Nutrient load modelling during floods in intermittent rivers: An operational approach

https://doi.org/10.1016/j.envsoft.2007.09.008Get rights and content

Abstract

In the Mediterranean area, lagoons and coastal waters are often under the influence of intermittent rivers that contribute the majority of pollutant loads during flood events. Reliable tools for evaluating pollutant loads from land sources are needed for managing the water quality of the receiving waters, but available water quality models are not suitable for simulating these flash flood events. An operational tool, D-PoL (Diffuse-Pollution Load), to estimate the pollutant loads transferred during rainfall events by small Mediterranean rivers is described. The tool is based on a rainfall-load semi-distributed approach and was developed with the aim of simulating temporal variations in fluxes of dissolved pollutants. The model simulates pollutographs at the outlet of the catchment during rainfall events: the rainfall triggers the processes.

The model can be described as a rainfall-load model. Three parameters describe the processes involved: the initial stock of pollutants on the hillslopes, the production parameter (which is related to the lag time of the catchment) and the routing parameter (which is related to the lag time of a basic river reach). First, sensitivity analysis demonstrated that each of the parameters controlled one key-feature of the pollutograph. The initial stock of pollutants influenced the amplitude of the pollutograph, the production parameter controlled the recession period, and the routing parameter influenced the rising phase. Second, the model was calibrated for dissolved inorganic nitrogen and dissolved phosphorus on a 20-flood data set. The sets of optimal parameters were fitted to empirical relationships for both elements. On the 20-flood calibration data set, the simulated total loads with adjusted parameters compared well with the observed parameters, except for very small events. Finally, the D-PoL model was checked against a new set of data, a 10-flood validation data set. This final validation of the model showed that the dynamics of the pollutograph were not perfectly reproduced, but that simulated total loads agreed with the observed loads.

Introduction

More than 40% of the Mediterranean basin corresponds to coastal river basins (Margat and Treyer, 2004). Most of the rivers are intermittent or even ephemeral. Because population densities are high in Mediterranean coastal areas, these basins concentrate dense urban areas, market gardening zones, industries and tourism. Thus, the rivers represent almost half the total water volumes and probably most of the pollutant loads that contribute to water quality problems in coastal Mediterranean waters (Najem et al., 2001).

Coastal Mediterranean rivers have hydrological regimes that are very similar to dry land rivers (Bull and Kirkby, 2002). Short rainfall–runoff events are separated by long dry periods. Little subsurface flow is available and the dry period flow is often mainly fed by effluents from Waste Water Treatment Plants (WWTP). Pollutants are transferred to the rivers by runoff during rainfall events or discharged directly and continuously in the river by sewage systems. Due to the long dry summer season, pollutants accumulate in the riverbed and in the catchment area during low flow periods. Intense rainstorms occurring at the beginning of autumn or during spring produce short flash flood events that remove the pollutants from the soils and the riverbed. It is widely recognised that during floods, pollutant fluxes can vary over several orders of magnitude (Eyre and Pont, 2003, Meybeck, 2005) and can represent the majority of the annual loads of suspended solids, nutrients and other pollutants (e.g. Cherifi and Loudiki, 1999). For small Mediterranean rivers, there is still a need for accurate evaluation of pollutant loads during flood events or over annual periods, through observation studies and modelling tools. Despite this need, observation campaigns are insufficient and existing water quality models do not accurately reproduce the behaviour of these temporary rivers.

An accurate model needs to be based on hypotheses that are suitable for the study site behaviour (Letcher et al., 2002). The model must be in accordance with both time and space variability of the processes (Donigian et al., 1995b, Leon et al., 2001, Merritt et al., 2003). Pollutant processes in a catchment are closely related to hydrological behaviour and to the heterogeneity of this behaviour with respect to geology, soil, land use and human activities (Jordan-Meille et al., 1998, Letcher et al., 2002, House, 2003, Bernal et al., 2005, Yuan et al., 2007, Shrestha et al., in press). Consequently, event-based water quality models are more suitable than continuous models for small Mediterranean catchments. In addition, a spatially distributed approach is needed to take into account different sources of pollutant in a given catchment.

The accuracy of the model is also linked to its level of complexity: incorporation of detailed processes does not necessarily lead to a better model (Bobba et al., 2000, Durand et al., 2002, Merritt et al., 2003). The design and practice of water quality modelling lags behind that of hydrological modelling (Donigian et al., 1995b). Most widely used process-based water quality models are based on existing hydrological models, e.g. HSPF (Donigian et al., 1995a), ANSWERS (Beasley and Huggins, 1985), AGNPS (Young et al., 1995), SWAT (Arnold et al., 1995), CatchMODS (Newham et al., 2004). These models are coupled models. In coupled modelling, both the hydrological aspect and the water quality aspect share similar model approaches and face similar problems in terms of uncertainties (Vandenberghe et al., 2007). Krysanova et al., 1998, Croke and Jakeman, 2001 and Letcher et al. (2002) concluded that under the same restrictive conditions, water quality prediction is much more uncertain than water yield. The errors that affect the outputs of the hydrological model lead to uncertain water quality results (e.g. Migliaccio et al., 2007, Polyakov et al., 2007). Furthermore, the coupled models include two sets of parameters, increasing the complexity of the model.

Water quality models can also run pollutant load dynamics independently from flow simulations by directly considering driving forces (e.g. rainfall and/or irrigation) and related conditions. Some examples can be found in the literature (Cassell et al., 1998, Payraudeau et al., 2002). Uncoupled modelling should reduce model uncertainty and provide simple, practical tools for the accurate evaluation of pollutant loads at catchment scale.

In this study, the objective was to develop an operational modelling tool able to simulate pollutant loads during flood events in Mediterranean basins with intermittent rivers. To reduce the complexity of the model and the number of parameters, an uncoupled and conceptual approach was chosen. A semi-distributed rainfall-pollutant load model is presented here. Sensitivity analysis was used to check the performance of the model and the influence of the parameters on the model outputs. The complete empirical process developed to characterize the parameters is presented. The model study was carried out on 30 flood events in three Mediterranean intermittent rivers.

Section snippets

Catchment description

Three small Mediterranean catchments located in the South of France, the Salaison, Vène and Pallas rivers (see Fig. 1) were studied. Annual average precipitation ranges from about 750 mm in the Salaison basin to less than 600 mm in the Pallas basin. Precipitation occurs as short intense storms mainly during autumn and spring.

The Salaison River flows into the Or lagoon and drains a 53-km2 area, of which 45% corresponds to natural areas (garrigue, pine trees) and 30% to agricultural land (orchards,

The D-PoL model

The D-PoL model presented here is based on the PoL model (Payraudeau et al., 2002). The PoL model (Pollutant Loads) is an uncoupled water quality model specifically developed for Mediterranean intermittent rivers. Here we describe how the model was adapted for use with dissolved elements.

Model application

The model was applied with the following main aims: to turn the model into an operational tool and to evaluate the error of the simulated total pollutant load. The set of 30 floods was divided into two sub-sets: the best sampled floods were kept in the calibration sub-sets, as Salles et al. (in press) demonstrated that a minimum of four samples a day are needed to obtain an accurate estimate of the total load value. The 10 other floods were put in a validation sub-set.

Conclusion

In this study, we proposed an operational tool to estimate the pollutant loads transferred by small Mediterranean rivers during rainfall events. This tool is based on a conceptual and uncoupled model in order to reduce its complexity and the number of parameters required.

The model scheme was built taking into account the main processes in the transfer of pollutants in such river basins. The model is semi-distributed in such a way that production on the hillslopes and transport in the river can

Acknowledgements

This study would not have been possible without the financial support of the Programme National Environnement Côtier (PNEC) through the Chantier Lagunes Méditerranéennes. This research has been also assisted by the European Ditty Project (Development of an Information Technology Tool for the Management of European Southern Lagoons under the influence of river-basin runoff). Thanks are also due to D. Goodfellow for her editing jobs and to the reviewers for their valuable scientific comments and

References (42)

  • K.W. Migliaccio et al.

    Evaluation of landscape and instream modeling to predict watershed nutrient yields

    Environmental Modelling & Software

    (2007)
  • J.E. Nash et al.

    River flow foresting through conceptional models, 1. A discussion of principles

    Journal of Hydrology

    (1970)
  • L.T.H. Newham et al.

    A framework for integrated hydrologic, sediment and nutrient export modelling for catchment-scale management

    Environmental Modelling & Software

    (2004)
  • V. Polyakov et al.

    Evaluation of a non-point source pollution model, AnnAGNPS, in a tropical watershed

    Environmental Modelling & Software

    (2007)
  • S. Shrestha et al.

    A framework for estimating pollutant export coefficients from long-term in-stream water quality monitoring data

    Environmental Modelling & Software

    (2008)
  • V. Vandenberghe et al.

    Evaluation of uncertainty propagation into river water quality predictions to guide future monitoring campaigns

    Environmental Modelling & Software

    (2007)
  • D. Yuan et al.

    Development of an integrated model for assessing the impact of diffuse and point source pollution on coastal waters

    Environmental Modelling & Software

    (2007)
  • J.G. Arnold et al.

    Watershed Modelling and GIS With SWAT and GRASS. SWAT User's Manual

    (1995)
  • D.B. Beasley et al.

    ANSWERS: Areal Nonpoint Source Watershed Environment Response Simulation – User Manual

    (1985)
  • S. Bernal et al.

    Seasonal variations of dissolved nitrogen and DOC:DON ratios in an intermittent stream

    Biogeochemistry

    (2005)
  • L.J. Bull et al.

    Dryland rivers: hydrology and geomorphology of semi-arid channels

  • Cited by (0)

    View full text