Search

In this study, we utilise Artificial Neural Network (ANN) models to estimate the 100- and 1500-year return levels for around 900,000 ungauged catchments in the contiguous USA. The models were trained and validated using 4,079 gauges and several selected catchment descriptors out of a total of 25 available. The study area was split into 15 regions, which represent major watersheds. ANN models were developed for each region and evaluated by calculating several performance metrics such as root-mean-squared error (RMSE), coefficient of determination (R2) and absolute percent error. The availability of a large dataset of gauges made it possible to test different model architectures and assess the regional performance of the models. The results indicate that ANN models with only one hidden layer are sufficient to describe the relationship between flood quantiles and catchment descriptors. The regional performance depends on climate type as models perform worse in arid and humid continental climates. Overall, the study suggests that ANN models are particularly applicable for predicting ungauged flood quantiles across a large geographic area. The paper presents recommendations about future application of ANN in regional flood frequency analysis.

In an assessment of how Arctic sea ice cover could be remediated in a warming world, we simulated the injection of SO2 into the Arctic stratosphere making annual adjustments to injection rates. We treated one climate model realization as a surrogate “real world” with imperfect “observations” and no rerunning or reference to control simulations. SO2 injection rates were proposed using a novel model predictive control regime which incorporated a second simpler climate model to forecast “optimal” decision pathways. Commencing the simulation in 2018, Arctic sea ice cover was remediated by 2043 and maintained until solar geoengineering was terminated. We found quantifying climate side effects problematic because internal climate variability hampered detection of regional climate changes beyond the Arctic. Nevertheless, through decision maker learning and the accumulation of at least 10 years time series data exploited through an annual review cycle, uncertainties in observations and forcings were successfully managed.

Solar geoengineering has been proposed as a method of meeting climate objectives, such as reduced globally averaged surface temperatures. However, because of incomplete understanding of the effects of geoengineering on the climate system, its implementation would be in the presence of substantial uncertainties. In our study, we use two fully coupled atmosphere–ocean general circulation models: one in which the geoengineering strategy is designed, and one in which geoengineering is implemented (a real-world proxy). We show that regularly adjusting the amount of solar geoengineering in response to departures of the observed global mean climate state from the predetermined objective (sequential decision making; an explicit feedback approach) can manage uncertainties and result in achievement of the climate objective in both the design model and the real-world proxy. This approach results in substantially less error in meeting global climate objectives than using a predetermined time series of how much geoengineering to use, especially if the estimated sensitivity to geoengineering is inaccurate.

Effective flood risk management depends on methods for estimating flood hazard and an appraisal of the dominant uncertainties in the analysis. Typically, hydraulic models are used to simulate the extent of flooding for an estimate of the flow in a particular reach for a chosen probability of exceedance. However, this definition causes problems at river confluences where flows derive from multiple sources. Here, a model-based approach was adopted to describe the multisite joint distribution of river flows for three rivers that converge on the city of Carlisle (UK). Monte-Carlo methods were used to generate flood events with realistic spatial dependence between tributaries which would occur over a 1000year period. To account for the uncertainty in the data used to create the event set, block bootstrapping was used to produce a further 100 runs of the event generator over notional 1000year periods. Each of the 20000 events created by this process was then simulated using a 10m resolution two-dimensional hydraulic model of the whole city to demonstrate the feasibility of the approach. Spatial dependence was found to be important because no single event caused the maximum flood extent at all locations and assuming perfect correlation between tributaries overestimated flood hazard. Uncertainty in estimates of inundation probability was significant to the extent that confidence intervals in risk estimates were larger than expected; however, the interaction of flows with the flood defences and valley topography gave a distinct structure to the inundation probabilities and risk.

This paper addresses the issue of identifying a state dependent input nonlinearity in a Data Based Mechanistic (DBM) flood forecasting model based on the data rather than some prior conceptualisation of nonlinearity in the system response. Four forms of nonlinear function are presented. A power law may be useful when the input non-linearity is simple. The Radial Basis Function (RBF) network method is appropriate for systems that exhibit well defined but complex input non-linearities. The Piecewise Cubic Hermite Data Interpolation (PCHIP) method also provides the flexibility to map complex input non-linearity shapes while providing the ability to maintain a natural curve. Overfit to the calibration data is a risk in both RBF and PCHIP methods when a large number of knots are used. The Takagi-Sugeno Fuzzy Inference method, together with interactive tuning, provides an alternative approach that allows human-in-the-loop interaction during the parameter estimation process but is not optimal in any statistical sense. Future work will explore the use of these methods with continuous time transfer functions and optimisation of the nonlinear function at the same time as the transfer function.

This paper considers the correction of deterministic forecasts given by a flood forecasting model. A stochastic correction based on the evolution of an adaptive, multiplicative, gain is presented. A number of models for the evolution of the gain are considered and the quality of the resulting probabilistic forecasts assessed. The techniques presented offers a computationally efficient method for providing probabilistic forecasts based on existing flood forecasting system output.

The data based mechanistic (DBM) approach for identifying and estimating rainfall to level, and level to level models has been shown to perform well for flood forecasting in several studies. The DELFT-FEWS open shell operational flood forecasting system provides a framework linking hydrological/meteorological real-time data, real-time forecast models, and a human/computer interaction interface. This infrastructure is used by the UK National Flood Forecasting System (NFFS) and the European Flood Alert System (EFAS) among others. The open shell nature of the FEWS framework has been specifically designed to make it easy to add new forecasting models written as FEWS modules. This paper shows the development of the DBM forecast model as a FEWS module and presents results for the Eden catchment (Cumbria UK) as a case study.

The January 2005 flood event in the Eden catchment (UK) has focused considerable research effort towards strengthening and extending operational flood forecasting in the region. The Eden catchment has become a key study site within the remit of phase two

All forms of hydrological prediction involve many different sources of uncertainty. Many of these sources of uncertainty involve knowledge (epistemic) uncertainties that are not necessarily easy to represent statistically. This can create problems for communication and interpretation between modeller and users when uncertain predictions are presented. One way of dealing with this problem is to define Guidelines for Good Practice in the form of a set of decisions that must be agreed and recorded for later evaluation and review. The Catchment Change Network (CCN) is a knowledge transfer project, funded by the UK Natural Environment Research Council, that aims to bring academic research and practitioners together to produce guidelines for good practice for uncertainty estimation in predicting the future in the areas of flood risk, water quality and water scarcity all of which involve important epistemic uncertainties.

The assessment and mapping of flood risk involves many different sources of uncertainty. Many of these sources of uncertainty involve epistemic uncertainties that are not necessarily easy to represent statistically. This can create problems for communication between analyst and users when uncertain flood risk maps are being prepared. It is suggested that one way of dealing with this problem is to define Guidelines for Good Practice in the form of a set of decisions that must be agreed and recorded for later evaluation and review. The Catchment Change Network (CCN) is a knowledge transfer project, funded by the UK NERC, that aims to bring academic research and practitioners together to produce guidelines for good practice for uncertainty estimation in predicting the future in the areas of flood risk, water quality and water scarcity all of which involve important epistemic uncertainties. The paper will set out the background to developing Guidelines for flood risk mapping and give an application to a site in Yorkshire, UK.

Water level predictions made with hydraulic models are uncertain and evaluating this uncertainty using Monte Carlo ensemble prediction is computationally very expensive. In this paper we show how a reduced order Dynamic Model Emulator (DME) can be used to reproduce, with high accuracy, the outputs of a large and complex 1-D hydraulic model (HEC- RAS) at specified cross-sections along the Montford to Buildwas reach of the River Severn in the U.K, together with estimates of uncertainty in the predictions. This emulation model is obtained by the application of Dominant Mode Analysis (DMA), involving the identification and estimation of nonlinear State-Dependent Parameter (SDP) transfer function models, using data generated by dynamic experiments conducted on the HEC-RAS model. The paper shows how this 'nominal' DME is able to emulate the distributed hydraulic model for a nominal set of its physically-defined parameters and it presents initial results from a complete DME that emulates the HEC-RAS model over a user-defined region of its parameter space.

Water level predictions made with hydraulic models are uncertain and evaluating this uncertainty using Monte Carlo ensemble prediction is computationally very expensive. In this paper we show how a reduced order Dynamic Model Emulator (DME) can be used to reproduce, with high accuracy, the outputs of a large and complex 1-D hydraulic model (HEC-RAS) at specied cross-sections along the Montford to Buildwas reach of the River Severn in the U.K, together with estimates of uncertainty in the predictions. This emulation model is obtained by the application of Dominant Mode Analysis (DMA), involving the identication and estimation of nonlinear State-Dependent Parameter (SDP) transfer function models, using data generated by dynamic experiments conducted on the HEC-RAS model. The paper shows how this `nominal' DME is able to emulate the distributed hydraulic model for a nominal set of its physically-dened parameters and it presents initial results from a complete DME that emulates the HEC-RAS model over a user-dened region of its parameter space.