Dieses Lehrbuch wendet sich an Studienanfänger und Studienanfängerinnen der Natur- und Ingenieurswissenschaften. Es gibt eine grundlegende Einführung in die Elemente des Wasserkreislaufs. Erläutert werden der Einfluss von Landnutzung und Klima, Hochwasser und Dürre sowie moderne Verfahren zur Quantifizierung hydrologischer Prozesse. Schließlich wird ein Einblick in die hydrologische Praxis und die Risikovorsorge gegeben. Rund 165 Abbildungen und 35 Tabellen veranschaulichen die komplexen Sachverhalte, in einem Glossar werden wichtige Fachbegriffe erklärt.

Study region: Lowland Ems catchment in North-west Germany Study focus: Alterations of streamflow caused by climate change can impact the riverine ecosystems, socio-economic development and the effectiveness of flood protection measures. Therefore, hydrological climate impact studies are crucial for foresighted water management. In addition to conceptual and physical hydrological models, machine learning based models are frequently used in streamflow modelling. However, the applicability of machine learning models for climate impact studies remains relatively unexplored. In this study, we use a standard ensemble of climate change projections of the German Weather Service for the Ems catchment. We compare the climate change signals of two conceptual hydrological models to those of two different long short-term memory model approaches. Based on long-term simulations, climate change impact on different streamflow indices is assessed. New hydrological insights for the region: The results indicate heterogenous alterations of streamflow depending on the applied model and its model-specific input data. Further analysis reveals that the application of the Haude formula for calculation of potential evaporation is questionable for climate change impact assessment despite its successful application in the study region in recent climate. For a successful application of LSTM in climate change impact studies, physical restrictions need to be defined to simulate a plausible catchment behavior also for climate conditions beyond the range of training data.

Dynamic phenology has so far been a modelling aspect that has received little attention. However, it has been shown that leaf emergence takes place earlier due to the shift in vegetation phase caused by climate change and is strongly dependent on temperature. Here, we demonstrate the calibration of a model for dynamic phenology within the water balance model WaSiM. Temperature sums and dormancy are used as controlling variables. The calibration of the respective parameters was realised using a shuffled complex evolution algorithm. ETa relevant parameters were calibrated based on MODIS data as a reference. Evaluation was done by comparing the ETa curves to MODIS ETa curves as well as a comparison of spatial ETa patterns based on Landsat ETa data. The evaluation shows that the dynamic phenology model used is capable of predicting the start of leaf emergence while also leading to better fitting evapotranspiration curves for the deciduous forest compared with the initial static parameterisation approach.

ABSTRACT Machine learning algorithms are increasingly applied in hydrological studies with promising results. However, these algorithms generally lack the ability for easy interpretability of the results by users. In this study, we compare six different explainable artificial intelligence (XAI) algorithms that help understand the effect of input data on the simulation results. The methods are explored on two distinct approaches for streamflow modeling using a long short-term memory (LSTM) model: a single model approach using only meteorological forcing data and a regional approach including also static catchment attributes. To gain further insight into the internal dynamics of the LSTM models, the relationship between cell states and soil moisture is investigated. A strong correlation suggests that the LSTM models inherently capture the concept of soil moisture as a catchment-scale storage mechanism. The XAI methods are applied to derive a timestep of influence, revealing how many days of input data are relevant for the model output. All XAI methods result in similar seasonal patterns in the timestep of influence, suggesting that the methods are comparable. Setting soil moisture dynamics in context to seasonal development of the timestep of influence suggests resetting LSTM as soon as soil moisture saturation occurs.

Bias in regional climate model (RCM) data makes bias correction (BC) a necessary pre-processing step in climate change impact studies. Among a variety of different BC methods, quantile mapping (QM) is a popular and powerful BC method. Studies have shown that QM may be vulnerable to reductions in calibration sample size. The question is whether this also affects the climate change signal (CCS) of the RCM data. We applied four different QM methods without subsampling and with three different subsampling timescales to an ensemble of seven climate projections. BC generally improved the RCM data relative to observations. However, the CCS was significantly modified by the BC for certain combinations of QM method and subsampling timescale. In conclusion, QM improves the RCM data that are fundamental for climate change impact studies, but the optimal subsampling timescale strongly depends on the chosen QM method.

The aim of this study was to simulate dominant runoff generation processes (DRPs) in a mesoscale catchment in southwestern Germany with the physically-based distributed hydrological model WaSiM-ETH and to compare the resulting DRP patterns with a data-mining-based digital soil map. The model was parameterized by using 11 Pedo-transfer functions (PTFs) and driven by multiple synthetic rainfall events. For the pattern comparison, a multiple-component spatial performance metric (SPAEF) was applied. The simulated DRPs showed a large variability in terms of land use, applied rainfall rates, and the different PTFs, which highly influence the rapid runoff generation under wet conditions.

Machine learning (ML) algorithms slowly establish acceptance for the purpose of streamflow modelling within the hydrological community. Yet, generally valid statements about the modelling behavior of the ML models remain vague due to the uniqueness of catchment areas. We compared two ML models, RNN and LSTM, to the conceptual hydrological model Hydrologiska Byråns Vattenbalansavdelning (HBV) within the low-land Ems catchment in Germany. Furthermore, we implemented a simple routing routine in the ML models and used simulated upstream streamflow as forcing data to test whether the individual model errors accumulate. The ML models have a superior model performance compared to the HBV model for a wide range of statistical performance indices. Yet, the ML models show a performance decline for low-flows in two of the sub-catchments. Signature indices sampling the flow duration curve reveal that the ML models in our study provide a good representation of the water balance, whereas the HBV model instead has its strength in the reproduction of streamflow dynamics. Regarding the applied routing routine in the ML models, there are no strong indications of an increasing error rising upstream to downstream throughout the sub-catchments.

<p>Machine Learning and Deep Learning have been proving their potential for streamflow modelling in various studies. In particular, long short-term memory (LSTM) models showed exceptionally good results. However, machine learning models often are considered “black boxes” with limited interpretability. Explainable artificial intelligence (XAI) comprise methods that analyze the internal processes of the machine learning network and allow to have a glance in the “black box”. Most proposed XAI techniques are designed for the analysis of images, and there is currently only limited work on time series data available.</p> <p>In our study, we applied various XAI algorithms including gradient-based methods (Saliency, InputXGradient, Integrated Gradient, GradientSHAP) but also perturbation-based methods (Feature Ablation, Feature Permutation) to compare their applicability for reasonable interpretation in the hydrological context. To our knowledge, only Integrated Gradient has been applied to a LSTM in hydrology so far. Gradient-based methods analyze the gradient of the output with respect to the input feature. Whereas perturbation-based methods gain information by altering or masking specific input features. The different methods were applied to a LSTM trained for the low-land Ems catchment in Germany, which has a major baseflow share of total streamflow.</p> <p>We analyzed the results regarding their “timestep of influence”, which describes the amount of past days having importance for the prediction of streamflow at a particular day. All of the algorithms applied result in a comparable annual pattern, characterized by relatively small timesteps of influence in spring (wet season) and increasing timesteps of influence in summer and autumn (dry season). However, the range of the absolute days of attribution varies between the methods. In conclusion, all methods produces reasonable results and appear to be suitable for interpretation purposes.</p> <p>Furthermore, we compare the results to ERA-5 reanalysis data and gained evidence that the LSTM recognizes soil water storage as the main driver for streamflow generation in the catchment: we found an inverse seasonality of soil moisture and timestep of influence.</p>

The land-use-specific calibration of evapotranspiration parameters in hydrologic modeling is challenging due to the lack of appropriate reference data. We present a MODIS-based calibration approach of vegetation-related evaporation parameters for two mesoscale catchments in western Germany with the physically based distributed hydrological model WaSiM-ETH. Time series of land-use-specific actual evapotranspiration (ETa) patterns were generated from MOD16A2 evapotranspiration and CORINE land-cover data from homogeneous image pixels for the major land-cover types in the region. Manual calibration was then carried out for 1D single-cell models, each representing a specific land-use type based on aggregated 11-year mean ETa values using SKout and PBIAS as objective functions (SKout > 0.8, |PBIAS| < 5%). The spatio-temporal evaluation on the catchment scale was conducted by comparing the simulated ETa pattern to six daily ETa grids derived from LANDSAT data. The results show a clear overall improvement in the SPAEF (spatial efficiency metric) for most land-use types, with some deficiencies for two scenes in spring and late summer due to phenological variation and a particularly dry hydrological system state, respectively. The presented method demonstrates a significant improvement in the simulation of ETa regarding both time and spatial scale.

Soil compaction by agricultural vehicles is regarded as a threat to soil functions. An important strategy to mitigate traffic-induced compaction might be avoidance of traffic on soils which are currently highly susceptible to compaction and adaption of machinery to site conditions. A spatial decision support system (sDSS) for farmers might help to reduce compaction risks by providing model-based information on site-specific, current compaction risk. As one part of the project "Smart Soil Information for Farmers", published models for compaction risk were assessed regarding their potential for implementation in an app-based-sDSS. As a first step, these models are evaluated based on wheeling experiments for selected sites and vehicles. Pre-selection of models resulted in two combinations that differ in terms of the required input data and the underlying modelling concept: * Combination 1 (C1) derives the precompression stress as a measure of soil strength parameter using pedotransfer functions and calculates compaction risk based on semi-analytical solutions for stress transmission (according to Keller et al., 2007). * Combination 2 (C2) derives the compaction risk according to Lorenz et al. (2016) as a combination of a susceptibility class (based on soil texture and moisture) and a load-input class from machinery parameters. Evaluation of modelling results is based on wheeling experiments on two test sites (loamy sand vs. clayey loam) and different agricultural vehicles (total mass 10 to 38 t). Compaction by vehicles was assessed by measuring soil physical and mechanical parameters before and after wheeling. Soil physical measurements included dry bulk density, pore size distribution, water and air conductivity. Mechanical parameters included in situ soil stress during passage of vehicles, precompression stress and shear strength. In all experiments, traffic had clear negative effects on physical properties in the topsoil (increase in bulk density, decrease in air capacity and water/air permeability). In the subsoil, only small effects were found for changes in physical and mechanical properties. This can presumably be explained by a “plough-pan” that increased load-bearing capacity. Comparing both models, it was found that C1 generally tends to predict higher compaction risks than C2. For the topsoil, C1 was able to predict the observed effects better than C2. For the subsoil, relatively small observed effects were generally better represented by model C2, which predicted lower risks than C1 for the subsoil. References * Keller et al.: SoilFlex: A model for prediction of soil stresses and soil compaction due to agricultural field traffic. Soil and Tillage Research 93 (2007), 2/391–411 * Lorenz et al: Anpassung der Lasteinträge landwirtschaftlicher Maschinen an die Verdichtungsempfindlichkeit des Bodens. Landbauforschung (2016), 66/101–144

Pedotransferfunctions (PTF) play a major role in physically based hydrological modeling, as they establish the relationship between soil properties and the water stress curve. The selection of the PTF thus has a great influence on the water balance of a catchment in general, as well as on the spatial distribution and intensity of runoff processes. In this study, these very influences of PTF's on the runoff processes will be investigated. The hydrological model "WaSiM-ETH", which was calibrated and validated, is used for the investigations. On the basis of this modeling, 11 further scenarios were created, which differ only in their PTFs. In all scenarios an artificial weather event is applied, which includes a constant heavy precipitation of 100 mm as well as input data, which excludes the generation of interfering processes e.g., evaporation or snowfall. In addition, these scenarios will be applied to different system conditions to determine the differences of dry, humid, and wet system preconditions. Ultimately, the precipitation intensity but not the amount of the artificial weather event will also be varied to be able to determine any change in the dominant runoff processes due to precipitation intensity. The results of the modeling will then be compared using the generated surface runoff, interflow, and deep infiltration. In addition, a check of the modeling with a runoff process map available for the catchment area as well as a pattern comparison using the spatial efficiency metric (SPEAF) will be performed. It is expected that the different PTF´s per se, as well as depending on the system precondition and precipitation intensity, will result in very different spatial distributions and dominance of the individual runoff processes. Thus, one goal is to find the PTF´s that provide comprehensible distributions and intensities of the considered runoff processes.

In physically based catchment modeling, one of the most crucial tasks is to parameterize the soils, where water balance components are highly sensitive to the spatial structure of soil hydraulic properties. To estimate these parameters, Pedo-Transfer functions (PTFs) are applied, which define the functional relationships that transfer available measurable soil properties into missing soil parameters. By selecting different PTFs, a wide range of different hydrological model behavior evolves. However, this might still result in an acceptable representation of the discharge hydrograph at the catchment outlet. Hence, model evaluation using streamflow measurements exclusively at catchment outlet may lead to implausible results and possibly to inappropriate decision making in water management, since this approach does not consider the spatiotemporal variability of hydrologic states and fluxes such as evapotranspiration (ET). Therefore, this study aims at aligning the soil parameterization towards mapping a plausible hydrological behavior. To carry this out, additional information on ET patterns (Landsat-8 satellite images, SSEBop method) are employed. Therefore. ET patterns simulated by hydrologic model (WaSiM-ETH) are evaluated with reference to the corresponding remotely sensed estimates using a spatial efficiency metric. Results show that simulated ET patterns are in reasonable accordance with satellite driven patterns, however, hydrologic model tends to overestimate the ET rates in summer. Moreover, individual PTFs demonstrate a distinct difference in patterns of ET, owing to the fact that estimation of different soil hydraulic parameters by different PTFs leads to a change in soil water storage capacity as well as water redistribution within the corresponding soils.

In every ecosystem, the calculation of water balance in the hydrological cycle requires accurate estimations of different processes and water components/fluxes. To address this need, this chapter first discusses the concept of the hydrological cycle and water balance at different spatial scales (from plot to global/continental). Then, the components/fluxes of the hydrological cycle are analyzed, with a specific focus on atmospheric, surface, and sub-surface waters. Finally, an overview of the different approaches to calculate the water balance is presented, including the in-situ measurement methods as well as the hydrological modeling.

Soil hydraulic properties, which are basically saturated and unsaturated hydraulic conductivity and water retention characteristics, remarkably control the main hydrological processes in catchments. Thus, adequate parameterization of soils is one of the most important tasks in physically based catchment modeling. To estimate these properties, the choice of the PTFs in a hydrological model is often made without taking the runoff characteristics of the catchment into consideration. Therefore, this study introduces a methodology to analyze the sensitivity of a catchment water balance model to the choice of the PTF. To do so, we define 11 scenarios including different combinations of PTFs to estimate the van Genuchten parameters and saturated hydraulic conductivity. We use a calibrated/validated hydrological model (WaSiM-ETH) as a baseline scenario. By altering the underlying PTFs, the effects on the hydraulic properties are quantified. Moreover, we analyze the resulting changes in the spatial/temporal variation of the total runoff and in particular, the runoff components at the catchment outlet. Results reveal that the water distribution in the hydrologic system varies considerably amongst different PTFs, and the water balance components are highly sensitive to the spatial structure of soil hydraulic properties. It is recommended that models be tested by careful consideration of PTFs and orienting the soil parameterization more towards representing a plausible hydrological behavior rather than focusing on matching the calibration data.

Laboratory landslide experiments enable the observation of specific properties of these natural hazards. However, these observations are limited by traditional techniques: frequently used high-speed video analysis and wired sensors (e.g. displacement). These techniques lead to the drawback that either only the surface and 2D profiles can be observed or wires confine the motion behaviour. In contrast, an unconfined observation of the total spatiotemporal dynamics of landslides is needed for an adequate understanding of these natural hazards. The present study introduces an autonomous and wireless probe to characterize motion features of single clasts within laboratory-scale landslides. The Smartstone probe is based on an inertial measurement unit (IMU) and records acceleration and rotation at a sampling rate of 100 Hz. The recording ranges are ±16 g (accelerometer) and ±2000∘ s−1 (gyroscope). The plastic tube housing is 55 mm long with a diameter of 10 mm. The probe is controlled, and data are read out via active radio frequency identification (active RFID) technology. Due to this technique, the probe works under low-power conditions, enabling the use of small button cell batteries and minimizing its size. Using the Smartstone probe, the motion of single clasts (gravel size, median particle diameter d50 of 42 mm) within approx. 520 kg of a uniformly graded pebble material was observed in a laboratory experiment. Single pebbles were equipped with probes and placed embedded and superficially in or on the material. In a first analysis step, the data of one pebble are interpreted qualitatively, allowing for the determination of different transport modes, such as translation, rotation and saltation. In a second step, the motion is quantified by means of derived movement characteristics: the analysed pebble moves mainly in the vertical direction during the first motion phase with a maximal vertical velocity of approx. 1.7 m s−1. A strong acceleration peak of approx. 36 m s−2 is interpreted as a pronounced hit and leads to a complex rotational-motion pattern. In a third step, displacement is derived and amounts to approx. 1.0 m in the vertical direction. The deviation compared to laser distance measurements was approx. −10 %. Furthermore, a full 3D spatiotemporal trajectory of the pebble is reconstructed and visualized supporting the interpretations. Finally, it is demonstrated that multiple pebbles can be analysed simultaneously within one experiment. Compared to other observation methods Smartstone probes allow for the quantification of internal movement characteristics and, consequently, a motion sampling in landslide experiments.

The files contain data of four Smartstone probes recorded during a landslide experiment. The MATLAB data files are named after the particular pebbles (1 to 4), which were equipped with the probes. The files contain acceleration (acc) and gyroscope (gyr) data as well as a time code (time).

Evapotranspiration is often estimated by numerical simulation. However, to produce accurate simulations, these models usually require on-site measurements for parameterization or calibration. We have to make sure that the model realistically reproduces both, the temporal patterns of soil moisture and evapotranspiration. In this study, we combine three sources of information: (i) measurements of sap velocities; (ii) soil moisture; and (iii) expert knowledge on local runoff generation and water balance to define constraints for a “behavioral” forest stand water balance model. Aiming for a behavioral model, we adjusted soil moisture at saturation, bulk resistance parameters and the parameters of the water retention curve (WRC). We found that the shape of the WRC influences substantially the behavior of the simulation model. Here, only one model realization could be referred to as “behavioral”. All other realizations failed for a least one of our evaluation criteria: Not only transpiration and soil moisture are simulated consistently with our observations, but also total water balance and runoff generation processes. The introduction of a multi-criteria evaluation scheme for the detection of unrealistic outputs made it possible to identify a well performing parameter set. Our findings indicate that measurement of different fluxes and state variables instead of just one and expert knowledge concerning runoff generation facilitate the parameterization of a hydrological model.

For impact assessment of interacting flood reservoirs in complex catchments it is required to generate hydrographs, which represent the peak processing of the flood in the entire catchment. Concerning unobserved or regulated (downriver of existing reservoirs) catchments a rainfall-runoff model - as presented in this analysis - is essential, to create these hydrographs. This paper concentrates on the question, how different recommended practices for generating hydrographs manipulate the results focusing on the precipitation sequence. A new method is developed to generate hydrographs using standardized observed rainfall sequences and method performance is discussed in comparison to common practice using synthetic precipitation sequences. Pre-analysis of flood generation and hydrographs are necessary, as realized for the example of the Sylvenstein reservoir in the alpine catchment of the river Isar on the basis of the three observed severest flood events. The observed hydrographs are compared with the hydrographs generated by the new method and those are discussed concerning peak overlap for the entire catchment. It can be demonstrated, that the new method using standardized observed rainfall sequences is more adapted to reproduce the peak overlap in the catchment, in case of weather situations with tracking rainfall, because choked flow at the alpine mountains is considered. For similar cases it can be recommended to include the new method as a variant, in order to represent the natural process more adequate. The uncertainties of the calculation can be shown via ensembles with different runoff coefficients. During calculation it turned out, that not only rainfall sequence and preconditions are highly relevant for calibration of the rainfall-runoff model runoff coefficient (how expected) but although the method of flood routing and the roughness coefficients.

This paper presents management of groundwater resource using a Bayesian Decision Network (BDN). The Kordkooy region in North East of Iran has been selected as study area. The region has been sub-divided into three zones based on transmissivity (T) and electrical conductivity (EC) values. The BDN parameters: prior probabilities and Conditional Probability Tables - CPTs) have been identified for each of the three zones. Three groups of management scenarios have been developed based on the two decision variables including "Crop pattern" and "Domestic water demand" across the three zones of the study area: 1) status quo management for all three zones represent current conditions; 2) the effect of change in cropping pattern on management endpoints and 3) the effect of future increased domestic water demand on management endpoints. The outcomes arising from implementing each scenario have been predicted by use of the constructed BDN for each of the zones. Results reveal that probability of drawdown in groundwater levels of southern areas is relatively high compared with other zones. Groundwater withdrawal from northern and northwestern areas of the study area should be limited due to the groundwater quality problems associated with shallow groundwater of these two zones. The ability of the Bayesian Decision Network to take into account key uncertainties in natural resources and perform meaningful analysis in cases where there is not a vast amount of information and observed data available – and opportunities for enabling inputs for the analysis based partly on expert elicitation,emphasizes key advantages of this approach for groundwater management and addressing the groundwater related problems in a data-scarce area.