Flood Generation Processes Research Articles

Compound and consecutive drought-flood events at a global scale

Flooding during or after droughts poses significant challenges to disaster risk management. However, interactions between droughts and floods are often overlooked as studies typically analyse these events in isolation. Here we explore historical occurrences of compound and consecutive drought-flood events and drought effects on flood severity and timing by analysing global datasets of hydrometeorological and biophysical variables for 8255 catchments worldwide. These data show that 24% of floods globally (anomalies above the 85th percentile) are preceded by, or happen during, drought conditions. Flood events occurring during drought conditions are typically of lower magnitude, especially in arid regions, while floods following drought events have a severity distribution comparable to single flood events. For most drought-flood events, flood timing appears relatively unaffected by drought conditions, but almost a quarter of the drought-flood events had flood timings occurring two to three months later than expected. These shifts in flood timing suggest droughts potentially affect flood-generating processes. As both drought and flood occurrences are projected to increase in a warming climate, interactions between them may become more common and need to be accounted for in flood risk assessment and management.

Modelling multidecadal variability in flood frequency using the Two-Component Extreme Value distribution

The Two-Component Extreme Value (TCEV) distribution is traditionally known as the exact distribution of extremes arising from Poissonian occurrence of a mixture of two exponential exceedances. In some regions, flood frequency is affected by low-frequency (decadal) climate fluctuations resulting in wet and dry epochs. We extend the exact distribution of extremes approach to such regions to show that the TCEV arises as the distribution of annual maximum floods for Poissonian occurrences and (at least two) exponential exceedances. A case study using coastal basins in Queensland and New South Wales (Australia) affected by low-frequency climate variability, shows that the TCEV produces good fits to the marginal distribution over the entire range of observed values without the explicit need to resort to climate covariates and removal of potentially influential low values. Moreover, the TCEV reproduces the observed dog-leg, a key signature of different flood generation processes. A literature review shows that the assumptions underpinning the TCEV are conceptually consistent with available evidence on climate and flood mechanisms in these basins. We provide an extended domain of the TCEV distribution in the L-moment ratio diagram to account for the wider range of parameter values encountered in the case study and show that for all basins, L-skew and L-kurtosis fall within the extended domain of the TCEV.

Neural networks for fast fluvial flood predictions: Too good to be true?

AbstractAccurate models with low computational times are required to predict the consequences of fluvial floods in real‐time. Even though detailed hydraulic models can predict flood water levels and corresponding inundation extents with high accuracy, their computational times limit their applicability to being used as a flood early‐warning system. Therefore, conceptual models were developed in the literature for many years. These types of models do not attempt to represent the complex dynamic flood generation processes but are based on simplified hydraulic concepts, generally only using a digital elevation model as input. However, a shift in this research field is currently present from conceptual models to data‐driven models, and more specifically to neural networks. This paper discusses the benefits and drawbacks of both modelling approaches and speculates which method is most promising to be used as a flood warning system to predict the consequences of fluvial floods in real‐time.

Shifts in flood generation processes exacerbate regional flood anomalies in Europe

Anomalies in the frequency of river floods, i.e., flood-rich or -poor periods, cause biases in flood risk estimates and thus make climate adaptation measures less efficient. While observations have recently confirmed the presence of flood anomalies in Europe, their exact causes are not clear. Here we analyse streamflow and climate observations during 1960-2010 to show that shifts in flood generation processes contribute more to the occurrence of regional flood anomalies than changes in extreme rainfall. A shift from rain on dry soil to rain on wet soil events by 5% increased the frequency of flood-rich periods in the Atlantic region, and an opposite shift in the Mediterranean region increased the frequency of flood-poor periods, but will likely make singular extreme floods occur more often. Flood anomalies driven by changing flood generation processes in Europe may further intensify in a warming climate and should be considered in flood estimation and management.

How uncertain are precipitation and peak flow estimates for the July 2021 flooding event?

Abstract. The disastrous July 2021 flooding event made us question the ability of current hydrometeorological tools in providing timely and reliable flood forecasts for unprecedented events. This is an urgent concern since extreme events are increasing due to global warming, and existing methods are usually limited to more frequently observed events with the usual flood generation processes. For the July 2021 event, we simulated the hourly streamflows of seven catchments located in western Germany by combining seven partly polarimetric, radar-based quantitative precipitation estimates (QPEs) with two hydrological models: a conceptual lumped model (GR4H) and a physically based, 3D distributed model (ParFlowCLM). GR4H parameters were calibrated with an emphasis on high flows using historical discharge observations, whereas ParFlowCLM parameters were estimated based on landscape and soil properties. The key results are as follows. (1) With no correction of the vertical profiles of radar variables, radar-based QPE products underestimated the total precipitation depth relative to rain gauges due to intense collision–coalescence processes near the surface, i.e., below the height levels monitored by the radars. (2) Correcting the vertical profiles of radar variables led to substantial improvements. (3) The probability of exceeding the highest measured peak flow before July 2021 was highly impacted by the QPE product, and this impact depended on the catchment for both models. (4) The estimation of model parameters had a larger impact than the choice of QPE product, but simulated peak flows of ParFlowCLM agreed with those of GR4H for five of the seven catchments. This study highlights the need for the correction of vertical profiles of reflectivity and other polarimetric variables near the surface to improve radar-based QPEs for extreme flooding events. It also underlines the large uncertainty in peak flow estimates due to model parameter estimation.

Spatial dependence of floods shaped by extreme rainfall under the influence of urbanization.

Weather system are spatially and temporally dependent, and these dependencies can result in flood events with similar behaviors. While it is well known that urbanization profoundly impacts the flood generation process, much less is known about the impacts of urbanization on the spatial dependence of floods, a major determinant of flood risk severity. To this end, a scheme was proposed to detect the flood dependence variations influenced by urbanization. Based on the scheme, we found that 1) the flood dependence can be weakened with extreme rainfall increasing from a short return period to a long return period; 2) The flood dependencies can be amplified in slightly urbanized regions and mitigated in highly urbanized regions due to intensifying urbanization. In addition, the change characteristics of the flood dependencies influenced by urbanization are first identified from the perspective of network structure. We found that urbanization can significantly affect the network structure (i.e., hub and connectivity) of flood dependence, especially in highly urbanized regions. The catchments with high hub and connectivity are prone to widespread floods and should be given more attention in flood warning and control management work, which can contribute to helping defend against floods in hazard-prone areas.

Uncertainty sources in flood projections over contrasting hydrometeorological regimes

ABSTRACT This study evaluates the uncertainty of four components of the hydroclimatic modelling chain on flood projections over 96 basins covering contrasting hydrometeorological regimes located in Canada and Mexico. Two ensembles of climate simulations are considered, a large ensemble of 22 global climate model simulations and a smaller ensemble of three high-resolution regional climate model simulations. The other components are two post-processing techniques, three lumped hydrological models and six probability distributions. These four sources are assessed through a method of variance decomposition applied to six flood indicators over a reference period and two future periods: 1976–2005, 2041–2070 and 2070–2099. Systematic differences are observed between basins with contrasting flood-generating processes. Snow-dominated basins consistently show larger variance contributions from hydrological models, while rain-dominated basins show climate simulations as their dominant source. These results underline the need to consider the variability of each component’s uncertainty contribution and its link to hydroclimatic conditions and dominant processes.

Classification of flood-generating processes in Africa

River flooding has large societal and economic impacts across Africa. Despite the importance of this topic, little is known about the main flood generating mechanisms in Africa. This study is based on 13,815 flood events that occurred between 1981 and 2018 in 529 catchments. These flood events are classified to identify the different flood drivers: excess rains, long rains and short rains. Out of them, excess rains on saturated soils in Western Africa, and long rains for catchments in Northern and Southern Africa, are the two dominant mechanisms, contributing to more than 75% of all flood events. The aridity index is strongly related to the spatial repartition of the different flood generating processes showing the climatic controls on floods. Few significant changes were detected in the relative importance of these drivers over time, but the rather short time series available prevent a robust assessment of flood driver changes in most catchments. The major implication of these results is to underline the importance of soil moisture dynamics, in addition to rainfall, to analyze the evolution of flood hazards in Africa.

Snow-influenced floods are more strongly connected in space than purely rainfall-driven floods

Widespread floods that affect several catchments are associated with large damages and costs. To improve flood protection, a better understanding of the driving processes of such events is needed. Here, we assess how spatial flood connectedness varies with the flood generation process using a flood event classification scheme that distinguishes between rainfall-driven and snowmelt-influenced flood types. Our results show that the dominant flood generation processes in Europe vary by region, season, and event severity. Specifically, we show that severe floods are more often associated with snow-related processes than moderate events. In addition, we find that snow-influenced events show stronger spatial connections than rainfall-driven events. The spatial connectedness of rainfall-driven events depends on the rainfall duration, and the connectedness decreases with increasing duration. These findings have potential implications for flood risk in a warming climate, both locally and regionally. The projected decrease in the frequency of occurrence of snowmelt-influenced floods may translate into a decrease in the frequency of severe and widespread floods in catchments where snowmelt processes are important for flood generation.

Flood generation: process patterns from the raindrop to the ocean

Abstract. This article reviews river flood generation processes and flow paths across space scales. The scale steps include the pore, profile, hillslope, catchment, regional and continental scales, representing a scale range of a total of 10 orders of magnitude. Although the processes differ between the scales, there are notable similarities. At all scales, there are media patterns that control the flow of water and are themselves influenced by the flow of water. The processes are therefore not spatially random (as in thermodynamics) but organized, and preferential flow is the rule rather than the exception. Hydrological connectivity, i.e. the presence of coherent flow paths, is an essential characteristic at all scales. There are similar controls on water flow and thus on flood generation at all scales but with different relative magnitudes. Processes at lower scales affect flood generation at larger scales, not simply as a multiple repetition of pore-scale processes but through interactions which cause emergent behaviour of process patterns. For this reason, when modelling these processes, the scale transitions need to be simplified in a way that reflects the relevant structures (e.g. connectivity) and boundary conditions (e.g. groundwater table) at each scale. In conclusion, it is argued that upscaling as the mere multiple application of small-scale process descriptions will not capture the larger-scale patterns of flood generation. Instead, there is a need to learn from observed patterns of flood generation processes at all spatial scales.

Identifying sensitivities in flood frequency analyses using a stochastic hydrologic modeling system

Abstract. This study employs a stochastic hydrologic modeling framework to evaluate the sensitivity of flood frequency analyses to different components of the hydrologic modeling chain. The major components of the stochastic hydrologic modeling chain, including model structure, model parameter estimation, initial conditions, and precipitation inputs were examined across return periods from 2 to 100 000 years at two watersheds representing different hydroclimates across the western USA. A total of 10 hydrologic model structures were configured, calibrated, and run within the Framework for Understanding Structural Errors (FUSE) modular modeling framework for each of the two watersheds. Model parameters and initial conditions were derived from long-term calibrated simulations using a 100 member historical meteorology ensemble. A stochastic event-based hydrologic modeling workflow was developed using the calibrated models in which millions of flood event simulations were performed for each basin. The analysis of variance method was then used to quantify the relative contributions of model structure, model parameters, initial conditions, and precipitation inputs to flood magnitudes for different return periods. Results demonstrate that different components of the modeling chain have different sensitivities for different return periods. Precipitation inputs contribute most to the variance of rare floods, while initial conditions are most influential for more frequent events. However, the hydrological model structure and structure–parameter interactions together play an equally important role in specific cases, depending on the basin characteristics and type of flood metric of interest. This study highlights the importance of critically assessing model underpinnings, understanding flood generation processes, and selecting appropriate hydrological models that are consistent with our understanding of flood generation processes.

Regionalisation of flood frequencies based on flood type-specific mixture distributions

The regionalisation of flood frequencies is a precondition for the estimation of flood statistics for ungauged basins. It is often based on either the concept of hydrological similarity of catchments or spatial proximity. Similarity is usually defined by comparing catchment attributes or distances. Here, we apply flood types in regionalisation directly to consider the type-specific aspects of similarity. The different flood types are classified according to their meteorological causes and hydrographs. Their probability distributions are modelled by type-specific distribution functions which are combined into one statistical annual mixture model afterwards. For regionalisation, we specified the parameters of each type-specific probability distribution separately with hierarchical clustering and regressions from catchment attributes. By selection of most relevant features, depending on the flood type, the specifics of flood-generating processes and meteorological causes were considered. The results demonstrate how this consideration of deterministic aspects can improve the transferability of distribution parameters to ungauged catchments. The type-specific regionalisation approach offers a higher degree of freedom for regionalisation as it describes the relationships between catchment characteristics, meteorological causes of floods and response of watersheds.

Managing flood risks in a changing climate

Managing flood risks in a changing climate

Identifying major drivers of daily streamflow from large-scale atmospheric circulation with machine learning

Previous studies linking large-scale atmospheric circulation and river flow with traditional machine learning techniques have predominantly explored monthly, seasonal or annual streamflow modelling for applications in direct downscaling or hydrological climate-impact studies. This paper identifies major drivers of daily streamflow from large-scale atmospheric circulation using two reanalysis datasets for six catchments in Norway representing various Köppen-Geiger climate types and flood-generating processes. A nested loop of roughly pruned random forests is used for feature extraction, demonstrating the potential for automated retrieval of physically consistent and interpretable input variables. Random forest (RF), support vector machine (SVM) for regression and multilayer perceptron (MLP) neural networks are compared to multiple-linear regression to assess the role of model complexity in utilizing the identified major drivers to reconstruct streamflow. The machine learning models were trained on 31 years of aggregated atmospheric data with distinct moving windows for each catchment, reflecting catchment-specific forcing-response relationships between the atmosphere and the rivers. The results show that accuracy improves to some extent with model complexity. In all but the smallest, rainfall-driven catchment, the most complex model, MLP, gives a Nash-Sutcliffe Efficiency (NSE) ranging from 0.71 to 0.81 on testing data spanning five years. The poorer performance by all models in the smallest catchment is discussed in relation to catchment characteristics, sub-grid topography and local variability. The intra-model differences are also viewed in relation to the consistency between the automatically retrieved feature selections from the two reanalysis datasets. This study provides a benchmark for future development of deep learning models for direct downscaling from large-scale atmospheric variables to daily streamflow in Norway.

Biases in national and continental flood risk assessments by ignoring spatial dependence

Recently, flood risk assessments have been extended to national and continental scales. Most of these assessments assume homogeneous scenarios, i.e. the regional risk estimate is obtained by summing up the local estimates, whereas each local damage value has the same probability of exceedance. This homogeneity assumption ignores the spatial variability in the flood generation processes. Here, we develop a multi-site, extreme value statistical model for 379 catchments across Europe, generate synthetic flood time series which consider the spatial correlation between flood peaks in all catchments, and compute corresponding economic damages. We find that the homogeneity assumption overestimates the 200-year flood damage, a benchmark indicator for the insurance industry, by 139%, 188% and 246% for the United Kingdom (UK), Germany and Europe, respectively. Our study demonstrates the importance of considering the spatial dependence patterns, particularly of extremes, in large-scale risk assessments.

Assessment of past flood changes across Europe based on flood-generating processes

ABSTRACT A rainfall–runoff model was employed to identify four major flood-generating processes corresponding to flood events identified from daily discharge data from 614 stations across Europe in the period 1961–2010: long-rain, short-rain, snowmelt, and rain-on-dry-soil flood events. Trend analyses were performed on the frequency of occurrence of each of the flood types continentally and in five geographical regions of Europe. Continentally, the annual frequency of flood events did not show a significant change over the investigation period. However, the frequency of both winter and summer long-rain events increased significantly, while that of summer snowmelt events decreased significantly. Regionally, the frequency of winter short and long-rain events increased significantly in Western Europe, while the frequency of summer snowmelt and short-rain events decreased in Northern Europe. The frequency of summer snowmelt events in Eastern Europe and winter short-rain events in Southern Europe showed a declining trend.

Flood frequency estimation in New Zealand using a region of influence approach and statistical depth functions

A region of influence approach in which statistical depth functions are used to select gauged catchments having similar flood generation behaviour to a given ungauged basin coupled with nonparametric regression on catchment characteristics is employed to estimate mean annual and a quantile flood of 100 year return period in New Zealand basins. This approach may be applied internationally. Relative and root mean square errors assessed by jack-knife resampling are too large to be useful for detailed design in ungauged catchments but are much smaller than those resulting from a conventional multiple nonlinear regression approach. The distribution of errors for samples of sites indicates that large error values result from contributions from a relatively small number of statistical outliers Although these are all natural catchments, the outlier basins have particular physical characteristics such as moraines, karst features and pumice sediments which affect their flood generation processes. Relative errors in estimates for an ungauged basin may be predicted using a relationship involving differences in area, rainfall and statistical depth between the ungauged catchment and gauged similar catchments. Improved analytical techniques and selection and measurement of catchment characteristics should reduce error magnitude but not perhaps sufficiently for use in detailed design. Parallel development based on rainfall-runoff models is desirable but the present lack of requisite information largely concerning rainfalls means that attainment of desired accuracy is a distant prospect.

The role of spatial and temporal model resolution in a flood event storyline approach in western Norway

We apply a physical climate storyline approach to an autumn flood event in the West Coast of Norway caused by an atmospheric river to demonstrate the value and challenges of higher spatial and temporal resolution in simulating flood impacts. We use a modelling chain whose outputs are familiar and used operationally, for example to issue flood warnings. With two different versions of a hydrological model, we show that (1) the higher spatial resolution between the global and regional climate model is necessary to realistically simulate the high spatial variability of precipitation in this mountainous region and (2) only with hourly data are we able to capture the fast flood-generating processes leading to the peak streamflow. The higher resolution regional atmospheric model captures the fact that with the passage of an atmospheric river, some valleys receive high amounts of precipitation and others not, while the coarser resolution global model shows uniform precipitation in the whole region. Translating the event into the future leads to similar results: while in some catchments, a future flood might be much larger than a present one, in others no event occurs as the atmospheric river simply does not hit that catchment. The use of an operational flood warning system for future events is expected to facilitate stakeholder engagement.

Regional flood frequency analysis: evaluation of regions in cluster space using support vector regression

Regional flood frequency analysis (RFFA) remains an area of active research around the world. Obtaining reliable RFFA estimates is critical for many hydrological and environmental studies. To this end, RFFA models based on multidimensional scaling (MDS) are coupled with support vector regression (SVR) to obtain improved flood quantile estimates at ungauged sites. MDS is used to form regions in catchment data space using the site characteristics and geographical distance from gauged sites. Then SVR models are applied to identify the functional relationships between flood quantiles and the catchment characteristics variables in MDS space, in particular SVR is used to capture the complexity and nonlinearity in flood generation processes that may not be easily captured in loglinear models. Four SVR models: linear, polynomial, radial basis function (RBF) and the sigmoid kernel models, are developed. The proposed approaches are applied to 202 catchments in the States of New South Wales and Victoria, Australia. The leave-one-out validation procedure is used to evaluate the performance of the proposed models. Results of the proposed models are compared with Bayesian generalised least squares regression which is also developed in MDS space. The results indicate that the MDS-RBF-based SVR model provides more consistent flood quantile estimates than the competing SVR models. The RBF SVR approach also demonstrates better generalisation ability than the competing SVR models examined. The MDS-based SVR RBF kernel generally shows the best performance among all the models in terms of prediction accuracy. Overall, the main strengths of the SVR are that it captures efficiently the nonlinearity between the dependent and predictor variables and when coupled with MDS for delineation of regions, the prediction ability of the model increases along with a reduction in bias and relative error.

The role of spatial dependence for large-scale flood risk estimation

Abstract. Flood risk assessments are typically based on scenarios which assume homogeneous return periods of flood peaks throughout the catchment. This assumption is unrealistic for real flood events and may bias risk estimates for specific return periods. We investigate how three assumptions about the spatial dependence affect risk estimates: (i) spatially homogeneous scenarios (complete dependence), (ii) spatially heterogeneous scenarios (modelled dependence) and (iii) spatially heterogeneous but uncorrelated scenarios (complete independence). To this end, the model chain RFM (regional flood model) is applied to the Elbe catchment in Germany, accounting for the spatio-temporal dynamics of all flood generation processes, from the rainfall through catchment and river system processes to damage mechanisms. Different assumptions about the spatial dependence do not influence the expected annual damage (EAD); however, they bias the risk curve, i.e. the cumulative distribution function of damage. The widespread assumption of complete dependence strongly overestimates flood damage of the order of 100 % for return periods larger than approximately 200 years. On the other hand, for small and medium floods with return periods smaller than approximately 50 years, damage is underestimated. The overestimation aggravates when risk is estimated for larger areas. This study demonstrates the importance of representing the spatial dependence of flood peaks and damage for risk assessments.