Filtern
Dokumenttyp
- Wissenschaftlicher Artikel (5)
- Postprint (1)
Sprache
- Englisch (6)
Gehört zur Bibliographie
- ja (6)
Schlagworte
- Catchment (1)
- Flood forecasting (1)
- Hydrological modelling (1)
- Improved calibration (1)
- Process identification (1)
- Rainfall-runoff response (1)
- Remote sensing (1)
- Soil moisture (1)
- Soil moisture measurement comparison (1)
- Soil-moisture (1)
The temporal dynamics of hydrological model performance gives insights into errors that cannot be obtained from global performance measures assigning a single number to the fit of a simulated time series to an observed reference series. These errors can include errors in data, model parameters, or model structure. Dealing with a set of performance measures evaluated at a high temporal resolution implies analyzing and interpreting a high dimensional data set. This paper presents a method for such a hydrological model performance assessment with a high temporal resolution and illustrates its application for two very different rainfall-runoff modeling case studies. The first is the Wilde Weisseritz case study, a headwater catchment in the eastern Ore Mountains, simulated with the conceptual model WaSiM-ETH. The second is the Malalcahuello case study, a headwater catchment in the Chilean Andes, simulated with the physicsbased model Catflow. The proposed time-resolved performance assessment starts with the computation of a large set of classically used performance measures for a moving window. The key of the developed approach is a data-reduction method based on self-organizing maps (SOMs) and cluster analysis to classify the high-dimensional performance matrix. Synthetic peak errors are used to interpret the resulting error classes. The final outcome of the proposed method is a time series of the occurrence of dominant error types. For the two case studies analyzed here, 6 such error types have been identified. They show clear temporal patterns, which can lead to the identification of model structural errors.
Low-cost monitoring of snow height and thermal properties with inexpensive temperature sensors
(2011)
Small, self-recording temperature sensors were installed at several heights along a metal rod at five locations in a case study catchment. For each sensor, the presence or absence of snow cover was determined on the basis of its insulating effect and the resulting reduction of the diurnal temperature oscillations. Sensor coverage was then converted into a time series of snow height for each location. Additionally, cold content was calculated. Snow height and cold content provide valuable information for spring flood prediction.
Good agreement of estimated snow heights with reference measurements was achieved and increased discharge in the study catchment coincided with low cold content of the snow cover. The results of the proposed distributed assessment of snow cover and snow state show great potential for (i) flood warning, (ii) assimilation of snow state data and (iii) modelling snowmelt process.
In this paper we investigate the use of hydrological models as learning tools to help improve our understanding of the hydrological functioning of a catchment. With the model as a hypothetical conceptualization of how dominant hydrological processes contribute to catchment-scale response, we investigate three questions: (1) During which periods does the model (not) reproduce observed quantities and dynamics? (2) What is the nature of the error during times of bad model performance? (3) Which model components are responsible for this error? To investigate these questions, we combine a method for detecting repeating patterns of typical differences between model and observations (time series of grouped errors, TIGER) with a method for identifying the active model components during each simulation time step based on parameter sensitivity (temporal dynamics of parameter sensitivities, TEDPAS). The approach generates a time series of occurrence of dominant error types and time series of parameter sensitivities. A synoptic discussion of these time series highlights deficiencies in the assumptions about the functioning of the catchment. The approach is demonstrated for the Weisseritz headwater catchment in the eastern Ore Mountains. Our results indicate that the WaSiM-ETH complex grid-based model is not a sufficient working hypothesis for the functioning of the Weisseritz catchment and point toward future steps that can help improve our understanding of the catchment.
The quest for improved hydrological models is one of the big challenges in hydrology. When discrepancies are observed between simulated and measured discharge, it is essential to identify which algorithms may be responsible for poor model behavior. Particularly in complex hydrological models, different process representations may dominate at different moments and interact with each other, thus highly complicating this task. This paper investigates the analysis of the temporal dynamics of parameter sensitivity as a way to disentangle the simulation of a hydrological model and identify dominant parameterizations. Three existing methods (the Fourier amplitude sensitivity test, the extended Fourier amplitude sensitivity test, and Sobol's method) are compared by applying them to a TOPMODEL implementation in a small mountainous catchment in the tropics. For the major part of the simulation period, the three methods give comparable results, while the Fourier amplitude sensitivity test is much more computationally efficient. This method is also applied to the complex hydrological model WaSiM-ETH implemented in the Weisseritz catchment, Germany. A qualitative model validation was performed on the basis of the identification of relevant model components. The validation revealed that the saturation deficit parameterization of WaSiM-ETH is highly susceptible to parameter interaction and lack of identifiability. We conclude that temporal dynamics of model parameter sensitivity can be a powerful tool for hydrological model analysis, especially to identify parameter interaction as well as the dominant hydrological response modes. Finally, an open source implementation of the Fourier amplitude sensitivity test is provided.
Flood generation in mountainous headwater catchments is governed by rainfall intensities, by the spatial distribution of rainfall and by the state of the catchment prior to the rainfall, e. g. by the spatial pattern of the soil moisture, groundwater conditions and possibly snow. The work presented here explores the limits and potentials of measuring soil moisture with different methods and in different scales and their potential use for flood simulation. These measurements were obtained in 2007 and 2008 within a comprehensive multi-scale experiment in the Weisseritz headwater catchment in the Ore-Mountains, Germany. The following technologies have been applied jointly thermogravimetric method, frequency domain reflectometry (FDR) sensors, spatial time domain reflectometry (STDR) cluster, ground-penetrating radar (GPR), airborne polarimetric synthetic aperture radar (polarimetric SAR) and advanced synthetic aperture radar (ASAR) based on the satellite Envisat. We present exemplary soil measurement results, with spatial scales ranging from point scale, via hillslope and field scale, to the catchment scale. Only the spatial TDR cluster was able to record continuous data. The other methods are limited to the date of over-flights (airplane and satellite) or measurement campaigns on the ground. For possible use in flood simulation, the observation of soil moisture at multiple scales has to be combined with suitable hydrological modelling, using the hydrological model WaSiM-ETH. Therefore, several simulation experiments have been conducted in order to test both the usability of the recorded soil moisture data and the suitability of a distributed hydrological model to make use of this information. The measurement results show that airborne-based and satellite-based systems in particular provide information on the near-surface spatial distribution. However, there are still a variety of limitations, such as the need for parallel ground measurements (Envisat ASAR), uncertainties in polarimetric decomposition techniques (polarimetric SAR), very limited information from remote sensing methods about vegetated surfaces and the non-availability of continuous measurements. The model experiments showed the importance of soil moisture as an initial condition for physically based flood modelling. However, the observed moisture data reflect the surface or near-surface soil moisture only. Hence, only saturated overland flow might be related to these data. Other flood generation processes influenced by catchment wetness in the subsurface such as subsurface storm flow or quick groundwater drainage cannot be assessed by these data. One has to acknowledge that, in spite of innovative measuring techniques on all spatial scales, soil moisture data for entire vegetated catchments are still today not operationally available. Therefore, observations of soil moisture should primarily be used to improve the quality of continuous, distributed hydrological catchment models that simulate the spatial distribution of moisture internally. Thus, when and where soil moisture data are available, they should be compared with their simulated equivalents in order to improve the parameter estimates and possibly the structure of the hydrological model.
The intention of the presented study is to gain a better understanding of the mechanisms that caused the bimodal rainfall-runoff responses which occurred up to the mid-1970s regularly in the Schafertal catchment and vanished after the onset of mining activities. Understanding, this process is a first step to understanding the ongoing hydrological change in this area. It is hypothesized that either subsurface stormflow, or fast displacement of groundwater, could cause the second delayed peak. A top-down analysis of rainfall-runoff data, field observations as well as process modelling are combined within a rejectionistic framework. A statistical analysis is used to test whether different predictors. which characterize the forcing. near surface water content and deeper subsurface store, allow the prediction of the type of rainfall-runoff response. Regression analysis is used with generalized linear models Lis they can deal with non-Gaussian error distributions Lis well its a non-stationary variance. The analysis reveals that the dominant predictors are the pre-event discharge (proxy of state of the groundwater store) and the precipitation amount, In the field campaign, the subsurface at a representative hillslope was investigated by means of electrical resistivity tomography in order to identify possible strata as flow paths for subsurface stormflow. A low resistivity in approximately 4 in depth-either due to a less permeable layer or the groundwater surface-was detected. The former Could serve as a flow path for subsurface stormflow. Finally, the physical-based hydrological model CATFLOW and the groundwater model FEFLOW are compared with respect to their ability to reproduce the bimodal runoff responses. The groundwater model is able to reproduce the observations, although it uses only an abstract representation of the hillslopes. Process model analysis as well Lis statistical analysis strongly suggest that fast displacement of groundwater is the dominant process underlying the bimodal runoff reactions.