed, where the first category converts rainfall into runoff by using one (constant or dynamic) runoff coefficient for the considered watershed. It is not the intention of this paper to give a full overview of hydrologic models. Nevertheless a more detailed analysis of the second category of approaches, which consider spatial heterogeneity, will be conducted here.

Usually models of this category follow the assumption, that the land surface is composed of hydrologically homogeneous areas of arbitrary shape and extension of a few square kilometres (Kunstmann et al., 2006, Leavesley et al., 1996, Lindström et al., 1997, Moussa et al., 2007, Wagner et al., 2006, Shrestha et al., 2007, Smithers and Schulze, 1995, Todini, 1996, Yu, 2000). They can be differentiated in hydrotopes or hydrological response units (HRUs) (Becker and Braun, 1999, Leavesley et al., 1983) or representative elementary areas (REA, Wood et al., 1988). Watersheds are composed of a connected set of HRUs, for which a more or less complex hydrologic model is solved. In many cases lumped parameters are used to describe the hydrologic properties of the HRUs and classical channel routing schemes are applied based on linear elements. In a recent publication Moussa et al. (2007) have used flow directions derived from a digital terrain model to delineate HRUs, which were then connected through a river network. A more flexible treatment of spatial heterogeneity by introducing variable source areas was provided with TOPMODEL (Beven and Kirkby, 1979, Beven et al., 1995). For a review of rainfall runoff models including variable source area models see Beven, 2001.

Another class of models assumes that the land surface heterogeneity can be represented through spatial grids (Bates and De Roo, 2000, De Roo et al., 2000, Fortin et al., 2000, Horritt and Bates, 2001, Jasper et al., 2002, Lee et al., 2005, Ludwig and Mauser, 2000, Mauser and Schädlich, 1998b, Niehoff et al., 2002, Refsgaard and Storm, 1995, Schulla and Jasper, 2007, Strasser and Mauser, 2001). A grid element is similar to a HRU, in the sense that it represents a defined area on the land surface, which is assumed to be hydrologically homogeneous. As soon as the area covered by each grid element becomes small enough, raster-grids allow to treat spatial (evapotranspiration, infiltration, water flow in unsaturated media, etc.) and linear hydrologic processes (channel network, etc.) with one consistent data set. At this point the raster- and HRU-concept converge. Much research has been carried out to adequately parameterize raster representations of watersheds and to identify the right process models and interfaces between them to best describe water flows on different scales. The question which solution is best suited for different raster resolutions and watershed sizes as well as purposes is still open.

The main advantages of raster-based hydrologic models against HRU-based models are, that they can more easily be coupled with atmospheric or groundwater models, which are usually also raster-based. Their spatial architecture also coincides with the organisation of spatial data fields derived from remote sensing data and therefore facilitates their use in hydrology.

None of the cited hydrological models explicitly states that they simultaneously obey two fundamental physical principles, conservation of mass and energy, throughout the whole modelling chain from rainfall to river runoff at the outlet gauge. Most conserve mass throughout the hydraulic description of the water flows in single sub-models (e.g. when treating channel flows) but they often at the same time violate conservation of energy when modelling, e.g. evapotranspiration.

The impacts of climate change on the hydrological cycle can be severe especially in mountainous watersheds with extended lowlands. In these regions the hydrologic regime of the whole watershed may shift considerably through temperature changes and changes in the amounts of rainfall and its spatial and temporal pattern. The consequences of climate change may also be large reductions in forest cover through management or burning, the expansion of agriculture, the introduction of irrigation, the reduction of snow cover as well as vanishing glaciers. These impacts may cause that any present calibration of a model may become invalid, which means that future states of the watersheds will virtually correspond to those of ungauged basins. Assuming that regional climate change will not introduce new hydrologic processes, which are not yet covered, models that are prepared for this situation should therefore at least demonstrate that they can describe the present hydrologic situation well even without calibration.

We conclude that the combination of the usual calibration procedures using measured streamflow, simplified process representations, lumped model parameters (especially in HRUs) and the fact that the simultaneous conservation of mass and energy is not guaranteed, makes it difficult and potentially risky to use these approaches to predict future states of regional hydrologic systems under changing boundary conditions with respect to climate. For the same reasons they can hardly be coupled directly with regional climate models, which are the physically most profound source of information on how regional climate will change in the future.

Therefore, the aim of this paper is to introduce the grid-based based large-scale hydrologic model PROMET, which is based on physical principles, a strict conservation of mass and energy and no calibration using measured streamflow records, as well as to use it to investigate the possible impact of climate changes on the future low-flow conditions in a complex large-scale mountainous watershed.

Section snippets

Model approach and principles

Following the boundary conditions introduced above, the distributed, physically based hydrologic model PROMET (Processes of Radiation, Mass and Energy Transfer) was specifically developed to study the impact of climate change on the water cycle of large scale, complex watersheds influenced by different hydrologic regimes. It targets at coupling with regional climate models. In order to proof its applicability it was built up and tested in the Upper Danube catchment in Central Europe, which due

The test catchment

In order to validate PROMET it is applied to the Upper Danube catchment in Central Europe. The Upper Danube catchment has an area of 76,653 km2 and covers parts of Southern Germany, Austria, Switzerland and Italy (Fig. 4). The catchment is characterized by its Alpine topography, the relief stretching from altitudes of 287 m a.s.l. at the discharge gauge Achleiten up to 4049 m a.s.l. at Piz Bernina in its Alpine headwaters. The Upper Danube catchment shows strong meteorological gradients with

Model validation

PROMET was run without calibration for the Upper Danube catchment for the period from 1.1.1970 to 31.12.2003. This simulation period was deliberately extended beyond the standard climate period from 1971 to 2000 to be able to give the model time to spin up and to include the extremely warm Central European Summer of 2003 into the analysis. PROMET was continuously run over the whole simulation period using the input data set described above and a modelling time interval of 1 h. Analysis of the

Climate change impact example: low flows

The validation of PROMET has proven that the historical low-flow conditions in the Upper-Danube watershed are well captured by the model. In order to exemplify the future use of PROMET for regional climate impact studies a possible change in annual low-flow is simulated for the upcoming 50 years from 2011 to 2060. Two scenarios are chosen for the future development of climate in the watershed:

  • 1.

    It is assumed that climate change will come to a stop in 2011 and climate will not change during the

Conclusions and outlook

The study shows, that the consideration of the guiding principles of “Introduction” during the design and implementation of PROMET results in a spatially distributed physically based hydrologic model, which describes well and in detail the variability of the annual water balance and the daily water fluxes in the complex, large scale, mountainous Upper-Danube watershed for an extended, climatologic period of 33 years. The water balance, daily discharges and return periods of peak and low-flow

Acknowledgements

The provision of meteorological and hydrological data by the German Weather Service DWD and the Austrian Weather Service and the Environmental Agency of the Free State of Bavaria is gratefully acknowledged. Our thanks go to all members of the GLOWA-Danube projects for many fruitful discussions, which considerably improved the PROMET modelling approach. We thank Markus Weber from the Glaciology Division of the Bavarian Academy of Science for valuable inputs on snow dynamics. Special thanks go to

References (94)

  • J.E. Nash et al.

    River flow forecasting through conceptual models, a discussion of principles

    Journal of Hydrology

    (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

    Journal of Hydrology

    (2002)
  • P. Racsko et al.

    A serial approach to local stochastic weather models

    Ecological Modelling

    (1991)
  • S. Shrestha et al.

    The assessment of spatial and temporal transferability of a physically based distributed hydrological model parameters in different physiographic regions of Nepal

    Journal of Hydrology

    (2007)
  • U. Strasser et al.

    Modelling the spatial and temporal variations of the water balance for the Weser catchment 1965–1994

    Journal of Hydrology

    (2001)
  • E. Todini

    The ARNO rainfall–runoff model

    Journal of Hydrology

    (1996)
  • E.F. Wood et al.

    Effects of spatial variability and scale with implications to hydrologic modeling

    Journal of Hydrology

    (1988)
  • J.H.M. Wösten et al.

    Development and use of hydraulic properties of European soils

    Geoderma

    (1999)
  • Bach, H., Verhoef, W., Schneider, K., 2000. Coupling remote sensing observation models and a growth model for improved...
  • Bach, H., Mauser, W., Schneider, K., 2003a. The use of radiative transfer models for remote sensing data assimilation...
  • H. Bach et al.

    The use of remote sensing for hydrological parameterisation of Alpine catchments

    Hydrology and Earth System Sciences

    (2003)
  • Barnes, H.H., 1967. Roughness Characteristics of Natural Channels. US Geological Survey Water Supply Papers No. 1849....
  • Barthel, R., Mauser, W., Braun, J., 2007. Integrated modelling of global change effects on the water cycle in the upper...
  • J.C. Bathurst et al.

    Future of distributed modelling: the Systeme Hydrologique Europeen

    Hydrological Processes

    (2009)
  • K. Beven

    Rainfall Runoff Modelling: The Primer

    (2001)
  • K. Beven et al.

    A physically based, variable contributing area model of basin hydrology

    Hydrological Sciences Bulletin

    (1979)
  • K. Beven et al.

    TOPMODEL

  • R.H. Brooks et al.

    Properties of porous media affecting fluid flow

    Journal of Irrigation and Drainage Division, American Society of Civil Engineering

    (1966)
  • BÜK, 1997. BÜK 1000 Bodenübersichtskarte von Deutschland 1:1000,000....
  • G.S. Campbell et al.

    An Introduction to Environmental Biophysics

    (1998)
  • CLC 2000, 2004. Corine Land Cover 2000 – Mapping a Decade of Change, Brochure No. 4/2004....
  • CLC 2000, 2008. <http://terrestrial.eionet.europa.eu/CLC2000> (visited...
  • J.A. Cunge

    On the subject of a flood propagation computation method (Muskingum method)

    Journal of Hydraulic Research

    (1969)
  • G. Czeplak et al.

    Parametrisierung der atmosphärischen Wärmestrahlung bei bewölktem Himmel

    Meteorologische Rundschau

    (1987)
  • C. Daly et al.

    A statistical–topographic model for mapping climatological precipitation over mountainous terrain

    Journal of Applied Meteorology

    (1994)
  • A.P.J. De Roo et al.

    Physically-based river basin modelling within a GIS: the LISFLOOD model

    Hydrological Processes

    (2000)
  • A. Denoth

    Structural phase changes of the liquid water component in Alpine snow

    Cold Regions Science and Technology

    (2003)
  • P.S. Eagleson

    Climate, soil and vegetation. 3. A simplified model of soil movement in the liquid phase

    Water Resources Research

    (1978)
  • G.D. Farquhar et al.

    A biochemical model of photosynthetic CO2 assimilation in leaves of C 3 species

    Planta

    (1980)
  • T.G. Farr et al.

    The shuttle radar topography mission

    Reviews of Geophysics

    (2007)
  • J.P. Fortin et al.

    Distributed watershed model compatible with remote sensing and GIS data, I: description of the model

    Journal of the Hydrologic Division, ASCE

    (2000)
  • C. Frei et al.

    A precipitation climatology of the Alps from high-resolution rain-gauge observations

    International Journal of Climatology

    (1998)
  • B. Früh et al.

    A pragmatic approach for downscaling precipitation in alpine scale complex terrain

    Meteorologische Zeitschrift

    (2006)
  • Garbrecht, J.D., Campbell, J., Martz, L.W., 2004. TOPAZ User Manual-Updated Manual. Grazinglands Research Laboratory...
  • G. Geisler

    Ertragsphysiologie von Kulturarten der gemäßigten Breiten

    (1983)
  • GLOWA-Danube, 2009....
  • Grell, G.A., Dudhai, J., Stauffer, D.A., 1995. A Description of the Fifth-Generation Penn State/NCAR Mesoscale Model...
  • Cited by (140)

    • Evaluation of ERA5 and WFDE5 forcing data for hydrological modelling and the impact of bias correction with regional climatologies: A case study in the Danube River Basin

      2022, Journal of Hydrology: Regional Studies
      Citation Excerpt :

      In this study, we chose to run PROMET on a 30″ grid (0.00833333° spatial resolution, ~1 km² at the equator) and with hourly temporal resolution. The model framework of PROMET comprises different components such as meteorology, soil, groundwater, vegetation, channel flow and agricultural management (with e.g. fertilization and irrigation options) modules, which are described in detail in Mauser and Bach (2009), Hank et al. (2015), Mauser et al. (2015) and Zabel et al. (2014). The vegetation module simulates net primary production, land surface energy balance and evapotranspiration in an iterative manner using the mechanistic plant-physiology approach of Farquhar et al. (1980) and Chen et al. (1994), dynamically computing transpiration and CO2 assimilation by C3 and C4 plant photosynthesis processes.

    View all citing articles on Scopus
    1

    Tel.: +49 89 5238 9802; fax: +49 89 5238 9804.

    View full text