The Challenge of Data Quality, Calibration Uncertainty and Freeboard. 28 th March Tim Jolley Technical Manager

Size: px
Start display at page:

Download "The Challenge of Data Quality, Calibration Uncertainty and Freeboard. 28 th March Tim Jolley Technical Manager"

Transcription

1 The Challenge of Data Quality, Calibration Uncertainty and Freeboard 28 th March 2017 Tim Jolley Technical Manager 0

2 Contents Introduction Case Study 1: Whitesands, Dumfries Case Study 2: Comrie, Perthshire Conclusions Questions 1

3 Introduction The aim of this presentation is show how sensitivity analysis can be used to quantify the uncertainty in modelled water levels and in particular the influence of hydrological data uncertainty The first case study, Whitesands in Dumfries, is an example of how a model re-calibration exercise was used to quantify uncertainty and to compare it with the W187 simple method for calculating freeboard allowance. The second case study, Comrie in Perthshire, is an example of a scheme located at the confluence of three rivers where scenarios have been used to assess alternative approaches to including the effect of joint probability on the design flow. The purpose of the presentation is not to promote the specific methodologies developed but to illustrate how data uncertainty can be quantified and to share our findings of two recent applications that focus on hydrological data. 2

4 Guidance Since the abstract for this paper was submitted in November 2016 the Fluvial Freeboard Guidance Note (W187) has been updated with Accounting for Residual Uncertainty. 3

5 The 5 Principles of Residual Uncertainty Allowance 4

6 Case Study 1 Whitesands, Dumfries

7 Whitesands, Dumfries Crown Copyright and database right All rights reserved. Ordnance Survey Licence number LA Greensands Whitesands meters 6

8 Whitesands, Dumfries The proposed scheme comprises a combination of embankments, walls, glass panels, demountable defences and flood gates The permanent defences will provide a 1 in 25 year Standard of Protection once all access openings have been sealed using a combination of floodgates and flip-up barriers. A 1 in 75 year Standard of Protection will be provided once demountable barriers have been installed on top of the glass panel system. To find out more and to get a VR tour of the scheme come and find us at Stand A39 7

9 Modelling Background A model was first developed in 1983 and subsequently migrated to ISIS in In 2007 it was reviewed and re-calibrated by Jacobs. In 2010, Mouchel carried out additional topographic surveys and converted the model to a 1D- 2D linked model. They calibrated the model for the flood events of March 2007 and November 2009 (estimated return periods of less than 2 and about 10 years respectively. Greensands_19 November 2009 Whitesands_19 November 2009 Stage, maod Time, hrs Observed_19 Nov 2009 Model with calibrated manning (n = ) Model with original manning (n = ) Stage, maod Time, hrs 19 Nov2009_observed" Model with calibrated manning (n = ) Model with original manning (n = ) 8

10 Modelling Re-calibration On 30th December 2015 the maximum water level recorded in November 2009 was exceeded when the peak flow at the Friars Carse Gauge was estimated to be 752 m 3 /s which is equivalent to a return period of between 21 and 25 years. Given the significance of this event the model calibration was checked and the hydraulic parameters were re-calibrated. 9

11 Potential Sources of Uncertainty Does the Model Represent the Current System? The topographic surveys were reviewed and sensitivity runs carried out The fixed head downstream tidal boundary was replaced with a time varying tidal boundary Model runs were carried out to check whether a reported embankment breach at Fishers Lodge during the 30th December 2015 event could explain the calibration error. Do Data Uncertainties Explain the Calibration Error? The datum at the SEPA gauging stations were checked for errors The SEPA rating at Greensands was compared with the modelled rating The modelled maximum water level and time were compared with photographs at a number of locations across the floodplain The accuracy of the hydrological inflows was reviewed 10

12 Hydrological Inflow Uncertainty Friars Carse Cluden Fiddlers Ford Greensands Whitesands Q/38.1 Rating curve 38 Comment Version 1 Last change 06/05/2016 Comment Validated limits (lower/upper) Validity periods for Q/38.1 From Transition from Valid since Valid to Transition to To 37 12/03/ :00:00 Created during 2015 rev iew by IM Addresses control change thought to have occurred during March 2015 spate. Lower segments based on gaugings. Upper segment retained f rom Hi-Flow rating m / m Extrapolated limits (lower/upper) m / m m <= SG <= m Q(SG) = * (SG )^2.14 [m³/s] m < SG <= m Q(SG) = * (SG )^ [m³/s] m < SG <= m Q(SG) = * (SG )^ [m³/s] 11

13 Inflow Uncertainty (The Nith at Friars Carse) 12

14 Additional Calibration Data Reducing uncertainty by seeking information on water levels and timing at as many new locations as possible. The objective was to find enough new records with accurate level and time data to balance the uncertainty of each measurement. Focus was also given to increasing confidence in the head loss coefficients at bridge structures. 13

15 Head Loss Coefficients (Afflux) 14

16 Residual Uncertainty 15

17 Freeboard Allowance Don t forget the calibration error.. Location Physical (m) Calibration (m) W187 (m) Value (m) The calibration error exceeded W187 at the upstream locations where inflow uncertainty has most leverage on water levels. W187 was large enough to capture all uncertainties tested for as we came downstream where hydraulic factors are more significant 16

18 Case Study 2 Comrie, Perthshire

19 Comrie, Perthshire The village of Comrie is located approximately 40 kilometres west of Perth at the confluence of 3 watercourses, the River Earn, the River Lednock and the Water of Ruchill. The village of Comrie and the Dalginross area of Comrie in particular, have historically suffered from repeat flooding from the watercourses flowing through the village. 18

20 Background The Ruchill Water has been gauged at Cultybraggan since June It is a natural control, all flows contained but flashiness and remoteness hinder flood gauging. There was a gauge at Aberuchill on the River Earn Station which closed in December The station re-opened in 1991 but as level only. SEPA installed level only stations on the River Earn at Dalginross and River Lednock upstream of the confluence in September 1992 SEPA have calculated an approximate rating curve for Dalginross but it is not issued. Post-flood surveys were carried out in 1993, 1995 and Lednock Aberuchill Dalginross Cultybraggan 19

21 The Challenge To establish the design flow at Dalginross downstream of the confluence and in particular to assess if the hydrological data is of sufficient quality to allow us to derive the design flow using joint probability or whether we should use a precautionary assumption with an allowance for uncertainty. Approach Calculate design flows for each river independently Calculate the design flow downstream of the confluence Establish potential scenarios that give the 200 year design flow downstream of the confluence Calibrate the hydraulic model and quantify the calibration error 20

22 Design Flow Scenarios Scenario Return Period (Years) Flow (m 3 /s) Ruchill Upper Earn Lednock Ruchill Upper Earn Lednock

23 Design Flows - Uncertainhty Lednock Aberuchill Dalginross Cultybraggan RP FEH Scenario 1 Scenario 5 Scenario 6 SS (Model) SS (SEPA) (i) (ii) (iv) (iii)

24 Model Calibration 23

25 Case Study 2: Conclusion We have confidence in the hydraulics so allow the model to determine the flow and level at Dalginross Hydrological uncertainty has to be accepted Limitations in the hydrological data and the paucity of calibration data meant that the effects of joint probability could not be included explicitly but had to be included as residual uncertainty In this case the W187 simple method was large enough to capture the uncertainty but this will have to be monitored as the scheme progresses 24

26 Conclusion

27 Challenge your understanding of the system Make the most of data and model calibration 26

28 WSP Parsons Brinckerhoff and Mouchel are at the Flood and Coast Conference for the next three days. Drop by Stand A39 to meet our technical experts and find out more about what we do in flood and coast and how we can help you. Using Virtual Reality, we will showcase Whitesands, a proposed scheme to counter flooding. Come and join us and get a live demo!