Accounting for seasonal dependence in hydrological model errors and prediction uncertainty

Robertson

et al. 2017

110

105

Abstract. Considerable research effort has recently been directed at improving and operationalising ensemble seasonal streamflow forecasts. Whilst this creates new opportunities for improving the performance of water resources systems, there may also be associated risks. Here, we explore these potential risks by examining the sensitivity of forecast value (improvement in system performance brought about by adopting forecasts) to changes in the forecast skill for a range of hypothetical reservoir designs with contrasting operating objectives. Forecast-informed operations are simulated using rolling horizon, adaptive control and then benchmarked against optimised control rules to assess performance improvements. Results show that there exists a strong relationship between forecast skill and value for systems operated to maintain a target water level. But this relationship breaks down when the reservoir is operated to satisfy a target demand for water; good forecast accuracy does not necessarily translate into performance improvement. We show that the primary cause of this behaviour is the buffering role played by storage in water supply reservoirs, which renders the forecast superfluous for long periods of the operation. System performance depends primarily on forecast accuracy when critical decisions are made -namely during severe drought. As it is not possible to know in advance if a forecast will perform well at such moments, we advocate measuring the consistency of forecast performance, through bootstrap resampling, to indicate potential usefulness in storage operations. Our results highlight the need for sensitivity assessment in value-of-forecast studies involving reservoirs with supply objectives.

Section: Actual Forecasts: Forecast Guided Stochastic Scenarios (Fogss)mentioning

confidence: 99%

Complex relationship between seasonal streamflow forecast skill and value in reservoir operations

Turner

Robertson

et al. 2017

110

105

“…To handle zero values in ephemeral catchments, we treat observations of zero as censored values in the likelihood, a technique established previously (Li et al, 2013).…”

Section: Estimating Parametersmentioning

confidence: 99%

Assessment of an ensemble seasonal streamflow forecasting system for Australia

Wang

Robertson

et al. 2017

Self Cite

Abstract. Despite an increasing availability of skilful longrange streamflow forecasts, many water agencies still rely on simple resampled historical inflow sequences (stochastic scenarios) to plan operations over the coming year. We assess a recently developed forecasting system called "forecast guided stochastic scenarios" (FoGSS) as a skilful alternative to standard stochastic scenarios for the Australian continent. FoGSS uses climate forecasts from a coupled oceanland-atmosphere prediction system, post-processed with the method of calibration, bridging and merging. Ensemble rainfall forecasts force a monthly rainfall-runoff model, while a staged hydrological error model quantifies and propagates hydrological forecast uncertainty through forecast lead times. FoGSS is able to generate ensemble streamflow forecasts in the form of monthly time series to a 12-month forecast horizon.FoGSS is tested on 63 Australian catchments that cover a wide range of climates, including 21 ephemeral rivers. In all perennial and many ephemeral catchments, FoGSS provides an effective alternative to resampled historical inflow sequences. FoGSS generally produces skilful forecasts at shorter lead times ( < 4 months), and transits to climatologylike forecasts at longer lead times. Forecasts are generally reliable and unbiased. However, FoGSS does not perform well in very dry catchments (catchments that experience zero flows more than half the time in some months), sometimes producing strongly negative forecast skill and poor reliability. We attempt to improve forecasts through the use of (i) ESP rainfall forcings, (ii) different rainfall-runoff models, and (iii) a Bayesian prior to encourage the error model to return climatology forecasts in months when the rainfallrunoff model performs poorly. Of these, the use of the prior offers the clearest benefit in very dry catchments, where it moderates strongly negative forecast skill and reduces bias in some instances. However, the prior does not remedy poor reliability in very dry catchments.Overall, FoGSS is an attractive alternative to historical inflow sequences in all but the driest catchments. We discuss ways in which forecast reliability in very dry catchments could be improved in future work.

“…The method of maximum likelihood is used to estimate the error model parameters θ E and the hydrological model parameters θ H . Using a similar derivation as given by Li et al (2013), the likelihood functions can be written as a. for AR-Norm…”

Section: Estimationmentioning

confidence: 99%

“…The probability density function is replaced by the cumulative probability function when evaluating events of zero flow occurrences (Wang and Robertson, 2011;Li et al, 2013). The shuffled complex evolution (SCE) algorithm (Duan et al, 1994) is used to minimise the log likelihood.…”

Section: Estimationmentioning

confidence: 99%

A strategy to overcome adverse effects of autoregressive updating of streamflow forecasts

Wang

2015

Self Cite

Abstract. For streamflow forecasting, rainfall-runoff models are often augmented with updating procedures that correct forecasts based on the latest available streamflow observations of streamflow. A popular approach for updating forecasts is autoregressive (AR) models, which exploit the "memory" in hydrological model simulation errors. AR models may be applied to raw errors directly or to normalised errors. In this study, we demonstrate that AR models applied in either way can sometimes cause over-correction of forecasts. In using an AR model applied to raw errors, the overcorrection usually occurs when streamflow is rapidly receding. In applying an AR model to normalised errors, the overcorrection usually occurs when streamflow is rapidly rising. In addition, when parameters of a hydrological model and an AR model are estimated jointly, the AR model applied to normalised errors sometimes degrades the stand-alone performance of the base hydrological model. This is not desirable for forecasting applications, as forecasts should rely as much as possible on the base hydrological model, with updating only used to correct minor errors. To overcome the adverse effects of the conventional AR models, a restricted AR model applied to normalised errors is introduced. We show that the new model reduces over-correction and improves the performance of the base hydrological model considerably.