Evaluating G2G for use in Rapid Response Catchments: Final Report

Cole, S.J.; Moore, R.J.; Robson, A.J.; Mattingley, P.S.; Black, K.B.; Kay, A.L.. 2013 Evaluating G2G for use in Rapid Response Catchments: Final Report. Bristol, UK, Environment Agency, 120pp. (Report SC110003/R2, CEH Project no. C02329)

Before downloading, please read NORA policies.
N533808CR.pdf - Published Version

Download (15MB) | Preview


Flood impacts can be severe for rapid response catchments (RRCs). Providing targeted flood warnings is challenging using existing methodologies and on account of the typical absence of river flow gauging. The Pitt Review of the Summer 2007 floods recognised the need for new alert procedures for RRCs able to exploit the new distributed flood forecasting capability being progressed from research into operations. Work on the G2G (Grid-to-Grid) distributed hydrological model was accelerated into operational practice to support 5-day countrywide flood outlooks, a major recommendation of the Pitt Review. The present study aims to explore the potential of G2G to support more frequent and detailed alerts relevant to flood warning in RRCs. Integral to this study is the use of emerging rainfall forecast products, in deterministic and ensemble form, which allow the lead-time of G2G flow forecasts to be extended and given an uncertainty context. This Report sets down the overall scope of the project, provides an introduction to G2G by way of background and then reports on the outcomes of the R&D study. This includes extensive preparatory work on collating historical datasets to support G2G model assessment, both relating to hydrometry and new rainfall forecast products. A framework is developed for assessing G2G in both simulation-mode and forecast-mode (as a function of lead-time) targeted at the RRC requirement. Relevant to the requirement is the RRC Register of points and areas of interest compiled by the Environment Agency, and the characteristics of RRCs (occurring in isolation or in combination): small catchment area, urban/sub-urban land-cover and steep slopes. The assessment framework is first applied assuming perfect knowledge of rainfall observations for past and future times, so as not to confound the analysis with errors from rainfall forecasts. Variability of performance measures across groups of sites is summarised through box and whisker plots, groups being differentiated on size of catchment area and nature of G2G run (simulation, and with the addition of state updating and flow insertion in turn). Skill scores judge how well the model performs in detecting a flood event exceeding a flow threshold, taken as the median annual flood (as an indicator of bankfull flow exceedance for natural channels) and fractional multipliers of it. The skill scores include POD (Probability of Detection) and FAR (False Alarm Ratio). Performance maps of R2 Efficiency, indicating the variability in the observations accounted for by the model, are used to portray the spatial variability of G2G accuracy across the country. G2G performance in small catchments, relevant to the RRC requirement, is best over South West, North East and North West regions; also median performance appears robust from one year to the next. Larger catchments benefit most in forecast-mode from flow insertion, whilst smaller headwater catchments gain particularly from ARMA (AutoRegressive Moving Average) error-prediction. An assessment is made of using deterministic rainfall forecasts from NWP UKV - the Numerical Weather Prediction UK Variable Resolution form of the Met Office Unified Model - in a full emulation of G2G in real-time, and using foreknowledge of rainfall observations as a reference baseline. Forecast quality can deteriorate strongly beyond 12 hours, especially for smaller catchments, whilst for some locations good performance is maintained even for long lead-times. Diagnostic analysis reveals that the UKV rainfall forecasts have patterns of overestimation in some lowland areas (e.g. over London) and leeward of high elevation areas (e.g. north and south Pennines). Overall performance is better in Scotland although there is evidence of UKV overestimating rainfall near the coast at Edinburgh and Elgin in the north. The assessment framework is extended to include rainfall forecast ensembles and probabilistic flood forecasting, using a combination of case-study and longer-term analyses. Blended Ensemble rainfall forecasts are assessed in two forms: forecasts out to 24 hours updated 4 times a day, and nowcasts out to 7 hours updated every 15 minutes. The 24 hour forecasts generally perform well as input to G2G in the case studies, the G2G flow forecasts typically signalling a flood peak 12 to 18 hours in advance and ahead of any observed response for small catchments. New regional summary map displays of the probability of flow threshold exceedances over a forecast horizon, and for increasing levels of severity, are developed to highlight evolving hotspots of flood risk over time. The first ever continuous assessment of G2G probability flow forecasts is reported using national maps of probabilistic skill scores - Relative Operating Characteristic (ROC) Skill Score and Brier Skill Score (BSS) - to spatially assess their performance. It is noted that the short periods available for assessment - a 7½ month period over England & Wales and 4 ½ months over Scotland - limit the analyses to low return period flow thresholds. Half the median (2-year) flood is used although a regional pooled analysis allows some assessment up to 5-year. The G2G probability forecast assessed is the probability of the chosen flow threshold being exceeded at any time over the forecast horizon (taken to be 24 hours). Comparison of these scores when applied to deterministic and probabilistic forecasts from G2G provides strong evidence of the value of G2G ensemble forecasts as an indicator of flood risk over Britain. Noticeably poorer performance indicated by the BSS across Scotland is in part attributed to the short, summer-dominated assessment period. Operational tools available to FFC and the SFFS for using G2G flow ensembles are reviewed and options for improvement identified drawing on the experience and findings of the study. This leads to identifying some work of an operational nature for consideration in Phase 3 of the project. The report closes with a summary of project achievements grouped thematically, a set of recommendations both of a general nature and specific to FFC and SFFS needs, and finally some proposals for consideration under Phase 3 of the G2G for Rapid Response Catchments project. Some key benefits arising from the project are summarised below. • Evidence has been produced that shows G2G has good skill in providing strategic forecasts for RRCs. The evidence is stratified by catchment type (area, urbanisation, headwater), form of forecast (simulation or forecast mode) and nature of rainfall input (raingauge, deterministic forecast, ensemble forecast). • Strong evidence has been presented on the advantage of using an ensemble rainfall forecast as input to G2G to obtain a probabilistic flood forecast for an RRC, relative to an approach where only a single deterministic rainfall and flood forecast is obtained. This indicates better guidance can be given on forecast flood risk for RRCs, improving the level of service provision for such catchments which are currently not well served. • An improved G2G model configuration, exploiting gauged flows from 912 sites and including new locally calibrated parameters, has been delivered and made operational for the FFC with England & Wales coverage. The benefit is improved operational flood forecast accuracy. For Scotland, an enhanced configuration will be delivered to SFFS in Spring 2014. • Detailed recommendations on how the visual presentation of G2G ensemble results could be improved are set down in this report. When further developed and implemented, these will prove of benefit to the preparation of Flood Guidance Statements issued by FFC and the SFFS across Britain.

Item Type: Publication - Report (Project Report)
UKCEH and CEH Sections/Science Areas: Reynard
Funders/Sponsors: Environment Agency, Flood Forecasting Centre, Scottish Environment Protection Agency, NERC/Centre for Ecology & Hydrology
Additional Information. Not used in RCUK Gateway to Research.: Please cite the report as: Cole, S.J., Moore, R.J., Robson, A.J., Mattingley, P.S., Black, K.B., Kay, A.L. (2013) Evaluating G2G for use in Rapid Response Catchments: Final Report. R&D Project Report SC110003/R2, Joint Defra/EA Flood and Coastal Erosion Risk Management, Research Contractor: CEH Wallingford, Environment Agency, Bristol, UK, 120pp.
Additional Keywords: Flood Forecasting, Ensembles, G2G, Rainfall-Runoff, Distributed, RRC, Rapid Response Catchments
NORA Subject Terms: Hydrology
Date made live: 05 Jan 2023 14:53 +0 (UTC)

Actions (login required)

View Item View Item

Document Downloads

Downloads for past 30 days

Downloads per month over past year

More statistics for this item...