Hydrological conditions for design in urbanising …eprints.hrwallingford.co.uk/920/1/SR302.pdfHR...
Transcript of Hydrological conditions for design in urbanising …eprints.hrwallingford.co.uk/920/1/SR302.pdfHR...
HR Wallingford
Hydrological Conditions for Design inUrbanising Areas and using Detention Storage
Report SR 302March 1992
31 lvlarch 1992
Address: Hydraulics Research Ltd, Wallingford, Oxfordshire OX10 88A, United Kingdom.Tclcphone: 0491 35381 lntemational +44491 35381 Telex: 848552 HRSWALC.Facsimile: A49l 32233Intemational + 44 491 32233 Registered in England No. 1622174
A:\TSRREP.WP
This repor t was prepared und.er contract No. PECD 7/6/193, funded by theDepartment of the Environment, for the research into Hydrological -
condi t ions for Design. in urbanis ing Areas and using oelent io i i to t .g" . Thework was carried out in the Resear6h oepartment. of HR warlingtora.
HR nominated project officeri Dr W R whiteSect,ion leader; M p Osborne
The DoE nominated officer is Mr Peter woodhead. This report is publishedon behalf of the.DoE but any opinions expressed are those of the authorsand not necessarily t,hose oi tne Department.
@ Crown copyright 1992
Published by permission of the Controller of Her Majesry's Srarionery office,and on behalf of the Deparunent of the Environment
A: \TSRREP.WP 31 March 1992
ABSTR,ACT
Most sewerage design :Ln the IIK uses slmthetic rainsto:ms derived fromrainfall statistics. These were originalJ-y developed to represent peakflow rates at average sucqEr conditioag. They may therefore not beapplicable to winter conditions, or !o systenrs where totat volune of runoffis i-nportant,' the design of detention storaete taaks, and predicting spiJ.J-from overflows.
One alternative method of design is to use J.ong timeseries of real rainfal.J.records. Ihis has the disadvantage of long corputation tines, anddifficuJ.ties i.rr obtal-ning the dati. This project aet out to develop amethod to rePtesent a 1ong tineseries with just a few slmthetic ator:m.s.
Anrrual rainfall ti-meseries representing a tlpical- year's rainfal-J- for threelocations in the IIK rere analysed to detemine the stonm characteristics.lhe stor:ms rere classified by depth of rain, catcbment rretness, peakednessand skew. lhis led to a new definition of peake&ress which gave higherpeak iatensitLes in 1ong durations.stores. Tbis work al-so gave an insightinto hor well the tLmeser:i.es represented the raLnfatJ. patterns in a t1p5.calyear.
Slmthetic stotms were generated using the averagie sto:m characteristics andwere found to give a good representation of the timeseries. Sensitivitystudies trere carried out to deternine hor rel'I the stor.ms would perfor:n inless than optinal conditions. Final testing of the slmthetic atolsts rrascarried out on uodels of real catchments incJ.uding detention Lanks andoverflows.
A : \ T S R R E P . W P 31. March 1992
CONTENTS
3 .
4 .
1 .
2 .
5 .
6 .
7 .
TNTRODUCTION
ANALYSIS OF RAINFALL CHARACTERISTICS
2.L Int roduct ion2.2 The lrlRc Annual Timeseries2.3 Analys is of the annual t imeser ies
2 . 3 . L S t o r m D e p t h2.3.2 Catchment weLness2 .3 .3 Peakedness2 . 3 . 4 S k e w
SYNTHETIC RAINFALL SERIES
3.1 Choice of storms for synthetic rainfall series3.2 Storage Prograrn3.3 Sensi t iv i ty test ing of skew3.4 Sensi t iv i ty test ing of s torm durat ion3.5 Conrparison with the annual t imeseries
TESTING ON REAL SYSTEI'{S
4. L Choice of model-s4 . 2 R e s u l t s
4 . 2 . 1 M o d e l 1 -4 . 2 . 2 M o d e l 2
4.3 Effect of sampling the annual t imeseries
CONCLUSIONS
FUTURE WORK
ACKNOWLEDGEMENTS
REF'ERENCES
Page
1
2
223
457
Ll_
12
t21 313l 41 5
1 6
1 61 8
1 81 8
l-9
1 9
2 L
2 2
2 3
A: \TSRREP.WP 31 March 1992
FIGURES
1 . Depth ratio/Return perioda South East , Durat ions L5 - 48 minutesb South East , Durat ions 60 - 1920 minutesc Yorkshi re, Durat ions 15 - 480 minutesd Yorkshi re, Durat ions 60 - L9Z0e South West , Durat ions L5 - 480 minutesf South west , Durat ions 60 - 1920 minutes
Not included in this report
Freguency/SkewSouth EastSouth East 20 largest stormsYorkshireYorkshire 20 largest stormsSouth WestSouth West 20 largest storms
4 . Frequency/UCwt ratioSouth East - annualYorkshire - annualSouth glest - annualSouth West - seasonaLYorkshire - seasonalSouth East - seasonal
2 .
3 .abcdef
abcdef
5 .6 .7 .8 .9 .l - u .
1 1 .L 2 .1 ?
L 4 .1 5 .l - o -
t 7 .' l q
Comparison with annual t imeseriesSouth East - 30? perv iousSouth East - 60? perv iousSouth East - 70? perv iousSouth West - 30t perv iousSouth West - 60? perv iousSouth West - 70% perv iousYorkshi re - 30? perv iousYorkshire - 60t perviousYorksh.ire - 702 perviousLayout of model 1Layout of model 2ModeL l Storage,/Return periodModel 1 Overflow,/Return periodModel 2 Overfl-ow Return period
Appendix A Skew Storms
A : \ T S R R E P . W P 31 March 1992
1. II{TRODUCTION
one of the longest runningr arguments in the use of
urban drainage models concerns whether the
behaviour of the system j.s best represented using
synthet ic s torms der ived f rom staList ics, or us ing
a t imeser ies of real s torms.
The argument in favour us using storms is that they
are readily available for aII locations in the UK,
and that they are easy to use and require only a
few storms. However the normal synthetic storms
represent only one set of rainfall conditions,
average surruner conditions, and have only been
proved to be correct in predicting peak flow rates'
not storage volumes. The argiument in favour of
timeseries is that they include a wider range of
conditions, and therefore are l ikely to contain the
condi t ions which are cr i t ica l on each catchment .
There are trdo crucial areas in which synthetic
storms may not adequately represent the response of
a catchment. The first of these is in the design
of detention storage tanks, where modest storms
fall ing on a very wet catchment may give a larger
volume of runoff than a larger storm with averag:e
catchment wetness. The second area is in
predicting spil l from overflows in small frequent
storms for pollution impact studies. The first
area is the main area of in terest of th is pro ject ,
although the second has been covered as well.
A : \ T S R R E P . W P 31 March 1992
2 . t
2 . 2
AI.IALYSIS OF
RAINFAjJIJ CITAR,ACTERI ST ICS
In t roduc t ion
The I{Rc Annual
Timeseries
The first slage of this study was to investigate
the variation that existed in long rainfall records
of real storms to see whether this variation could
be represented by a few synthetic storms. To do
this a representative timeseries !{as required. The
$lRc annual ,Timeseries were used for this. These
were chosen, because although they did not include
the more extreme storma which were also of interest
to this project they were a convenient source of
representative rainfalt data. This work is
reported more fulty elsewhere IHR wal]-ingford
1 9 9 1 1 .
Timeser ies ra infa l l is a sequence of h is tor ic
ra infa l l eventa stat is t ica l ly representat ive of the
precip i ta t ion pat terns for a g iven locat ion. The
hlRc annual t imeseries are series of storm events
representing typical years for each of three
regions of the country. They have a duration of
one year and are sequences of f ine resolution rain
data suitable for input to WASSP-SIM. The storms
are available in three forms, a chronological
ser ies of a l l s igni f icant events in a t lp ica l year ,
the same eventa ranked in order of severity and a
pair of similarly ranked series for summer and
srinter periods. Annual t imeseries are available
for the South East , South West and Yorkshi re. Each
ser ies consist of approximately 99 storms.
A: \TSRREP. WP 31 March 1992
2 . 3 Anal-ysis of the
annual t imeseries
In deriving the timeseries a prelj-minary seltection
of month ly ra infa l l to ta ls was made. The tota ls
were screened and any months with values within the
range of the mean plus or minus one standard
deviation were accepted. Approximately one third
of alt the monthly rainfalls were outside this
rang'e, representing very wet or vary dry months;
these were re jected. f rom the analys is . This
selection of representative months is not thought
to be significant, ' however t,he analysis of the fuII
rainfall record may provide better results. The
annual t, imeseries are currently ranked by severity;
this involved using the timeseries rainfall for
simulation in a number of models in order that the
Iargest discharge volumes could be identif ied.
This project attempts to rank the series by
compar ison wi th s tat is t ica l ra infa l l data.
The storms in the annual t imeseries r.tere analysed
for four parameters: storm depth, catchment wetness
(UCWI), and two parameters describing the storm
shape, peakedness and skew. For each storm these
parameters were compared to the standard values
used for synthetic storms for the locations from
which the timeseries data was derived.
The data assumed for the t imeser ies locat ions is :
A: \TSRREP.TIP 31 March 1992
South
East
6 0 3
Yorkshi re
L 2 2
Average annual rainfal l mm
M5 60 mm
Standard UCWI winter tnm
Standard UCWI surruner nm
Location lndex
5 1
2 . 3 . t Storm Depth
Each storm in the series was analyzed to find the
maximum depth in each of a series of durations.
The durat ions used where: 1-5, 30 ' 60, L20, 240r
480, 950, 1920 minutes. The depth rat io was
defined as the maximum depth divided by the depth
in a storm of one year return period of the sarne
durat ion. The cr i t icat durat ion was def ined as the
duration with the largesL depth ratio.
The depth ratios were then ranked in ascending
order for each duration, and the rank converted to
a return period, 1/ (number of storms - rank) and a
graph of depth ratio against return period was
p l o r r e d ( F i g L a - 1 f ) .
The results for each duration are similar' although
for all of the regions the long storms with
durations of 240 minutes and above show a decrease
of depth rat io wi th increasing durat j -on. This is
particularly noticeable for the South West region
i f the out l ier s torms ax !2 and L20 minutes
duration are iqnored. This indicates that long
duration storms are under-represented in the
t imeser ies. For durat ions up Lo 244 minutes the
depth ratio is independent of duration.
A: \TSRREP.WP 31 March 1992
The variation of depth ratio with return period for
the South East and Yorkshire regions are similar in
that the depth ratio steadily increases up to a
return per iod of about 0.5 then r iees more s lowly.
The graph for the South West region rises much more
rapid ly to a value of 0.8 then more s lowly to a
maximum of approximately 1 (Fig 1e). The
difference in these graphs is indicative of two
dist inct ra infa l l pat terns. South East and
Yorkshire regions representing the rainfall
patterns in Eastern EngJ-and and the South West
representing Western England.
The curves of depth ratio were redrann so as to
give a smooth curve which passed through a depth
rat io of l - .0 at a return per iod of 1 year . ( fh is
is requi red by the def in i t ion of depth rat io . ) The
averalJe of the values for durations of 60 minutes
to 240 minutes are tabulated bel-ow.
2.3.2 Catchment hretness
Catchment wetness is defined in the Wallingford
Procedure by the Urban Catchment Wetness Index
(UCWI). Standard values of UCWI are g iven for a1l
locations in the UK for summer and winter
ReturnPer iod yrs
South East South West Yorkshi re
1 : 1 L . 0 0 0 1 . 0 0 0 I n n n
2 : t 0 . 8 1 3 0 . 7 7 5 0 . 8 5 5
4 . 1 0 . 6 3 5 0 . 6 5 0 0 . 7 1 0
8 : 1 0 . 4 0 3 0 . 5 6 3 0 . 4 8 8
1 6 : 1 0 . 3 0 0 0 . 4 1 6 0 . 3 0 2
3 2 z L 0 . 1 9 6 0 . 2 9 3 0 . 1 8 5
6 4 . 1 0 . 1_05 0 . L 7 7 0 . L 1 0
A : \ T S R R E P . W P 31 March 1992
conditions. Surwner is taken as April to September
and winter as October to March. These standard
values nrere derived by taking the average of the
values calculated for the last day of each month in
the season over a per iod of many years.
An UCWI ratio was calculated for each storm as the
UCWr for the storm divided by the standard UCWI for
the appropriate season. (Not.e the original
timeseries was derived for version 6 of WASSP which
does not allow UCWI values greater than 170. The
latest version of the timeseries which has
corrected values was therefore used.)
No correlation was found between UCWI ratio and
ret,urn period.
The results are surprising. There is a very large
spread of values but most results are above the
sLandard values. The results are shown in
Figures 4A, 48 , 4C and the mean and standard
deviation are tabulated belo$t.
South East South West Yorkshi r
Mean 1 . 5 7 9
Standard Deviat ion 0 . 4 5 2 0 . 3 3 2 0 . 6 3 7
If summer and winter.storms are considered
separately (Figrs 4D-4F) a different pattern
emerges. In winter there is a smaller spread of
values, and the average is close to the standard
value. Thj -s is l ike1y to be because in winter
A: \TSRREP . WP 31 March 1992
catchments may remain consistently wet.. fn sunmer
there is a large spread of values including some
very much higher than the standard. ?he result has
been noted previously by other researchers using
other rainfall series. As the standard suruler UCWI
is the basis of most urban drainage design' and
these higher values wil l increase the amount of
runof f f rom a stormr. th is is of great concern. A
hlpothesis for this variation can be put forward.
The pattern of rainfall in summer is generally to
have of dry weather followed by a series of sununer
st,orms. There is therefore a greater chance that a
storm has been recently preceded by another storm
than if the stoxms !{ere uniformly distributed
throughout the suruner. The average UCwf within
these periods of rainy weather wil l- be greater than
the overall suruner average which is given by
averaging the last day of each month.
This phenomena is worthy of further investigation,
but these results show that the higher, ninter'
standard. values of UCWI are in fact appropriate for
both sununer and winter.
2 . 3 . 3 P e a k e d n e s s
The rainfalJ- anal-ysis in the flood studies report,
did not explicitJ-y investigate the variation of
peakedness with durations greater than l- hour. The
f i rs t analys is of the t imeser ies a lso looked at
peakedness independent of duration. Peakedness was
or ig inal ly def ined as:
cL/c2
where C2 depth of rain in the crit, ical duration
of the storm
A: \TSRREP.WP 31 March 1992
C1 maximum depth of rain in half the
cr i t ica l durat ion of the storm.
This showed similar values of peakedness to
synthet ic s torms generated ;us ing the Flood Studies
Report.s standard 50 percentile summer or 75
percentile winter profi les. The standard storms do
not give a constant .value of peakedness using this
definit ion, but have values varying by about 20
percent .
No correlation was found between peakedness and
return period of storm.
A better definit ion of peakedness was then used'
which allowed for the variation of peadkedness wit,h
durat ion. This was def ined as:
r I / 1 2
where r1 the maximum depth in a given duration
in the storm
12 the maximum depth in half that duration
in the storm.
The resul ts f rom th is analys is are tabulated below.
They show that peakedness increases with storm
durat ion.
A: \TSRREP.WP 3 L M a r c h 1 9 9 2
Average Peakedness
Using this definit ion of peakedness t.here are
physical l imit.s that Lhe value must l ie between 0-5
and 1.0. However there is another bray in which the
peakedness at J-ong durations can be calculated.
This is by us ing the rat io of f ive year return
period 60 minute and 2 day storms which are mapped
for the whole of the UK.
The 2 day depth vras assumed to represent the depth
in 48 hours. (There is in pract ice a smal l
d i f ference between the two but th is was ignored.)
The peakedness factor must be applied for each
doubl ing of the durat ion. From one hour to 48
hours there are n doublinqs where:
4 8 : 2 "
n - 5 . 5 8
Durat ion minutes South East South 9{est Yorkshi re
3 0 0 . 6 9 3 0 . 6 7 7 0 . 6 6 9
6 0 0 . 7 5 2 0 . ? 0 9 0 . 7 1 8
L20 0 . 8 1 5 0 . 7 5 7 0 . 7 7 8
240 0 . 9 0 4 0 . 8 2 3 0 . 8 4 2
4 8 0 0 . 9 8 8 0 . 9 8 5 1 . 0 0 0
9 6 0 0 . 9 9 ? 1 . 0 0 0 1 . 0 0 0
L920 1 . 0 0 0 r . . 000 1 . 0 0 0
A: \TSRREP . WP 31 March 1992
Therefore assuming an average peakedness P:
P r . t , = r
where r = M5_60
M5_60
/ r45-2DAt
t hour depth at 5 year return
period
2 day depth at 5 ;year return
per iod
M5 zDAY
These value were taken as maximum limits for the
peakedness at durations of four hours and above.
For durations less than L hour the peadkedness was
taken as the min imum observed value of 0.67. A
Iinear variation was used betri leen one hour and for
hou rs .
The resulting storm profi le was compared with the
standard 50? sunurner profi les which are generally
used for urban drainagre design. At short
durations, less than one hour, the new profi le has
much lower peaks than the 50t summer storm.
However as the duration increases the new profi le
shows almost no decrease in peak intensity, whereas
the 50t suruner profi le shows a rapid reduction in
peak intensi ty . At long durat ions, greater than 4
hours, the new profi le therefore has a higher peak
intensi ty
Yorkshire
A: \TSRREP . '{P
1 031 March 1992
2 . 3 - 4 S k e w
Skew is a measure of the position of the peak
rainfa l l in tensi ty wi th in the storm. Skew was
expected to have a significant effect on the
response of the system to ra infa l l . The usual
synthetic stonns are syrrtrnetrical (ie central skew)
Several different definit ions
but the following was adopted
usefu l
skew were tried,
being the most
o f
as
The crit ical duration (D) of the storm and the
start and end of this duration were identif ied.
This period was then taken as representing the
storm, and rainfall before and after this was
ignored. The time for half of the rainfall depth
within this duration (T50) !,ras then caLculated.
Skew is def ined as:
S k e w = T 5 0 / D
A skew of 0 would represent a storm with most of
the ra infa l l a t the star t o f the storm, a skew of
0.5 would have most of the ra infa l l in the centre
of the strom, and a skew of 1 would have most of
the ra infa l l a t the end of the storm.
The resul - ts (F ig 3a - 3c) show that there is a
variation of skew, \dith most storms having fairly
centa l skew. The average values of skew are: 0.5
in the South East , s l ight ly above 0.5 in the South
West and 0.5 in Yorkshi re. To invest igate the
relationship between skew and storm size the skews
of the twenty largest storms for each series were
p lo t t ed (F ig 3d - 3 f ) . The resu l t s a re s im i l a r .
A: \TSRREP.WP
1131 March 1992
3 .
3 . L
STIIXHETTC R,AINFAIII
SERIES
Choice of storms
for synthet.ic
ra infa l l ser ies
From th is analys is i t is reasonabLe t .o use storms
wi th a centra l skew, but the s igni f icance of skew
was investigate further in the sensitivity testing
carr ied out in the next phase of the pro ject .
The storms for the synthetic rainfall series were
chosen to have the average characteristics of the
t imeser ies stonns. Separate character is t ics were
used for each region. In sumfirary these
characteristics ri lere :
The depth of rain was calculated using the depth
rat ios g iven in sect ion 2.
The catchment wetness was taken as the standard
winter va lue.
The peakedness was varied with duration.
The skew was centra l .
Sensi t iv i ty test ing of these storms and in i t ia l
comparisons of the synthetic storms and the
timeseries were carried out using a simple storage
simulation program. This work is reported more
fu l ly e lsewhere [Rainey & Osborne L991] .
A: \TSRREP,WP
L23 l - March 1992
3.2 Storag,e Program
3 . 3 Sensi t iv i ty
test ing of skew
A simple storage simutation program was developed.
This represented a catchment as a lumped hydrologic
runoff model, and the sewerage system as a single
storage tank hrith a l imit on the continuation flow.
The results of the model showed the maximum volume
of water stored in t.he tank during a storn.
The hydrologic model uses the runoff model from the
WALLRUS program where the rrunoff is defined by the
proportion of the area which is covered by
impermeable surfaces, a soil index and the
catchment rdetness (UCWI) .
The storage tank i-s of infinite size and has an
outflow limit on the continuation flow. Any flow
above this l imit wil l be stored unti l such time
that the inflow drops below the outflow limit. Any
stored ri later can then flow from storagte to bring'
the tota l out f low up to the l imi t . The out f low
limits can be varied to represent different
catchment character is t ics. The storage model was
validated by comparison with a simple WALLRUS
model-.
From our analys is of the t imeser ies, no real
tendency for skew was found. Hordever the effect of
skew was investigated using the storage model and
checked using a 2 pipe test system on WALLRUS.
Various catchment characteristics were used in the
model . a cr i t ica l skew of 0.52 was most common
with l i t t le var iance. The ef fect of th is smal l -
skew compared to a centra l skew of 0.5 was
A : \ T S R R E P . n P
1 33 1 M a r c h 1 9 9 2
3 . 4 Sensi t iv i ty
test ing of
storm duration
negl ig ib le. Even using largrer skew values of 0.6
and 0.7 a maximum increase in s torage of 10t was
found.
The crit, ical duration of storm will carry with
catchment characteristics, in part.icular the
catchment. size and the amount of attenuation
storage in the catchment. The crit ical duration
may also vary with the storm size. The sensitivity
of the calculated storage vol-umes to the ueing
durations other than the crit ical duration was
therefore invest igated.
Synthet ic s torms of f requency of 2:1 year and
durat ions of 1, 2, 4 8 and 16 hours were s imulated
using the storage program to find the crit ical
durat ion for a set of catchment character is t ics.
Storm durations in excess of 16 hours were not
normally considered, as for most systems the
cr i t ica l durat ion was found to be less than th is .
The storm duration with the largest storage volume
was then taken as the cr i t ica l durat ion. Using the
cr i t ica l durat ion the storage volume for a ser ies
o f s to rms o f r e tu rn pe r i ods o f l - : 8 yea rs t o 64 :1
year were calculated. The var iat ion of cr i t ica l
duration wLth return period was investigat.ed by
simulat ing the fu l l set of return per iods (1:8
years to 64:1- year) for durat ions of t hour to
1 6 h o u r s .
A: \TSRREP.WP
1 431 March 1992
3 , 5
The crit ical durations were found to vary with
return period. However, the error introduced by
choosing the cr i - t ica l durat ion f rom the 2:1 year
storm was less than 10t . This procedure wi l l
t .herefore normal ly be sat is facLory.
Comparison with
the annual t. imeseries
The effect of the synthetic rainfall series was
compared srith t,hat of the annual t imeseries using
the storage program. The synthetic series was
using 10 storms to repreaent the complete range of
return periods, whereas the timeseries uses 99.
The comparisons were done for each of the three
t imeser ies.
Three different sets of catchment characteristics
were used. These had perv ious areas of 60*, 30t
and ?0t. Some tests were done with a range of soil
ind ices, but most test used a value of 0.4. For
each set of catchment character is t ics two d i f ferent
outflow limits were used. These represented
catchments with small amounts of attenuati-on
storage, and vrith very large amounta of attenuation
s to rage .
For the synthet ic ra infa l l - ser ies the cr i t ica l
duration was first identif ied, and then storms of
1 : 8 , L : 4 , L z 2 , 1 : 1 , 2 z L , 4 z L r 8 : 1 - , 1 5 : 1 , 3 2 : 1 a n d
64:1 year were analysed. Al t 99 storms of the
appropriate timeseries were analysed. The two sets
of storage volume results were plotted together for
compar ison.
The comparisons are generally good (Figs 5-13)
The South East synthet ic ra infa l l ser ies
A: \TSRREP . WP
1531 l ' tarch 1992
4 .
A 1
TESTING ON
REAI. SYSTEMS
Choice of models
consistently indicate a slight under-prediction of
storage compared to the timeseries. The Yorkshire
synthet ic ra infa l l ser ies g ives a good compar ison
for all of the catchments at the low outflow limit
with some over-prediction on the more pervious
catchments at high outflow limits.
The South west synthetj-c rainfall series did not
perform as well with a LO-20? under-prediction of
storage at low outflow limits and a slight over-
predict ion,at h igh out f low l imi ts .
These results sho!i led that the accuracy of the
synthetic rainfall- series was generally insensitive
to catchment character is t ics.
The f ina l s tage of th is s tudy was to test the
synthetic series against the annual t imeseries
using WALLRUS models of real- caLchments.
The Water Service Companies suppl-ied a number of
models. Two suitably representative models were
selected as there was insuf f ic ient t ime to test the
synthet ic ra infa l l ser ies on a l l the models.
However the test ing of synthet ic ra infa l l ser ies
using the storage model had indicated that the
results were not verv sensitive to catchment
cha rac te r i s t i cs .
The models were selected to have typical values of
the fo lJ .owing parameters: s ize, perv ious area,
tpaved, t roof and soi l index. The deta i l -s of the
A: \TSRREP . WP
1 631 March 1992
models are given below, and the details of the
storage tanks and over f lows are g iven in F igures L4
a n d 1 5 .
Model 1 Model 2
Permeable area 4 5 4 . 6 2 L . 9
Paved area 4 7 . 0 5 . 9
Roof area 0 . 0 6 . 1
Soi l index 0 . 3 0 0 . 4 5
Pipes 8 5 1 0 9
Tanks L 2
Volumes 3 3 0 0 . 3
0 . 3
Locat ion D11 Yorks
Cr i t i ca l
duration
1 5 1 5
A timeserJ-es appropriate to the region of the
catchment was then selected. A synthet ic ser ies of
s t . o rms o f r e tu rn pe r i od re tu rn pe r i od 1 :8 , 124 ,
L z 2 , L : 1 , 2 : 1 - , 4 : ! , 8 : 1 , 1 - 6 : 1 , 3 2 : L . 6 4 : 1 - y e a r w a s
generated using the standard data for the
t imeser ies locat ion. Al though durat ions of 1, 2,
4, 8 and 16 hours were s imulated, the resul ts
presented here are a l l for durat ions of 16 hours as
the error in making this simpLification was found
to be small. The synthetic series was compared
$rit,h the annual t imeseries by simulation using the
WATLRUS program. The results of storage and
overf low, against return per iod for the synthet ic
ser ies and the annual t imeser ies were compared.
A: \TSRREP.WP
L731 March 1992
4 . 2 R e s u l t s
4 . 2 . 1 M o d e l 1
The graph of storage against return period (Fig 16)
shows significant storage for the return periods
considered. a good comparison between the
synthetic rainfall series and the annual t imeseries
is obtained for storms which occurred less
frequently than 5:1 year. For storms which
occurred more frequentJ-y than 6:1 year the
synthetic rainfall series under predicted the
atorage by up to 20t. The effect of storm duration
on this was tested, and a storm of 32 hours
duration and return period 32:1 year provided a
storage value of 183.6m2 which reduced the error at
th is return per iod by 5-6t .
The overflow from the storage tank spil led in many
of the storms and the comparison of spil l volume is
shown in Figure L7. The predictions of the annual
t imeser ies and the synthet ic ra infa l l ser ies are
s im i l a r .
4 . 2 . 2 M o d e l 2
The overflow at tank 1 did not operate in any of
the storms which were used.
The graph of overflow spil l against return period
for tank 2 is given in Figure 18. Good agreement
is shown between the annual t imeseries and the
synthet ic ra infa l l ser ies.
A: \TSRREP.WP
1 831 March 1992
4 . 3 Effect of sampling
the annual t imeser ies
5 . coNcrusroNs
The annuat timeseries contains 99 stonns and it is
therefore not usual to analyse all of these because
of the long computing times which this would
enta i l . Instead a sample of the storms is
analysed, and the results for the whole series
ext,rapolated from this. The most usual sampling is
to use the first f ive large storms, and then every
fifth storrn for the rest of the series. The effect
of this on the predictions for the two models was
checked. In both cases the synthetic rainfall
series rrras a better representation of the full
t imeseries than the sampLe. It should be
remembered that the sampled timeseries st,i l l
consis ts of 23 storms compared to 10 for the
synthet ic ser ies.
The annual t imeseries does not contain enough longr
duration storms to correctly represent the true
ra infa l l pat tern.
The variation of storm depth with return period for
return periods less than one year appears to be
independent of storm duration, but does vary
regional ly .
The published average sununer values of UCWI appear
to be too Iow, as many suruner storms have much
higher values
If storm peakedness is defined to vary with
duration then it gives much higher intensities for
A: \TSRREP.wP
1 931 March 1992
long duration storms, although lower peak
intensi t ies for shor t durat ion storms.
For models using the constant runoff coefficient
model of the Wallingrford Procedure, the skew of the
storm profi le has l itt le effect on storage volume
or over f low spi I I , and centra l skew can be used.
Design of detention storage in urban drainage
systems should use the standard ldint.er value of
UCWI and the new storm profi le with peakedness
varying with duration. A range of durations
including very long durations should be analysed.
To predict overflow spil1, the use of the
ra infa l l ser ies, us ing the new storm
characLer is t ics, g ives resul ts as good as
annual t imeseries and usually better than
sampled timeseries, and uses fewer storns
e i t he r o f t hese .
synthetic
the
the
than
A: \TSRREP.WP
2 031 March 1992
5. E"UII'RE WORK
The work carried out i-n this study has made several
imporLant advances, but has raj-sed more questions
than iL has answered. Some of these should be the
subject of fu ture research work.
The work has indicated that the annual t imeseries
is not a good representat ion of ra infa l l pat terns.
The derivation of the synthetic rainfall series
should therefore be repeated using longer rainfall
records which wil l be more representative.
The proposed changes to UCWI and storrn profi le
should apply to the design of pipe systems as weLl
as detention storage. As pipe systems are normally
designed for short duration storms the tero changes
are l ike ly to cancel out . However the ef fect
should be tested-
The analysis carried out here has not looked at all
of the differences between sunmer and winter
condi t ions. As many types of s tudies, for example
overf low spi l l , requi re separate considerat ion of
summer and winter conditions, future analyses
should be done separate ly for the two seasons.
A modified runoff model is being introduced for the
Wallingford Procedure which uses a different
measure of catchment wetness. The analysis should
be updated to provide results for this model. This
model wil l change the catchment $retness and runoff
during a storm. This may therefore make skewed
storms more s igni f icant .
The use of the synthet ic ra infaLl ser ies should be
analysed for water qual i ty modelJ- ing. I t is again
A: \TSRREP.WP
2 L31 March 1992
7. ACKNO}]II.EDGEMENIIS
l ikely that storm skew will be more significant for
t h i s app l i ca t i on .
The analysis has ahosrn significant regional
var iat ions in ra infa l l character is t ics, and future
work should look at more regions.
The definit ions of skew and peakedness are sti l l-
clumsy and consideration should be given to
developing better definit ions.
The authors are grateful to glRc for providing the
timeseries rainfalL on which the eynthetic rainfall
series have been based. To Thames wat.er Util i t ies'
Yorkshire water, Norfolk District Council and
Hertsmere Drainage for supplying real models which
were used for test ing.
A: \TSRREP.WP
2231 March 1992
8. REFERENCES
HR gfa l l ingford. The c lass i f icat ion of s torms in
the annua l t imese r i es . Ex 2429 . Oc tobe r 1991 - .
Rainey C M L t Osborne M P. Design storms.
HR Wal l ingford Report No SR 291. November 1991.
A: \TSRREP.wP
2 331 March 1992
Figures
t-{
tr
o0
ooctm
-*-
O.rF
N
oN
O(o
O1,r^)
qo
r..o
T)o.C()o_cL-f
c)v.
ooL-
_c+Jo-q)
O
+roo
L-r-l
_C+:loo
]C,o .odb
o-c
n L _c i=
-+-tCIv.
af
o
rr)o
Nci
rf,)r..
r
o
loct
o!+Dr r-, l ld ac
F
c)${
' o 0.Fl
O t rm$
o=fN
III
]Coo'cN( l ) r -o- l
c il - r:l
-lJ
0)v.o
ro
qo
\o
roci
loci
to
rrlo
clo
-oo'coo-cf
.|J
otr
oot*
-c+Jo-c)o
-ano
t-r-J
-C+:lo
(,t') oN
3
,r- oci lO
Ol
lO
oln
I!"j
I
iIIi
IIi-"._ __*_ I *-_*- * *_ .l
o!+or qldec
r-l
C)lr
o0.rl
oroctm
$
C)+N
!
3f3O- ,
' o [ Ic i :
+)O-rx.6
13o.Cq)o-cL-:J
-+-)()M.
o^l-J
ot-
_c.+)o-c)o
C)t_
o -
-Cn
-Yt_o
oI'O
:
rot-
rOcj
lo
IL-
o!+Dr q+dec
q)tr
@
o6
o l$O l
l
qoo
+N
oroII
lnEo:
+J()E.
{cj
I t \lc ;
T)o' f0)o-ct-:l
C)tr
ooL-
-co-oc]
C)L-
o - -
-Cn
-Yt_o
'ro
8fi FO - 1
r?ooN
Ol
oci3
No
II
I!IIIIII
IL*-.-*_*L*_t q F
__*-"-*----.--*Lu)o
o!+Dr r., l ldec
Flr-{
ot{
o0'-h
om.f,
I
o=fN
1'
3'f; sO - l
Er:l
1-,
&B
on
IiI
rO!
, O li oI
I
Il
i
ioc;
\o
T'o'c(uo-cL:t()E.
o1J
oL.
-co-c)c)
p
U)a)
=
_C+_lon
rOct
to
t?o
c!o
l F
l o
I-1
III
t1it-{
I
I- J O
oo
o t @ t , . r o l o $ n Nctc ic tc tdctc ic i
o!+Dr r.{}dec
hF{
c)l-l
@
ot6=f,
o.sf,N
o(o
oN
3
1l
q,,$3
olo
qo
t .c;\
\\
\\
\1f
\
t',
\. \, \
. \i
\
\
\1\I\II\i
'lco'cc)o-ct-J0)tr
ooL
-co-c)c)
-+rn()
=
-C-p
:)
on
^cul t-o: l
Pq)
Mlf
cj
i/)
o
No
III-1
oci3
r r l_ l-_ *___ | _.__
' . F ( l t ON
o!+Dr L{}d oc
35
30
25
20
l5
1 0
)
0
Skew DistributionSouth East
0 0.1 o.2 0.3 0.4 0.5 0.6 01 0-8 0.9 I
Distribution
Figure 3A
Skew Distirbution, Top 20 stormsSouth East
() 0. l 0.2 0 3 0.4 0.5 0.6 0.1 0.8 0.9 I
Distribution
F iqu re 38
30
25
20
l5
l0
5
0
Skew DistributionYorkshire
I
Figure 3C
0.90.80.70.30 .1 0.4 0.5 0.6
Distribution
Skew Distirbution, Top 20 stormsYorkshire
0.4 0 .5 0 .6
Distr ibution
30
25
20
l5
l0
5
0
UCWI ratioSouth East
2.42.20.4 1.2 1 .4 1 .6 1 .8
Distribution2.8 3
Figure 4A
35
30
25
20
l 5
l 0
)
0
UCWI ratioYorkshire
o 0 .2 0 .4 0 .6 0 .8 1 t -2 I .4 1 .6 I .8 2
Distr ibution2.4 2.6 2.8 3
Figure 48
i-*-III
.it
c)k
@.Fl
l-
0)EE:Jn
I
LCI
4-'c
=
I
co
t-
r=- ( J:)
c!
oot-
=():f
+o()
=
_C+
_)
oo q
o
toc;
1o
c!o
lnoN
III
IIiL.._*-osf
loN
Xr uan bar3
q
.9oaoEFo
.nEoCNc.9CN(I)oCN(gI,U
=oa
ootr)r
(^ur) aOe:orq' u
30% peruious catchment winter UCWI, peakedness varying with duration, Design storms 16hrs
FIGURE 5
p.9q)
:tq)
ff
oooaoEFo
oEL
oCN
.9oooo(sUJ
5oo
(ru)a0e:o1g
60% pervious catchment winter UCWI, peakedness varying with duration, Design storms 16
FIGURE
q
.9oaoEtrU'
Q
Eo
@c.9,o(1)o@(UUJ
foa
E.9o(L
q)tf
(ru:) aOerotg
7O/" pervious catchment winter UCWI, peakedness varying with duration, Design storms 16hrsFTGURE 7
oooooEt-o
oEl-
oac.9o(l)oo(l)=
oo
E.9
o_
(l)u
(rLu)a6eto13
30% pervious catchment winter UCWI, peakedness varying with duration, Design storms 16hFIGURE
aooaoEtro
oEo
@
.9@oo@o
oa
o.9(t)
(L
(DE
(rur) a6elolg
60% pervious catchment winter UCWI, peakedness varying with duration, Design storms 16hrsFIGURE 9
!oo
4
oE.
@.gl-
oooEFog,
El-
o@
.9oooCNo=
oa
oop
(ru)e6eto15
ooo
70"/" pewious catchment winter UCWI, peakedness varying with duration, Design storms 1
.n.9l-oaoEFan
aEl-
o@g
.9oooo.=
ol(t-o
o.9
n
rf
(rtu) a6etolS
30% pervious catchment winter UCWI, peakedness varying with duration, Design storms 16hrs
FIGURE 11-
o(l)
oaoEFoq
EoCN
o)ctooo.=
al<o
aoo
q)(E
(ru)aOetotg
60% pervious catchment winter UCWI, peakedness varying with duration, Design storms 1
o.9q)
o_
q)at-
q
.gr-ocnoEFCI'
aEo
@c.9oooo.=
Q.!<r-o
c)oI
(ru)eOerolg
70"/" pewious catchment winter UCWI, peakedness varying with duration, Design storms 16hrsFIGURE 1-3
Example I
TANK
Bed level 84.450Overflow crest level 86.100
Plan area 2fi0m2
SOFFITLEVEL AD AREAha
PIPE DIAMETER LENGTH UPSTREAM DOWNSTREAM PA\IED ROOF PERVIOUS
500
685
900
450
300
525
8
35
30
5m
2fr
160
84.960
85.715
85.600
86.450
85.300
86.63s
84.930
85.135
85.3s0
84.900
84.750
84.975
47.672
23.652
18.454
1.336
0.99
3.24
453.603
39.8
4r3.136
0.014
0.01
0.ff3
0
0
0
0
0
0
Figure 14
Example 2a
Example 2a
PIPE DIAMETER
TANK2Bed level 214.8n
Overflow crcstlevel 2L4.952Planarc,a2.49hn2
SOFFIT LEVEL AD AREA ha
LENGfi{ UPSTREAM DOWNSTREAM PAVED
Example 2b
TANK IBedlevel 25l.MOOverflow crest level 251.390
Plan area 0.866m2
ROOF PERVIOUS
| 225
2 750
3 750
4 225
95 215.052 210.811
r0 2r5.45r 215.580
r15 2r7.940 2r5.45r
19 21836t 215.055
5.332 4.784 16.837
5.332 4.784 16.837
5.312 4i64 16.777
0.020 0.020 0.060
Example 2b
SOFFITLEVELAD AREAhA
PIPE DIAMETER LENGTH IJPSTREAM DOWNSTREAM PAVED ROOF PERVIOUS
1 300 75 25r.390 28.900 0.877 0.79s 2.949
2 300 13 252.44 25r.390 0.787 0.755 2.879
Figure 15
a
t-
-
n
- .c . '
@
€-->
<L.>
<-
€>
@
rr't
E
. :OJ
o-(9c { :
(uE
4
<I_>
-<t-->
e/->
<-5
tL<
<-
c7=
€l_>
cE
><
S S €s r J ) €c\
A P o L o c n o J E { - . )FIGURE 16
ea
r-e
i--
+ -
e,--
<t->
:€
t-<
<I->
<L>
qg
-;-=
p
. :(l,
s o -c\t c
a)E
U=-
4
<l->
L-
<L->'n
€,.->
-<=>
E_
<L->-
(SS€s
ssstt)
sSsc\
G)€s=
ss€
€€S€c!
€ €ISsLN
FIGURE T7O > < U L + - O - E a r )
C->
a<
€
4
a
<l->
F-
-<-->-
n<->
L
<l->4
A-
<3-->
E-<
i=:-
--
=>
-\2
<l->
€g
;'<
ss€
s
o
Ol
€ o -c\
E
O > o g s _ e , - E C f )
FIGURE 18
APPENDIX A
A: \TSRREP.T'P
2 5
31 uarch 1992
Annual Series Rainfall Data for SOUTH EAST U.K. **'t'k IUNKED IEAII ****
315FF3 .0
ucltll=9o ** I ** EVENT
0
1 4 : 4 1 4 / B / 7 LI
1
67 .000 66 .000 65 .000 64 .000 63 .000 62 .00057 .000 56 .000 55 .000 54 .000 53 .000 52 .00047 .000 46 .000 45 .000 44 .000 43 .000 42 .ooo37 .000 36 .000 35 .000 34 .000 33 .000 32 .00027 .000 26 .000 25 .000 24 .000 23 .000 22 .o0017 .000 16 .000 15 .000 14 .000 13 .000 12 .0007 .000 6 .000 5 .000 4 .000 3 .000 2 .0000 .000 0 .000 0 .000 0 .000 0 .000 0 .0000 .000 0 .000 0 .000 0 .000 0 .000 0 .000
08:01 25/9/67 610 vAtUES
3 .000 4 .000 5 .000 6 .000 7 .000 8 .00013 .000 14 .000 15 .000 16 .000 17 .000 18 .00023 .000 24 .000 25 .000 26 .000 27 .000 28 .00033 .000 34 .000 35 .000 36 .000 37 .000 38 .00043 .000 44 .000 45 .000 46 .000 47 .000 48 .00053 .000 54 .000 55 .000 56 .000 57 .000 58 .00063 .000 64 .000 65 .000 66 .000 67 .000 68 .00067 .000 66 .000 65 .000 64 .000 63 .000 62 .00057 .000 56 .000 55 .000 54 .000 53 .000 52 .00047 .000 46 .000 45 .000 44 .000 43 .000 42 .00037 .000 36 .000 35 .000 34 .000 33 .000 32 .00027 .000 26 .000 25 .000 24 .ooo 23 .000 22 .00017 .000 16 .000 15 .000 14 .000 13 .000 12 .0007 .000 6 .000 5 .000 4 .000 3 .000 2 .0000 .000 0 .000 0 .000 0 .000 0 .000 0 .0000 .000 0 .000 0 .000 0 .000 0 .000 0 .000
F
4 (AUG)
90320 VATUES
0 .3090 60 350 600
9 .000 0 .000 1 .000 2 .000 3 .000 4 .000 5 .000 6 . o0o 7 . oo0 8 .00019 .000 10 .000 11 .000 12 .000 13 .000 14 . o0o l s .000 16 .000 17 .000 18 .00029 .000 20 .000 21 .000 22 ,ooo 23 .ooo 24 .OOO 25 .000 26 .000 27 .OOO 28 .00039 .000 30 .000 31 .000 32 .000 33 .000 34 . OOO 35 .000 36 .OOO 37 .000 38 . OOO49.000 40 .000 41 .000 42 .Ooo 43 .000 44 .000 45 .ooo 46 .ooo 47 .ooo 48 .00059 .000 50 .000 51 .000 52 .000 53 .ooo 54 .000 55 .000 56 .000 57 ,000 58 .00069 .000 60 .000 61 .000 62 .000 63 .000 64 .000 65 .000 66 .000 67 .000 68 .0000 .000 0 .000 0 .000 0 .000 0 .000 0 .000 0 .000 0 .000 o .o0o 0 .0000 .000 0 .000 0 .000 0 .000 0 .000 0 .000 0 .000 0 .000 0 .000 0 .000
ucwr=146 ** 2 ** EVENT 3 (APR) 01:07 5/4/sB 380 vALUEs0 .30 146
90 60 360 60061 .000 70 .000 69 .000 68 .00051 .000 60 .000 59 .000 58 .00041 .000 50 .000 49 .000 48 .00031 .000 40 .000 39 .000 38 .00021 .000 30 .000 29 .000 28 .00011 .000 20 .000 19"000 18 .0001 .000 10 .000 9 .000 8 .0000.ooo o.o0o o.ooo o.ooo0 .000 0 .000 0 .000 0 .000
ucl^il=65 ** 3 ** EVENT 15 (SEP)0 .30 65 I
160 60 480 6009 .000 0 .000 1 .000 2 .000
19 .000 10 .000 11 .000 12 .00029 .000 20 .000 21 .000 22 .00039 .000 30 .000 31 .000 32 .00049 .000 40 .000 41 .000 42 .000s9 .000 50 .000 51 .000 52 .00069 .000 60 .000 61 .000 62 .00061 .000 70 .000 69 .000 68 .000s1 .000 60 .000 59 .000 58 .00041 .000 50 .000 49 .000 48 .00031 .000 40 .000 39 .000 38 .0002 r .000 30 .000 29 .000 28 .00011 .000 20 .000 19 .000 18 .0001 .000 i 0 .000 9 .000 8 .0000 .000 0 .000 0 .000 0 .0000 .000 0 .000 0 .000 0 .000