.1

' A R CH I VE :

PLEA SE D O N OT DE S TR OY

ii Thames Water Tidal Thames Defence Levels Stage 1 Report on Data Assembly and Preliminary Analyses

June 1987

Thames Water Thames Water Tidal Thames Defence Levels Stage1Report on Data Assembly and Preliminary Analyses

June 1987

Sir William Ha !crow & Partners Ltd Institute of Hydrology Slr Wil liam Halcrow & Partners Ltd HALCROW Burderop Park, Swindon, Wiltshire SN4 00D, England Consulting Engineers Telephone (0793) 812479 •International Telephone +44 793 812479 Telex 44844 Halwil G Fax 0793 812089 International Fax + 44 793 812089 And at Vineyard House, 44 Brook Green, L.ondon W6 7BY, England Telephone 01-602 7282 'NEr A P L Smith Thames Water Nug ent House Vastern Road Reading BERKSHIRE RG1 8DB

19 June 1987 Our ref winG/TTm/11 Your rM

Dear Mr Smith

TIDAL THANES DEFENCE LEVELS

We have pleasure in submitt ing 20 cop ies of our Stage 1 repor t on Eiata Assembly and Pr eliminary Analy ses .

Please do not hesitate to con tac t me or Max Beran at IH if you have any queries .

Your s sincerely

E P Evans DIRECTOR

Enc

EPE/sgb

Directors Consultants $ Baxter BSc. FICE Chairman S Gray FICE J L Beaver FICE I C Pnce FICE Sir Alan Muir Wood FRS FEng ACE C L D arks MA A CE A R Hardy FICE J C Buswell BSc FICE N A Trenter MSc MIGEOL N 1 Cochrane BSc FICE E J D Mansfield FRIBA H J Ambler FCCA F E Chasemore BSc F/CE J Weaver PnD MICE E Loevw BSc FICE A R Kopec ACE D Buckley FICE P A S Ferguson MASM MICE J P Wood BSc AcE K Arnold PhD FICE W Rotnwell MA ACE N G Jot/neon ELSc ACE C A Fleming PhD MICE J Ahmed BSc FIE A I Robertson BSc FICE D S Kennedy BSc MICE G 0 Hilber BSc FICE R N Craig BSc MICE Registered in England No I 722541 SUMMARY

Th is. report describes the progress made and conclusions drawn dur ing Stage 1 of the Tidal Thames De fence Levels study and makes recommendations on the scope of work and methodology for the next stage of the project . The broad aims of the study are to establish the reliability of the tidal de fences in the Thames and to define downstream boundary conditions for various named tributaries for use in future studies.

Th e River Crane was the only tributary included explicitly in the or iginal scope of work but the follow ing additional work ha s eeen requested during Stage 1.

Flow—frequency ana lysis for the River Crane Similar studies for River Lea as required for River Crane Extension of ONDA model down to Southend and up to Molesey Crane and River Lea Joint probability distributions of levels and discharges on River Extension of the anlaysis up to 1 in 200 year return period .

A revision to the agreement and revised cost estimates to include these items has been submitted to Thames Water for approva l.

Data collection ha s been a ma jor component of Stage 1. Data acquired comprises tida l leve ls (including 47 years of diurnal tida l max ima), fluvial discharges (annual maxima , daily means and selected event data) topograph ical data for the com putational hydraulic model, and general information .

A report on preliminary anlaysis of flows and leve ls at the m outh of the River Crane was submitted on 24 Apr il 1987. The results were required urgently for use in the com putational hydraulic mode l of the River Crane being developed by Thames Wa ter. The 50 year return period flow was estimated to be 20 .6 m3/s for current conditions rising to 22 .5 m3/s for future 75% urbanisation conditions. The 50 year re turn period tide heigh t was found to be 5 .50 m based on histor ic data . At this probability level the Thames Barr , which has operat ing rules which aim to keep leve ls at Tower Pier below 4 .85 m , could reduce the Crane mouth level to about 5.06 m . However this value is a very pre lim inary estimate based solely on pre—barrier conditions and w ill be revised during the ensuing stage s o f the study .

Preliminary ana lysis of the River Lea is nearing completion and results w ill be presented imm inently under separate cover in a similar form to the preliminary report for the River Crane. The River Lea channel network is much more complex than the River Crane and a much greater effort has been necessary in the preliminary stages to der ive sensible conclusions . However this should be compensated by a lessening of the workload in the main River Lea anlaysis proposed for Stage 4 of the project . The complex network of channels and regulating structures and the numerous operating procedures m ake the accurate prediction of floods within specific channels impossible without the application of a riv er model.

Probability anlaysis of the primary data series is well under way but will overlap into Stage 2. Investigation of correlation be tween the primary data series indicates a we ak tendency for interaction between surge and riverflow wh ich w ill probably be modelled adequately by a seasona l or other decomposition of the data .

The ana lysis of trend in fluvial flows indicates that it can be safely excluded from the study . Trend analysis of annua l maximum tide levels indicate a nearly uniform pattern of trend throughout the tideway wh ich is compatible with va lues obtained by IOS for mean sea level (2 to 3 mm /annum).

Following a review of previous model studies and re ports the ONDA mode l has been recommended to be adopted for the study as the most cost effective , re liable and ver satile option ava ilable. On verbal approval from Thames Water the model has been extended to encompass the whole of the Tida l Thames from Molesey Lo c k to Southend , including the Thames Barrier . Additiona l topographic data w as taken from the PLA hydrographic charts and data acquired from the Thames Water Barrier office and the surveys section .

The Thames barrier has changed the probability distributions of water leve ls in the tideway quite dramatically . A strategy for including this effect into the ana lysis by combiåing the bivariate probability distr ibution of fluvial flows and Southend leve ls w ith structure functions at specific locations has been defined . The structure function is a diagram show ing lines of equa l water level at a selected location in the tideway for di fferent combinations of fluvial flows and downstream tide levels.

It is proposed that Stage 2 of the study w ill continue much in the same manner as envisaged in the origina l Consultancy Brief, although it will be necessary to take much more careful consideration o f the dom inating effec t o f the Thames Barrier . TIDAL TH AM ES DEFENCE LEVEL S

STAGE 1 REPORT

Page No

INT ROD UCTION 1 1 .1 Gener al 1 1 .2 Scope 1 1 .3 Addi tional Work 1 1 .4 Progre ss 2

DATA COLLEa ION 4

HYDROLO GY 6 3 .1 Prelim ina ry Study o f Hy dro logy 6 of River Crane 3 .1 .1 Flood Freq uency Rel ationsh ip 6 3 .1 .2 Tidal An alysi s 6 3 .1 .3 High Return Per iod F loods 7 3 .2 Prelim ina ry Stu dy o f Hy drolo gy at River Lee Ana lysis 8 3 .2 .1 Objectives 8 3 .2 .2 The L ow er L ee System 8 3 .2 .3 Data Availability 8 3 .2 .4 Tida l Water L eve l Data 9 3 .2 .5 Floo d Data fo r th e L ow er Lee and Trib ut aries 10 3 .2 .6 Analysis o f Flow s and Levels 10 3 .3 Probability Ana lysis o f Tham es L evels 10 an d Flow s 3 .3 .1 Stati sti cal Ana lysis o f Th am es Di scharge Data 11 3 .3 .2 Statistical Ana lysis o f Tide Data 12 3 .3 .3 Co rrel ation Be tween Primary Data serie s 13 3 .3 .4 Effect o f Tren d o n Re cord Breaking 13 Event s 3 .3 .5 Req uirem ent s for Fur ther Analysis 15

COMP UT ATIO NAL HYDRA IL IC MODEL 16 4 .1 Review of PreviousM odels 16 4 .1 .1 LOS Mo del 17 4 .1 .2 Th e O NDA M odel 17 4 .1 .3 Ch oice o f Mo del 18 4 .2 Ex tension o f Tidal Tham es O NDA m odel 19 4 .2.1 To pogra phi cal Data 19 4 .2 .2 Th am es Barrier 19 a Pag e No

a

II

a

II e a es m a a m a n an e o n INTRODUCTION

1 .1 General

This interim re port de scribe s th e pro gress o f w ork carried out to date and con clusio ns drawn d uring Stage 1 of the Ti dal Thames Defence L eve ls study . Th e obje ctives o f th e study are a s de scr ibed in the Cons ultan cy Bri ef, wh ich w as confirmed in th e letter o f appoi ntm ent by Tham es W ater on 27 Februa ry 1987 and form s th e ba sis for th e agr eem en t w ith Tham es Water . Th e wo rk is being undertaken by Sir W ill iam Halcrow and Partner s in association w ith th e Institute of Hydro lo gy . Sin ce commenceme nt o f the study the sco pe has cha nged, w ith req ue sts from Th ames W ater for furth er w ork to incorpo rat e analy sis o f floo d flows in th e River Crane an d als o to ca rry out a sim ilar inve stiga tion o f flows and l eve ls for the river Lee .

1 .2 Sco pe

Th e m ain scope o f Stage 1 were a s follows :

(a ) Assemble an d asse ss river flow data at Tedding ton an d at th e tribut aries, downstream , and l evels a t th e down stream tidal boun da ry and interme diate tide gauge s.

(b) Pro bability ana lyse s o f flow s at Te dding ton and level s at downstream boundary including degre e of dependen cy be tw een the two series.

Prel im ina ry estim ation of sta ge- freq uency rel ationsh ip at th e river Cran e m outh .

Study ex- GL C com put ational hydraul ic m odels and m ake recomm en dations o n h ow th ey can be use d in co njunc tion w ith th e Teddington- Gallion ONDA m ode l .

Optio nal stage la to extend ON DA model to the down stream tida l bo undary and include all struc ture o perating rul es.

Addi tional Work

D ur ing Stage 1 sever al item s o f additional work w ere req ue sted by Th am e s Water . Th ese were o utlined in th e pro po se d revisio n to th e Consultancy Br ief wh ich w as subm itted for a pproval to Th ames W ater on 11 M ay 1987 together w ith re vised co st estim ate s. Th e revision invo lve d:

additiona l analyses to incl ude flow- freq uency for th e River Crane, a sim ilar study for the River L ea as is to be ca rried out for the River Cran e, extension of the ONDA model down to Southend including the Thames barrier operating rules , joint probability distribution of levels and di scharges on both the River Crane and River Lea . extend the analysis up to 1 in 200 years return period

Approva l from Thames Water for these amendments to the scope of work is awaited , although verbal approval was given in the progress meet ing of 14 May 1987 for both extension of the ONDA model and preliminary river Lee analysis to commence immediate ly.

1.4 Progress

Such items as fall under Stage 1 of the study are now complete:

The main bu lk of data required for the study has been collected and input to the computer da tabase prior to anlaysis . Outstand ing data requested from Thames Water include river and tributary flows for ca libration of the hydraulic model to be undertaken in Stage 2. Data acquisition is covered in Chapter 2.

Preliminary analys is of flows and leve ls at the m outh of the River Crane was the most urgent requirement during Stage 1 of the study , since the results were required for immediate use by Thames Water in the computational hydraulic mode l of the River Crane . The preliminary report for this component of the study was submitted on 24 April 1987 and a summary of results is presen ted in Section 3 .1.

Preliminary ana lysis of flows and levels for the downstream reaches of the River Lea has a lso been carried out during Stage 1 of the study . The results are to be presented under separate cover and a brief description is presented in Section 3 .2.

A review of the available com putationa l models and relevant reports from previous studies has been carried out . A comparison of the ONDA model and the IOS m odel of the tidal Thames is presented in Section 4 .1. The recommendation has been made to adopt the ONDA model as the most cost effective option for the study .

After receipt of verbal approval, given by Thames Water at the progress meet ing of 14 May 198 7, the ONDA mode l has been extended to include the complete reach from Mo lesey Locks down to Southend tide gauge including the Thames Barrier. This is described in Section 4.2. Th e influence of th e Tham es Barrier on levels in th e tidew ay is m uch mo re pronounced than originally anticipated . The procedure for including the effect of the barrier into the probability an d hydraulic anlayse s has been identified and is de scribed in Section 4 .3 .

Probability analysis of the primary data series (flow s at Kingston and levels at Southend) has begun, particularly with respect to trends in discharge and lev els, distribution of tidal surges and correlation be tw een the two primary var iables. Progress to date is described in Section 3 .3 . Th is com ponent w ill extend into Stage 2 of the study. 1 1 i a

M I W I In n l M I la I N I W O I lli In i n 11 OS a n MN 4 DATA COLLECTION

Da ta collection h as been a m ajo r com ponent of th e Stage 1 work . Data r eq uired for th e stu dy com prise s tidal lev el s, fluvial di scharge s, topo gra ph ical da ta and gener al inform ation:

(a) 47 year s o f the diurnal tidal m ax im a fo r tide gauge s at South end , Richm ond , Tow er Pier and Gallions (12 year s only ) have been colle cted from PLA an d entere d to the com puter .

(b ) 104 years o f daily m ean flows at Kingston/Teddington gau ging station have been ext racte d from the data ar chive h eld at Ili.

(c) Annual m ax im a and th reshold ex cee dences for sta tions in (a ) and (b ) have been obtained, in th e case o f (a ) from 1911 .

(d ) Ann ual m ax im a and POT flood di scharge da ta for M ar sh Farm and Cranford Park (R Crane ) were ext racte d from m icro film da ta held at IH and brough t up to date from charts a t Thame s Water 's W alth am Cross o ffice.

(e ) Annual m ax im a and POT flood disch arge da ta for Fielde s Weir , L ow Hall an d five trib utar ies (R Lea ) have bee n taken from th e IH da ta arch ive and W altham Cro ss. Annua l m ax im a have also bee n collecte d for tide gauge s at Bow Locks and Brunsw ick Wharf on the River L ea .

(f ) Tide Charts have bee n collected for tide gauge s at Richm ond, Ch elsea , Tow er Pier , Charl ton , Sil vertown , Gallio ns, Erith Dee p Water, T ilb ury , Co ry to n an d So uthend for sever al events w ith va rious combina tions o f sur ge tide s, h igh fluvial flow s and Barrier clo sures for use in calibration o f th e hydr aul ic m odel . Concurrent flow da ta have not yet bee n re ceived from Th ame s Water. Info rm a tion on gate m ovem ent s, particul arly at Ted dington Lock, are not yet com plete .

(g) Topo graph ica l da ta fo r ex ten sion o f th e hydraul ic m ode l have bee n take n from hydrographic cha rts o btaine d from PL A and cross sectiona l data have bee n input to th e com puter . Cro ss se ctio nal data fo r th e Mole sey Loc k to Teddin gton reach were t aken from th e River Th am es Mode l da taba se .

(h ) Gate dimensio ns for Thames barrier and Ted dington loc ks have also been obtained and e ntered to th e m odel . (i ) General information collected include s statutory and interim defence levels, informatio n on barrier operation, Storm Tide W arning Service (STWS) analysis o f forecast error s etc. en

a fl a n e e fl a a n n HYDROLOGY

3.1 Preliminary Study of Hydrology of the River Crane

The prelim inary study of the River Crane (Paragraph 2.1.4 of the revised terms of reference) wa s required to provide TWA with a rap id assessment of design conditions for an ongo ing pro ject invo lving a computationa l hydraulic model of the River Crane. The objectives are to provide a stage frequency relationsh ip for the Thames tideway at the Crane confluence and a flood frequency relationship for the Crane at Marsh Farm. At the time the work was carried out little was known of the possible impact on the river Thames stage frequency relationsh ip of the Thame s barr ier operation , so the va lues presented represent very much an upper bound to that which is poss ible . The follow ing represen ts a brief summ ary of the results prevented in the preliminary report , which was submitted in April 198 7.

3.1.1 Flood Frequency Re lationship

Annua l Max imum and Peak over Thresho ld series were extracted for the full period of record (1939-date) at Marsh Farm gauging station . The effect of increasing urbanisation is very apparen t in the increasing frequency of exceedences over the years. For example the 194 7 flood was the largest up to that date but was exceeded in 1960 and on eight occasions since 1967 . The effect of urbanisation on the mean flood conformed close ly w ith the method of Flood Studies Supplementary Report No 5 (FSSR5 ) which predicts an increase of 60% over the period of record.

The data for the period from 1972 to 1986 was used to estimate the current mean ann ua l flood . This was then scaled us ing the appropriate regional flood frequency relationship adjusted for the influence of urbanisation . This estimation strategy yie lded a 50 year return period flood peak of 20 .6m 3/s for current conditions rising to 22 .5 m3/s for future 75% urbanisation cond ition s .

3 .1.2 Tida l Ana lys is

The short stage record (1976-da te) kept for the Crane tide gates prov ided a certa in amount of information on the re lationship between the much longer Richmond level records (1911-date) and tho se at the Crane m outh . The difference between the two sites was typically 10 cm although quite large departures w ere on occasion observed . The tendency for a dim inish ing difference w ith increas ing tide height was confirmed after ana lysing Richmond and Tower Pier high tidal levels. Sea leve ls in the Thames estuary are known to be rising relative to land levels and a figure of 3 mm/yr has been quoted as a working hypothesis in the prelim inary Crane report. Th is figure was applied to Richmond too although the annual maximum data themselves suggest lower trends for the upper estuary . A preliminary trend ana lysis of the tida l records led to a va lue for the 50 year re turn period tide he ight of 5.50 m .

Information received subsequently showed that the barrier shou ld lim it leve ls in the tideway to a max imum he ight of 4 .85 m at London Bridge . Based on comparison of the fitted probability distributions this is approximately equ ivalent to 5.06 m at the mouth of the Crane. Subsequent ana lysis should reveal whether this equiva lence wh ich is based on pre-barrier data wou ld ho ld for current conditions .

3.1.3 High Return Period Floods in the Tributaries

Subsequent Requests from Thames Water for estima tes of h igh re turn period flood magnitudes and for an analys is of the Lea basin data has lead to the definition of a genera l analytical strategy for tributary streams to the tideway wh ich tries to make maximum use of the relative ly long flow records that are available in the London area . Th is can be specified as follows :

(i) fit the Gumbel distr ibution to the ann ua l maximum series of sites w ith due regard to heterogeneity due to urban isation ; (ii) use this fit to estimate floods up to a maximum return period of 2N years where N is the length of record used to fit the distribution ; (iii) extrapolate to higher return periods from this point using internal ratios der ived from the regiona l flood frequency curve (with due regard to the influence of urbanisation on the shape of the growth curve).

Th is strategy wa s applied to the Crane da ta . The parame ters of the Gumbel fit to the recent portion of the record is a = 2.4483, u = 9 .157 m 3/s . It was considered prudent to restrict extrapolation with the se parameter va lues to 20 year re turn period (16 .42 m3/s) beyond which the method of FSSR5 wa s applied . As explained in that report the method of extrapolation beyond 50 years was constructed in a consistent fashion but was not based on recorded data . The extrapolation was reviewed briefly taking advantage of the additiona l years of record now ava ilable . Data were assembled from 12 stations with 151 station years of record in the London area all with urban fractions between 50 and 81 per cent. The average growth factors for th is data set were 2 .13 and 2 .37 at 100 and 200 year return periods . This accords quite closely with the values shown in Table 3 .1 computed from FSSR5 of 2.19 and 2.44 . Table 3 .1 R Crane at M arsh Farm- Flood Frequency (m3/s)

AM return period - years 2 5 10 25 50 100 200

10 .1 12 .8 14 .7 16 .9 18 .5 20 .7 23 .2

3 .2 Prelim inary River Lea Analysis

3 .2 .1 Objectives

Th e objectives of th is prelim inary study were to ascertain th e freq uencies of flood s an d peak water levels in th e Lower Lea and the freq uency of peak w ater levels in the Tham es near the Lea confluen ce. Th e low er Lea reache s in question are the tidal sections of th is river sy stem up to Lea Bridge on the Naviga tion channel and nearly up to the Flood Relief Channel river gauging station at Low Hall, and the sections immediately upstream of the tidal influence (see Fig ure 3 .1).

Preliminary analysis of th e River Lea is nearing completion and results w ill be presented imm inently under se perate cover in a sim ilar form to th e prelim inary report for the River Crane.

Th e River Lea channel network (Fig ure 3 .1) is m uch more com plex than th e River Crane and a much greater effort h as been necessary in the prelim inary stages to derive sensible con clusions . Th e main problem stem s from th e facts that Low Hall ga uge has a short record, for wh ich deficiency Feildes We ir record had to be ex am ined, and no records ex ist on the second branch of th e Lea (the Navigation Channel) fo r wh ich flood freq uencie s have had to be synthesised . How ever th is w ill be com pensated by a lessening o f the work load in th e main River Lea analysis propose d for Stage 4 of the project.

3 .2 .2 The Low er River L ea System

Th e Lea river system is very com plex in it s lower reaches downstream of Feildes W eir (See Figure 3 .1) . Until Feildes Weir the Lea ha s a predom inantly rural and chalk catchmen t an d generally natural channels, though signi ficant offtakes ex ist ups tream which supply various water undertakings. At Feildes Weir the Le a is regulated through two ch annels flanking its flood pl ain, the western channel supplying and later becom ing th e Lea Navigation Channel , and the eastern ch annel being th e Lea Flood Relief Channel. Upstream of the Py nmes Brook confluence, relatively steady flows are maintained w ithin th e Navigation Channel so far as WIl II III a 12111 0 0 M IS US • IIIII IS IS M I M I S IS

R.Lea Rye Needs R.Stort R.Lynch Glen Tuba Brook Woolens Brook I Feildes Weir C.S.1 Sp ita I Brook

Enf ield Edmonton L., The Ching Silver Street &SI

• Chalk Bridge

I Low Hall GI I CoppermIll Stream

Bogen ham Brook _

Layton

R.Lea Grand Union Canal

Bow Back River Three Wills . • • • Wall River Homeill Al Bow Locks • ' Prescott Channel Limehouse Cut Bow Locks T.& Bow Creek Brunswick Lea Tidel Barrier Wharf I C RIVER THANES possible by diverting flood flows eastw ard through several channels to the Floo d Relief Channel. A number o f large raw water storage and service reservoirs ex ist w ithin th e flood plain be tw een the two main channels , an d are linked wi th th e 1 at ter by a n umber of addi tional channels to a f fect supply and to accommodate spill age . The com plex netwo rk of channels and reg ulating st ructure s, and the num erous al ternative operating procedur es, make the accurate prediction o f floods within spe cific channels im possible without the appl ication o f a river model .

As the L ea channels pass southw ards from Feildes Weir, the cat chments o f incom ing Crib utaries from east an d we st are progressively m ore urbani sed , th e lowest such as Dagenham Brook and Th e Moselle being almost en tirely b uil t-up. The degree of urbani sation in the Low er Lea tributaries has increased significantly in recent decades, and this infl uence wo uld be ex pected to increase flood peaks unl ess bal ancing pond de velopm ent ha s kept pace.

3 .2 .3 Da ta Avail abil ity

Water levels have been recorded at a number of stations along the tidal Thames from Teddington to Southen d. Th e neare st Thame s w ater level recording station to th e m outh of the River Lea i s at Brunsw ic k Wharf (Figure 3 .1) , wh ich is 0 .5 km upstream of the confluence . Th is station ceased to operate in 1983 when the Thames Barrier was comm issioned, but it s effec tively continuous record provides an annual series of water level peaks of 30 years. Upstream o f the confluen ce with the Thame s b ut within the tidal reach of the River Lea, the Bow Locks w ater 1 evel recording sta tion offers a data set o f 27 years o f ann ual peak water levels since its start in 1934 until its data were a ffected by th e comm issioning o f the Lea Barrier in 1972. This data serie s include s the exceptionally severe events of 1953 and 1949 . The nex t nearest Thame s long term water level recor ding stations upstre am and downstream of th e L ea confluence are at Tow er Pier and N Woolwich (Gallions) . Data for th e latter stations extend back t o 1912 and 1915 respectively . . Trends are also de tectabl e in the tidal w ater level records . Howe ver, the m ajor infl uence on Lea confl uence wa ter level freq uencies is now the Tham es Barrier, the precise ef fect s o f which w ill only become apparent in the course of th e ful 1 anlaysis in Part 2 of the study . .

Floods on the m ain River Lea are measured at Feildes Weir and at Low Hall on the Floo d Relief Chann el. No river gauging station ex ists on th e Navigation Channel , wh ich , al though floods are diverted from it to th e Floo d Rel ief Channel for much o f its course, act s as th e flood collecting channel for a number o f fl ood- prone trib utaries south of the Turkey Brook confluence, including am ong o thers th e Salmon Brook, Pymm es Brook and Sadl er s M ill stre am . Dire ct estim ate s are th erefore possible o f flood freq uencies on the Flood Rel ief Ch annel us ing the short L ow Hall record supplemen ted by Feildes Weir longer record. H owever, flood freq uencies on the low er Navigation Ch an nel m ust be deduce d from flood and catchment cha rac teristics o f gauge d trib utaries such as Salm on Brook, Pymm es Brook and Turkey Bro ok.

3 .2 .4 Tidal Water Level Data

Water level recor ds relating to th e Th am es/Lea and tida l low er Lee have been collecte d from Thame s Water for Brunsw ick Wha r f, Bow Locks and the L ee Barrier . These da ta are in th e form o f m icro film . Ann ual m axim um peak s an d the date s o f these peaks have bee n ext racted from th e m icro film s . Further tida l Th am es w ater level data fo r oth er stations upstream and downstream o f the Thame s/L ea confl ue nce have been presented in th e Pre l im ina ry Repo rt on River Crane Flow s and Le vels, an d som e o f the se h ave been use d in this study .

3 .2 .5 Flood Data for the L ow er Lea and Tributaries

Floo d peaks, gen erally in th e fo rm o f ann ual m ax im um (AM ) ser ies and som etim es peak- over- threshold (POT ) series, have been extracted from th e records o f usa ble river gauging sta tion data fo r th e Lower L ea and tributa ries. In the cour se o f ex tracting th ese da ta, cer tain data set s h ave been rejected as bein g of too poo r q uali ty .

The ana lysis o f flood flow s and tida l levels in the low er reach es o f th e river Lea is almo st com plete . Result s and con clusio ns w ill be presented under se parate cover .

3 .3 Probability Analysis of Primary Data Series

The fitting o f statistical di stributions to tide and fluvi al flow serie s an d the study o f th eir interactio n is a pre req uisite for the deriva tion o f the di strib ution o f m ax imum levels at interm edi ate point s in the tidew ay . Con ven tio nally fo r floo d de sig n pur pose s the sta tistics are ex pressed as return per iods be tw een annua l m ax im um even ts . H owe ver in or der to com bine th e statistic al distrib utio n w ith th e hydraul ic m ode lling it is ne cessary to use a tim e ste p app ropria te to th at over wh ich events nat urally interac t. Thus it is necessary to relate annual m ax imum statistics to those a t daily and sem i-month ly time scale s. Th e follow ing sec tions illustrate th e form of ana lysis th at w ill be required during stage 2 . Sufficient n umerical info rmatio n is give n to perm it an apprecia tio n of som e o f th e numeri cal problem s that w ill have to be solve d.

10 3 .3 .1 Statistical Ana lysis o f Thames Discharge Data

Th e Thames at Ted dington/Kingston prov ides over 100 years of data for annual m ax im um and daily analysis . Two series are available: gauged flows and naturalised flows (which in clude upstream abstractions etc). The latter series are req uired for statistical purpo ses such as trend analysis. However it is th e gauged flows wh ich are experienced w ith in th e estuary so these are used in structural relation ships of estuary levels . a ) Relationsh ip between daily and annual quanti ties:

Three internal statistical relation ships are to be inve stigated concerning the probabil ities of ex ceeding given di scharges expressed in ann ual m axim um , daily an d sem i-month ly term s. b) The 1983 Surface W ater Y earbook presents some evidence for a climatically induced trend in ann ual runo ff total s. Fluctuations in the annual runoff are seen also by inspection o f successive 20-year time slices from 1883 - 62 , 81 , 88 , 76 and 85 m3/s - wh ich differ by amounts considerably in ex cess o f th eir standard errors . The graph ical presen tation used in the Surface Water Yearbook, cumulative departure diagram , is not well eq uipped for detecting trend h owe ver, so a mo re fo rm al analysis was used.

A linear regression of the logarithms of mean flows on tim e was considered. This indicated a sm all trend, eq uivalen t to one q uarter per cent per annum , in th e annual runoff which w as at th e lim it of signi ficance. T o in ve stiga te further more te sts were carried out on the m onthly runoff totals which indicate tha t only in th e low- flow season is the trend sign ifican t. Th is may be d ue to a tendency towards drier summ ers (no t reported elsewhere ) or to da ta error, eg low flow rating changes or overestimation of th e withdraw als for supply in earlier years.

Further insigh t into these suppositions was obtained by analysing the m inim um an d m ax im um daily values each mon th and over th e wh ole year . As shown on Table 3 .2 the trend in m inim a is most m arked during the summer months. Th ere appear s to be no significant trend in th e annua l m ax ima and that in the summ er months is probably a reflection of the low flow trend . More sophisticated tests are pl anned w ith in the M aidenhead Flood Study to furth er discriminate the clim atic influence b ut for present purpose s it is felt that fluvial trend can be neglected . Table 3 .2 Kingston/Teddington trend in per cent per year in mean , minim um an d maxim um daily flows in each month

M onth Mean M inimum Max imum Trend t Trend t Trend

Jan 0 .56 1 .38 0 .52 1 .22 0 .55 1 .43 Feb 0 .73 1 .75 0 .79 2 .13 0 .84 1 .92 M ar 0 .57 1 .48 0 .70 1 .99 0 .51 1 .18 Apr 0 .76 2 .19 0 .80 2 .59 0 .89 2 .20 M ay 0 .93 2 .62 0 .96 3 .01 1.17 2 .71 Jun 1 .00 2 .70 0 .98 3 .37 1.21 2 .53 Jul 0 .83 2 .66 0.95 3 .11 0 .70 1 .91 Aug 0 .99 2 .97 0 .94 3 .12 1 .51 3 .36 Sep 1 .17 3 .28 0 .93 3 .19 1.91 3 .88 Oct 0 .85 1.91 1 .3 1 3 .87 0 .96 1 .66 Nov 0 .46 0 .89 0 .54 1 .31 0 .73 1 .16 Dec 0 .62 1.37 0 .54 1 .18 1.02 2 .32

Year 0 .64 2 .63 0 .98 3 .59 0 .39 1 .53

Note: t measures the statistical significance of the trend. The th resho ld for significance is 2 .70 at 1% level .

3 .3 .2 Statistical Analy sis o f Tide Data

The Tham es tidew ay is very well suppl ied w ith tide gauge s as illustrated in Table 2 .1 and Figure 2 .1 o f the Crane Prelim inary report .

(a) Annual m ax im a data for tide stations for the pre-Barrier period have been presented in the Crane Preliminary report . Oppo rtunity is being taken to correct som e errors that have been revealed in the raw data given in Appen dix A of the Crane Preliminary report .

(b ) It is proposed to work w ith th e di stribution o f max imum levels achieved d uring each nea p-spring-neap t ide cycle giving an rox imately two values each m onth.

(c ) Th e ann ual m ax im a have bee n analysed fo r trend as described in Section 3 .4 of the Crane preliminary report . The data corrections m ade thus far indicate a more nearly unifo rm pattern of trend behaviour th roughout the tideway , but still com patible w ith values obtained by IOS for mean sea level.

(d ) Knowledge of the surge distribution m ay be im portant in order to mo del the operation of the barrier , wh ich is based upon a forecast of the surge added to th e predicted high water level

12 (Section 4 .3) . Figure 3 .2 shows a prelim inary assessm ent of the magnitude of surges tha t have occurred at the time of h igh sea level . The histogram indicates a surge distribution wh ich has a positive mean and with a long tail to the right.

3 .3 .3. Correlation Between the Primary Data Series

Figure 3.3 indicates that there is a weak tendency for interaction between surge and riverflow . This may be seen by contrasting the surge magnitudes corresponding to discharges in excess of 100 m3/s - a preponderance of values between .2 and .5m - with that for lower flows where there is a preponderance of va lues between 0 and .4m . This level of correlation will probably be m odelled adequately by a seasonal or other decomposition of the data.

3.3 .4 E ffect on Trend o f Record Breaking Ev ents

Section 3.4 of the Crane Prelim inary report refers to Horner 's ana lysis of record breaking events. It seems that the trend value so derived , 760 mm per century , played an important par t in the origina l design calculations for the barrier . Because this va lue is considerably larger than in the results presen ted in the River Crane report (between 200 and 300 mm per century ) and because of theoretical difficulties with approaches based on record-breaking events , the relationship between record breaking events and general trend has been investigated.

(a) Problem s w ith the Approach

The theoretical difficulty can be seen very simply by considering an indefinitely long time series plot of a random ly varying but trend free series . In such a series recor ds w ill inevitably be broken and hence an upward slop ing line must alway s be obtained if such values alone are considered . Such a method clearly opens up some questions:

Is a trend of 760 mm per century in the record breakers compatible with a 250 mm or indeed 0 mm per cen tury? trend in annua l maximum w ater levels

How robust is such a procedure for estimat ing a trend?

There is no simple ana lytical answer to these questions so a simulation approach was adopted . The rules of the simulation were :

Random numbers are generated from a population w ith known distribution and trend ;

13 Figure 3.2

0 0 0 0 0 r0 — Number of Surges i n Class

HISTOGRAM OF SOUTHEND SURGE MAGNITUDES 1953- 1966 Figure 3.3

400

• 300

• 200

• • • • • • • • • S • • • • S• • s • 100 • 90 • •

•

10 0 +.2 4•4 +.6 +.8 +1.0 Sout hend Surge height —rn

CORRELATION BETWEEN SOUTHEND SURGE AND TEDDINGTON DISCHARGE The ser ies of record br eakers w as started at the first incidence of a value wh ich exceeded the 0 .8 quantilee (5 year return period );

Series w ith four or more record breakers only were accepted;

The trend w as estimated as th e coefficient of a linear regression on time of th e record breaking events.

(b ) Resul ts o f Simulation Experiments

One thousand sim ulated trend-free 200 year record s were generated for Tower Pier using the be st fit parameters o f the General Extreme Value distrib ution for th at location. For each 200 year realisation it was possible to extract the record breaking events and then fit a linear trend to th em . There were 145 cases where the th ird rule above led to its exclus ion leaving 855 simulated trend values. Fig ure 3 .4 shows the se coefficients and in dicates a w ide range o f values are possible from near zero to over 0 .13 m per an num . The m ajority of the values fall in th e interval from zero to 1 .2 mm per annum with a mode close to 3 .5 mm per annum . The mean trend of 9 .2 mm per ann um lies to th e right because of th e skewne ss in the distribution. High values o f the trend w ill tend to occur in those series where the last record breaking event occ urs well before the end of the record .

The imposed true trend in th is simulation is o f course zero yet the observed trend based upon record breaking event s of 7 .6 mm per' annum lies well within th e m ain region of the da ta. Th is answ ers part of the first que stion above; an observed trend in th e record breaking of 7 .6 mm per annum is entirely com patible w ith a truly trend-free population; mo re than one sample in three would have given even h igher apparent trends .

Concerning the second question above the bi ased and h ighly va riabl e results and the paradox of a reducing expectation of sim ulated trend w ith increasing population trend both lead to the conclusion that this method provides very poo r estimates of trend.

(d ) Conclusions

Th e histogram of estimated trend va lues indicates a very wide spread o f possible value s which can with ease encompass values as h igh as 7 .6 mm per annum from an entirely trend free population . The me thod is h ighly biased and rather dependent upon the particular set of ground rules used in the simulation .

14 a m m o — a s as m o a n la as a S = I es

TOwER PIER

MEAN = 0 . 0 0 9 2 S. D. O. 0 1 33 2 0 0 10 0 0

DI SC. 14 5

0 . 0 4 54

NO TREND

0

0 0 0.02 0 .0 4 0.0 6 0.0 8 0 . 1 Simulated Trend M I M N IMO 111fl l MN MN W I NM M O O a I=

I TOWER PIER —1 0 MEAN = 0 . 0 0 75 rn / year GI X/ S. 0 0 . 0 0 9 2 * > K --t 150 N 100 0 I 0 -n K 0 . 0 4 54 -o- cn 0 R U 4 . 58 4 0 -o c C r AL PHA = 0 . 2 19 1 r > -n-, > --I a) - I 171 .2 loo_ TREND = O. 00 27 rn/ year 0 0 03 z —1 ,.<0 H z rna) ri z z ° , 0 >< r9 I—C 50 —.1rn 3 w 3 -.. s< x i m91 8 rn* o Xi -u o 0 .0 2 0.04 0 .06 0 .08 0.1 Simulat ed Tr end 1 1 The trend value obtained m ore directly from annual max ima or m ean sea level is m uch to be preferre d to an approach based upon record breakers.

3 .3.5 Requirements for Fur ther Analyses

Further work is necessary in order to establish th e link between the flow duration curve and the annual m aximum ser ies.

The fluvial input to the tideway downstream of Ted dington should not be neglected and a strategy for incorporating it into the distribution must be identified.

Further work is necessary to establish the statistical distributions for tide stations, especially to allow for the impact of the barrier operation . An optimal stratification is required to accommodate the weak dependence that can be observed between surge and tide .

15 a m in o w e n n li n wil a m a n e 4 COMPUT ATIONAL HYDRA ULIC MODEL

This sec tion contains a re view of previous models of the tidal Tham es. Recommendations have already been given to Th am es Water opting for th e application of the ON DA m odel. Following verbal agreement to proceed the ONDA model has been ex tended to include the complete tidew ay from Molesey Locks down to Southend. The Tham es barrier has been included in the m odel and its m ode of operation is described below together with the identified strategy for incorporating the effect of barrier operations into th e probability analysis.

4 .1 Review of Previous Mode ls

The consul tancy brief req uired a study of ex isting models to be carried out d uring Stage 1 of the study togeth er with recomm endations on th e com putational hydraulic m odel to be used. The options for th e model were either the ON DA mo del which was used by Halcrow for investigating the operating rules for Richm ond sluice, the LOS model which is curren tly used in o peration of the Thames ba rrier , or a combination o f bo th .

Several reports provided by Thame s Water concerning hydraulic modelling studi es for the tidal Thames have been reviewed dur ing Stage 1 . The se reports tend to concentrate on the Thame s Barr ier, its operating rules and the effect it has on the river . Attention in previous studie s has generally focussed on tidal rather than fluvial influences, since the former have a m ore dom inant e ffect on water levels , particularly in the seawar d re aches. Little is said of m ethods applied by GLC to se t de fence levels particularly for the upper tidal reaches.

Several hydraulic models have been der ived for the tidal Thames of which three were physical scale models (two constructed by HRS and one at Wimpey Laboratories). Other scale models, such as the BH RA model, were b uilt to study localised effects and covered only a short section o f the tidal river . The results from physical m odel studies are of interest both as background info rmatio n and for comparative purposes, but could not be used for the present study since the se models, of co urse, have long been dism an tled.

Of th e various numerical models which have been derived, the HRS model was intended on ly to study sedimentation effects and is not relevant to th is study . The two models wh ich are o f direct interest are the IOS numerical model , wh ich originated in the early 1970's and is now used at the Tham es barrier, and the more recent computational hy draulic model (ONDA ) by Halcrow wh ich was used for two recent studies concerning Teddington weir and the operation o f Richmond sluice.

16 4 .1.1 IOS Mod el

The IOS model was origina lly developed by Rossiter and Lennon in 1965 and received various refinements dur ing the early 1970 's. In 1974 the model was interfaced to a two dimensional model of the southern North Sea . Since then on ly m inor re finemen ts to the barrier equations and to the calibration hav e been made. Th is model is now used at the Thames barrier a lthough not in real time mode. Operation of the barrier is based on the Meteorologica l Office 's forecasts from the ir Nor th Sea models (see Section 4 .3.1). The IOS model uses an explicit finite difference schem e for which there are more strict lim itations on time and distance steps than are required with imp licit schemes . The time step adopted in the present mode l formulation is 3 m inutes and the longitudina l distance step varies from 1 mile above Tow er Pier increasing to 6 m iles at Southend . The model covers the complete length of river from Kingston down to Southend in 44 steps and includes the lower Medway estuary . The Chezy friction formula is used and the friction coefficients are allowed to vary with stage . The convective term is om itted from the momentum equation in the finite difference scheme and an allowance for th is is made by adjusting the friction coefficients. Staff at the Thames Barrier are confident that the mode l reproduces conditions satisfactorily in the vicin ity of the barrier but are less sure of accuracy towards the upstream boundary . There is no quan titative ana lysis of ac curacy ava ilable for the LOS model.

4 .1.2 The ONDA Model

The ONDA model , is a very flex ible ana lytical tool and incorporates an implicit finite difference scheme which has much less stringent stabi lity constraints than for explicit schemes , the complete St Venant equations are used . It has a sparse matrix routine to speed matrix solution and is da ta steered . The Manning friction equation is used and a range of options enable a var iety of complex hydraulic structures , bank overspills, boundary conditions etc to be included readily in a particular model formulation . The model is at present calibrated to reproduce a complete month of normal tides covering the complete neap-spr ing-neap tida l cycle coupled w ith low fluvial flows . The model encompasses the reach from Teddington Weir to Gallions tide recorder near Woolw ich in 50 distance steps averaging about 1 km each and has used time steps of 10 m inutes . Reproduction of observed conditions is very satisfactory . For the current application it w ill be necessary to recalibrate and check the model for high fluvial flows coupled w ith a h igh surge tide , since under these conditions the channe l roughness could change due to changes in bed forms . 4 .1 .3 Cho ice of Model

As far as a choice between the two m odels is concerned , the ONDA model has been recomm ended since it is potentially more accurate , includes the latest developments for one dimensional com putational hydraulic river mode ls, and covers a w ider range of options than could be achieved with the IOS models .

From a techn ical standpo int it is preferable because :

it will permit tributaries to be jo ined to the main Thames in case interaction s have to be studied , it is of known and satisfactory ac curacy in its low flow ca libration , IOS model is an unknown quantity in upper tida l reaches , ONDA incorporates the full dynam ic flow equations wh ich is important in tidal flow , it permits a longer time step , it easily permits different contro l rules and tributary barr iers to be included .

Logistically it is preferable because:

use of the IOS mode l at the Barrier would be impractica l, transfer to Ha lcrow and the associated learning curve more time consuming than extending ONDA.

M inor modifications and extension required to the existing ONDA model are no t considered to be significant disadvantages compared w ith the positive points . The sm all cost of extending ONDA down to Southend wou ld a lso be offset agains t the time taken to transfer and fam iliarise with the IOS m odel. In short , adoption of the ONDA model would be the most cost effective option .

The third option which was ava ilable , operation of the two models in tandem , has been discounted both from a logistical standpoin t and because of the difficulties which would be encountered at the boundary between the two mode ls . The most obvious choice for the interface is the barrier site, but it is now seen that this could not be achieved since differen t boundary conditions prevail at different stages of the tide .

In conclusion , it has been recommended that the ONDA model is the most cost effective option , and should be extended and modified to include the whole reach from Southend up to the tida l lim it above Kingston .

18 4 .2 Extension of the Tidal Thames ONDA Model

Verbal approval was given at the progress meeting o f 14th May 1987 to go ahead w ith optional stage la to extend th e ONDA m odel and to include the Thame s barrier in preparatio n for application in subsequen t stages. This com ponent is now com plete and work is now con tinuing to recalibrate th e model under Stage 2 of the study . Details of work which w as required to extend th e model are given below .

4 .2 .1 Topogra phical Data

Extension of the ONDA model required additional cross sectional dat a in the re ach between Southend and Gallion s tide recorder . Initially it w as intended to use da ta from th e IOS m odel at th e Thames barrier; however, cross sectional data were not readily available although hydraulic properties o f the river (Cross section al areas and hy draulic depths ) for a range of levels were obtained and used in preliminary runs of the m odel . More up to date data were then taken from the Port of London Authority hydrograph ic charts, which date from surveys made in 1983 and 1984 . Cross sections were taken at spacings varying from 1 .0 km at Gallion 's reach to 2 .5 km at the seaw ard limit . Locations of cross sections are shown in Figure 4 .1.

Spring tides of insufficient severity to close the barrier quite o ften pass over Teddington we ir, which was previously nom inated as the upstream boundary of the model. As a result the upstream lim it will now be moved to Molesey Locks . Cross sectional data for th is reach has been taken from the data base for the river Tham es m odel which is being developed conc urrently in a separate study by th e Consultant . The upper Thames modelling team has downloaded and transcribed th is data which has now been in cluded in th e tidal Thames model .

4 .2 .2 Thames Barrier

The Consultants have recently developed a new routine for the ONDA model to include structures w ith multiple radial or vertical sluice gates. Th is routine enables the Thames barrier to be included in the model. The main gates for the Thames barrier are rising sector gates wh ich di ffer from conventional radial gates in two respects: the gates rise from the bed rather th an fall from an overhead position, and have a permanent gap beneath so that both overshot and undershot flow occurs sim ultaneously dur ing closure. These conditions are now included in the ONDA m odel .

19 * TID E GA UGES

A Teddmgton Lock E Towner Pier I Er it h

B Richmond Lock F Charlton J Tilbur y 2 3 4 5 km C Hurlinghom Jett y G Si Ivertoe n K Coryton

D Chelseo Bridge H Go Mons L Southend

318 iiee Cross Section Figure 4 I PLAN OF TIDAL THAMES SHOWING LAYOUT OF MODEL During Stage 2 of the study the extended ONDA model w ill be recalibrated using a variety of recorded event s with various combinations of upstream flows and tidal boundary conditions . The model w ill then be applied to define the structure functions illustrated in Figure 5 .1 at various locations thorughout the tidew ay .

20 I I I I I I I I I a 5 . OPERATION AND EFFECT OF THE THAMES BARRIER I I I I I I I I I I I OPERAT ION AND EFFECT OF THE THAMES BARRIER

5 .1 Operation of Barrier

The Thames tide barrier was commissioned in late 1982 with the object of excluding high surge tides from the tideway thr ough London . It has a dominating influence on the up per parts of the tidal level probability distr ibution . The barrier consists of four falling radial gates and six rising sector gates wh ich can be closed in advance of a forecast high sea level. Its operation is at present determined primarily by high water at Southend and up land flows past Teddington/Kingston . The operating criteria ar e that the barrier should limit the maximum water level at London Bridge to 4 .85 m . Its design criteria is that it should cope w ith the 10 ,000 year return period event (assessed for the year 2030).

As a matter of routine the barrier operations centre rece ives a surge _forecast for Southend from the Storm Tide Warning Service (STWS ), part of the Meteoro logical Office , 12 hours prior to each high tide . The STWS uses two methods : a mathematical model of the North Sea area and a regression equation us ing in put data rece ived from more northerly sites and me teorological ob serva tions . An average of the two is normally used as the basis of adv ice sent to users although judgement may be applied to weight in favour of one or the other forecasting proced ures. The particular forecast leve l used in the barrier operation decision is that for the time of high tide itse lf, although hourly information is available from the North Sea model. The STWS also issues a seven hour and a three hour forecast , the latter based upon Lowestoft observed residua l. However only the 12 hour forecast is usable in view of the requirement to close near to low tide and to serve notice to PLA and other interested author ities . The decision to close the barrier is based on the forecast 'tide peak , perhaps adjusted using the forecast errors at Southend prior to the curren t point in real time . The barrier gates are closed in pairs , each pair taking 10 m inutes to close , and full closure of the barrier is usua lly spread over about one hour . Closure w ould norma lly be ach ieved within 2 to 3 hour s after low tide , but under very severe conditions barrier staff would aim to complete the operation as soon after low tide as possible .

Wh en the barrier is closed water continues to en ter the tideway benea th the ma in rising sector gates through the permanent 20 cm gap . The discharge benea th the gates (on closure) is of the order 100 m3/s for a difference in wa ter leve ls of 3 metres across the barrier . Water also enters from the up land catchments, principally the Thames upstream of Teddington , the Lea and other tr ibutaries within the London area .

21 The total upland catchment area is about 1000 km2 . The barrier operation proced ure allows for this by adjusting the closure level at Southend according to the current di scharge at Teddington/ Kingston. Thus , for example when the Teddington discharge is 10 ,000 MGD (525 m3/s) the action level at Southend is 3 .35 m , whereas when the discharg e is only 2000 MGD (105 m3/s ) th e action level at Southend increases to 3 .80 m .

It has to be appreciated that th e action level at Southend is itself a forecast quantity and hence entails an inevi table error . The degree of error has not been ful ly investigated but indications from STWS reports are that a 0 .28 m nns error is to be expected. Th is is borne out by experience wh ich also indicates that considerably larger underestimates are possible. Factors such as ti past experien ce, forecast error, height o f low tide , and trend in upland di sch arge can vary th e time and the rate of closure. Barrier operations staff attemp t to take forecast error s into account by comparing forecast and recorded levels at low w ater at Southend prior to m aking the decision to close the barrier .

5 .2 Effect of Barrier on W ater Levels

The im pact of the barrier can be viewed in two ways: (a ) the impact on th e course of events over a specific h igh tide cycle, and (b ) the influen ce on the statistical distrib ution of levels upstream .

5 .2 .1 Effect During a High Tide Event

The Thames barrier has been closed in earnest on thr ee occasion s up to date: on 1 February 1983 , 26 Decem ber 1985 and 27 March 1987. In formatio n for th e first two closure s has been obtained from FLA and Thames W ater . The barrier is currently operated to protect Central London from abnormal tides with a surge com ponen t. Normal spring tides (astronom ic) without surges are allowed to pass through . Thame s Water barrier staff h ave stated that for events for wh ich the barrier is closed the resulting upstream water levels are certainly not h igher than m igh t be reached in con dition s wh ich would not req uire barrier closure under statutory req uirements.

Examples of the effect of Thames barrier closur e is demo nstrated in Table 5 .1 .

22 II Table 5 .1 Effect of Barrier Closure

Location I Maximum Levels

27/12/85 1/2/83 26/12/85

Southend 3 .1 3 .7 3 .3

Tow er Pier 3 .2 1.5 4 .6

Richmond 3 .7 2.0 4 .9

Fluvial Flow I 408 219 319 (mean daily in m3/s)

Closing time (hrs 3 .5 1 .0 after low tide at barrier )

Barrier state closed 1 open 1 1 1

From the three events presented in Table 4 .1 it seems clear that the highest levels experienced now that the barrier is operational do arise from non— closure con ditions. For example , the monthly report for December 1986 of the London pl anning section of the Thames Wa ter Rivers Division noted floo ding along the upstream section of the tidal Thames under such non—closure conditions .

The impact of the barrier clearly diminishes as one considers points further upstream . For ex ample the figures o f Appendix C of the Crane Prelim inary report indicates the larger influence of the fluvial contribution at Richmond com pared w ith Hamm ersmith . However the relative contribution o f fluvi al and tidal components to level in the whole tideway is not certain at present.

5 .2.2 Effec t on Statistical Distributions

Figure 5 .t illustrates di agrammatically the general procedure for obtaining the probability distribution o f a derived q uantity , such as the stage at a point along the tidew ay , from prim ary variables such as sea level and upland flow. Two sets of relation ships are required . The first relationship, termed the structure function and illustrated by the broken lines on Figure 5 .1a, is the loc us of

23 Figure 5 1 (a) WITHOUT BARRIER

Probability Level at Southend (x )

4.».

Density of ± ,f (x ) 2 a.

Region in which Bivariate density h > h P( s ty)

Lines of equal h value (water level at particular location)

( b) WITH BA RRIER

DI A h - h 46

-44***•

Level at Southend

ILLUST RAT ION OF METHOD OF COMBINING PROBABILITIES constant levels. The second set, the full lines on Figure 5 .la , shows the bivaraite distribution of sea level and upland flow and are lines of equal probability density . If the two va riables are independent the bivariate probability density function is obtained as the product of the two "marginal" densities. The volume under the bivariate surface beyond some chosen level gives the probability of exceeding that level . Note that the bivariate surface is a con stant for an estuary wh ereas the structure function varies from point to point .

Th is last point is illustrated in Figures 5 .2a and 5 .2c where the "without barrier" lines for Richmond are m uch more nearly parallel to the tide ax is than those for Tow er Pier wh ere th e tidal influence is more dominant. The barrier o peration influen ce s th e structure function dram atically, as illustrated in Figur es 5 .2b and 5 .2d . A literal interpretation of th e o perating rule would be to cut o ff the curve at 4 .85 on Figure 5 .2a as shown on Figure 5 .2b . In application the structure function migh t be mo re complex to allow for uncertainty in the forecast leading to unnecessary or delayed closure . One way the forecast uncertainty co uld be incorporated would be to attach probab ilities o f barrier operation across the range o f tide :flow combination s, as shown in Fig ure 5 .3 . The effect o f variations in timing o f closure would be to change the pattern of the structure function above the barrier closure line .

It is likely that the situation at Richmond is rather different than further downstream as the flatter structure curv e lead s to the barrier operations permitting h igher levels for h igh flow events .

This is illustrated in Figure 5 .2d . Application o f the ON DA model will assist in the definition o f the shape o f the "barrier closed" curves. Difficulty w ill be experien ced in assessing th e effect of uncertainties in the forecast as illustrated in Figure .6 3 , on the operation and hence on th e cho ice of "barrier closed" or "barrier open" curves for particular cases.

The Thames barrier has changed the probability distributions o f water levels in the tideway dramatically . The new distrib ution s were shown above to depend on a com plex interaction wh ich is dominated by the Thames barrier o peration . It is essential th at barrier operations are simulated , and the three barrier closure event s will there fore feature in the calibration proced ure .

24 Figure 5.2

WITHOUT BARRIER pier %6 l (a) e 4105 7 \o r 4 4 (4 5 3 \ TOWER Level at Southend PIER

c "• WITH BARRIER ( b) gsi'6\ 2 3 •4 4 0 ii 3 -5 3

Level at Southend

7 WITHOUT BARRIER (c) .ce 4 (>0° - 3.5 vets 3 Ntiolef Leve l at Southend RICHMOND

WITH BARRIER 5 4 3 (d ) 4 ""•- .„ 3.5 • 2 3 Level at Southend

Barrier closure line 5 — Max. water level at specif ied location

NOTE Water levels shown are illustrative and do not relate to the actual situation.

ILLUSTRATION OF POSSIBLE EFFECTS OF THAMES BARRIER ON WATER LEVE L S Figure 5 •3

Probability of closure

CO 0 O . 0 .6 . so as (P r o b a b iof) l i t y .c 0/ 0 vs... \ a\ non—closure ro. Kr) o. roo 3S-/ o) 0

Level at Southend

ILLUSTRATION OF PROBABILIT Y OF BAR RIER CLOSURE • • I I I a • UN • N I IS • NM MN M I MO III II I NP NM OM • • 111 I N 1 1

0 CONCLUSIONS AND RECOMMENDATIONS

6 .1 Completed Analyses

The purpose of the stageing o f the study is to permit Thames Water to review its needs for further analysis in the light of what has been discovered to date . This report describes the data that have 1 been collected or identified as available for both tide levels and tributary discharges, and presents prelim inary statistical results for two trib utaries, the Crane and Lee, regarding the ir discharge and level freq uency relationships (sections 3 .1 and 3.21 .

The basic procedure that would be followe d in order to derive stage frequency relationsh ips for intermediate points has also been iden tified. This consists of combining a hydraulic model o f the estuary , in wh ich intermediate levels can be predicted from tide and fluvial inputs, with a statistical description o f th e combined occurrence of those two m ain variables as illustrated in Chapter 5 . Sample preliminary statistical description s o f the primary variables are shown in section 3 .3 .

A re view of hydraulic modeling req uirement s has been m ade and is summarised in Section 4 . The ONDA model is to be used and the necessary mo difications and extensions to th e model are complete , although the model has not yet been recalibrated .

6 .2 Recommendations for Stage 2 Analysis

In the terms o f reference Stage 2 of the analysis would be devoted to the derivation o f stage freq uen cy distrib utions for interme diate points along the tidew ay in order to de fine defence levels up to 1,000 year return period.

At the outset it was supposed that the barrier 's impact on th is computation o f intermediate stag e freq uency would be accommodated relatively simply by m aking the necessary adjustm ent to th e tide frequency distrib ution to allow for post barrier con ditions. This may not be the case . The barrier operation is a rather complex funtion of predicted (astronomic component ) h igh tide at Southend , forecast surge residual 12 hours ahead for Southend, and Teddington/Kingston discharge . To this must be added rather inderterm inate elements to allow for the incorporation of forecast errors plus th e influen ce of o perational constraints on barrier closure . Consequently the work to be carried out dur ing Stage 2 will not deviate significantly from the original pro po sal and w ill include , the following:

1

25 Calibrate functional relationship to predict w ater levels at intermediate tide gauges from the primary boun dary variables, by model simulation and anlaysis o f stage records, allowing for the dominant effect o f the Thames ba rrier . Determine influence of supposed secondary effects eg tributary inflow , hydrograph and tide and surge shapes .

Derive probability distributions of daily stage at intermediate tide gauges and convert to ann ual m axim a. Th is w ill give the frequency of overtopping o f defences o f specified heights at tide gauges from Teddington to the Thames Barrier .

Calibrate a simple relationship between the prim ary variables and stage at intermediate points to reduce the number of hydraulic model runs. Th is m ay be ba sed on a regression relationship or on a simplified hydraulic formulation. Use this to pr e pa r e longitudinal profiles through the tidal reach down to the Barrier.

Probability distributions o f water levels at a given point in the tideway can be derived using me thods illustrated by Fig ur e s 5.1 to 5 .3. At present the ex act form of th e se diagrams is not known , especially in the way the probability distribution overlays the structure function . Also, for the ex isting case (with Barrier closures ) the form o f the structure function above the Barrier closure line cannot be de fined clearly without simulation results from runs o f the hydraulic model. To clari fy these points two interesting scenarios would be analysed using the hydraulic model as early as possible in the Stage 2 analysis:

(a ) An extreme combination of a very high surge tide (say 1 in 1000 year) with a very h igh fluvial flow and barrier closure

(b ) A very high fluvi al flow coupled w ith the highest tide which is expected to be allowed to propogate upstream without ba rrier closure .

The results o f these two runs, coupled w ith the analyses o f w ater levels at intermediate tide gauges (particularly of Richmond and Tower Pier ) would clarify the picture. One possibility is that the difference be tween water levels at low return period and those at high return period m ay be so small, now that the barrier is operational, that designs co uld be carried out using a "worst case" combina tion of tides and fluvial flows . This m ay very well reduce the wo rk scheduled for subsequen t analyses, if only with respect to Thames de fence levels and not to boundary conditions and flows for

26 tributaries. At present the indications are tha t this "worst case" would be of type (b ) above. However the interaction s are o f such complex ity that it is not possible at the present time to m ake an objectively based recommendation.

27