Comparative predictions of discharge from an artificial catchment (Chicken Creek) using sparse data (original) (raw)
Related papers
Ten conceptually different models in predicting discharge from the artificial Chicken Creek catchment in North-East Germany were used for this study. Soil texture and topography data were given to the modellers, but discharge data was withheld. We compare the predictions with the measurements from the 6 ha catchment and discuss the conceptualization and parameterization of the models. The predictions vary in a wide range, e.g. with the predicted actual evapotranspiration ranging from 88 to 579 mm/y and the discharge from 19 to 346 mm/y. The predicted components of the hydrological cycle deviated systematically from the observations, which were not known to the modellers. Discharge was mainly predicted as subsurface discharge with little direct runoff. In reality, surface runoff was a major flow component despite the fairly coarse soil texture. The actual evapotranspiration (AET) and the ratio between actual and potential ET was systematically overestimated by nine of the Correspondence to: H. M. Holländer (hartmut.hollaender@tu-cottbus.de) ten models. None of the model simulations came even close to the observed water balance for the entire 3-year study period. The comparison indicates that the personal judgement of the modellers was a major source of the differences between the model results. The most important parameters to be presumed were the soil parameters and the initial soilwater content while plant parameterization had, in this particular case of sparse vegetation, only a minor influence on the results.
2011
Due to recent environmental change and subsequent hydrological change, there is an increasing demand for hydrological predictions, as well as improved understanding of processes . Hydrological models can serve as useful tools for both purposes. Conceptually different models have been developed and applied for prediction and scenario analysis. Previous model comparison studies revealed that conceptually different models showed similar sensitivities to environmental change scenarios , while the modeller often made the difference between each model application, rather than the choice of the model . The question re-mains to which degree the influence of the modeller depends on data availability and whether a gradual improvement of the data base reduces differences between predictions based on different models.
Discharge predetermination in ungauged basins
Predetermination of peak discharges and flood volumes of ungauged basins presents important society issues: management of surface waters, protection against floods, water supply etc. Thus we propose a methodology the predetermination of discharges from rainfall. It’s an association of effective rainfall obtained from Monte Carlo Simulations (MCS) with a unit hydrograph based on geomorphology. The unit hydrograph (UH) based on geomorphology is selected knowing that the geomorphological parameters can be obtained from topographic charts, soils charts, the ground occupation charts and soil data. The UH used is produced from the Nash cascade model in which the scale and shape parameters are those proposed by Rosso (1984). These parameters depend on the hydrographical network, the Horton ratios and the average peak flow velocity assumed constant throughout the network and time invariant. The average peak flow velocity can be expressed as a function: 1) of geomorphological parameters such as the total surface of the basin, the slope of the highest order stream, the coefficient of Manning-Strickler, the width of the channel, the parameter of the kinematic wave of the highest order stream and the length of the main channel; and 2) of the effective rainfall intensity and its duration. For the effective rainfall intensities, the idea is to consider the effective rainfall as a vector of parameters of the hydrological model then to use the MCS method to generate the corresponding components. The proposed simulation framework includes: The specification of the data: the geomorphological parameters and the time increments are fixed for all simulations, while the duration of the total rainfall and the effective rainfall volume vary from one event to another, and constitute constraints determining simulations to be rejected. The analysis of the simulated hydrographs deals with aspects: the first relates to the hydrographs simulated for each event, while the second analyzes the whole simulations seeking to highlight indicators to characterise outputs. In order to statistically interpret the simulated hydrographs, the generated peak discharges are classified for each event, and their percentiles 25th, 50th and 75th as their mode are analyzed. The same treatment is applied to the simulated time to peak of the hydrographs. The percentiles 25th and 75th make it possible to evaluate the extent of the 50% interval of the simulated discharges whereas the median and the mode make it possible to position values representative of the distribution of the generated discharges. The hydrographs containing are assumed have the same “recurrence” as their peak discharge. In conclusion, the process of hydrographs generation by the MCS method includes primarily two steps: 1) the generation of the intensities of effective rainfall supposing that its total volume is resulting from the observations and 2) the convolution of the unit hydrograph resulting from each interval of effective rainfall. The case under study is a small catchment area: Saddine1, having a surface of 384 hectares. It is located beside Makthar in Tunisia (northern latitude 35°48'06'' and longitude is 9°04'9'') in a mountainous zone, monitored from 1992 to 1999. This catchment is controlled by a small headwater dam. For five minutes durations, the maximum intensity observed is 260 mm/h and the minimal one is 10 mm/h whereas the maximum total rainfall recorded for an event was of 106mm. The longest duration for an event was of approximately 5 hours (299mn) and shortest was 12 minutes. In addition a great disparity of volumes is also to note: the maximum observed is 67200 m3 and the minimum is 1275 m3. The peak discharges of the recorded hydrographs are very variable with a Minimum/maximum ratio about 1/1370. Indeed the maximum discharge observed is of 85,6m3/s while the minimum is of 0,062m3/s. The times to peak for the events vary from 15 to 120 minutes. For Horton overland flows, the determination of effective rainfall intensities is still complex, despite the important number of works, since the thirties. The infiltration phi-index (f-index) method provides an approximation of the infiltration process estimation and is often used in practice. Several studies show that effective rainfall depends on total rainfall intensities, soil characteristics, antecedent conditions and moisture index. Thus the effective rainfall intensities are calculated by the infiltration index method f, which remains a method still largely used in spite of its rudimentary character. The effective rainfalls estimated for each event vary from 0,38mm with 17,5mm. Before the use of the MCS, the model is calibrated. The calibration results analysis show that the calculated hydrographs compare in a reasonable way with the observed hydrographs. As well the shape, the peak discharge and the peak time reconstitutions are satisfying. A total of 44 simulations were carried out for each of the 15 events observed, of which 13 allowed the identification of the distributions of effective rainfall intensities and durations. The remaining two events were used for the validation of the approach. The analysis of the generated hydrographs showed a rather weak dispersion of the peak output from one simulation to another, for a given event. Moreover, the discharges and times to peak resulting from the generated hydrographs follow a dissymmetrical distribution. The observed values of the peak discharges and times to peak represent realisations of outputs simulations with different probabilities of occurrence. In order to capitalize on the model, relationships between simulated peak discharges, time to peak, base time and volumes were built. Another aspect is investigated: the sensitivity analysis (SA) which must be an integral part of model. SA can in particular help to apprehend the effect of each parameter on model outputs and thus to classify parameters on a hierarchical impact. This knowledge is of primary importance for the model user, who is consequently informed about the degree of attention and precision to be given to each parameter. A new approach is proposed to carry out the SA of a GIUH’s parameters. It is proposed to study the model sensitivity to a given parameter, using a bivariate analysis of parameter and output, with Kendall-plots and Kendall’s tau. It is suggested to use the latter to quantify and classify the impact of model parameters in hierarchical order. To achieve SA, Monte Carlo Simulations (MCS) are used and different sampling methods can be adopted and the distributions thus obtained are compared using Q-Q plots. The results show that duration has the most significant impact on peakflow; the highest-order stream length and Manning roughness coefficient have the same weight, maximum and average intensities play an identical role, and finally the variability of event and Horton’s ratios have the least significant role. Moreover peak time is mainly sensitive to Manning roughness coefficient and the variability of the event, fairly sensitive to duration and Horton’s ratios and not very sensitive to average and maximum intensities. It is worth noting that base time has an equivalent behaviour except toward Manning roughness coefficient which has not an important influence. In addition, peakflow is the most sensitive output compared with peak time and Base time. The previous chapters the rainfall volume is known, consequently in this chapter we assume that the rainfall volume is unknown. We propose a methodology to apply GIUH to ungauged basins using Monte Carlo Simulations and copulas. The effective rainfall, input of GIUH is assumed to be unknown; it is estimated with infiltration index method (f-index). The two correlations detected between this index and the characteristic rainfall intensities: maximum intensity and average intensity, modelled with Gumbel copula, are used to generate effective rainfall hyetographs. Thus resulted hydrographs from the two f conditioned distributions are analyzed, and give statically the same results: dispersion and variability for the entire studied characteristics (volume, peak discharge, peak time and base time). However only the ones derived from f conditioned to maximum intensity distribution allow reconstituting the observed hydrographs. Moreover the comparison between the series of order statistics corresponding to interest output and those corresponding to observed one leads to decide on the representative catchment hydrograph. Finally empirical relationships between the main characteristics of hydrograph are found which may be used in practice. To build a methodology for predetermination of peak discharges and flood volumes of ungauged basins, we suggest identifying and modeling the relationship between this index and the maximum intensity of rainfall (Imax). The investigation concerns 22 catchments (areas between 1 km² to 10 km²) situated in Tunisia (from 35°N to 37°N, from 8°E to 11°E), in a semi – arid climate zone (average annual rainfall between 280 mm to 500 mm). For this purpose, we propose to adopt Kendall-plots which are a recent graphical method based on rank statistics, and which allow the detection of nonlinear dependence between two variables. In addition, the use of rank correlation Kendall’s tau (t) allows measuring the degree of dependence. This exploration has showed a positive correlation between f-index and Imax. Indeed, the estimated t vary from 0.47 to 0.91. This reveals the important dependence between the studied variables. Therefore to model this correlation, Gumbel copula has been adopted, with parameter ranging between 1.9 and 11.1. The key question is: what catchment’s characteristics factors explain the inter-site variability of the copula parameter? Thus the further stage of this study concerns the regionalization of copula parameter using the physiographic and geographical characteristics of basin. The idea is to achieve a b...
Physics and Chemistry of the Earth, Parts A/B/C, 2011
The 1-D process-based model SIMULAT was applied to the 6 ha large artificial catchment ''Chicken Creek'' in Lausatia, Germany. Within the framework of a model intercomparison study, data availability was improved step by step, starting from sparse data conditions. Initially, the model was parameterised based on transfer functions (e.g., soil hydraulic properties were estimated from pedotransfer functions) and literature (e.g., plant parameters, boundary conditions), only. Then parameterisation was revised based on field inspection and additional quantitative data (e.g., from point measurements). Finally, soil moisture data were used for validation and calibration purposes. During this parameterisation process, model results became increasingly plausible although calibration and validation against observed discharge were not feasible because discharge data were not available to the modellers. Simulated discharge dynamics changed from an initially base flow-dominated and continuous flow regime to a system in which different flow components contribute similarly to the event-based total discharge, better conforming to the hydrological process understanding with respect to the development of a gully network. Qualitative information (=soft data) gained from a field visit particularly contributed to this improvement in process understanding towards a flow regime dominated by surface runoff, while additional quantitative information on system characteristics rather served the purpose of verifying (or revising) of model parameterisation and defining appropriate initial conditions. An evaluation of simulated surface runoff rates based on event-based discharge information for a subcatchment revealed that the model overestimated the surface runoff generation for all advanced modelling steps. A final validation of model results is not yet feasible as continuous discharge data at the catchment outlet are not available so far. However, the model application indicated that integration of soft and quantitative information can considerably increase the plausibility of model results in the case of poorly gauged basins while applying transfer functions developed under natural conditions might fail in artificial catchments.
Hydrology and Earth System Sciences Discussions, 2014
The Wageningen Lowland Runoff Simulator (WALRUS) is a new parametric (conceptual) rainfall-runoff model which accounts explicitly for processes that are important in lowland areas, such as groundwater-unsaturated zone coupling, wetness-dependent flowroutes, groundwatersurface water feedbacks, and seepage and surface water supply (see companion paper by Brauer et al., 2014). Lowland catchments can be divided into slightly sloping, freely draining catchments and flat polders with controlled water levels. Here, we apply WALRUS to two contrasting Dutch catchments: the Hupsel Brook catchment and Cabauw polder. In both catchments, WALRUS performs well: Nash-Sutcliffe efficiencies obtained after calibration on one year of discharge observations are 0.87 for the Hupsel Brook catchment and 0.83 for the Cabauw polder, with values of 0.74 and 0.76 for validation. The model also performs well during floods and droughts and can forecast the effect of control operations. Through the dynamic division between quick and slow flowroutes controlled by a wetness index, temporal and spatial variability in groundwater depths can be accounted for, which results in adequate simulation of discharge peaks as well as low flows. The performance of WALRUS is most sensitive to the parameter controlling the wetness index and the groundwater reservoir constant, and to a lesser extent to the quickflow reservoir constant. The effects of these three parameters can be identified in the discharge time series, which indicates that the model is not overparameterised (parsimonious). Forcing uncertainty was found to have a larger effect on modelled discharge than parameter uncertainty and uncertainty in initial conditions.
2002
The search for generalized theories to cope with the inherent spatial scale problem of hydrological prediction has been largely unsuccessful to date. The modelling of flow processes in catchments is hampered by this scale problem, and therefore results in poor predictions in the ubiquitous ungauged catchment (a catchment devoid of any streamflow measurements). Until now, observations of streamflow have been required to calibrate the rainfall–runoff models that are central to most strategies for hydrological prediction. In ungauged catchments other methods must be used. Catchment characteristics play a central role in this. The IAHS Predictions in Ungauged Basins (PUB) initiative aims at the development of science and technology to provide data and/or knowledge for ungauged or poorly gauged basins. The Hydrology and Quantitative Water Management (HWM) group of Wageningen University is very active in this field. In this paper some examples show how the group develops new measuring tec...
Efficacy of a hydrologic model in simulating discharge from a large mountainous catchment
Journal of Hydrology, 2006
Three sets of climatic input data from weather station observations and reanalysis products were compared for use in the simulation of streamflow for a large mountainous basin in subarctic Canada. These data sets are statistically different or show biases for most months. Yet, when they were used in conjunction with specific suites of parameters optimized for individual data sets, the hydrological model (SLURP) was able to simulate flows from these data sets which compare satisfactorily with measured discharge of the 275,000 km 2 Liard catchment. The progressive downstream change in simulated discharge was scrutinized to reveal how and why, despite using inputs that are different, the model can simulate comparable basin outflows. It was found that through the effects of overestimating or under-adjusting the flow for various sub-basins, the model can simulate discharge that matches the measured Liard flow. This compensating mechanism enhances the flexibility of the model in producing acceptable outflow for a large catchment.
Physics and Chemistry of the Earth, 2005
Although societal hazard related to hydrological events has increased dramatically over time, our understanding of the discharge generating processes is still deficient. The main challenges appear to be related to the scale of the processes. Hydrological processes are highly heterogeneous, non-linear and interconnected. Upscaling from micro-to catchment scale and subsequent parametrisation appears undoable because of data requirements, model complexity, computational time requirements, and equifinality (implying that different combinations of parameters generate equally good results). A new generation of hydrologists is looking for answers to match the observed complexity at the plot-scale, with the apparent simplicity that arises at the catchment scale. One of the key processes in the watersheds of the Meuse and Rhine is the rapid sub-surface flow, a process which is still poorly understood.
Hydrologic modeling for the determination of design discharges in ungauged basins
Issue 3, 2013
All stormwater management projects in Greece are required to get environmental permit before construction. The design return period is often determined by the environmental permit. Determination of design discharges is an important parameter for the design. Design discharge for a given return period is not uniquely defined and may vary considerably depending on the selection of the parameters and methodologies involved. Using a rainfall height distribution that tends to maximize the peak discharge (worst profile distribution) in essence corresponds to a lower probability of occurrence that is not quantified at present. There are publications, based on data from the US and UK that show that center-loaded storms are appropriate for design of stormwater systems. In this paper a test case study is presented. Possible variation of the estimated flood peak that can result from variation of rainfall distribution of given total height and duration is shown. Comparisons are shown between the...
A practicable approach for evaluating runoff changes in ungauged basins
In hydrology a basic task is the estimation of design discharges and runoff changes in ungauged catchments. However, traditional empirical rules of thumb as well as regionalization of measured discharges are subject to uncertainty. It seems that precipitation-runoff modelling is the only comprehensible way to predict discharge alterations due to changes in ungauged basins, even though the results are perhaps not less uncertain. In order to minimize this uncertainty we supplemented a methodology for discharge estimation in ungauged basins by introducing runoff coefficients derived from field assessment, an adapted precipitation-runoff model (ZEMOKOST) and routines for a plausibility check. Subsequently ten gauged Austrian catchments were used as hypothetical ungauged catchments for application and verification of this method. Except for special questions in karst- and glacier-hydrology the procedure showed satisfying results. In addition, the approach has been tested in catchments that have been intensively impacted by human use in the last decades; in this regard variations in discharge and future runoff characteristics have been analyzed.