UK Rivers Research Articles

Development and application of eDNA-based tools for the conservation of white-clawed crayfish

eDNA-based methods represent non-invasive and cost-effective approaches for species monitoring and their application as a conservation tool has rapidly increased within the last decade. Currently, they are primarily used to determine the presence/absence of invasive, endangered or commercially important species, but they also hold potential to contribute to an improved understanding of the ecological interactions that drive species distributions. However, this next step of eDNA-based applications requires a thorough method development. We developed an eDNA assay for the white-clawed crayfish (Austropotamobius pallipes), a flagship species of conservation in the UK and Western Europe. Multiple subsequent in-situ and ex-situ validation tests aimed at improving method performance allowed us to apply eDNA-based surveys to evaluate interactions between white-clawed crayfish, crayfish plague and invasive signal crayfish. The assay performed well in terms of specificity (no detection of non-target DNA) and sensitivity, which was higher compared to traditional methods (in this case torching). The eDNA-based quantification of species biomass was, however, less reliable. Comparison of eDNA sampling methods (precipitation vs. various filtration approaches) revealed that optimal sampling method differed across environments and might depend on inhibitor concentrations. Finally, we applied our methodology together with established assays for crayfish plague and the invasive signal crayfish, demonstrating their significant interactions in a UK river system. Our analysis highlights the importance of thorough methodological development of eDNA-based assays. Only a critical evaluation of methodological strengths and weaknesses will allow us to capitalise on the full potential of eDNA-based methods and use them as decision support tools in environmental monitoring and conservation practice.

A newly developed genetic sex marker and its application to understanding chemically induced feminisation in roach (Rutilus rutilus).

Oestrogenic wastewater treatment works (WwTW) effluents discharged into UK rivers have been shown to affect sexual development, including inducing intersex, in wild roach (Rutilus rutilus). This can result in a reduced breeding capability with potential population level impacts. In the absence of a sex probe for roach it has not been possible to confirm whether intersex fish in the wild arise from genetic males or females, or whether sex reversal occurs in the wild, as this condition can be induced experimentally in controlled exposures to WwTW effluents and a steroidal oestrogen. Using restriction site-associated DNA sequencing (RAD-seq), we identified a candidate for a genetic sex marker and validated this marker as a sex probe through PCR analyses of samples from wild roach populations from nonpolluted rivers. We also applied the sex marker to samples from roach exposed experimentally to oestrogen and oestrogenic effluents to confirm suspected phenotypic sex reversal from males to females in some treatments, and also that sex-reversed males are able to breed as females. We then show, unequivocally, that intersex in wild roach populations results from feminisation of males, but find no strong evidence for complete sex reversal in wild roach at river sites contaminated with oestrogens. The discovered marker has utility for studies in roach on chemical effects, wild stock assessments, and reducing the number of fish used where only one sex is required for experimentation. Furthermore, we show that the marker can be applied nondestructively using a fin clip or skin swab, with animal welfare benefits.

Inclusion of seasonal variation in river system microbial communities and phototroph activity increases environmental relevance of laboratory chemical persistence tests

Regulatory tests assess crop protection product environmental fate and toxicity before approval for commercial use. Although globally applied laboratory tests can assess biodegradation, they lack environmental complexity. Microbial communities are subject to temporal and spatial variation, but there is little consideration of these microbial dynamics in the laboratory. Here, we investigated seasonal variation in the microbial composition of water and sediment from a UK river across a two-year time course and determined its effect on the outcome of water-sediment (OECD 308) and water-only (OECD 309) biodegradation tests, using the fungicide isopyrazam. These OECD tests are performed under dark conditions, so test systems incubated under non-UV light:dark cycles were also included to determine the impact on both inoculum characteristics and biodegradation. Isopyrazam degradation was faster when incubated under non-UV light at all collection times in water-sediment microcosms, suggesting that phototrophic communities can metabolise isopyrazam throughout the year. Degradation rate varied seasonally between inoculum collection times only in microcosms incubated in the light, but isopyrazam mineralisation to 14CO2 varied seasonally under both light and dark conditions, suggesting that heterotrophic communities may also play a role in degradation. Bacterial and phototroph communities varied across time, but there was no clear link between water or sediment microbial composition and variation in degradation rate. During the test period, inoculum microbial community composition changed, particularly in non-UV light incubated microcosms. Overall, we show that regulatory test outcome is not influenced by temporal variation in microbial community structure; however, biodegradation rates from higher tier studies with improved environmental realism, e.g. through addition of non-UV light, may be more variable. These data suggest that standardised OECD tests can provide a conservative estimate of pesticide persistence end points and that additional tests including non-UV light could help bridge the gap between standard tests and field studies.

Persistent contaminants in adipose fins of returning adult salmonids to the river Tees (UK)

We report on concentrations of polybrominated diphenylethers (PBDEs), polychlorinated biphenyls (PCBs), dichlorodiphenyldichloroethylene (p,p′-DDE) and hexachlorobenzene (HCB) measured in the adipose fins of returning adult Atlantic salmon (Salmo salar) and sea trout (Salmo trutta) to the river Tees in the Northeast of England. Overall, higher concentrations of these contaminants were found in sea trout samples, where detected congeners reflected the more widely used commercial formulations, in particular for the PBDEs. Our results suggest that these fish could be bioaccumulating persistent organic pollutants via diet during their migratory routes (North Sea and the Norwegian Sea) and, in addition, some level of re-mobilisation of these compounds could still be occurring in the UK eastern coastal areas. The use of adipose fin of returning salmonids could be further developed as a non-lethal approach to assess whether persistent contaminants are being accumulated during the juvenile to adult phase of salmonids originating from UK rivers.

Exploring the potential of Google Earth as a communication and engagement tool in collaborative Natural Flood Management planning

This paper considers the development and evaluation of a Google Earth “virtual globe” tour for communicating spatial data and engaging stakeholders in the early stages of a natural flood management (NFM) planning scenario, using a rural UK river catchment that suffered significant flooding in 2007. With a range of diverse stakeholder interests to consider, early engagement and the development of trust before decision‐making are essential for the long‐term success of such catchment‐wide projects. A local catchment group was consulted to identify key information requirements, and from this a “virtual globe” tour was created. The process involved specialist skills and expert leadership, but the end result was accessible to a range of audiences. User evaluation indicated that it was easy to navigate and can be used to stimulate interest and engage stakeholders. Participants trusted the content and valued the interactivity of the tour. It was helpful for communicating and educating participants about the catchment, the issues it faces, and the potential to incorporate NFM, particularly for those with little or no prior knowledge. More abstract information was harder to convey and there were limitations in the availability of suitable data for some variables and the in quality of satellite imagery. This exploratory research found that a Google Earth “virtual globe” tour can be a valuable tool in the initial stages of an NFM project, but there are also opportunities to use this technique in the more advanced stages of the planning process. The approach could be used as part of a wider toolkit for communication and engagement and has potential as a decision support tool in other environmental management scenarios with requirements for public participation, enabling the views of a range of participants to be captured through online distribution and to generate discussion in workshop settings.

Seeking river restoration appraisal best practice: supporting wider national and international environmental goals

AbstractWith growing investment in river restoration, we increasingly need to justify costs by demonstrating success and wider benefits of measures. To aid practitioners, the UK River Restoration Centre has worked with experts to develop a practical monitoring guidance that links objectives to specific monitoring to demonstrate achievable outcomes. Feedback, however, via an online questionnaire highlighted the need to rationalise the guidance contents for a new growing audience, taking advantage of new developments and incorporating the evaluation of social and economic aspects of river restoration. With these potential improvements, it is hoped that practitioners will follow this guidance, improve the quality of monitoring undertaken and share evidence of success and lessons learnt. This paper outlines how this guidance has been adopted as best practice. We discuss why we need to embed this guidance into wider monitoring protocols that can feed into national and international environmental policy and targets.

A multi-objective ensemble approach to hydrological modelling in the UK: an application to historic drought reconstruction

Abstract. Hydrological models can provide estimates of streamflow pre- and post-observations, which enable greater understanding of past hydrological behaviour, and potential futures. In this paper, a new multi-objective calibration method was derived and tested for 303 catchments in the UK, and the calibrations were used to reconstruct river flows back to 1891, in order to provide a much longer view of past hydrological variability, given the brevity of most UK river flow records which began post-1960. A Latin hypercube sample of 500 000 parameterisations for the GR4J model for each catchment were evaluated against six evaluation metrics covering all aspects of the flow regime from high, median, and low flows. The results of the top ranking model parameterisation (LHS1), and also the top 500 (LHS500), for each catchment were used to provide a deterministic result whilst also accounting for parameter uncertainty. The calibrations are generally good at capturing observed flows, with some exceptions in heavily groundwater-dominated catchments, and snowmelt and artificially influenced catchments across the country. Reconstructed flows were appraised over 30-year moving windows and were shown to provide good simulations of flow in the early parts of the record, in cases where observations were available. To consider the utility of the reconstructions for drought simulation, flow data for the 1975–1976 drought event were explored in detail in nine case study catchments. The model's performance in reproducing the drought events was found to vary by catchment, as did the level of uncertainty in the LHS500. The Standardised Streamflow Index (SSI) was used to assess the model simulations' ability to simulate extreme events. The peaks and troughs of the SSI time series were well represented despite slight over- or underestimations of past drought event magnitudes, while the accumulated deficits of the drought events extracted from the SSI time series verified that the model simulations were overall very good at simulating drought events. This paper provides three key contributions: (1) a robust multi-objective model calibration framework for calibrating catchment models for use in both general and extreme hydrology; (2) model calibrations for the 303 UK catchments that could be used in further research, and operational applications such as hydrological forecasting; and (3) ∼ 125 years of spatially and temporally consistent reconstructed flow data that will allow comprehensive quantitative assessments of past UK drought events, as well as long-term analyses of hydrological variability that have not been previously possible, thus enabling water resource managers to better plan for extreme events and build more resilient systems for the future.

Immunoassays are not immune to errors: Examples from two studies of steroid output from freshwater trout farms

A “reproducibility crisis” is widespread across scientific disciplines, where results and conclusions of studies are not supported by subsequent investigation. Here we provide a steroid immunoassay example where human errors generated unreproducible results and conclusions. Our study was triggered by a scientific report citing abnormally high concentrations (means of 4–79 ng L−1) of three natural sex steroids [11-ketotestosterone (11-KT), testosterone (T) and oestradiol (E2)] in water samples collected from two UK rivers over 4 years (2002–6). Furthermore, the data suggested that trout farms were a major source because reported steroid concentrations were 1.3–6 times higher downstream than upstream. We hypothesised that the reported levels were erroneous due to substances co-extracted from the water causing matrix effects (i.e. “false positives”) during measurement by enzyme-linked immunoassay (EIA). Thus, in collaboration with three other groups (including the one that had conducted the 2002–6 study), we carried out field sampling and assaying to examine this hypothesis. Water samples were collected in 2010 from the same sites and prepared for assay using an analogous method [C18 solid phase extraction (SPE) followed by extract clean-up with aminopropyl SPE]. Additional quality control (“spiked” and “blank”) samples were processed. Water extracts were assayed for steroids using radioimmunoassay (RIA) as well as EIA. Although there were statistically significant differences between EIA and RIA (and laboratories), there was no indication of matrix effects in the EIAs. Both the EIAs and RIAs (uncorrected for recovery) measured all three natural steroids at <0.6 ng L−1 in all river water samples, indicating that the trout farms were not a significant source of natural steroids. The differences between the two studies were considerable: E2 and T concentrations were ca. 100-fold lower and 11-KT ca. 1000-fold lower than those reported in the 2002–6 study. In the absence of evidence for any marked changes in husbandry practice (e.g. stock, diet) or environmental conditions (e.g. water flow rate) between the study periods, we concluded that calculation errors were probably made in the first (2002–6) study associated with confusion between extract and water sample concentrations. The second (2010) study also had several identified examples of calculation error (use of an incorrect standard curve; extrapolation below the minimum standard; confusion of assay dilutions during result work-up; failure to correct for loss during extraction) and an example of sample contamination. Similar and further errors have been noted in other studies. It must be recognised that assays do not provide absolute measurements and are prone to a variety of errors, so published steroid levels should be viewed with caution until independently confirmed.

Distribution of benthic macroinvertebrate communities and assessment of water quality in a small UK river catchment

Communities of benthic macroinvertebrates obtained from a river catchment impacted by point source pollution were investigated from March 2013 to April 2014. The main purpose of this study was to describe the taxon richness of benthic macroinvertebrate communities obtained at upstream and downstream of a wastewater treatment works and combined sewer overflows, assess water quality and to determine major physicochemical variables that affect the distribution and richness of benthic macroinvertebrate communities. The biotic index Whalley Hawkes, Paisley and Trigg revealed distinction between the sites upstream of the wastewater treatment works from the downstream sites. Between and within sampling locations, the dominance of certain taxonomic groups was recorded including Oligochaete, Baetidae and Chironomidae. Predominant in the river were insects which constituted 72% of the overall benthic invertebrate communities followed by Oligochaeta (28%). Distribution of benthic macroinvertebrate communities was not associated with sewage variables but related more with altitude, catchment area, slope and discharge. Altitudes range from 31 to 140 m. The sites at higher altitude had better quality compared to the lower sites and therefore suggests that altitude could provide a spatial proxy for anticipated future ecological changes in the river over time. Classification of the river based on the European Union’s Water Framework Directive indicated that the river had moderate ecological status.

Model‐based inference of conditional extreme value distributions with hydrological applications

AbstractMultivariate extreme value models are used to estimate joint risk in a number of applications, with a particular focus on environmental fields ranging from climatology and hydrology to oceanography and seismic hazards. The semi‐parametric conditional extreme value model of Heffernan and Tawn involving a multivariate regression provides the most suitable of current statistical models in terms of its flexibility to handle a range of extremal dependence classes. However, the standard inference for the joint distribution of the residuals of this model suffers from the curse of dimensionality because, in a d‐dimensional application, it involves a d−1‐dimensional nonparametric density estimator, which requires, for accuracy, a number points and commensurate effort that is exponential in d. Furthermore, it does not allow for any partially missing observations to be included, and a previous proposal to address this is extremely computationally intensive, making its use prohibitive if the proportion of missing data is nontrivial. We propose to replace the d−1‐dimensional nonparametric density estimator with a model‐based copula with univariate marginal densities estimated using kernel methods. This approach provides statistically and computationally efficient estimates whatever the dimension, d, or the degree of missing data. Evidence is presented to show that the benefits of this approach substantially outweigh potential misspecification errors. The methods are illustrated through the analysis of UK river flow data at a network of 46 sites and assessing the rarity of the 2015 floods in North West England.

Artificial substrate experiments to investigate potential impacts of invasive quagga mussel (Dreissena rostriformis bugensis, Bivalvia: Dreissenidae) on macroinvertebrate communities in a UK river

Artificial substrate experiments to investigate potential impacts of invasive quagga mussel (Dreissena rostriformis bugensis, Bivalvia: Dreissenidae) on macroinvertebrate communities in a UK river

Stochastic modelling of flow sequences for improved prediction of fluvial flood hazards

Abstract The availability of historical streamflow data of the desired length is often limited and, in these situations, the ability to synthetically generate statistically significant datasets becomes important. We previously developed a highly efficient stochastic modelling approach for the synthetic generation of daily streamflow sequences using the systematic combination of a hidden Markov model with the generalized Pareto distribution (the HMM-GP model). Daily streamflow sequences provide limited information on various significant small duration flooding events exceeding the peak over threshold values, but these are averaged out in the daily datasets. These small duration intense flooding events are often capable of causing significant damage and are important in conducting thorough flood risk management and flood risk assessment studies. This paper presents upgrades to our HMM-GP stochastic modelling approach and examines its efficiency in simulating streamflow at a temporal resolution of 15 minutes. The potential of the HMM-GP model in simulating a synthetic 15-minute streamflow series is investigated by comparing various statistical characteristics (e.g. percentiles, the probability density distribution and the autocorrelation function) of the observed streamflow records with 100 synthetically simulated streamflow time series. The proposed modelling schematics are robustly validated across case studies in four UK rivers (the Don, Nith, Dee and Tweed).

Trends in the use of large wood in UK river restoration projects: insights from the National River Restoration Inventory

AbstractLarge Wood (LW) is increasingly employed in river restoration to promote physical habitat heterogeneity and ecological diversity. To explore how LW has been used in restoration schemes across the United Kingdom in recent decades, we analysed data on 912 LW projects archived in the UK’s National River Restoration Inventory (NRRI). The number of LW schemes has continued to increase following the earliest records in the 1990s, largely tracking overall trends in river restorations. LW projects have been predominantly located in lowland, rural streams, although there has been a notable cluster in and around London. LW projects have mainly revolved around the desire to deliver hydromorphological improvements and specifically the creation of fish habitat. Most schemes used LW in simple deflector forms despite the growing scientific evidence of the benefits of using structurally complex LW. Post project monitoring has been limited and mostly restricted to photographic records. This type of database analysis can provide important insights and help guide future restoration practice.

Modelling spatial extreme events with environmental applications

Spatial extreme value analysis has been an area of rapid growth in the last decade. The focus has been on modelling the spatial componentwise maxima by max-stable processes. Here, we will explain the limitations of these modelling approaches and show how spatial models can be developed that overcome these deficiencies by exploiting the flexible conditional multivariate extremes models of Heffernan and Tawn (2004). We illustrate the benefits of these new spatial models through applications to North Sea wave analysis and to widespread UK river flood risk analysis.

Which commonly monitored chemical contaminant in the Bohai region and the Yangtze and Pearl Rivers of China poses the greatest threat to aquatic wildlife?

The present study assessed the relative risk of 29 chemical contaminants to aquatic wildlife in the Bohai region and the Yangtze and Pearl Rivers of China. River monitoring data from 2010 to 2015 for metals, pesticides, plasticizers, surfactants, polyaromatic hydrocarbons, flame retardants, and ammonia were collected. For each chemical, ecotoxicity data were compiled for Chinese-relevant aquatic species. The chemicals were ranked by relative risk either by comparing the ratios of the median river concentration divided by the median ecotoxicity concentration or by the percentage of river measurements which exceeded the lower 10th percentile ecotoxicity value. To provide context, these results were compared with the same analysis for rivers in the United Kingdom. From this collection of chemicals in Chinese rivers, the highest risks appear to be from Cu, closely followed by Zn, Fe, and Ni together with linear alkyl benzene sulfonate, nonylphenol, and NH3 . This risk, particularly from the metals, can be several times higher than that experienced in UK rivers when using the same analysis. Ammonia median concentrations were notably higher in the Pearl and Yangtze than in UK rivers. The results suggest that China should focus on controlling metal contamination to protect its aquatic wildlife. Environ Toxicol Chem 2018;37:1115-1121. © 2017 SETAC.

Asynchronicity of fine sediment supply and its effects on transport and storage in a regulated river

PurposeA disconnected ephemeral tributary was reconnected to the regulated River Ehen (NW England) as part of a river restoration initiative, providing a renewed delivery of sediment to a highly stable and armoured channel. This paper (1) assesses spatial and temporal dynamics of suspended and stored sediments in the Ehen, (2) characterises the composition of stored sediment, (3) develops fine sediment budgets for downstream river reaches, and (4) assesses the controls on the storage of fine sediment in the riverbed.Materials and methodsA 3-km study section in the upper part of the River Ehen was divided into two reaches. Suspended sediments were monitored at the downstream limits of each reach over a 2-year period. In-channel storage was measured in three morphological units within the upper reach, on 13 occasions over the same period. Samples were used to assess changes in volumes of stored fine sediment, as well as the grain sizes and organic content of the material. A time-lapse camera facing the confluence of the tributary was used to conceptualise different flow scenarios. These scenarios reflect the degree of synchronicity between flows in the main-stem and those in the tributary. Fine sediment budgets were developed for each reach to assess the relative contribution of different sources of sediment.Results and discussionThe reconnection significantly affected suspended sediment loads in the Ehen. Bed storage increased twofold, with changes most evident in the slow-flowing morphological unit. Changes in the composition of stored sediment were less marked than changes in the quantity of material. Changes in bed storage were controlled by the degree of synchronicity between flows in the Ehen and those in the newly reconnected tributary. Results show that three generalised flow scenarios occur, with total asynchronicity between flows in the tributary and the Ehen being responsible for the main episodes of fine sediment deposition. Overall, the estimated sediment budgets provide insights into the importance of non-perennial sources of sediment in supply-limited systems such as the Ehen. Although bed storage values are within the range of those published for UK rivers, the increase observed since the reconnection, together with the persistence of a static pavement, highlights the ecologically critical conditions of the regulated main-stem River Ehen.ConclusionsIntermittent sources control fine sediment transport dynamics in the upper River Ehen. In this regulated river, ongoing deposition associated with increased low- and medium-sized flow events exerts more of a control on bed storage than large but rare floods. Management actions to limit delivery of material from lateral sources could help prevent further deterioration of habitat conditions for biota sensitive to fine sediment. Given the ongoing adjustment in the newly reconnected tributary, continued monitoring is needed to capture further morphosedimentary response in the main-stem.

Does river restoration work? Taxonomic and functional trajectories at two restoration schemes

Rivers and their floodplains have been severely degraded with increasing global activity and expenditure undertaken on restoration measures to address the degradation. Early restoration schemes focused on habitat creation with mixed ecological success. Part of the lack of ecological success can be attributed to the lack of effective monitoring. The current focus of river restoration practice is the restoration of physical processes and functioning of systems. The ecological assessment of restoration schemes may need to follow the same approach and consider whether schemes restore functional diversity in addition to taxonomic diversity. This paper examines whether two restoration schemes, on lowland UK rivers, restored macroinvertebrate taxonomic and functional (trait) diversity and relates the findings to the Bradshaw's model of ecological restoration. The study schemes are considered a success in terms of restoring physical processes, longitudinal connectivity and the resulting habitat composition. However, the rehabilitation of macroinvertebrate community structure and function was limited and inconsistent, varying over time, depending on the restoration measure applied and the taxonomic or functional index considered. Resampling of species pools at each site revealed a role for functional redundancy, meaning that increases in functional diversity are more difficult to achieve than outcomes based on taxonomic analyses. Our results highlight the usefulness of applying functional traits alongside taxonomic indices in evaluating river restoration projects.

Temporal variability in lotic macroinvertebrate communities associated with invasive signal crayfish (Pacifastacus leniusculus) activity levels and substrate character

Invasive signal crayfish (Pacifastacus leniusculus) are considered to be the most prevalent non-native crayfish species in Europe. Where large populations become established they have significant and long-term effects on benthic macroinvertebrate communities. However, much less is known about how community effects associated with crayfish invasion change in the short-term as a function of varying activity levels during the summer months. We examined the macroinvertebrate community composition of two lowland UK rivers, one which supported a well-established non-native crayfish population (invaded) and one in which crayfish had not been recorded (control). Colonisation cylinders were deployed which recorded community composition over a 126-day time period. Results indicate that once the activity period commences, invasive crayfish consistently altered macroinvertebrate community structure regardless of substrate character. Invaded communities displayed reduced beta-diversity compared to control sites. However, effects on the macroinvertebrate assemblage varied over the period when crayfish were active probably reflecting the behavioural activity of crayfish (which intensifies with increasing water temperature and during the spawning season) and life histories of other macroinvertebrates. The results indicate that crayfish invasions modify macroinvertebrate community composition, but over shorter timescales, the effects vary associated with their activity levels.

Hydrogeological Controls on Regional-Scale Indirect Nitrous Oxide Emission Factors for Rivers.

Indirect nitrous oxide (N2O) emissions from rivers are currently derived using poorly constrained default IPCC emission factors (EF5r) which yield unreliable flux estimates. Here, we demonstrate how hydrogeological conditions can be used to develop more refined regional-scale EF5r estimates required for compiling accurate national greenhouse gas inventories. Focusing on three UK river catchments with contrasting bedrock and superficial geologies, N2O and nitrate (NO3-) concentrations were analyzed in 651 river water samples collected from 2011 to 2013. Unconfined Cretaceous Chalk bedrock regions yielded the highest median N2O-N concentration (3.0 μg L-1), EF5r (0.00036), and N2O-N flux (10.8 kg ha-1 a-1). Conversely, regions of bedrock confined by glacial deposits yielded significantly lower median N2O-N concentration (0.8 μg L-1), EF5r (0.00016), and N2O-N flux (2.6 kg ha-1 a-1), regardless of bedrock type. Bedrock permeability is an important control in regions where groundwater is unconfined, with a high N2O yield from high permeability chalk contrasting with significantly lower median N2O-N concentration (0.7 μg L-1), EF5r (0.00020), and N2O-N flux (2.0 kg ha-1 a-1) on lower permeability unconfined Jurassic mudstone. The evidence presented here demonstrates EF5r can be differentiated by hydrogeological conditions and thus provide a valuable proxy for generating improved regional-scale N2O emission estimates.

Residual uncertainty estimation using instance-based learning with applications to hydrologic forecasting

Abstract. A non-parametric method is applied to quantify residual uncertainty in hydrologic streamflow forecasting. This method acts as a post-processor on deterministic model forecasts and generates a residual uncertainty distribution. Based on instance-based learning, it uses a k nearest-neighbour search for similar historical hydrometeorological conditions to determine uncertainty intervals from a set of historical errors, i.e. discrepancies between past forecast and observation. The performance of this method is assessed using test cases of hydrologic forecasting in two UK rivers: the Severn and Brue. Forecasts in retrospect were made and their uncertainties were estimated using kNN resampling and two alternative uncertainty estimators: quantile regression (QR) and uncertainty estimation based on local errors and clustering (UNEEC). Results show that kNN uncertainty estimation produces accurate and narrow uncertainty intervals with good probability coverage. Analysis also shows that the performance of this technique depends on the choice of search space. Nevertheless, the accuracy and reliability of uncertainty intervals generated using kNN resampling are at least comparable to those produced by QR and UNEEC. It is concluded that kNN uncertainty estimation is an interesting alternative to other post-processors, like QR and UNEEC, for estimating forecast uncertainty. Apart from its concept being simple and well understood, an advantage of this method is that it is relatively easy to implement.