Physical Catchment Descriptors Research Articles

Regionalization of hydrological models for flow estimation in ungauged catchments in Ireland

Study RegionThe study area consists of 44 catchments across Ireland. Study FocusWe regionalize two hydrological models (GR4J and GR6J) to produce continuous discharge simulations and compare performance in simulating high, median and low flow conditions with other established approaches to prediction in ungauged basins and a simple benchmark of using the median parameter set across all catchments. These include K-nearest neighbor (KNN) and statistical methods for predicting flow quantiles using catchment characteristics. Different objective functions were selected for different parts of flow regime and the success of different methods for regionalizing hydrological model parameters; including multiple linear regression (MLR), non-linear regression (NL) and random forests (RF) were evaluated. New Hydrological insights for the RegionAll regionalization approaches perform well for average flow conditions. The GR4J model regionalized using RF performs best for simulating high flows, though all regionalized models underestimate the median annual flood. GR6J regionalized using RF performs best for low flows. While KNN and statistical approaches that directly leverage physical catchment descriptors provide comparable median performances across catchments, the spread in relative error across our sample is reduced using regionalized hydrological models. Our results highlight that the choice of hydrological model, objective functions for optimization and approach to linking model parameters and physical catchment descriptors significantly influence the success of regionalization for low and high flows.

Benchmarking seasonal forecasting skill using river flow persistence in Irish catchments

ABSTRACT This study assesses the seasonal forecast skill of river flow persistence in 46 catchments representing a range of hydrogeological conditions across Ireland. Skill is evaluated against a climatology benchmark forecast and by examining correlations between predicted and observed flow anomalies. Forecasts perform best when initialized in drier summer months, 87% of which show greater skill relative to the benchmark at a 1-month horizon. Such skill declines as forecast horizon increases due to the longer time a catchment has to “forget” initial anomalous flow conditions and/or to be impacted by “new” events. Skill is related to physical catchment descriptors such as the baseflow index (correlation ρ = 0.86) and is greatest in permeable high-storage catchments. The distinct seasonal and spatial variations in persistence skill allow us to pinpoint when and where this method can provide a useful benchmark in the future development of more complex seasonal hydrological forecasting approaches in Ireland.

Streamflow prediction in ungauged basins: analysis of regionalization methods in a hydrologically heterogeneous region of Mexico

ABSTRACTThe ability of three regionalization methods – multiple linear regression (MLR), spatial proximity (SP) and physical similarity (PS) – to predict streamflow in ungauged catchments in Mexico is estimated. Three hydrological models (GR4J, HMETS and MOHYSE) are calibrated on 30 diverse catchments in Mexico. Implementation of leave-one-out cross-validation enabled the regionalization skill at each of the 30 study sites to be estimated, or the site to be considered as being ungauged. This study shows that regionalization in a hydrologically heterogeneous area, such as the Mexican area under study, poses problems to regionalization approaches that depend on physical catchment descriptors, such as MLR and PS. The transfer of complete parameter sets from a neighbouring catchment provided the most robust method to estimate streamflow in semi-arid and humid ungauged basins. The model performance for arid catchments was worse in the context of regionalization, with GR4J being more robust than the other models due to its simpler structure.

Assessing flood probability for transportation infrastructure based on catchment characteristics, sediment connectivity and remotely sensed soil moisture

Flooding may damage important transportation infrastructures, such as roads, railways and bridges, which need to be well planned and designed to be able to withstand current and possible future climate-driven increases in flood frequencies and magnitudes. This study develops a novel approach to predictive statistical modelling of the probability of flooding at major road-stream intersection sites, where water, sediment and debris can accumulate and cause failure of drainage facilities and associated road damages. Two areas in south-west Sweden, affected by severe floods in August 2014, are used in representative case studies for this development. A set of physical catchment-descriptors (PCDs), characterizing key aspects of topography, morphology, soil type, land use, hydrology (precipitation and soil moisture) and sediment connectivity in the water- and sediment-contributing catchments, are used for the predictive flood modelling. A main novel contribution to such modelling is to integrate the spatiotemporal characteristics of remotely-sensed soil moisture in indices of sediment connectivity (IC), thereby also allowing for investigation of the role of soil moisture in the flood probability for different road-stream intersections. The results suggest five categories of PCDs as especially important for flood probability quantification and identification of particularly flood-prone intersections along roads (railways, etc.) These include: channel slope at the road-stream intersection and average elevation, soil properties (mainly percentage of till), land use cover (mainly percentage of urban areas), and a sediment connectivity index that considers soil moisture in addition to morphology over the catchment.

Flow Regionalization Under Limited Data Availability – Application of IHACRES in the Western Ghats

Long-term measurements of river stream flow are essential for numerous applications in water resources. In many parts of the world, rivers remain ungauged; stream flow prediction for such ungauged catchment requires information transfer from gauged catchments that are perceived to be hydrologically similar to them. Prediction of stream flows at ungauged catchments is typically performed through the transfer of hydrologic information (e.g., model parameters, hydrologic indices, stream flow values) from gauged to ungauged catchments. This procedure is commonly referred to as regionalization. Hydrologic models have been used extensively for the purpose of transfer of hydrologic model parameters from gauged to ungauged catchments and these models include HBV, IHACRES and PDM.In the present study, for the purpose of regionalisation, a lumped conceptual model namely, IHACRES has been used. IHACRES is a parsimonious model having six numbers of parameters and requires small data input. The model parameters are used to characterise the daily stream flow of the catchment and are related to the landscape attributes of the catchments. Further, as an offshoot, IHACRES provides for base-flow separation and development of unit hydrograph, which becomes an input in other hydrological works.In the current paper, the model IHACRES is applied at daily time step to six catchments (Malathi, Hemavathi, Lakshmanthirtha, Yetthinahole, Kadumanehalla and Kumaradhara) in the region of Western Ghats, Karnataka. IHACRES, being parsimonious, requires precipitation and temperature as the only data-input. Information is compiled for physical catchment descriptors using the elevation map of the area (DEM) and relationships are developed between the parameters and the descriptors. These relationships are validated by modelling daily stream flow of a gauged catchment (Nethravathi). This work proves that as an alternative model, IHACRES does well in the domain of flow regionalization with least data inputs.

A method for mapping flood hazard along roads

A method was developed for estimating and mapping flood hazard probability along roads using road and catchment characteristics as physical catchment descriptors (PCDs). The method uses a Geographic Information System (GIS) to derive candidate PCDs and then identifies those PCDs that significantly predict road flooding using a statistical modelling approach. The method thus allows flood hazards to be estimated and also provides insights into the relative roles of landscape characteristics in determining road-related flood hazards. The method was applied to an area in western Sweden where severe road flooding had occurred during an intense rain event as a case study to demonstrate its utility. The results suggest that for this case study area three categories of PCDs are useful for prediction of critical spots prone to flooding along roads: i) topography, ii) soil type, and iii) land use. The main drivers among the PCDs considered were a topographical wetness index, road density in the catchment, soil properties in the catchment (mainly the amount of gravel substrate) and local channel slope at the site of a road-stream intersection. These can be proposed as strong indicators for predicting the flood probability in ungauged river basins in this region, but some care is needed in generalising the case study results other potential factors are also likely to influence the flood hazard probability. Overall, the method proposed represents a straightforward and consistent way to estimate flooding hazards to inform both the planning of future roadways and the maintenance of existing roadways.

Regionalisation of the parameters of a hydrological model: Comparison of linear regression models with artificial neural nets

In many practical applications of hydrological models, a case-specific parameter optimisation is not feasible due to lack of data. Model parameters may then be derived through regionalisation, i.e. by linking model parameters to physical catchment descriptors. Most studies reported in literature rely on linear regression techniques to establish the regionalisation scheme although it is well known that the underlying assumptions, e.g. the linearity of the relationship between physical catchment descriptors and model parameters, are often violated. Artificial neural networks (ANNs) are distribution-independent non-linear model structures and are therefore potentially useful tools for regionalising model parameters. This paper compared linear regression analysis and ANNs for regionalising the most sensitive parameters of the semi-distributed hydrological model SWAT (Soil and Water Assessment Tool) with an application in the Flemish part of the Scheldt river basin. The uncertainty on both regionalisation procedures was assessed with a non-parametric bootstrap method. ANNs delivered more accurate parameter estimates than linear regression equations if the non-linearities simulated by the ANNs have a physical meaning and if the physical catchment descriptors of the catchment under study lie within the range of the descriptor values of the sites used for the construction of the ANNs. For extrapolations outside this range, regression analysis yielded a better result. The uncertainty on the regionalised parameters was generally higher for ANNs than for regression equations, however, intervals for stream flow predictions were only slightly broader. Finally, the results of previous analyses were interpreted in function of the parameterisation of virtual land use scenarios. Because five out of seven model parameters depend on land use, it can be expected that regionalisation can enhance the estimation of non-crop related parameters of virtual land use scenarios.

Predicting daily flows in ungauged catchments: model regionalization from catchment descriptors at the Coweeta Hydrologic Laboratory, North Carolina

AbstractRegionalization approaches to daily streamflow prediction are investigated for 13 catchments in the Coweeta Hydrologic Laboratory using a conceptual rainfall‐runoff model of low complexity (six parameters). Model parameters are considered to represent the dynamic response characteristics (DRCs) of a catchment. It is demonstrated that all catchments within the region cannot be assumed to have a similar hydrological behaviour, and thence a regionalization approach considering differences in physical catchment descriptors (PCDs) is required. Such a regionalization approach can be regarded as a top‐down method, in the sense that factors controlling parameter variability are identified first within the entire region under study, and then such information is exploited to predict runoff in a smaller sub‐region. Regionalization results reveal that consideration of interrelations between dependent variables, which here are the parameters of the rainfall‐runoff model, can improve performance of regression as a regionalization method. Breaking the parameter correlation structure inherent in the model, and exploiting merely relationships between model parameters and PCDs (no matter how weakly related they are), can result in a significant decrease in regionalization performance. Also, high significance of regression between values of PCDs and DRCs does not guarantee a set of parameters with a good predictive power. When there is a reason to believe that, in the sense of hydrological behaviour, a gauged catchment resembles the ungauged catchment, then it may be worthwhile to adopt the entire set of calibrated parameters from the gauged catchment instead of deriving quantitative relationships between catchment descriptors and model parameters. Copyright © 2003 John Wiley & Sons, Ltd.

Relationships between dynamic response characteristics and physical descriptors of catchments in England and Wales

A regionalisation methodology has been applied to catchments in England and Wales enabling estimation of daily flows for any catchment in the region for which physical data and records of rainfall and temperature are available. The rainfall-runoff model IHACRES has been calibrated to 60 catchments to obtain a set of dynamic response characteristics (DRCs) describing the hydrological behaviour within the region. Physical catchment descriptors (PCDs) indexing topography, soil type, climate and land cover were collated and linked to the hydrological model by overlaying catchment boundaries with a geographical information system. Relationships were derived to describe the DRCs in terms of the PCDs so that the model may be used to simulate flow for any catchment in the region, given the driving variables, i.e. rainfall and temperature. In the England and Wales region, rainfall loss parameters have been defined in terms of land use, climate and soil type, whilst hydrograph separation parameters were characterised using topographical and soil variables. The set of DRC–PCD relationships, which were obtained by balancing the dual objectives of hydrological integrity and statistical significance, has been satisfactorily validated on two additional catchments within the region. Analysis of calibration errors was aided by sensitivity tests at one of these catchments in which flow response to variations in DRCs was assessed. Finally, a simple land use scenario demonstrates an application of the methodology in which variation in PCDs may be used to assess the impacts of environmental change.

Dynamic response characteristics of the plynlimon catchments and preliminary analysis of relationships to physical descriptors

Assessment of the impacts of environmental change on hydrological systems at regional or national scale requires a modelling technique capable of spatial extrapolation. Few such techniques, providing a regional or national picture of hydrological impact, are currently available. The proposed regionalization methodology, outlined in Jakeman et al. (1992), is being applied to up to 100 gauged UK catchments of various size, geology, topography and climate using a rainfall-runoff model IHACRES (a program for Identification of unit Hydrograph And Component flows from Rainfall, Evapotranspiration and Streamflow data). Dynamic response characteristics (DRCs), such as the proportion of slow flow, identified by the model are being related to physical catchment descriptors (PCDs) such as soil type, slope and area, obtained from a geographical information system (GIs). Extrapolation to ungauged catchments should then be possible and the resulting hydrological GIS of the UK would allow spatial analysis of the hydrological impact of environmental changes in climate and land use. This paper presents a preliminary application of the methodology using the Plynlimon experimental catchments in central Wales for which good hydrological and physical descriptor records exist. Eight catchments are modelled: the headwaters of the Wye and Severn rivers and the three main subcatchments of each, the Gwy, Cyff and Iago and Hafren, Hore and Tanllwyth, respectively. Relationships between DRCs and PCDs are investigated and interpreted in the context of the wider project.

How much complexity is warranted in a rainfall‐runoff model?

Development of mathematical models relating the precipitation incident upon a catchment to the streamflow emanating from the catchment has been a major focus of surface water hydrology for decades. Generally, values for parameters in such models must be selected so that runoff calculated from the model “matches” recorded runoff from some historical period. Despite the fact that the physics governing the path of a drop of water through a catchment to the stream involves complex relationships, evidence indicates that the information content in a rainfall‐runoff record is sufficient to support models of only very limited complexity. This begs the question of what limits the observed data place on the allowable complexity of rainfall‐runoff models. Time series techniques are applied for estimating transfer functions to determine how many parameters are appropriate to describe the relationship between precipitation and streamflow in the case where data on only precipitation, air temperature, and streamflow are available. Statistics from an “information matrix” provide the clues necessary for determining allowable model complexity. Time series models are developed for seven catchments with widely varying physical characteristics in different temperate climatic regimes to demonstrate the method. It is found that after modulating the measured rainfall using a nonlinear loss function, the rainfall‐runoff response of all catchments is well represented using a linear model. Also, for all catchments a two‐component linear model with four parameters is the model of choice. The two components can be interpreted as defining a “quick flow” and “slow flow” response of the given catchment. The method therefore provides a statistically rigorous way to separate hydrographs and parameterize their response behavior. The ability to construct reliable transfer function models for describing the rainfall‐runoff process offers a new approach to investigate empirically the controls of physical catchment descriptors, land use change, climate change, etc., on the dynamic response of catchments through the extensive analysis of historical data sets.

A systematic approach to modelling the dynamic linkage of climate, physical catchment descriptors and hydrologic response components

A systematic approach to modelling the dynamic linkage of climate, physical catchment descriptors and hydrologic response components