HSE Rr740 MonteCarlo

Health and Safety
Executive
A Monte Carlo approach to joint

probability of wave, tide and surge
in extreme water level calculations
Prepared by PhysE Limited
for the Health and Safety Executive 2009
RR740
Research Report
Health and Safety
Executive

Joseph P Fox MPhys PhD
PhysE Limited
The Harbour Offices
The Quay
Yarmouth
Isle of Wight
PO41 0NT
Extreme Water Level is the total level of the water arising from wave crest, storm surge and tidal level
acting in combination with each other throughout an extreme event. Individually, each of these components
is relatively well understood. However, the combination of these components is not straightforward and
depends on the conditions of the location in question. This study addresses the behaviour of the individual
components, their interaction, and their joint probability, leading to the development of a Monte Carlo
based approach to the estimation of Extreme Water Level. The study concludes with a review of latest ISO
guidance on estimating extreme water levels and a comparison between extreme water level results derived
using existing methods, ISO guidance and Monte Carlo simulations.
This report and the work it describes were funded by the Health and Safety Executive (HSE). Its contents,
including any opinions and/or conclusions expressed, are those of the author alone and do not necessarily
reflect HSE policy.
HSE Books
© Crown copyright 2009
First published 2009
All rights reserved. No part of this publication may be

reproduced, stored in a retrieval system, or transmitted
in any form or by any means (electronic, mechanical,
photocopying, recording or otherwise) without the prior
written permission of the copyright owner.
Applications for reproduction should be made in writing to:

Licensing Division, Her Majesty’s Stationery Office,
St Clements House, 2-16 Colegate, Norwich NR3 1BQ
or by e-mail to hmsolicensing@cabinet-office.x.gsi.gov.uk
ii
TABLE OF CONTENTS
EXECUTIVE SUMMARY ................................................................................... 7

1. BACKGROUND ........................................................................................ 1
2. DATA ........................................................................................................ 4
2.1 MEASURED DATA .................................................................................................... 4
2.2 MODEL DATA .......................................................................................................... 4
2.3 OVERVIEW OF MAXIMA AND ASSOCIATED VALUES .................................................... 5
3. WAVE CREST........................................................................................... 6
3.1 KROGSTAD/BORGMAN INTEGRAL ............................................................................. 7
3.2 TROMANS & VANDERSCHUREN METHOD .................................................................. 9
4. TIDES...................................................................................................... 12
4.1 INFLUENCE OF TIDE ON SURGE AND W AVES ........................................................... 13
5. STORMS ................................................................................................. 21
6. RELATIONSHIP BETWEEN SURGE AND WAVE HEIGHT ................... 29
7. MONTE CARLO SIMULATIONS............................................................. 31
7.1 THE PROCESS....................................................................................................... 31
7.1.1 Wave heights ............................................................................................................ 31
7.1.2 Tidal heights............................................................................................................. 33
7.1.3 Surge heights............................................................................................................ 33
7.2 MONTE CARLO OUTPUT STATISTICS ...................................................................... 35
8. RESULTS................................................................................................ 36
8.1 100 YEAR RETURN PERIOD EXTREME W ATER LEVELS ........................................... 36
8.2 10,000 YEAR RETURN PERIOD EXTREME W ATER LEVELS ...................................... 37
8.3 COMMENTS ON 100 AND 1,000 YEAR EXTREME W ATER LEVELS ............................. 38
8.4 MEASURED EXTREME W ATER LEVELS .................................................................... 39
8.5 CALIBRATING THE SIMULATION .............................................................................. 40
8.6 100 AND 10,000 YEAR RETURN PERIOD EXTREME W ATER LEVELS......................... 41
9. CONCLUSIONS ...................................................................................... 43
10. APPENDIX: SOFTWARE TESTING..................................................... 44
10.1 SOFTWARE TESTING .......................................................................................... 44
10.1.1 Input Data............................................................................................................. 44
10.1.2 Hs ......................................................................................................................... 45
10.1.3 Surges ................................................................................................................... 45
10.1.4 Tides ..................................................................................................................... 48
10.1.5 Still water level ..................................................................................................... 48
10.1.6 Summary............................................................................................................... 51
iii
FIGURES
Figure 1 Location of measured data sources used in this study........................................................................ 4
Figure 2 Exceedence Distribution of Extreme Crest Elevation .......................................................................... 8
Figure 3 Illustration of mean excess plot............................................................................................................. 10
Figure 4 Probability of occurrence of measured tidal level (m) at Auk, K13 and Euro. Relative to MSL ... 12
Figure 5 K13 measured: Probability of occurrence of maximum surge and maximum wave height relative
to high water.......................................................................................................................................... 14
Figure 6 K13 Model: Probability of occurrence of maximum surge and maximum wave height relative to
high water. ............................................................................................................................................. 14
Figure 7 Euro Measured: Probability of occurrence of maximum surge and maximum wave height
relative to high water............................................................................................................................ 15
Figure 8 Euro model: Probability of occurrence of maximum surge and maximum wave height relative to
high water. ............................................................................................................................................. 15
Figure 9 Auk Measured: Probability of occurrence of maximum surge and maximum wave height relative
to high water. Based on measured Auk data .................................................................................. 16
Figure 10 Auk Model: Probability of occurrence of maximum surge and maximum wave height relative to
high water. ............................................................................................................................................. 16
Figure 11 Auk Measured: Maximum Hs and SWL throughout the positive tide. ............................................. 18
Figure 12 Auk Model: Maximum Hs and SWL throughout the positive tide. .................................................... 18
Figure 13 K13 Measured: Maximum Hs and SWL throughout the positive tide. ............................................. 19
Figure 14 K13 Model: Maximum Hs and SWL throughout the positive tide..................................................... 19
Figure 15 Euro Measured: Maximum Hs and SWL throughout the positive tide............................................. 20
Figure 16 Euro Model: Maximum Hs and SWL throughout the positive tide.................................................... 20
Figure 17 Auk Measured: Surge duration analysis .............................................................................................. 22
Figure 18 Auk Model: Surge duration analysis..................................................................................................... 22
Figure 19 K13 Measured: Surge duration analysis.............................................................................................. 23
Figure 20 K13 Model: Surge duration analysis based on model K13 data ...................................................... 23
Figure 21 Euro Measured: Surge duration analysis ............................................................................................ 24
Figure 22 Euro Model: Surge duration analysis ................................................................................................... 24
Figure 23 Auk Measured: Time series plots of largest 5 surges........................................................................ 25
Figure 24 K13 Measured: Time series plots of largest 5 surges........................................................................ 26
Figure 25 Euro Measured: Time series plots of largest 5 surges ..................................................................... 27
Figure 26 Auk Measured: Hs-Surge (positive only) scatter plot......................................................................... 30
Figure 27 K13 Measured: Hs-Surge (positive only) scatter plot ........................................................................ 30
Figure 28 Euro Measured: Hs-Surge (positive only) scatter plot ....................................................................... 30
Figure 29 Monte Carlo: Input Specifications for Significant Wave Height ........................................................ 32
Figure 30 Monte Carlo: Input Specifications for Tidal level ................................................................................ 33
Figure 31 Monte Carlo: Input Specifications for Surge........................................................................................ 34
Figure 32 Monte Carlo: Identification of the Surge ‘Scaling factor’.................................................................... 35
Figure 33 Variation of 20 year EWL with decreasing scaling factor. The red dashed line represents the
measured 20 year extreme water level at K13................................................................................. 40
Figure 34 Comparison between simulated and theoretical 100 year Hs values.............................................. 45
Figure 35 Distribution of surges associated with the maximum crest+surge. .................................................. 46
Figure 37 Distribution of surges associated with the maximum crest+surge ................................................... 47
Figure 39 Distribution of tides associated with the maximum crest+tide. ......................................................... 48
Figure 40 Distribution of SWL associated with the maximum crest+tide+surge.............................................. 49
iv
TABLES
Table 1 Key statistics from the measured and model data sets .....................................................................5
Table 2 Comparison of Extreme Crest Elevation – Borgman & Monte Carlo ...............................................8
Table 3 Height and Duration of the 5 largest Surges at Auk, K13 and Euro...............................................28
Table 4 Comparison of 100 year Extreme Water Levels constituents calculated using various
techniques ..............................................................................................................................................36
Table 5 Summary of 100 year Extreme Water Levels found using different techniques...........................37
Table 6 Comparison of 100 year Extreme Water Levels constituents calculated using various
techniques ..............................................................................................................................................37
Table 7 Summary of 10,000 year Extreme Water Levels found using different techniques .....................38
Table 8 Summary of extrapolated total water levels at K13 ..........................................................................39
Table 9 Summary of Monte Carlo Simulation inputs at K13..........................................................................40
Table 10 Comparison of 100 and 10,000 year return periods at K13 ............................................................41
v
vi
EXECUTIVE SUMMARY
Measured data from sites across the North Sea are used to investigate the individual and
combined behaviour of the wave crest, storm surge and tidal level. The joint probability of the
parameters is discussed and Monte Carlo simulations are proposed as a method of rigorously
accounting for the joint probability of the individual components.
Software was developed and the results of Monte Carlo simulation compared with existing
methods of estimating the extreme water level at the 100 and 10,000 year return periods.
Measured total water level data was used and comparisons made between the predicted 100
and 10,000 year extreme water level from the measured data and those predicted via Monte
Carlo simulation and other methods, including those described in ISO-19902:2007.
Comparisons revealed that the Monte Carlo simulations provided a reliable method with which
the extreme water level could be estimated. Furthermore, ISO-19902 describes ‘best’ and
‘worst’ cases which provide an envelope into which the predicted extreme water levels from
all the other techniques used in this report fall. The techniques used in this report represent
both historical and the most commonly used contemporary techniques for estimating extreme
water levels. However, the list may not be exhaustive and other techniques may exist.
vii
viii
1. BACKGROUND
Extreme Water Level (EWL) is a key parameter in the design and safe operation of offshore
installations, particularly with respect to bottom founded structures. A wave impact on the
topsides produces a significantly greater overturning moment than a wave passing through the
platform legs1. EWL therefore determines the elevation of the base of the topsides.
EWL is defined as a combination of three components;

• Wave crest
• Tidal elevation
• Surge (the variation in sea level induced my meteorological influences).
It is important to differentiate between EWL and Still Water Level (SWL). EWL includes the
wave crest and is therefore a transient condition associated with water particle motions induced
by waves. SWL does not include waves and defines the level due only to tide and surge acting
together.
Between 1974 and 1996 offshore structures may have been designed according to the UK
Guidance Notes2. Guidance for the derivation of extreme water levels given therein was that
EWL may be derived through the direct superposition of crest heights calculated empirically
from a sea state and surge amplitude (both with a 50 year return period) and the level Spring
Tidal Amplitude (defined as M2 + S2). Hence:
EWL50 = (M2 + S2) + S+50 + Hcrest50

Where:
EWL50 = Extreme Water Level with a 50-year return period
M2 = Amplitude of the Lunar semi-diurnal tidal constituent
S2 = Amplitude of the Solar semi-diurnal tidal constituent
S+50 = Amplitude of the surge with a 50-year return period
Hcrest50 = Amplitude of the wave crest for the maximum wave height with a 50
year return period
For design, an air gap of 1.5 m between EWL and the base of the lowermost deck was
recommended.
This approach might be described as unsophisticated because:
• The calculation of height of the wave crest does not take account of the possibility
that the maximum wave might occur in a sea state which, although high, might not
necessarily be the highest sea state.
• The derived value of Hcrest50 is the most probable value in the distribution of maxima.
As such, there is a significant probability (nominally 63%, subject to the particular
form of crest distribution applied) that the derived Hcrest50 will be exceeded.
• The tidal range may not be a mean spring tide.
1
Airgap Workshop
HSE/E&P Forum
Imperial College, London. 14-15 June 1999
2
The Offshore Installations (Construction and Survey) Regulations 1974
HMSO,1974. SI 1974 / 289
[Replaced by SI 1996 / 913 – The Offshore Installations and Wells
(Design and Construction etc.) Regulations, 1996 – ISBN: 0 110 54451 X].
1
• The maximum crest may not occur at the time of high water.
• The surge may not be a 50 year event, simply because the waves are a 50 year event.
• The peak of surge may not be coincident with the highest wave.
• The value of the 1.5 m air gap varied with the general exposure of the location – 1.5 m
provided a significantly smaller margin of safety in the northern North Sea than in the
southern North Sea.
It should be noted that by the early 1980’s the larger oil and gas exploration and production
companies were designing facilities to a return period of 100 years and applying variations on
the 1974 recipe to account for joint probability of tide and surge, for example:
EWL100 = (M2 + S2) + S+50 + Hcrest100
In this case the implicit assumption was:
SWL100 = (M2 + S2) + S+50
…and some evidence to support this assumption was available from the analysis of long term
sea level data sets at standard ports.
As a result of significantly greater temporal and spatial resolution in acquired oceanographic

data, together with advances in computing and recent developments in formulating short-term
statistics, potential shortcomings in the 1974 approach may be addressed. International
Standard ISO 19901-1 published in 20053 addresses improvements to the estimation of the
wave crest in some detail, but does not provide guidance with respect to the computation of the
corresponding tide and surge components.
Latest guidance with respect to the incorporation of tide and surge may be found in ISO
19902:20074 which states
-----Beginning of quote-----
“…the deck elevation, h, above the mean sea level can be estimated [as follows]:
If a storm surge is not expected to occur at the same time as the abnormal wave crest:
h = √(a 2 + s 2 + t 2) + f
3
Petroleum and natural gas industries – Specific requirements for offshore structures
Part 1: Metocean design and operating considerations
BS EN ISO 19901-1:2005
4
Petroleum and natural gas industries – Fixed steel offshore structures
ISO 19902:2007
2
If a storm surge is expected to occur at the same time as the abnormal wave crest:
h = √((a + s) 2 + t 2) + f
where:
a is the abnormal wave crest height
s is the extreme storm surge
t is the maximum elevation of the tide relative to mean sea level
f is the expected sum of subsidence, settlement and sea level rise over the design
service life of the structure.
For deep and intermediate water depths ‘a’ can be approximated to
a > 1.3 a100
a > a100 + 1.5m
where :
a100 is the extreme wave crest height with a return period of 100 years.
The estimate for h obtained with this procedure is indicative and suitable for conceptual design
studies. The owner should review the deck elevation prior to detailed design.
In general, no platform processing elements, piping, or equipment should be located below the
lower deck in the designated air gap. However, when it is unavoidable to position such items
as minor sub-cellars, sumps, drains, or production piping in the air gap, provisions should be
made for the actions due to waves developed on these items.
NOTE: 1.5m is a traditional value used for air gap, but analysis of metocean data has shown
that it does not always allow sufficient reliability in certain geographical areas.”
-----End of quote-----
The aim of this work in this report is therefore to:
a) Investigate the joint probability relationships between the key components of extreme
water level; namely the tide, surge and wave data, and against that background to…
b) Establish a mathematically rigorous method for the estimation of EWL that can be
implemented within the constraints of the data actually available within the public
domain.
The work concludes that a Monte Carlo technique is appropriate to the task, this being
particularly advantageous in that it can be run using data supplied as frequency distributions.
In this form data are more readily available and less costly that data as time series.
In closing, the veracity of methods for the calculation of extreme water level are assessed,
including past techniques (upon which the design of many existing structures is based) present
techniques (developed in response to ISO 19901-1) and future techniques (as recently
proposed in ISO 19902:2007).
3
2. DATA
There are two types of data used in this study: measured data and computer-generated model
data.
2.1 MEASURED DATA

The measured data have been extracted from the Rijkswaterstaat website5 and comprise
measurements of significant wave height (Hs) and still water level (SWL). The Hs data are
recorded every 3 hours while the SWL data are presented at 10 minute intervals. Data from
the available North Sea sites (shown below) were extracted. The location and duration of the
datasets are summarised below.
Name: Auk platform

Location: 56°23’59”N 02°03’56”E
Water depth: 85m
Data recording: 1989-1996
Name: K13 platform
Location: 53°13’04”N 03°13’13”E
Water depth: 30m
Name: Euro platform
Location: 51°59’55”N 03°16’35”E
Water depth: 32m
Figure 1 Location of measured data sources used in this study
For convenience the Euro Platform is hereafter abbreviated ‘Euro’. The SWL from each data
set was obtained by harmonic analysis to identify the constituents of tide and the residual
(surge) component. Thus, following processing, the measured data sets comprise Hs, tide and
surge.
2.2 MODEL DATA

Model data, from the NEXTRA/DHI hindcast was extracted for same locations as the
measured data. The model data includes Hs, tide and surge, and is presented at 1-hour
intervals between 1964 and 1995. Experience has shown that the model performs least well in
shallow regions, for example the southern North Sea and, in particular, regions close to the
coast. However, the model data will be used in this study as a direct comparison with the
measured data and interpretation of the model data results will take account of potential
limitations of the model.
5
http://www.rijkswaterstaat.nl/
4
2.3 OVERVIEW OF MAXIMA AND ASSOCIATED VALUES
The following table presents a summary of the data from each of the data sets.
Table 1 Key statistics from the measured and model data sets
Auk K13 Euro

Parameter Auk K13 Euro
[model] [model] [model]
Duration of Data Set 7 yrs 20 yrs 15 yrs 30 yrs 30 yrs 30 yrs
Maximum Hs 12.73 7.98 6.64 11.97 8.64 7.39

Associated Surge 1.22 1.23 0.45 1.15 1.48 1.33
Associated Tide 0.48 -0.31 0.39 -0.12 -0.34 0.30
Associated SWL 1.70 0.92 0.84 1.04 1.14 1.63
Maximum Surge 2.13 1.98 2.79 1.46 2.04 2.18

Associated Hs 7.81 4.36 5.05 9.66 7.25 6.13
Associated Tide -0.63 -0.56 -0.16 -0.22 -0.83 -0.03
Associated SWL 1.50 1.42 2.63 1.24 1.21 2.15
Maximum SWL 2.20 2.51 2.65 1.66 3.19 2.58

Associated Hs 7.35 4.95 4.62 8.92 7.16 6.32
Associated Surge 1.60 1.81 1.96 1.28 1.62 1.78
Associated Tide 0.60 0.70 0.69 0.38 1.57 0.79
Maximum Tide 0.90 1.26 1.36 0.58 1.85 0.89
Notes: Levels and amplitudes are specified in metres

The maximum tidal level was reached on many occasions and thus associated values
are not applicable to the Maximum Tidal level.
Tidal levels and SWL are specified relative to Mean Sea Level (MSL)
It is worth noting that without exception:
• The tidal level associated with the maximum surge at all locations is negative
• The maximum surge did not occur coincident with the maximum Hs
• The maximum Hs did not occur at a high tide.
• The SWL associated with the maximum Hs was significantly less than the maximum
possible SWL - i.e. the maximum surge plus the maximum tide
• The SWL associated with the maximum Hs was significantly less than the maximum
surge.
• At the time of maximum SWL neither surge nor tide was at a maximum.
These observations will be further discussed in the following sections.
5
3. WAVE CREST
ISO 19901-1 states in Section 8.5:
“The required long-term, individual wave height, HN, shall be established by convolution of
long-term distributions derived from the data with a short-term distribution that accounts for
the distribution of individual wave heights in a sea state.”
Note especially the use of the word ‘shall’ which makes application of this technique
obligatory if the resulting Metocean criteria are to be ISO-compliant. The phrase the “long
term” distribution as used here is the distribution of significant wave height (Hs) based on
multiple years of data. ISO 19901-1 states in Section A.5.7:
“…at least a 25 year data set should be used to estimate the 100 year storm parameters”
The required mathematical convolution of the two distributions is complex 6,7,8. The
methodology is not discussed here and readers are referred to the corresponding references for
further information. Since the publication of ISO 19901-1 in November 2005 two alternative
methods for calculating long term individual wave heights have been proposed. These have
been encoded by PhysE following the method set down in the corresponding papers:
• Method one (as described by Krogstad and based on the Borgman Integral) is based on
a fit to the entire distribution of significant wave height data, rather than to storm data.
This method requires no user intervention other than to approve the quality of the fit of
the ‘long term’ Hs data to the Weibull 3-parameter distribution, and to state which
distribution is to be used for the short-term distribution of wave crests. A fundamental
weakness of this approach is that it concerns wave crest elevation only and does not
address the concurrent values of tide and surge.
• Method two (as described by Tromans & Vanderschuren) is based on the use of storm
data only. This method requires a greater degree of user intervention, in particular the
selection of ‘storm’ data via the specification of a threshold value for Hs, specification
of the short-term distribution to be used for wave crests and subsequent optimization
of the quality of the fit of the distribution of ‘most probable crest heights’ (Hmp) to the
Weibull 3-parameter distribution. This latter process may require binning of the data
and/or restriction of the fit to the upper tail. Whilst it is possible to include the
contribution from tide and surge by adjusting each Hmp value for the associated SWL
as it is calculated, the resulting fit to the Hmp distribution is degraded such that the
results of the calculation become unpredictable. That is to say, when this technique is
applied to two data sets having only small geographical separation, it is not
6
Krogstad H. E. (2004)
Analysis of Wave Data from the Valhall Field – A Comparison of Methodologies
SINTEF ICT Applied Mathematics, October 2005
STF F04409
7
Borgmann, L. (1973)
Probabilities for the highest wave in a hurricane
J. Waterways Harbours and Coastal Eng., Vol 99 (WW2), pp. 185-207.
8
Tromans P.S. and Vanderschuren L. (1995)
Response Based Design Conditions in the North Sea – Application of a New method
Proceedings OTC, Houston, OTC 7683
6
impossible that the inclusion of tide and surge may cause EWL to increase at one site
and to fall at the adjacent site.
In view of the fact that neither the Krogstad/Borgman approach, not the Tromans and
Vanderschuren approach can satisfactorily include the contribution from tide and surge into the
calculation, a third approach was developed here. This is based on Monte Carlo simulation
and includes full convolution of the long and short term distribution of wave data, plus the
contributions from tide and surge. The background, development, testing and results of this
software are the subject of this report.
3.1 KROGSTAD/BORGMAN INTEGRAL

In order to make an estimate of extreme wave crest elevation for any given return period the
user must specify:
• Shape, scale and location parameter of the 3-parameter Weibull Distribution.

• The Short Term distribution to be used for the calculation of wave crests. There are
several alternative distributions that may be selected by the user for wave crest
analysis (e.g. Forristall 2D, Forristall 3D, Jahns & Wheeler) and the resulting extreme
wave crest elevations will vary accordingly.
• The relationship between significant wave height and mean wave period (this
determines the number of waves deemed to occur in the selected return period).
• The required return period for the output value.
Having specified the above, there user is given no subsequent opportunity for intervention in
the calculation.
Note that the result of an analysis of this type is the mode of the Borgman Integral. This is the
most probable value of the extreme wave crest elevation which occurs and is thus associated
with a probability of non-exceedence of 0.37. This means that if the 100-year sea state were to
occur 100 times, then on 63 of those occasions, the design maximum crest elevation would be
exceeded.
Figure 2 illustrates a distribution of crest maxima generated using Monte Carlo code, based on
100 simulations of the extreme crest elevation with a return period of 10,000 years. Note that
for the particular test data set in question the extreme crest elevation for design would be
calculated as 21.2 m. The Monte Carlo simulation gives access to the spread of results and
while the value of 21.2 m (63% exceedence) is normally used, other larger values could be
used for different applications – indeed the simulations show that the largest 10,000 year value
reaches 28.2m.
7
Most Probable Maximum
Crest occurs at P = 0.37
Figure 2 Exceedence Distribution of Extreme Crest Elevation
The Borgman Integral is an elegant mathematical technique for calculating the mode of the
distribution of most probable maximum crest elevations in extreme sea states. It is therefore
possible to verify this technique against Monte-Carlo simulations, providing that the Monte-
Carlo result is derived at the point where probability of non-exceedence = 0.37.
The available data sets were therefore processed and the following results obtained:
Table 2 Comparison of Extreme Crest Elevation – Borgman & Monte Carlo
10,000 Year Crest Elevation (m)

Method Euro
Auk K13
Platform
Borgman Integral
Based on the method described by 21.36 13.96 12.65
Krogstad
Monte Carlo 21.36 13.92 12.79
Mean of 300 10,000-year simulations
8
These two completely independent approaches give virtually the same result. It is to be
anticipated that slight differences may occur because of the randomness built into the Monte-
Carlo technique. The Borgman Integral and Monte Carlo techniques are thus verified against
each-other and give the same result to within an acceptable level of accuracy.
3.2 TROMANS & VANDERSCHUREN METHOD

The Tromans & Vanderschuren (TVM) method of convolving the long and short term
distributions is a storm-based approach. Indeed, ISO 19901-1 states:
“The statistically correct methods are based on storms. Storms are obtained from a time series
of significant wave height by breaking it into events that have a peak significant wave height
(Hsp) above some threshold.
The long-term uncertainty in the severity of the environment is treated using the probability
distribution of the severity of the storm, measured either in terms of its peak significant wave
height or the most probable maximum value of the individual waves in the storm (Hmp). The
uncertainty in the height of the maximum wave of any storm is estimated as a probability
distribution conditional on Hsp or Hmp. Convolution of the two distributions gives the
distribution for any random storm and, thereby, the complete long-term distribution for the
heights of individual waves.”
The fist stage in a TVM analysis is therefore to select the threshold that defines storm
conditions. A method for doing this based on the ‘mean excess’ is described in the
proceedings of OMAE079. Briefly, the mean excess is defined for a long time series of wave
data as the average difference between the selected threshold and every Hs value in excess of
it. Plotting the mean excess against threshold produces a curve (see Figure 3 that has been
extracted from Reference 9).
Plotting the mean excess curve enables the operator to identify the zone within which the mean
excess changes linearly with threshold, and it is within this region that the appropriate storm
threshold is defined. Tests reported in Reference 9 indicate that an appropriate storm threshold
should return approximately 3 events per month when averaged over 12 months.
9
Leggett IM, Bellamy NF, Fox JP and Sheikh R.
A Recommended Approach for Deriving ISO-Compliant 10,000 Year Extreme Water Levels in the
North Sea
26th International Conference on Offshore Mechanics and Arctic Engineering
June 10-15, 2007, San Diego, California.
OMAE2007-29559
9
3
2.5
Mean Excess (m)
Threshold Range
2
1.5
0.5
0
0 2 4 6 8 10 12 14
Threshold (m)
Figure 3 Illustration of mean excess plot
Having selected an appropriate threshold the operator then filters the input time series data to
extract only those periods when Hs is in excess of that threshold. During this procedure it may
be appropriate to apply some form of storm separation algorithm. For example, users may
define individual storms as those separated by ‘n’ hours, where ‘n’ takes a value between 24
and 72 hours. Alternatively, or in addition, users may define storms as being separated by
periods within which Hs fell to less than ‘n%’ of the peak value of the preceding storm, in this
case 80% has been proposed as an appropriate value.
The value of storm separation in hours, and the storm separation in terms of reduced Hs are
user defined variables. Individual metocean specialists are, at the time of writing, using storm
separation algorithms with differing levels of sophistication and therefore the selected ‘storms’
frequently (indeed, usually) differ from one analyst to the next. In practice the fine detail of
the selected period may not have a significant affect on the resulting estimation of the most
probable maximum crest elevation in the individual storm (Hmp). However, the storm
separation algorithm does influence the number of storms identified and hence the ‘storm rate’
which in turn determines the probability levels corresponding to the return periods of interest.
Having defined the individual storms the analyst will then calculate the most probable
maximum values for each storm event (the Hmp values). In order to do this the analyst must
state the particular short term distribution that will be used for the calculation of Hmp. If a
distribution that is appropriate to individual trough-to-crest waves is chosen (e.g. Rayleigh)
then the Hmp values will be trough-to-crest wave heights. Alternatively, if the selected short
term distribution is one that corresponds to wave crest distributions, then the Hmp values will
be crest elevations relative to mean water level. There are several alternative distributions that
may be selected by the user for wave crest analysis (e.g. Forristall 2D, Forristall 3D, Jahns &
Wheeler) and the resulting Hmp values will vary accordingly.
10
Once the Hmp values have been calculated, they must be formed into a frequency distribution
prior to extrapolation. Here again there are several options. The unprocessed frequency
distribution may be extrapolated, or the data may be classified into class bins prior to
extrapolation. In the later case the binning routine tends to add weight to the upper tail and
reduce the influence of the lower tail, and the greater the bin size the greater this effect will be.
During the extrapolation of the Hmp data the user may also restrict the fit of the Weibull
extrapolation to the upper tail only. So doing excludes any contribution from lower values of
Hmp and thus, in effect, reduces the number of storms analysed, and hence the storm rate.
Great care should be taken when restricting fits to the upper tail in order to ensure that the
storm rate remains valid.
Note also that the parameters of a fit that has been constrained to the upper tail of a distribution
do not necessarily provide a fair indication across the entire distribution. Indeed, the
parameters of the fitted tail may be significantly in error with respect to regions of that
distribution which fall below the tail.
11
4. TIDES
Tides are controlled by the position of celestial bodies, are predictable and are not influenced
by weather conditions. In their most simplified approximation, they can be regarded as a
sinusoidal wave with a period equal to approximately 12.5 hours. As such, the probability of
the tidal level being at its extreme is less than the probability of it being between its extremes
(that is to say, the amount of time that the tidal level is actually at high water is relatively small
compared with the amount of time the tide is in the process of rising or falling). Figure 4
shows the probability of occurrence of tidal level at the selected study sites, taken from
measured data. Note that the area under each of the three lines shown in Figure 4 is equal, and
represents a Probability of 1.0.
K13
Euro
Auk
Probability of Occurance
1.5 1.0 0.5 0.0 0.5 1.0 1.5

Tide (rel msl) [m]
Figure 4 Probability of occurrence of measured tidal level (m) at Auk, K13 and
Euro. Relative to MSL
Figure 4 clearly shows that the least likely tidal levels are the extremes (both high and low)
while the most likely levels are either side of mean sea level. The shape of the Auk and K13
lines show that at both locations the tide is relatively even in that the time spent above mean
sea level is approximately the same as time spent below mean sea level. At Euro, however, the
time spent above mean sea level is significantly less than below mean sea level (Probability of
being > MSL = 0.45, Probability of being < MSL = 0.55).
While the tides remain un-influenced by the severity of the sea conditions, they can have an
influence on them. This is discussed in the next section.
12
4.1 INFLUENCE OF TIDE ON SURGE AND WAVES
It has been reported10 that the tidal level can influence the local surge conditions with greatest
effect being felt in shallow waters. This is due to the fact that water depth plays a role in the
severity of surges with deeper water lessening the effect of a surge. Thus, in very shallow
water, such as the southern North Sea, the tide can alter the water depth sufficiently to alter the
surge. However, in deep water, the change in water depth brought about by the tide is slight
(as a fraction of the total depth) and so the influence that the tidal level has on surge in deep
water will be less.
To investigate the influence of tide on surge the data sets were split into tidal cycles,
concentrating on the up-tides. For each positive tide (i.e. that part of the tidal cycle when SWL
is above MWL) the time of the maximum recorded surge relative to the peak of the tide is
recorded. Data were collected for the entire duration of each data set and a probability
distribution was created to show when, during the elevated part of the tidal cycle, the
maximum surge peak occurred. The task was repeated for the maximum wave height in each
positive tide. The results are shown in the Figures 5 to 10. Note that the resolution of the
model data is one record every hour whereas the measured data are recorded at 10-minute
intervals.
10
Horsburgh K.J. and Wilson C.
Tide-surge interaction and its role in the distribution of residuals in the North Sea
Proudman Oceanographic Laboratory. In press.
13
20 years of data from K13a [ 9,303 surge peaks 11,183 Hs peaks ]
0.1 80
Black is Surge Peak
Green is Hs Peak
0.09
Red is Tide average 60
0.08
40
0.07
Average Tidal Level [cm]

20
0.06
% of occurance
0.05 0
0.04
-20
0.03
-40
0.02
-60
0.01
0 -80
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Time after high water (hours)
Figure 5 K13 measured: Probability of occurrence of maximum surge and

maximum wave height relative to high water.
NEXTRA/DHI K13a [ 8,031 surge peaks 11,260 Hs peaks ]

0.35 150
Black is Surge Peak
Green is Hs Peak
Red is Tide average
0.3
100
0.25
50
% of occurance
0.2
0.15
-50
0.1
-100
0.05
0 -150
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 6 K13 Model: Probability of occurrence of maximum surge and maximum

wave height relative to high water.
14
15 years of data from EuroPlatform [ 7,176 surge peaks, 9,127 Hs peaks ]
0.1 120
Black is Surge Peak
Green is Hs Peak 100
0.09
Red is Tide average
80
0.08
60
0.07

40
0.06
% of occurance
20
0.05
0
0.04
-20
0.03
-40
0.02
-60
0.01 -80
0 -100
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 7 Euro Measured: Probability of occurrence of maximum surge and

NEXTRA/DHI EuroPlatform [ 6,961 surge peaks, 9,152 Hs peaks ]

0.3 80
Black is Surge Peak
Green is Hs Peak
Red is Tide average
60
0.25
40
0.2
% of occurance
20
0.15
0.1
-20
0.05
-40
0 -60
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 8 Euro model: Probability of occurrence of maximum surge and maximum

wave height relative to high water.
15
7 years of data from Auk [ 2,744 surge peaks, 3,579 Hs peaks ]
0.1 80
Black is Surge Peak
Green is Hs Peak
0.09
Red is Tide average
60
0.08
40
0.07

0.06
% of occurance
20
0.05
0
0.04
0.03
-20
0.02
-40
0.01
0 -60
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 9 Auk Measured: Probability of occurrence of maximum surge and

maximum wave height relative to high water. Based on measured Auk
data
NEXTRA/DHI Auk [ 8,440 surge peaks, 11,392 Hs peaks ]

0.35 40
Black is Surge Peak
Green is Hs Peak
0.3
20
0.25
10
% of occurance
0.2
0.15
-10
0.1
-20
0.05
-30
0 -40
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 10 Auk Model: Probability of occurrence of maximum surge and

16
All of the previous figures show a consistent theme: the maximum surge value in any positive
tide tends to occur at a time other than the time of the high water. The same is true for Hs
value – the maximum Hs in each positive tide tends to occur at a time other than the time of
the high water.
The shape of the Auk plots (Figure 9 and Figure 10) is the most straightforward: as the tidal
level increases, the probability of experiencing the maximum surge or maximum Hs smoothly
decreases – due to the increasing water depth. The shape of the remaining plots, at Euro
platform and K13 are less straightforward. In these plots there are periods on the tidal cycle
when the maximum surge or maximum Hs is significantly more likely to occur – the peaks in
the probability distributions. These results are similar to those in the work by K.J.Horsburgh
and C.Wilson10 who studied tide/surge interaction at various ports around the UK.
The above does not imply that the surge will never be large when the tide is at its peak – it
simply states that it is most likely for the maximum surge in any given positive tide to occur
somewhere other than at high tide. The surge in a positive tide that occurs at high tide could
still be large but it is unlikely to be the largest. To understand more about the interaction
between tide and surge, it is useful to plot the maximum value of the Still Water Level (SWL)
that occurs at each part of the positive tide. Thus, the data was again split into positive tides
and for each time increment (10 minutes for measured data, 1 hour for model data) the largest
SWL value (tide+surge) was recorded. The maximum of all the recorded SWL for each time
increment in the positive tide was then plotted. The results are shown in the following figures.
In the following figures, the vertical axis on the right corresponds to the tidal level while the
vertical axis on the left corresponds to the SWL and the Hs height.
17
7 years of data from Auk [ 2,744 surge peaks, 3,579 Hs peaks ]
1400 80
Black is SWL Peak
Green is Hs Peak
Red is Tide average
1200 60
1000 40
Average Tidal Level (cm)

800 20
Level (cm)
600 0
400 -20
200 -40
0 -60
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 11 Auk Measured: Maximum Hs and SWL throughout the positive tide.
NEXTRA/DHI Auk [ 8,440 surge peaks, 11,392 Hs peaks ]

1400 40
Black is SWL Peak
Green is Hs Peak
Red is Tide average
30
1200
20
1000 Average Tidal Level (cm)
10
800
Level (cm)
600
-10
400
-20
200
-30
0 -40
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 12 Auk Model: Maximum Hs and SWL throughout the positive tide.
18
20 years of data from K13a [ 9,303 surge peaks 11,183 Hs peaks ]
900 80
Black is SWL Peak
Green is Hs Peak
800 Red is Tide average
60
700
40
600

20
Level (cm)
500
0
400
-20
300
-40
200
100 -60
0 -80
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 13 K13 Measured: Maximum Hs and SWL throughout the positive tide.
NEXTRA/DHI K13a [ 8,031 surge peaks 11,260 Hs peaks ]

1000 150
Black is SWL Peak
Green is Hs Peak
900 Red is Tide average
100
800
700
50
600
Level (cm)
500 0
400
-50
300
200
-100
100
0 -150
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 14 K13 Model: Maximum Hs and SWL throughout the positive tide.
19
15 years of 10 minute data from EuroPlatform [ 7,176 surge peaks, 9,127 Hs peaks ]
700 120
Black is SWL Peak
Green is Hs Peak
600
80
500 60

40
400
Level (cm)
20
0
300
-20
200 -40
-60
100
-80
0 -100
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 15 Euro Measured: Maximum Hs and SWL throughout the positive tide.
NEXTRA/DHI EuroPlatform [ 6,961 surge peaks, 9,152 Hs peaks ]

800 80
Black is SWL Peak
Green is Hs Peak
Red is Tide average
700
60
600
40 Average Tidal Level (cm)
500
20
Level (cm)
400
0
300
-20
200
-40
100
0 -60
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
Figure 16 Euro Model: Maximum Hs and SWL throughout the positive tide.
20
The preceding plots are based on extracting maxima – hence they will be rather noisy.
However, the trend is the same across all the plots – the maximum Hs and maximum surge
values appear to be independent of tidal level. If the tide had no influence on the surge then
the maximum surge would be equally likely to occur at any point on the tidal cycle and
therefore the maximum SWL (which is the sum of the maximum surge and the tidal level)
would have approximately the same shape as the tidal level. The preceding plots show that
this is not the case; in fact the SWL is relatively constant across the positive tide indicating that
the tide and the surge interact with the tide effectively moderating the effect of the surge such
that the maximum SWL is unvarying.
This is emphasized by the data contained in Table 1 which shows that the maximum surge
from the three study sites actually occurred when the tide was negative and that the maximum
SWL comprises neither the maximum surge nor the maximum tidal level. Furthermore, the
maximum Hs values in all the data sets occurred at moderate tidal levels (all less than 54 % of
the maximum tide).
5. STORMS
The previous section has shown that tides can, and do, have an effect on when the largest Hs
and surge are most likely to occur in a tidal cycle. However, as will be shown in this section, a
storm can last much longer than a complete tidal cycle and so while the tides may affect when
the largest surge value occurs within a tidal cycle, it is true to say that the surge can be large
for long periods of time – sometimes the surge remains large over many complete tidal cycles.
5.1 SURGE DURATION
To investigate the duration of storms, the surges were analysed and the duration which the
surge level persisted above a given threshold was recorded. The threshold was varied from 0
cm to 100 cm in increments of 10 cm. The following figures present the results of the surge
durations analysis.
Figures 17 to 22 show that while the majority of surges are relatively short (less than around 6
hours) there are still surges, both small and large, which persist for a considerable time. To
understand the duration and shape of the storm surges, the top 5 surges were extracted from
each data set and plotted. The plots and the discussion appear after Figure 22.
21
Auk measured (7 years) percentage frequency
20
18
16
Frequency as a percentage of total surges
Surge > 0
14
Surge > 10
Surge > 20
12
Surge > 30
Surge > 40
10 Surge > 50
Surge > 60
8 Surge > 70
Surge > 80
Surge > 90
6
Surge > 100
0
0 2 4 6 8 10 12 14 16 18 20
Surge duration (hrs)
Figure 17 Auk Measured: Surge duration analysis
Auk modelled (31 years) percentage frequency
20
18
16
Frequency as a percentage of total surges
Surge > 0
14
Surge > 10
Surge > 20
12
Surge > 30
Surge > 40
10 Surge > 50
Surge > 60
8 Surge > 70
Surge > 80
Surge > 90
6
Surge > 100
0
0 2 4 6 8 10 12 14 16 18 20
Figure 18 Auk Model: Surge duration analysis
22
K13 Measured (20 years) percentage frequency
10
8
Frequency - as a percentage of total surges
Surge > 0
7
Surge > 10
Surge > 20
6
Surge > 30
Surge > 40
5 Surge > 50
Surge > 60
4 Surge > 70
Surge > 80
Surge > 90
3
Surge > 100
0
0 2 4 6 8 10 12 14 16 18 20
Figure 19 K13 Measured: Surge duration analysis
K13 Modelled (31 years) percentage frequency
10
8
Surge > 0
7
Surge > 10
Surge > 20
6
Surge > 30
Surge > 40
5 Surge > 50
Surge > 60
4 Surge > 70
Surge > 80
Surge > 90
3
Surge > 100
0
0 2 4 6 8 10 12 14 16 18 20
Figure 20 K13 Model: Surge duration analysis based on model K13 data
23
Euro measured (15 years) percentage frequency
7
6
Surge > 0
Surge > 10
5 Surge > 20
Surge > 30
Surge > 40
4 Surge > 50
Surge > 60
Surge > 70
3 Surge > 80
Surge > 90
Surge > 100
2
0
0 2 4 6 8 10 12 14 16 18 20
Figure 21 Euro Measured: Surge duration analysis
Euro modelled (31 years) percentage frequency
7
6
Surge > 0
Surge > 10
5 Surge > 20
Surge > 30
Surge > 40
4 Surge > 50
Surge > 60
Surge > 70
3 Surge > 80
Surge > 90
Surge > 100
2
0
0 2 4 6 8 10 12 14 16 18 20
Figure 22 Euro Model: Surge duration analysis
5.2 INDIVIDUAL SURGE PLOTS
The measured data from all three sites were used in order to investigate surge shapes. From
each data set the highest 5 surges [see Table 3 on page 5] were identified and plotted. In the
following figures, the vertical axis on the left corresponds to the tide and surge level while the
vertical axis on the right corresponds to the Hs height. All values are in centimetres.
24
Auk plots:
Surge 1: Peak 212.59cm, Duration 139 hrs Surge 2: Peak 189.3cm, Duration 118hr 10min Surge 3: Peak 184.81cm, Duration 93hrs 50
250 900 250 900 200 1200
800 800
200 200
150 1000
700 700
150 150
600 600 100 800
100 100
Height (cm)
Height (cm)
Height (cm)
500 500
50 600
400 400
50 50
300 300 0 400

0 0 12/02/1993 14/02/1993 16/02/1993 18/02/1993 20/02/1993 22/02/1993 24/02/1993 26/02/1993 28/02/1993 02/03/1993
18/02/1990 20/02/1990 22/02/1990 24/02/1990 26/02/1990 28/02/1990 02/03/1990 04/03/1990 06/03/1990 08/03/1990 10/03/1990 10/09/1990 12/09/1990 14/09/1990 16/09/1990 18/09/1990 20/09/1990 22/09/1990 24/09/1990 26/09/1990 28/09/1990 30/09/1990 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00200 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00200
-50 200
-50 -50
100 100
-100 0 -100 0 -100 0

Time stamp Time stamp Time stamp
Surge 4: Peak 178.73cm, Duration 50hrs Surge 5: Peak 178.22cm, Duration 41hrs
200 1200 200 1400
1200
150 1000 150
1000
100 800 100
800
Height (cm)
Height (cm)
50 600 50
600
0 400 0
09/10/1991 11/10/1991 13/10/1991 15/10/1991 17/10/1991 19/10/1991 21/10/1991 23/10/1991 25/10/1991 27/10/1991 29/10/1991 03/12/1990 05/12/1990 07/12/1990 09/12/1990 11/12/1990 13/12/1990 15/12/1990 17/12/1990 19/12/1990 21/12/1990 23/12/1990
400
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
-50 200 -50

200
-100 0 -100 0
Time stamp Time stamp
Figure 23 Auk Measured: Time series plots of largest 5 surges
25
K13 Plots:
Surge 1: Peak 197.83cm, Duration 123hrs 10 Surge 2: Peak 197.21cm, Duration 37hrs 50 Surge 3: Peak 177.32cm, Duration 92hrs 40
250 600 250 800 200 700
700
200 200 150 600
500
600
150 150 100 500
400
500
100 100 50 400
Height (cm)
Height (cm)
Height (cm)
300 400
50 50 0 300
300 09/10/1991 11/10/1991 13/10/1991 15/10/1991 17/10/1991 19/10/1991 21/10/1991 23/10/1991 25/10/1991 27/10/1991 29/10/1991
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
200
0 0 -50 200
10/12/1991 12/12/1991 14/12/1991 16/12/1991 18/12/1991 20/12/1991 22/12/1991 24/12/1991 26/12/1991 28/12/1991 30/12/1991 12/02/1993 14/02/1993 16/02/1993 18/02/1993 20/02/1993 22/02/1993 24/02/1993 26/02/1993 28/02/1993 200
02/03/1993
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
100
-50 -50 -100 100
100
-100 0 -100 0 -150 0

Time Stamp Time stamp Time stamp
Surge 4: Peak 173.26cm, Duration 20hrs Surge 5: Peak 165.22cm, Duration 29hrs 20
200 200 900
800
150 150
700
100 100 600

Height (cm)
Height (cm)
500
50 50
400
0 0 300
05/02/1989 07/02/1989 09/02/1989 11/02/1989 13/02/1989 15/02/1989 17/02/1989 19/02/1989 21/02/1989 23/02/1989 25/02/1989 03/12/1990 05/12/1990 07/12/1990 09/12/1990 11/12/1990 13/12/1990 15/12/1990 17/12/1990 19/12/1990 21/12/1990 23/12/1990
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
200
-50 -50
100
-100 -100 0
No Hs data available during above storm
Figure 24 K13 Measured: Time series plots of largest 5 surges
26
Euro plots:
Surge 1: Peak 279.45cm, Duration 39hrs Surge 2: Peak 205.37hrs, Duration 35hrs 30 Surge 3: Peak 200.04cm, Duration 19hrs 40
350 600 250 700 250 600
300
200 200
600
500 500
250
150 150
200 500
400 400
150 100 100
400
Height (cm)
Height (cm)
Height (cm)
100
300 50 50 300
50
300
0 0 0
24/12/1994 26/12/1994 28/12/1994 30/12/1994 01/01/1995 03/01/1995 05/01/1995 07/01/1995 09/01/1995 11/01/1995 200
13/01/1995 12/02/1993 14/02/1993 16/02/1993 18/02/1993 20/02/1993 22/02/1993 24/02/1993 26/02/1993 28/02/1993 02/03/1993 05/02/1989 07/02/1989 09/02/1989 11/02/1989 13/02/1989 15/02/1989 17/02/1989 19/02/1989 21/02/1989 23/02/1989 25/02/1989
200
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00200 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
-50
-50 -50
-100
100 100
100
-100 -100
-150
-200 0 -150 0 -150 0

Time stamp Time stamp Time stamp
Surge 4: Peak 183.83cm, Duration 31hr 20 Surge 5: Peak 143.87cm, Duration 30hrs 40
200 700 200 700
150 600 150 600
100 500 100 500
50 400 50 400
Height (cm)
Height (cm)
0 300 0 300
03/12/1990 05/12/1990 07/12/1990 09/12/1990 11/12/1990 13/12/1990 15/12/1990 17/12/1990 19/12/1990 21/12/1990 23/12/1990 05/11/1993 07/11/1993 09/11/1993 11/11/1993 13/11/1993 15/11/1993 17/11/1993 19/11/1993 21/11/1993 23/11/1993 25/11/1993
00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00 00:00
-50 200 -50 200
-100 100 -100 100
-150 0 -150 0
Figure 25 Euro Measured: Time series plots of largest 5 surges
27
Table 3 Height and Duration of the 5 largest Surges at Auk, K13 and Euro
Surge Peak Height [m] Surge Peak Duration [hours]

Auk K13 Euro Auk K13 Euro
Surge 1 2.13 1.98 2.79 139 123 39
Surge 2 1.89 1.97 2.05 118 38 36
Surge 3 1.84 1.77 2.00 94 93 20
Surge 4 1.78 1.73 1.84 50 20 31
Surge 5 1.78 1.65 1.44 41 29 31
Figures 23 to 25 show that large surges and large Hs values are generally coincident –
although as shown earlier in Table 1 on page 5, the maximum surge is not coincident
with the maximum Hs. They also show that large surges can persist for considerable
times – occasionally measurable in days. In terms of extreme water level, the previous
plots are relevant in that they show that large Hs values should be accompanied by large
surges, but that the relationship is not linear because the largest surge does not coincide
with the largest Hs.
To further investigate the relationship between Hs and surge, scatter plots of Hs and
surge were generated for the 3 locations. These are shown in the following section.
28
6. RELATIONSHIP BETWEEN SURGE AND WAVE
HEIGHT
In this section, the relationship between surge and wave height is examined. For each
location, the positive surges were plotted against the wave height to create a scatter plot.
If the relationship between surge height and wave height were linear, one would expect
to see a scatter of points lying approximately on a line of constant gradient of 1.
However, as the following plots show, the relationship between surge and wave height is
not a simple linear relationship. [Note that there was no appreciable difference in the
shape of the plots when using measured or model data and so, for brevity, only the
measured data plots are included below].
In the less shallow waters at Auk (85 m depth), the surges remain small for larger values
of wave height than at Euro platforms or K13, as seen in Figure 26. However, large
wave heights are not necessarily associated with large surges.
Figure 27 shows that, at K13, as the wave height (labeled as H4RM0 on the x-axis)
increases, the surge also increases. However, for any given wave height, there are many
instances of surges below the maximum surge for that wave height. Thus, a large wave
height does not automatically imply that the surge will be large. Even when the waves
are particularly severe – say 7 m – surges still occur which are very small. Experiencing
a large wave does not automatically mean that the surge will be large.
At the Euro platform, Figure 28, the situation is very similar to that of K13 with the
largest wave heights having associated surges which are distributed across the range of
surge values.
In summary, the degree of association between wave height and surge is not
straightforward and is affected by location (depth). In general, as the wave height
increases, the range of surges that can be associated with the wave also increase but it
does not follow that large waves are associated with only large surges.
29
Figure 26 Auk Measured: Hs-Surge (positive only) scatter plot
Figure 27 K13 Measured: Hs-Surge (positive only) scatter plot
Figure 28 Euro Measured: Hs-Surge (positive only) scatter plot
30
7. MONTE CARLO SIMULATIONS
Whilst it is possible to convolve two distributions mathematically, the task becomes

increasingly complex as more distributions are involved. In such cases Monte Carlo
simulation (named after a location famous for its games of chance) is a recognized
technique. In broad terms, to perform a Monte Carlo simulation one has to define each
of the input parameters in terms of its frequency distribution. Sequential random
selections are then made from the several distributions describing each input parameter
in order to construct a new distribution that represents all parameters acting in
combination. This is relatively easy when the input parameters are entirely independent
of each-other, but special attention must be given to partially dependent parameters such
as, in this case, wave and surge.
Measured data sets are, at best, only tens of years in duration. This is sufficient to
identify the distributions of wave height, tide and surge and, by using Monte Carlo
methods, longer durations can be simulated.
A Monte Carlo method has been adopted in order to convolve the short and long term
distributions of waves, the tidal and surge signals. After a brief description of the
software, the results of the study are presented and discussed.
7.1 THE PROCESS
Each analysis consists of multiple individual simulations, each being run over a number
of years. For example, to extract statistics for 100 years one would perform a simulation
of 100 years duration and repeat the simulation, say, 1,000 times – thus the 100 year
statistics would be based on 1,000 entries. The more repetitions of a simulated period,
the smoother the resulting frequency distribution (such as that plotted in Figure 2) and
the more reliable will be the statistics.
7.1.1 Wave heights
The Monte Carlo simulation process works by randomly choosing Hs values for all the
sea states within the specified duration of the simulation. For the particular application
under consideration, the specified duration is normally the return period for which
extreme values are required (100 or 10,000 years). The Hs values are chosen at random
from the Hs distribution which is entered by the user in terms of the scale, shape and
location parameters of the selected distribution. If a single Hs distribution is used then
all the sea state Hs values are chosen at random from a Weibull distribution.
A multiple wave distribution option has been built into the software for this process.
This provides the option of ensuring that the distribution of sea states is not perfectly
described by any one function. If multiple distributions (of which there are 4: Weibull,
Fisher-Tippett Type 1, Fisher-Tippett Type 3 and Exponential) are used then the
distribution from which the sea state will be drawn is first chosen at random and then the
Hs is chosen at random from the selected distribution.
31
The chosen Hs for a sea state is then used, in conjunction with the (fixed) Tz or
steepness11 to determine how many wave crests occur in that sea state.
The correct number of crests are then chosen, at random, from a distribution of crests
based on Hs, Tz, depth using either the Forristall 2D, Forristall 3D or Rayleigh crest
distribution (user selectable).
The various options are entered via the input screen, as illustrated in Figure 29.
Figure 29 Monte Carlo: Input Specifications for Significant Wave Height
11
Steepness is defined as 1/S = (2*π*Hs) / (9.81*Tz^2)
32
7.1.2 Tidal heights
Tidal level variations may optionally be included in any simulation. If tides are to be
selected then the user must define the tidal regime for the study location in terms of the
harmonic constituents of tide (phase and amplitude) for that location – up to a maximum
of 62 harmonics. A time is then randomly picked from an 18.6 year (representing the
duration in which one would see the HAT – highest astronomical tide) window and
calculates the tidal level at that time. Then, as each random selection of an individual
crest elevation is made, the tidal time is incremented by the corresponding wave period
and the level at that instant is calculated. In this way only one random selection of tidal
level is made for each selected Hs value, but that tidal elevation is then allowed to
change throughout the duration of that sea state (as defined by the user) so the individual
crests within that Hs are associated with a simulated tidal signal.
An alternative description of the tide may be applied by fitting tidal elevations to a

Weibull distribution. This technique is an approximation only and is intended for use
only when harmonic constituents of tide are not available.
Figure 30 Monte Carlo: Input Specifications for Tidal level
7.1.3 Surge heights

Prior to analysis the surge signal must be separated into positive and negative surges, and
the fit of each to a Weibull distribution then optimised and described in terms of the
scale, shape and location parameter. During this process the relative proportions of up-
and down-surges in the database is identified.
One of five surge options can be chosen for a simulation:
1. Exclude surge
2. Random Normal: each crest is assigned a surge randomly chosen from a normal
distribution with a mean value which is assigned to each sea state at random
(user defined variance and maximum mean value)
33
3. Linked Normal: each crest is assigned a surge randomly chosen from a normal
distribution with a mean value which is assigned to each sea state using
probability of exceedence which is linked to that used in calculating the Hs for
the sea state (user defined variance and maximum mean value)
4. Random Weibull: for each sea state a surge is chosen at random from (user
input) Weibull distributions (one for positive and one for negative surges) of
surges and all crests in the sea state are assigned the same surge
5. Linked Weibull: the probability of exceedence which was randomly chosen to
determine the Hs is linked to the probability of exceedence of the surge (chosen
from a user input Weibull distribution). All crests in the sea state are assigned
the same surge.
Figure 31 Monte Carlo: Input Specifications for Surge
The scaling factor accounts for the partial correlation between waves and surge [see
Section 6]. It is empirically derived by evaluating the Monte Carlo output against a
suitable measured data set that includes measured EWL. The Monte Carlo simulation is
run for the duration of the measured data using a range of scaling factors. The results are
then plotted (as shown in the following figure) to reveal the scaling factor that results in
the most accurate output value of EWL. For the data set illustrated in the following
figure the measured EWL in 20 years was 10.5m; the figure reveals that a scaling factor
in the range 0.92 to 0.95 is appropriate to this data set. Where a range is applicable it
may be prudent to select the most conservative option, hence in this case, 0.95. If
measured EWL is not available for the data set being processed, then the scaling factor
maybe derived for a nearby alternative.
34
12.5
12.0
20 year EWL [m]

11.5
11.0
10.5
10.0
1.000 0.950 0.900 0.850 0.800 0.750 0.700 0.650 0.600 0.550 0.500
Scaling Factor
Figure 32 Monte Carlo: Identification of the Surge ‘Scaling factor’
7.2 MONTE CARLO OUTPUT STATISTICS

For each sea state, statistics are calculated and recorded as follows (assuming that the
user has chosen to include tide and surge):
• Maximum crest height in sea state

• Maximum crest+surge in sea state
• Surge associated with the maximum crest+surge in sea state
• Maximum crest+tide in the sea state
• Tide associated with the maximum crest+tide in sea state
• Maximum crest+tide+surge in the sea state
• Surge associated with the maximum crest+tide+surge in sea state
• Tide associated with the maximum crest+tide+surge in sea state
• Maximum surge in sea state
• Maximum tide in sea state
• Maximum tide+surge in sea state
The statistics are then stored in four 2d frequency distribution tables (Hs and crest, Hs
and tide, Hs and surge, Hs and tide+surge). The statistics can be plotted and exported
upon completion of the simulation.
The simulation is performed over a number of years and is repeated many times to gather
statistics. For example, a 100 year simulation may be repeated 1,000 times. For each
100 year interval, the maximum Hs, crest, crest+tide, crest+surge, crest+tide+surge are
recorded and, after the 1,000 iterations are complete, the statistics for the 100 year
extreme values of these parameters are entered into a probability of exceedence table.
The 100 year values can then be read off at the 37% exceedence level if the historical
interpretation of design value is required – of course other exceedence levels are now
available. Plots of the probability of exceedence and probability of occurrence are also
presented and can be exported upon completion of the simulation.
Further details of the software testing can be found in Section 10 at the end of this report.
35
8. RESULTS
Simulations of 100 and 10,000 years were performed for each of the three locations.
Each simulation was repeated 600 times to acquire reliable statistics. The simulation
inputs for the distribution of Hs and surge were found by fitting a Weibull 3-parameter
distribution to the measured Hs and surge data from each location (the data was sub-
sampled to 3 hours so that the number of surge entries matched the number of Hs
entries).
‘Worst case’ scenario simulations were run whereby the Hs and surge were linked such
that the surge that accompanied each Hs had the same probability of exceedence as the
Hs – this is equivalent to combining the n-year Hs with the n-year surge, for example..
The results from these simulations were compared with existing extreme water level
methods.
8.1 100 YEAR RETURN PERIOD EXTREME WATER LEVELS
The following table summarises the results at the 100 year return period.
Table 4 Comparison of 100 year Extreme Water Levels constituents calculated

using various techniques
100 Year Crest [m] Associated SWL [m]
Method K13 Euro Auk K13 Euro Auk
Krogstad ISO-compliant 10.71 10.61 16.35 - - -
TVM ISO-compliant 10.73 10.13 19.11 -1.34 -0.62 -5.09
Shell Method 2 10.56 9.35 15.40 1.78 1.83 1.56
Guidance Notes 9.20 8.14 14.47 2.59 3.05 1.94
MC Sim Linked Surge 10.57 9.52 16.21 2.74 2.86 2.82
MC Sim Random Surge 10.54 9.61 16.15 0.64 0.63 0.15
Notes:
• The Krogstad/Borgman approach does not include SWL.

• The TVM method used here employs storm thresholds as follows:
o Auk - 5m
o K13 - 3 m
o Euro - 3 m
.
• For TVM, all Hmp values were used in the fit without binning. Work is ongoing
to establish the most applicable technique for using the TVM. In the above
cases, the predicted total water level is less than the predicted extreme crest,
indicating a negative associated SWL and also indicating that the results from
TVM method using these data sets may not be reliable.
• The ‘Shell method 2’, described in the reference in Footnote 9, scales the 100
year Hs and adds a 1 year surge for Auk and a 3 year surge for Euro and K13.
No tide is added.
36
• The method described in footnote 2, commonly referred to within the oil and gas
industry as the ‘Guidance Notes’ method is found by scaling the 50 year Hs and
using SWL values from the guidance notes using nearest standard port data
• The MC simulations used either a random or linked surge from the Weibull
distributions and a tide value equal to the largest tide from each measured data
set.
The 100 year extreme water levels, relative to mean sea level, are as follows:
Table 5 Summary of 100 year Extreme Water Levels found using different
techniques
100 Year Total Water Level [m]
Method K13 Euro Auk
TVM ISO-compliant 9.39 9.51 14.02
Shell Method 2 12.34 11.18 16.96
Guidance Notes 11.79 11.19 16.41
MC Sim Linked Surge 13.31 12.38 19.03
MC Sim Random Surge 11.18 10.24 16.30
8.2 10,000 YEAR RETURN PERIOD EXTREME WATER LEVELS
The following table summarises the results at the 10,000 year return period.
Table 6 Comparison of 100 year Extreme Water Levels constituents calculated

using various techniques
10,000 Year Crest [m] Associated SWL [m]

Method K13 Euro Auk K13 Euro Auk
Krogstad ISO-compliant 13.96 14.38 21.36 - - -
TVM ISO-compliant 14.66 14.24 28.78 -3.31 -2.65 -11.34
Shell Method 2 14.21 12.58 19.88 1.78 1.83 1.56
Guidance Notes 12.10 10.71 19.03 2.92 3.44 2.19
Shell Method 1 11.48 10.69 17.63 2.00 2.29 2.01
MC Sim Linked Surge 13.92 12.75 21.36 3.50 3.70 4.18
MC Sim Random Surge 13.88 12.79 21.16 0.52 0.56 0.34
37
Notes:
• See the notes below Table 5.
• The ‘Shell Method 1’, sometimes referred to as the ‘Interim Method’ is only
defined for 10,000 year return periods and involves combining the 1,000 year
steep sea crest with a ¾ tide and a 1 year surge.
• The TVM associated SWL are again negative and, particularly at Auk, do not
look realistic.
The 10,000 year extreme water levels, relative to mean sea level, are as follows:
Table 7 Summary of 10,000 year Extreme Water Levels found using different
techniques
10,000 Year Total Water Level [m]
Method K13 Euro Auk
TVM ISO-compliant 11.35 11.59 17.44
Shell Method 2 15.99 14.41 21.44
Guidance Notes 15.02 14.15 21.22
Shell Method 1 13.48 12.98 19.64
MC Sim Linked Surge 17.42 16.43 25.67
MC Sim Random Surge 14.40 13.35 21.50
8.3 COMMENTS ON 100 AND 1,000 YEAR EXTREME WATER LEVELS
The data sets do not appear to provide reliable results when analysed using the TVM
method. Results from ongoing work may reveal procedures to refine the parameters
chosen within the TVM analysis. Leaving the TVM analysis to one side, all remaining
methods are within the bounds of the most and least conservative simulations – apart
from the Shell Method 1 which is significantly less than the other methods. Note that the
tide and surge values for the Guidance Notes method are extracted from contour maps
whose resolution is insufficient to cope with the rapidly changing conditions
approaching the coast.
The extreme water level methods imply some correlation between the wave height and
the surge although this correlation is not a strong as is used in the simulations which
enforce the Hs and surge to occur with the same probability of exceedence.
While the above comparisons are useful in that they compare the extreme water level
methods with theoretical most and least conservative extremes, they do not allow for the
comparison of the methods with the true extreme water level. In an effort to address
this, the following section describes work to establish the ‘true’ extreme water level
using measured data sets.
38
8.4 MEASURED EXTREME WATER LEVELS
Note: the method described below to convert a measured data set of Hs, tide and surge
into a data set of crest, tide and surge is not as desirable as having a measured data set
of crest, tide and surge. However, given that such a data set does not exist, it is
necessary to create the data set manually. This should be remembered when reading the
next section.
The total water level (crest + tide + surge) for each sea state is rarely measured and long
duration data sets of such measurements are not readily available. However, it is
possible to engineer the total water level by using the Hs value for each sea state together
with a short term crest distribution. The following steps were carried out to convert the
measured data at K13, Euro platform and Auk into total water level data sets:
• For each Hs
o Calculate Tz based on a steepness of 1:16
o Calculate the number of waves, n, in the sea state
o Randomly select n crests from a Forristall 3D distribution for the sea
state
o Select the largest crest
o Add the largest crest to the tide and surge to create the total water level
for each sea state
The data set of total water level can then be compared with the total water level
predicted via the different techniques used in Section 8 to establish which is the most
faithful prediction of the true extreme water level.
Using the longest data, at K13, provides 20 years of total water level ‘measurements’
based on the Forristall 3D crest from each sea state added to the still water level. The
maximum total water level in the 20 year data set is 10.50 m. Extrapolating the total
water level measurements to various return periods results in the following table:
Table 8 Summary of extrapolated total water levels at K13
Return Period [years] Extrapolated Total Water Level

[m, rel MSL]
20 [duration of measured data] 10.47
100 11.60
10,000 14.65
Table 8 shows that for the duration of the data set, the extrapolated total water level
(found by fitting a Weibull distribution to the total water level records) is very close to
the maximum recorded in the data set. The very small difference arises due to the fact
that the extrapolated values are based on a best fit to the data.
39
8.5 CALIBRATING THE SIMULATION
To calibrate the Monte Carlo simulations against the measured data, the scaling and
offset factors were varied. However, the sum of the scaling and offset factors was
always maintained at 1.0 to ensure that the possibility of the largest crest occurring
coincident with the largest surge was not excluded. The 20 year extreme water level
from the simulations was then compared with the measured 20 year extreme water level,
10.5 m (relative to MSL). The following figure shows the variation of the 20 year Monte
Carlo simulation extreme water level for various different scaling factors [note that the
offset factor will be equal to 1.0 – scaling factor]. A value of 0.95 was chosen as a
scaling factor – with the offset factor equal to 0.05.
12.5
12.0
20 year EWL [m]
11.5
11.0
10.5
10.0
1.000 0.950 0.900 0.850 0.800 0.750 0.700 0.650 0.600 0.550 0.500
Scaling Factor
Figure 33 Variation of 20 year EWL with decreasing scaling factor. The red
dashed line represents the measured 20 year extreme water level at K13.
Simulations of 100 and 10,000 years were performed at the K13 locations (each
simulation was repeated 600 times to acquire reliable statistics). The simulation inputs
for the distribution of Hs and surge were found by independently fitting a Weibull 3
parameter distribution to the measured Hs and surge data (the data was sub-sampled to 3
hours so that the number of surge and tide entries matched the number of Hs entries).
The tidal harmonics were found by performing a harmonic analysis of the tidal data. The
scale and offset factors were set to 0.95 and 0.05 respectively. A summary of the input
parameters is given in the table below.
Table 9 Summary of Monte Carlo Simulation inputs at K13

Weibull Parameters Shape Scale Location
Hs 1.3518 1.3266 0.2419
Surge 0.9247 0.1608 0.0326
Harmonic Coefficients Harmonic Phase [deg] Amp [m]

M2 208 0.6
S2 266 0.2
N2 190 0.1
K1 342 0.1
O1 183 0.1
40
8.6 100 AND 10,000 YEAR RETURN PERIOD EXTREME WATER LEVELS
The following table summarises the crest and SWL results at the 100 and 10,000 year
return period.
Table 10 Comparison of 100 and 10,000 year return periods at K13

100year EWL 10,000 year EWL
Method
TVM ISO-compliant 9.39 11.35
Shell Method 2 12.34 15.99
Guidance Notes 11.79 15.02
Shell Method 1 --- 13.48
MC Sim Linked Surge 11.86 15.57
Extrapolation of measured data 11.60 14.65
ISO 19902 ‘Associated’ --- 17.17
ISO 19902 ‘Independent’ --- 14.87
Notes:
• The ISO19902 ‘Independent’ case is defined as
h = √(a 2 + s 2 + t 2) + f
and the ISO19902 ‘Associated’ case is defined as
h = √((a + s) 2 + t 2) + f
where:
a is the abnormal wave crest height [10,000 year Crest]
s is the extreme storm surge [100 year Surge]
t is the maximum elevation of the tide relative to mean sea level [HAT]
f is the expected sum of subsidence, settlement and sea level rise over the
design service life of the structure [set to zero in this example]
• The ISO19902 methods are only defined for the 10,000 year return period
Table 10 shows that for both the 100 and 10,000 year return periods, the extrapolated
measured data is in good agreement with the Monte Carlo simulations although the
Monte Carlo results are slightly higher than the extrapolated measured data results at
both return periods. The TVM method predicts extreme values which are well outside
range of all other techniques which leads to it being considered less applicable. Results
from ongoing work may reveal procedures to refine the parameters chosen within the
TVM analysis and enhance its applicability. At the 100 year return period, the results,
excluding the TVM analysis, are within 75 cm of each other. At the 10,000 year return
period, again excluding the TVM analysis, the results range from the Shell Method 1
41
[13.48 m] to the ISO Associated Method [17.17 m], with most falling within one metre
of 15 m. Given the very different approaches to the various techniques, it is remarkable
that so many [Shell, Guidance Notes, MC Sims, Measured data] are in reasonable
agreement, even at this extremely long return period.
The ISO-19902 methods combine the crest, tide and surge in two ways: one which
ignores any possibility of the extreme crest and abnormal surge occurring together and
the other which stipulates that they must occur together. Thus, it provides a best and
worst case scenario for a given location. It is reassuring to see that, ignoring the Shell
Method 1 and TVM methods, the best and worst case ISO-19902 methods provide an
upper and lower bound into which all of the predicted 10,000 year EWL from all the
various techniques fall. Again this is noteworthy given that each approach uses unique
methods to estimate the 10,000 year extreme water level.
42
9. CONCLUSIONS
This work has examined the individual components of total water level, namely the wave
height, tide and surge. The relationship between the components was explored and
Monte Carlo simulations method proposed to combine the individual components, taking
account of their joint probability. Calibration of the Monte Carlo simulations was used
to ensure that the joint probability was tailored to the location. The predicted total water
level for the 100 and 10,000 year return periods, as found by various methods, were
compared.
The following is a bullet point list of the key observations:
• Monte Carlo simulations can be used to create synthetic time series of total
water level but they must be calibrated for sites with different sea conditions.
They can be used to establish a range of results derived from modeling the
interactions between the total water level components; these results range from
total dependence of surge on wave height to complete independence of the two
parameters.
• Crest is the overwhelming component of Extreme Water Level
• Surges last longer than tides so surge can be large throughout a tidal cycle, but
the peak of the surge tends to occur at times other than high tide
• Measured data shows a link between wave and surge – dependent on location
The following is a bullet point list of the further observations:
• Tides remain at their extremes for very little of the time

• Tides moderate surge heights – the way this moderation takes place is dependent
on location
• Largest Hs is NOT found at largest surge or largest tide.
• Largest SWL is NOT made up of largest surge or largest tide
• The n-year Crest is NOT associated with n-year surge
Extreme Water Level and joint probability are interesting and challenging topics for
research. Many factors control the development of an extreme water level and each
plays an important role. The following is a bullet point list of key topics that may be
interesting topics for future work. The list is not exhaustive.
• Tide interaction with other parameters: further exploration and understanding

of the physical processes that control the tide interaction with other parameters.
• Measured Total Water Level data: few data sets exist which contain measured
crest, tide and surge. It would be worthwhile to locate further data sets which do
have the three total water level components and use them as a direct comparison
with existing EWL techniques and Monte Carlo simulations.
• Mapping: a map of contours of 100 year Hs exists for the water around the UK.
It would be interesting to produce a similar map of 100 and 10,000 year extreme
water levels, as calculated by various different techniques, to allow for a quick
comparison of the severity of different sites.
• Shallow water sites: the range of water depths in the North Sea is large. The
applicability of the usual extreme water level techniques may be limited very
close to the shore where the water is very shallow. It would be interesting to
explore this region in more detail to observe if any changes are brought about by
the water depth or proximity to shore.
• ISO 19902:2007: it would be an interesting task to compare the results of both
the Independent and Associated forms of the total water level equation in ISO
19902 with the existing extreme water levels at all sites across the North Sea.
43
10. APPENDIX: SOFTWARE TESTING
10.1 SOFTWARE TESTING

First, simulations were run to ensure that the software was performing correctly. 100
years of data were simulated 1,000 times. A cutoff was employed and any sea states
whose Hs values were below the cutoff were not processed. As the outputs from the
simulation are the extreme values, the cutoff does not affect the results so long as the
cutoff is not too high. For the central North Sea, a cutoff of 7 m was used in the 100 year
simulations. Simulation inputs are detailed in the next section.
10.1.1 Input Data

The simulation was tested against a measured 25 year data set from the central North
sea. The data set was chosen because it provided a source of data which is not otherwise
used in this report – thus it provides ‘external’ verification of the simulation software.
The data set comprises 25 years of measured significant wave height which was fitted
using a 3 parameter Weibull distribution with the following parameters:
Shape: 1.1915
Scale: 1.5727
Location: 0.5741 100 year Hs: 13.14 m
The surge and tide distributions are again based on measured data from the central North
Sea and are fitted as follows:
Normal Surge Weibull Surge:

Maximum central value: 0.70 Shape: 1.543
Variance: 0.25 Scale: 0.214
Location -0.034 100 year surge: 1.03 m
Tide:
Harmonic Amplitude [m] Phase [deg]
M2 0.7 218
S2 0.2 357
N2 0.1 34
K1 0.1 197
O1 0.1 287
44
10.1.2 Hs
The results below show that the software is correctly predicting the 100 year Hs values
and the distribution of Hs and crest heights is realistic. The simulated 100 year Hs value
is 13.15 m.
1
Simulated Hs
Theoretical [Weibull]
0.9
Simulated Crest
0.8
0.7
Probability Non Exceedence
0.6
0.5
0.4
0.3
0.2
0.1
0
10 12 14 16 18 20 22 24
100 Year value [m]
Figure 34 Comparison between simulated and theoretical 100 year Hs values
10.1.3 Surges
The different options for including surges in the simulation were tested and the
distribution of surges associated with the largest crest+surge for each sea state were
plotted. The results are shown in the following figures and confirm that the surge
distributions are being implemented correctly. Note that only statistics for Hs values
above the cutoff are shown in the following plots and that the striped effect on the plots
is simply due to the resolution of the images and is not a feature of the underlying data.
10.1.3.1 Random normal Surges
Figure 35 shows the results of using a surge which is normally distributed around a mean
value which is chosen, at random, for each sea state. The maximum mean value is set at
0.7 m with the variance of the normal distribution set at 0.25.
45
Figure 35 Distribution of surges associated with the maximum crest+surge.
100 year simulation repeated 1,000 times. Random normal surge with the mean of the normal distribution
ranging from -0.7 and 0.7 m
10.1.3.2 Linked normal

Figure 35 shows the results of using a surge which is normally distributed around a mean
value for each sea state which is chosen using a probability of exceedence that is linked
to the probability of exceedence of the Hs for the sea state. The maximum mean value is
set at 0.7 m with the variance of the normal distribution set at 0.25.

100 year simulation repeated 1,000 times. Normal surge with the mean of the normal distribution ranging from
-0.7 and 0.7 m and being selected with a probability of exceedence linked to the probability of exceedence of
Hs for each sea state
46
10.1.3.3 Random Weibull
Figure 36 shows the results of using a surge randomly chosen from a Weibull
distribution – parameters given in Section 10.1.1. The surge is the same for every crest
in a sea state and the surge is chosen (at random) for each sea state.
Figure 37 Distribution of surges associated with the maximum crest+surge

100 year simulation repeated 1,000 times. Surge selected at random from a Weibull distribution
10.1.3.4 Linked weibull

Figure 37 shows the results of using a surge chosen from a Weibull distribution –
parameters given in Section 10.1.1. The surge is the same for every crest in a sea state
and the surge is chosen for each sea state using a probability of exceedence that is linked
to the probability of exceedence of the Hs for the sea state.

100 year simulation repeated 1,000 times. Surge selected from a Weibull distribution with a probability of
exceedence linked to the probability of exceedence of Hs for each sea state
47
10.1.4 Tides
As discussed in Section 7.1.2 the tide is modelled as a combination of harmonic
constituents [up to 62 different harmonics] each having a user-defined phase and
amplitude. The first crest in a sea-state is assigned a random tide and all subsequent
crests in the sea state have their “true” tide, based on the time difference from the first
tide in the sea state. The results of using a tide as described in Section 10.1.1 is shown in
Figure 39. There is no discernable pattern to the tidal levels at for increasing values of
Hs – in fact, as was the case in Table 1, the tidal levels associated with the extreme sea
states are all much lower than the maximum possible tidal level.
Figure 39 Distribution of tides associated with the maximum crest+tide.

100 year simulation repeated 1,000 times. For details of tidal harmonics see Section 7.1.2
10.1.5 Still water level

The still water level is the sum of the tide and the surge. The tide can be combined with
either a linked or random surge selected from a Normal or Weibull distribution. Each
case is discussed below. In the following section, the “standard tide” refers to a tide
described by the harmonic constituents defined in Section 7.1.2.
10.1.5.1 SWL using random Weibull surge
The still water level (tide + surge) associated with the maximum crest+tide+surge for
each sea state is plotted in Figure 39 using a standard tide and a surge selected at random
from a Weibull distribution (100 year surge = 1.03 m).
48
Figure 40 Distribution of SWL associated with the maximum crest+tide+surge.
100 year simulation repeated 1,000 times. Standard tide and random Weibull surge. See text for further details
10.1.5.2 SWL using linked Weibull surge
each sea state is plotted in Figure 40 using a standard tide and a surge selected from a
Weibull distribution (100 year surge = 1.03 m) with a probability of exceedence linked
to the probability of exceedence of the Hs for each sea state.

100 year simulation repeated 1,000 times. Standard tide and linked Weibull surge See text for further details
49
10.1.5.3 SWL using random normal surge
each sea state is plotted in Figure 41 using a standard tide and a random normal surge
distribution (maximum mean value = 0.7 m, variance = 0.25).

100 year simulation repeated 1,000 times. Standard tide and random Normal surge See text for further details
10.1.5.4 SWL using linked normal surge
each sea state is plotted in Figure 42 using a standard tide and a surge chosen from a
normal distribution (maximum mean value = 0.7 m, variance = 0.25) with a probability
of exceedence linked to the probability of exceedence of the Hs for each sea state.

100 year simulation repeated 1,000 times. Standard tide and linked Normal surge See text for further details
50
10.1.6 Summary
Max crest + “anything” is dominated by the crest – which often occurs in a sea state
OTHER than the one with the largest Hs. The associated “anything” is less than its
largest possible value for the extreme crest + “anything”. Even when the surge is forced
to occur with a similar probability of exceedence as the Hs for a given sea state, the
SWL associated with the extreme crest+tide+surge is much smaller than its largest
possible value.
51
Published by the Health and Safety Executive 10/09
Health and Safety
Executive

Extreme Water Level is the total level of the
water arising from wave crest, storm surge and
tidal level acting in combination with each other
throughout an extreme event. Individually, each of
these components is relatively well understood.
However, the combination of these components is
not straightforward and depends on the conditions
of the location in question. This study addresses
the behaviour of the individual components, their
interaction, and their joint probability, leading to the
development of a Monte Carlo based approach to
the estimation of Extreme Water Level. The study
concludes with a review of latest ISO guidance on
estimating extreme water levels and a comparison
between extreme water level results derived using
existing methods, ISO guidance and Monte Carlo
simulations.
This report and the work it describes were funded

by the Health and Safety Executive (HSE). Its
contents, including any opinions and/or conclusions
expressed, are those of the author alone and do not
necessarily reflect HSE policy.
RR740
www.hse.gov.uk

HSE Rr740 MonteCarlo

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HSE Rr740 MonteCarlo

Uploaded by

Copyright:

Available Formats

Health and Safety

A Monte Carlo approach to joint

A Monte Carlo approach to joint

First published 2009

All rights reserved. No part of this publication may be

Applications for reproduction should be made in writing to:

EXECUTIVE SUMMARY ................................................................................... 7

EWL is defined as a combination of three components;

EWL50 = (M2 + S2) + S+50 + Hcrest50

This approach might be described as unsophisticated because:

EWL100 = (M2 + S2) + S+50 + Hcrest100

In this case the implicit assumption was:

SWL100 = (M2 + S2) + S+50

As a result of significantly greater temporal and spatial resolution in acquired oceanographic

For deep and intermediate water depths ‘a’ can be approximated to

a > 1.3 a100

a > a100 + 1.5m

The aim of this work in this report is therefore to:

2.1 MEASURED DATA

Name: Auk platform

Figure 1 Location of measured data sources used in this study

2.2 MODEL DATA

Auk K13 Euro

Maximum Hs 12.73 7.98 6.64 11.97 8.64 7.39

Maximum Surge 2.13 1.98 2.79 1.46 2.04 2.18

Maximum SWL 2.20 2.51 2.65 1.66 3.19 2.58

Maximum Tide 0.90 1.26 1.36 0.58 1.85 0.89

Notes: Levels and amplitudes are specified in metres

It is worth noting that without exception:

These observations will be further discussed in the following sections.

ISO 19901-1 states in Section 8.5:

3.1 KROGSTAD/BORGMAN INTEGRAL

• Shape, scale and location parameter of the 3-parameter Weibull Distribution.

Figure 2 Exceedence Distribution of Extreme Crest Elevation

Table 2 Comparison of Extreme Crest Elevation – Borgman & Monte Carlo

10,000 Year Crest Elevation (m)

3.2 TROMANS & VANDERSCHUREN METHOD

Figure 3 Illustration of mean excess plot

­1.5 ­1.0 ­0.5 0.0 0.5 1.0 1.5

Average Tidal Level [cm]

Figure 5 K13 measured: Probability of occurrence of maximum surge and

NEXTRA/DHI K13a [ 8,031 surge peaks 11,260 Hs peaks ]

Figure 6 K13 Model: Probability of occurrence of maximum surge and maximum

Average Tidal Level [cm]

Figure 7 Euro Measured: Probability of occurrence of maximum surge and

NEXTRA/DHI EuroPlatform [ 6,961 surge peaks, 9,152 Hs peaks ]

Figure 8 Euro model: Probability of occurrence of maximum surge and maximum

Average Tidal Level [cm]

Figure 9 Auk Measured: Probability of occurrence of maximum surge and

NEXTRA/DHI Auk [ 8,440 surge peaks, 11,392 Hs peaks ]

Figure 10 Auk Model: Probability of occurrence of maximum surge and

Average Tidal Level (cm)

NEXTRA/DHI Auk [ 8,440 surge peaks, 11,392 Hs peaks ]

Average Tidal Level (cm)

NEXTRA/DHI K13a [ 8,031 surge peaks 11,260 Hs peaks ]

Average Tidal Level (cm)

NEXTRA/DHI EuroPlatform [ 6,961 surge peaks, 9,152 Hs peaks ]

5.1 SURGE DURATION

Figure 17 Auk Measured: Surge duration analysis

Auk modelled (31 years) percentage frequency

Figure 18 Auk Model: Surge duration analysis

Figure 19 K13 Measured: Surge duration analysis

K13 Modelled (31 years) percentage frequency

Figure 21 Euro Measured: Surge duration analysis

1.5 1.0 0.5 0.0 0.5 1.0 1.5