Importance of soil moisture measurements for inferring parameters in hydrologic models of low-yielding ephemeral catchments

https://doi.org/10.1016/S1364-8152(02)00038-5Get rights and content

Abstract

Low-yielding catchments with ephemeral streams provide a stern test of the capability of conceptual catchment models for predicting the hydrologic response of the natural landscape. Sustained periods of little or no flow mean that the information content of the streamflow time-series for parameter estimation is limited. During periods with no streamflow, such ephemeral catchments also offer no information on a catchment’s soil moisture status. As a result, parameters estimated solely from streamflow data are often poorly identified and span a wide range of the feasible parameter space. These general observations were confirmed by an application of the conceptual VIC model in a 6 ha experimental catchment in eastern Australia. Using a Monte Carlo style assessment of parameter uncertainty, it was shown that the simple three-parameter model was ill-posed when calibrated solely to the streamflow response. Failure of the calibration procedure to distinguish unique antecedent moisture storage conditions prior to large rainfall events meant that the observed streamflow response could be replicated from a large envelope of potential parameter combinations. The inclusion of an estimated time-series index of areal soil moisture status into the calibration procedure, however, significantly reduced the number of feasible parameter combinations, and resulted in predictions that confirmed Bowen ratio measurements of actual evapotranspiration. Attempts to further reduce parameter uncertainty by including the measured evapotranspiration data into the joint calibration procedure were unsuccessful. The shortness of the measurement record was seen as a major factor inhibiting improvement. The results of this study highlight the critical importance of antecedent moisture conditions on streamflow yields in ephemeral catchments and point to the desirability of spatio-temporal soil moisture accounting. Future research efforts are discussed in terms of establishing the appropriate spatial and temporal resolution of soil moisture measurements needed to extend the results observed for this small experimental study to larger catchments.

Introduction

Catchment models are hypotheses of the dynamic water balance at the catchment scale. The identification of such models requires validating the model hypotheses and, as part of that process, making inferences about model parameters. In this article, the issue of parameter identification is considered in the application of conceptual catchment models in low-yielding ephemeral catchments.

Conceptual models typically involve a configuration of interconnected stores with mathematical transfer functions used to direct the movement of water between stores or into the stream. Although a mass balance is enforced for each store, the flux equations defining flows into and out of the stores are typically conceptual rather than physically based (Nash and Sutcliffe, 1970). This conceptual nature means that many of the parameters, state variables, and fluxes are not directly measurable and usually represent spatially and temporally lumped catchment characteristics. Consequently, although rather parsimonious and not very data intensive, one of the distinguishing characteristics of conceptual models is that the process of parameter inference relies heavily upon calibration via inverse reasoning, typically to an observed time-series of streamflow.

In low-yielding ephemeral catchments, parameter identification by calibration to a streamflow record is hampered by the fact that the number of non-zero data points in the streamflow time-series may be quite small, even though the length of record is large. Thus, the information content of the streamflow time-series for parameter identification is small. This presents particular problems for models that generate surface runoff through a threshold process such as a spilling bucket. During calibration the exceedence of the threshold may rarely occur and thus the bucket size is unidentifiable. This can lead to problems such as the existence of multiple optima within the feasible parameter space and the presence of high interaction or correlation between subsets of fitted model parameters (see Duan et al., 1992, Freer et al., 1996). In more humid catchments this problem is often not as severe, as the information contained in the streamflow series is likely to be rich enough to activate every model process several times during calibration (Ye et al., 1997).

A key outcome of ill-defined model parameters is that it can no longer be assumed that accurate streamflow simulation at the catchment outlet reflects accurate simulation of internal catchment states and responses. This situation arises from the large number of model parameter sets that produce virtually indistinguishable simulated streamflow time-series even though the relative contributions of the fluxes that make up the streamflow vary greatly.

One obvious and well-documented way to increase the information content available for parameter estimation is to augment streamflow data with other kinds of hydrologic information relevant to the prediction task (see for example, Mroczkowski et al., 1997; Franks et al., 1998). Examples of multiple responses include streamflow and stream chemical tracer data at different locations within a catchment and measurable internal hydrologic fluxes or states such as soil moisture, saturated areas, piezometric levels, and evapotranspiration at selected locations. Such data represent a much richer source of information about the catchment water balance dynamics than do streamflow data alone. General statements relating multiple data sources with improved parameter identification, however, have been shown to be not universal. It has been shown, for example, that augmenting streamflow with ‘point’ groundwater measurements does little to reduce parameter and predictive uncertainty (e.g. Seibert et al., 1997; Kuczera and Mroczkowski, 1998). Areal soil moisture, however, would appear to provide a valuable source of additional information, especially for ephemeral catchments during periods with no streamflow, and thus no information on catchment-average soil moisture status.

Soil moisture content is a major control on hydrological processes for both storm and interstorm periods. During storm periods it influences the partitioning of precipitation into infiltration and runoff (for saturation excess processes). For interstorm periods, soil moisture determines whether the soil column can meet the atmospheric demand for moisture; either at the surface (bare soil evaporation) or in the root zone (transpiration) and it thus affects the partitioning between latent and sensible heat fluxes. In this way, the soil moisture content is the link between the surface energy and water balances.

In most conceptual models there is some representation of soil moisture status, but validation against field data is often difficult because of at least two problems. Firstly, field measurements of soil moisture content are made at the point scale while conceptual models provide an estimate for a specified area, producing a disparity in scales. Secondly, soil moisture is highly variable in space, meaning that individual point measurements rarely if ever represent the spatial average of even small areas. This necessitates that areal values are estimated from many point measurements.

The hydrological literature contains few examples of catchment studies where distributed measurements of soil moisture values have been compared with values simulated by conceptual catchment models. Johnston and Pilgrim (1976) showed a comparison between soil moisture modelled with a simple conceptual model and soil moisture data obtained from field measurements, providing an independent assessment of model performance. Kuczera (1983) used soil moisture and throughfall measurements with a conceptual rainfall-runoff model. He noted that the use of data on runoff, soil moisture and interception with catchment models can yield substantial reductions in the uncertainty of model parameters. Kalma et al. (1995) described a comparison between simulated soil moisture resulting from both a fixed and variable storage conceptualisation and a soil moisture index based on point measurements to show the potential of conceptual models to make useful predictions of soil moisture status at the catchment scale. Western et al. (1999) demonstrated that simulated time-series of spatially average soil moisture storage achieved with a quasi-distributed conceptual model was consistent with the observed soil moisture characteristics. The statistical distribution of soil moisture storage assumed in the model, however, was shown to differ from that observed.

This article aims to re-examine the usefulness of conceptual models for soil moisture prediction at the catchment scale. This is done via a case study application of the conceptual variable infiltration capacity (VIC) model (Wood et al., 1992) in a 6 ha experimental catchment located in eastern Australia. The low-yielding catchment, which is representative of a large number of catchments in semiarid regions of Australia, was chosen to be a stern test of the capability of the VIC model. The VIC model uses a statistical distribution to characterise the spatial variation in soil moisture storage. For the current study, this distribution is determined explicitly by calibration against combinations of surface runoff, soil moisture and evapotranspiration data. Monte Carlo based assessment of parameter uncertainty resulting from individual and joint calibrations leads to the main contribution of the article, namely to provide insight into the value of field measured soil moisture, evapotranspiration and surface runoff data for parameter inference and hydrological prediction in low-yielding ephemeral catchments.

Section snippets

Study area

The 6 ha Nerrigundah experimental catchment is located in the Williams River catchment, approximately 11 km north-west of Dungog, New South Wales, Australia (Fig. 1). The catchment runs east to west with a relief of 27 m. Hillslopes range from 3 to 22%, with the main drainage line having an average slope of 9%. Average annual rainfall is 1000 mm and areal potential evapotranspiration is 1600 mm. The soil type is a moderately well drained clay-loam duplex with an A horizon of approximately 30–40

Measurements

For the 908-day period of investigation (28.10.1996–07.04.1999) undertaken in this study, a variety of hydro-meteorological variables were measured. A weather station continuously measured net radiation, atmospheric pressure, wind speed and direction, relative humidity, air temperature, rainfall, soil heat flux and soil temperature at various depths. Apart from rainfall, all measurements were made at 1-min intervals, with the average taken every 10 min. Rainfall was recorded for each tip of the

Soil moisture analysis

The 13 spatially distributed measurements of soil moisture were discontinuous in time (i.e. approximately one measurement every 2 weeks) while the Virrib soil moisture sensors at the weather station provided a continuous soil moisture time series for an individual point. In order to recover a continuous record of areal soil moisture, a merging of the two data sets was performed. The idea was to utilise the spatial measurements to obtain instantaneous catchment average soil moisture estimates,

Description of the VIC model

The conceptual water balance model used here is the single layer VIC model (Wood et al., 1992, Sivapalan and Woods, 1995, Kalma et al., 1995). Fig. 6a provides a schematic illustration of the soil moisture distribution approach that forms the basis of the VIC model. The VIC model assumes that scaled infiltration (i.e. storage) capacity is a random variable with its cumulative distribution function given by the Xinanjiang distribution (Zhao et al., 1980). The distribution function allows for a

VIC simulations

For the current application of the VIC model a daily time step was used. Results are based on the 908-day period between 28.10.1996 and the 07.04.1999. Potential evapotranspiration (Ep) for the period was calculated with the Penman–Monteith model, following the methodology outlined by Smith et al. (1991). Net radiation, temperature, humidity and wind data were obtained from the Nerrigundah weather station. Daily rainfall was taken as the average of the measured volumes obtained from the two

Calibration to streamflow

In a general sense, calibration of the VIC model to streamflow involved optimising the model parameters to ensure that the minimum soil moisture deficit (sminv) following a prolonged dry period would generate the correct amount of streamflow for the next large rainfall event. This could be achieved by adjusting either the threshold depth to overland flow, smin, or the evaporation parameter, β (which would then change v). Table 1a presents the Metropolis-sampled posterior mean and standard

Conclusions and recommendations

This study has illustrated how a simple three-parameter version of the conceptual VIC model is ill-posed when calibrated to the streamflow time-series for a 6 ha ephemeral catchment. The fact that a streamflow response is an integrated result of both quick-flow and slow-flow processes, combined with the fact that extended periods with no streamflow offer no information on the catchment’s soil moisture status, makes the process of inferring the compartmentalisation of storage within the

Acknowledgements

Special thanks are made to John Russell, the owner of the Nerrigundah catchment. John provided free access to his land and willing cooperation throughout this project. A grateful acknowledgement is also made to George Kuczera for his insightful comments and assistance with the parameter-fitting package nlfit (Kuczera, 1994). Improvements to the current version of the manuscript were aided by the constructive suggestions of two anonymous reviewers.

References (27)

  • P.R. Johnston et al.

    Parameter optimisation for watershed models

    Water Resour. Res.

    (1976)
  • J.D. Kalma et al.

    Predicting catchment-scale soil moisture status with limited field measurements

  • G. Kuczera

    Improved parameter inference in catchment models. 2. Combining different kinds of hydrologic data and testing their compatibility

    Water Resour. Res.

    (1983)
  • Cited by (49)

    • Post-wildfire recovery of water yield in the Sydney Basin water supply catchments: An assessment of the 2001/2002 wildfires

      2014, Journal of Hydrology
      Citation Excerpt :

      This would provide a clearer indication on how the response of vegetation regrowth influences the subcatchments within the Sydney Basin. Furthermore, other predictor variables such as; sea surface temperatures (Van Ogtrop et al., 2011), wind speed and direction (Kletti and Stefan, 1997), evapotranspiration (McMichael and Hope, 2007), soil moisture (Wooldridge et al., 2003), NDVI (Sun et al., 2008), Leaf Area Index (LAI) (Ha et al., 2010) and catchment wetness index (CWI) (Jakeman and Hornberger, 1993) may improve model accuracy and expand the knowledge on the interactions between each subcatchment. The use of such predictor variables would allow the model to account for additional variation in the data especially if shorter prediction models are required immediately after a wildfire.

    View all citing articles on Scopus
    View full text