Displays

Catchments are complex dynamical systems exposed to highly-variable inputs (rainfall). Despite this complexity, it is uncommon to model these systems as stochastic ones. Previous works offer a large number of examples where deterministic (conceptual or physics-based) models are used to describe hydrological basins in spite of the fact that, in some cases, the output of the model shows substantial deviations from the observed data even after meticulous calibration.
There are different ways to include stochasticity in the hydrological modeling of catchments. With this contribution we explore a systematic way to improve our knowledge of the system at hand by using time-dependent parameters, which are driven by suited stochastic processes. The fundamental idea, which dates back to seminal works carried out about ten years ago, is to correlate the evolution of the selected time-dependent parameters to catchment features, input variables, or possible changes over time within the catchment area, to improve the structure of the model in a data-driven fashion, rather than to merely resort to including a bias term on the output of the model.
In doing so for different catchments, we make use of a newly-developed inference framework called SPUX, which is particularly suited to deal with non-linear stochastic models as it enables the usage of high-performance computing clusters for (Bayesian) inference coupled with the particle filter method. This allows us to explore and show our approach at work on different settings, such as models of different complexity and data-sets of different resolutions, lengths, and relevant to catchments with different characteristics, which have (or not) changed over time.

How to cite: Bacci, M., Fenicia, F., and Sukys, J.: Stochastic time-dependent parameters to improve the modeling and characterization of catchments, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10073, https://doi.org/10.5194/egusphere-egu2020-10073, 2020.

The authors explore uncertainty associated with the quantitative precipitation forecasts (QPF) and its implication to the predictability of real-time streamflow forecasts. Including rainfall forecasts into real-time streamflow forecasting system extends the forecast lead time. As rainfall is a key driver of rainfall-runoff models both past and future rainfall estimates should be used in streamflow and flood forecasting. Since both QPE and QPF are subject to substantial uncertainties, questions arise on the trade-off between the time horizon of the QPF and the accuracy of the streamflow forecasts. Particularly QPF is notorious for its significant uncertainty with respect to location, timing and magnitude. Operational hydrologic services often limit their use of the QPF to one or two days into the future. The authors study this problem systematically using operational models and QPF. Their focus is on scale-dependence of the trade-off between the QPF time horizon and streamflow accuracy. To address this question, the authors first perform comprehensive independent evaluation of QPF at about 140 basins with wide range of spatial scales (10 - 40000 km2) corresponding to U.S Geological Survey (USGS) streamflow monitoring stations over the state of Iowa in Midwestern United States. High Resolution Rapid Refresh (HRRR) is an hourly short-medium range rainfall forecast of up to 18 hours updated every hour with spatial resolution of about 3 km by 3 km. Six-hourly rainfall forecasts are available for up to seven days ahead. Since basins are hydrologically relevant, the authors perform HRRR skill verification for the years 2016-2019 using conventional verification techniques and mean areal precipitation (basin scale rainfall volume) with respect to multi-radar

multi-sensor (MRMS) QPE (gauge-corrected) rainfall. The authors show that the QPF errors/uncertainties are scale-dependent. The QPF skills show increase as the basin scale and lead time of the forecast increases at short-medium range. In the second part of the study, both QPE and QPFs are forced separately to the hydrologic model called hillslope-link model (HLM) used at the Iowa Flood Center for real-time streamflow forecasting for Iowa. The objective is to understand the contribution of QPF uncertainty structure on the skill of streamflow forecasts. Since real-time streamflow observations (15 minutes resolution) are available at USGS sites, the authors incorporate them using a simple data assimilation framework. Several scenarios of forecasts, such as open-loop combined with QPF, persistence-based approach (using streamflow observations) combined with QPF, and open-loop combined with QPF for more than 18 hours horizon is explored. The authors report the contribution of QPF errors on hydrologic predictions across scales and suggest a forecasting scenario that shows the most enhanced predictability of streamflows.

How to cite: Krajewski, W. and Ghimire, G.: Scale-Dependent Worth of QPF for Real-Time Streamflow Forecasting, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-10958, https://doi.org/10.5194/egusphere-egu2020-10958, 2020.

As the importance of Soil Moisture (SM) has been recognized in various fields, including agricultural practices, natural hazards, and climate predictions, ground-based SM sensors such as Frequency Domain Reflectometry (FDR), Time Domain Reflectometry (TDR) are being widely used. However, gaps in in-situ SM data are still unavoidable due not only to sensor failure or low voltage supply, but to environmental conditions. Since it is essential to acquire accurate and continuous SM data for its application purpose, the gaps in the data should be handled properly. In this study, we propose a physically based gap-filling method in a mountainous region, in which in-situ SM measurements and flux tower are located. This method is developed only with in-situ SM and precipitation data, by considering variation characteristics of SM: increases rapidly with precipitation and decreases asymptotically afterward. SM data from the past is used to build Look-Up-Tables (LUTs) that contains the amount and speed of increment and decrement of SM, with and without precipitation, respectively. Based on the developed LUTs, the gaps are filled successively from where the gaps started. At the same time, we also introduce a machine learning-based gap-filling framework for the comparison. Ancillary data from the flux tower (e.g. net radiation, relative humidity) was used as input for training, with the same period as in the physically based method. The trained models are then used to fill the gaps. We found that both proposed methods are able to fill the gaps of in-situ SM reasonably, with capabilities to capture the characteristics of SM variation. Results from the comparison indicate that the physically based gap-filling method is very accurate and efficient when there’s limited information, and also suitable to be used for prediction purposes.

How to cite: Lee, S., Jeon, H., Park, J., and Choi, M.: Performance of a Physically Based Gap-Filling Technique of in-situ Soil Moisture, in Comparison with Machine Learning, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-13671, https://doi.org/10.5194/egusphere-egu2020-13671, 2020.

The onset and generation of runoff, and the overall rainfall-runoff transformation, resulting in hillslope and catchment runoff response, are controlled by multiple interacting small-scale processes. Small scale features such as surface microtopography -small variations around the average terrain shape- can govern large scale signatures of runoff dynamics. This is the net result of local heterogeneities in the flow paths and ponding which in turn control the development of the surface water layer and how it connects and flows downslope. It is therefore relevant to understand which microtopographic features may play a governing role in runoff generation dynamics. Given that it is very difficult to assess such processes experimentally in the field, we turn to computational modelling to assess different features, hydrological conditions and the overall response.

In this work, we numerically solve a physically-based surface water model (based on the Zero-Intertia approximation of the shallow-water equations) on an idealised hillslope domain, forced by a single pulse of rain. To explore different topographies and microtopographies, we model 1460 surfaces, based on 10 sloping planes (from 0.1% to 10%) on which a sinusoidal microtopography of various amplitudes (from 1 to 10 cm) and wavelengths (from 15 to 200 cm) is overlaid. In a previous proof-of-concept work, we showed how these microtopograhies have an impact on rainfall-runoff-infiltration partitioning and generate different runoff regimes from disconnected flow to steady sheet flow. In this contribution, we extend our analysis to include a more realistic, time-dependent infiltration capacity, and therefore explore the effects this has in the process of ponding and establishing surface flow connectivity. We extend the number of surfaces (within the same ranges) to better observe the different runoff regimes. We quantitatively assess the results mainly in terms of the increase in total infiltration in the presence of microtopography relative to a smooth plane, and qualitatively in terms of the generated runoff regimes.

The results show that microtopography increases total infiltration (up to six times in our simulations) over the whole domain relative to a smooth plane and there is a strong non-linear dependency of infiltration and runoff on slope and on the ratio of the characteristic wavelength and amplitude of microtopography. Moreover, three characteristic regimes of influence of microtopography exist: one in which microtopography plays a negligible role, another in which microtopography increases infiltration, but the particular microtopography features are not very relevant, and one regime in which small changes in microtopography generate significant variations on infiltration. Such regimes are the result of the interplay between small (microtopography) and large scale (slope) system features. Finally, the results also show that the time-dependent infiltration capacity can enhance the effect of microtopography on infiltration. From a modelling perspective, these results hint that neglecting microtopography and time-dependent infiltration in hydrological modelling can lead to an underestimation of infiltration and an overestimation of runoff. The coupled analysis of spatial hydrodynamics and hydrological signatures suggests that the latter can be interpreted and explained by the spatiotemporal variations triggered by surface connectivity.

How to cite: Ahmadinia, E., Caviedes-Voullième, D., and Hinz, C.: Coupled effects of microtopography and time-dependant infiltration capacity on rainfall-runoff-infiltration partitioning on a hillslope, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-18200, https://doi.org/10.5194/egusphere-egu2020-18200, 2020.

We present a novel particle based numerical method, which supports regions of particle agglomeration, for simulating water flow in hydrological applications. This is done to tackle the difficulties that arise while modeling due to diverse operation scales, both in space and time. In this study we aim to concentrate only on multiple spatial scales. Smoothed particle hydrodynamics (SPH) is a mesh free method and it enables the interactions of particles in different media, for example media with different porosities. We use a technique of agglomerating particles based on parameters like velocity and treat the agglomerated mass as a single particle. With the presented method, the SPH method can be extended to rainfall-runoff models with multi-phase soil properties. First, the numerical method associated with SPH to solve the shallow water equation (SWE) is introduced. Then the way in which the mass term is replaced during agglomeration is derived. Calculating the modified parameters of a newly agglomerated particle to satisfy the continuity criteria is also introduced and derived. In order to validate the method, benchmark cases that align with our target application with experimental data were chosen from literature study. These include, uniform rainfall falling on an one-dimension flat slope channel, non-uniform rainfall with different duration over an one-dimension flat slope. In order to explore the extent of method a three-dimension test case, where water particles are allowed to pass through a different medium stacked on top of one another with different porosity, is chosen. The three-dimension benchmark case is not inspired from a real time application like the one-dimension test cases, but the results can be scaled and deployed into a flood-forecasting simulation. Also, the proposed method was proven robust and the one-dimension test cases show good agreement with experimental results. The three-dimension test case shows decent improvement in computational time and can provide new possibilities for simulating practical hydrological applications.

How to cite: Somasundaram, M., Gelleszun, M., and Meon, G.: Multi scale smoothed particle hydrodynamics using particle agglomeration for simulating rainfall-runoff processes, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-18951, https://doi.org/10.5194/egusphere-egu2020-18951, 2020.

The results of precision monitoring of the groundwater level are used as the main criterion in study of the mechanism of deformation of a water-saturated reservoir under dynamic impact. The purpose of the field investigation is a registration of the hydrogeological responses to the passage of seismic waves from mining of the iron ore deposit. An instrument-measuring complex of autonomous synchronous registration of seismic signals and hydrogeological responses to mass explosions during underground and surface mining was installed at the site of the iron ore deposit. For the first time, the amplitudes and frequency ranges of the hydrogeological responses of the different aquifers were determined in the near field zone. During mass explosions has been previously established that the reaction of aquifers to explosions is faster than the displacement of soil on the surface. The hydrogeological responses in the water-saturated sands, slates and quartzites are registered at the different frequencies. The change of the reservoir filtration properties under the mass explosions are not possible to judge due to the lack of a long series of observations. Only the continuation of the precise hydrogeological monitoring allows receiving a new data of the deformation of a water-saturated reservoir of pore and pore-fracture types under high-intensity exposure have been obtained. Probably these results can be used for the understanding of the hydrogeological and hydrogeomechanic processes in the near field of earthquakes. The reported study was funded by RFBR according to the research project № 19-05-00809.

How to cite: Gorbunova, E. and Besedina, A.: Study of the hydrogeological responses to mass explosions during mining at the iron ore deposit, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21353, https://doi.org/10.5194/egusphere-egu2020-21353, 2020.

Prediction of rainfall-runoff response in Alpine catchments is complex because hydrological processes vary strongly in space and time, they are elevation and temperature dependent, subsurface water stores are heterogeneous, snow plays an important role, and runoff response is fast. As a result, the transformation of rainfall into runoff is highly nonlinear. Machine Learning (ML) methods are suitable for reproducing such nonlinearities between input and output data and have been used for streamflow prediction. Recurrent Neural Networks (RNNs) with memory states, such as Long and Short-Term Memory (LSTM) models, are particularly suitable for hydrological variables that are dependent in time. An example of a recent application of LSTM to the rainfall-runoff transformation in many catchments in the USA showed that the LSTM model can learn physically meaningful catchment embeddings from precipitation-temperature-streamflow data, and performs comparably to widely used conceptual hydrological models (Kratzert et al., 2019).

In this study, we tested the LSTM approach on high-quality daily data from 23 Alpine catchments in Switzerland with three goals in mind. First, the LSTM model was trained and validated using daily climate variables (precipitation, air temperature, sunshine duration) and streamflow data on all catchments individually and the performance was compared to a distributed hydrological model (PREVAH). The performance of the LSTM model was in many (but not in all) cases better than the hydrological model. Second, a single LSTM model was trained in all catchments simultaneously, embedding terrain attributes extracted from the Digital Elevation Model (DEM). In this way differences between catchments related to the elevation and temperature dependent hydrological processes, such as snow accumulation and melt, evapotranspiration, runoff generation, etc., can be captured. We show the performance of this model and evaluate the regionalization potential provided by it. Third, the LSTM model was applied in an ensemble forecasting context, and we discuss the benefits and limitations this application brings compared to forecasting with a process-based hydrological model.

How to cite: Anand, M., Molnar, P., and Peleg, N.: Daily streamflow prediction using an LSTM neural network in Alpine catchments, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-21737, https://doi.org/10.5194/egusphere-egu2020-21737, 2020.