Dunboyne Combined Residents Association
Home - About DCRA - Contact DCRA -Archive - News- Politicians Directory - Planning - Environment - Links Page - Newsletters - DCRA Meetings - Soapbox - Waste & Recycling
An Bord Pleanala Oral Hearing: 20/21.04.04
( Appeal Ref:PL 17.204690 P.A.Reg Ref DA/30236)
Presented by Douglas Dennehy
I will first deal with the subject of flood risk assessment.
It has become the custom to measure the risk of flooding by the peak flow in
the Tolka at Glasnevin.
It also appears that a peak flow having a frequency of once in 100 years has
become accepted as a suitable standard for the level of protection to be provided.
The once in one hundred year return flow given in the Final Report of the
River Tolka Study is 90 m3/sec. This figure is based on flawed analysis and
I will demonstrate that that a more realistic estimate will be seen to be of
the order of 130
Such a flow implies that the widely believed estimated is underestimated by
some 45%. If confirmed, this finding would have significant implications for
the proposed development Among other things, as one can no longer regard the
levels that were observed for a flood of 98 m3/s in 2002 to be valid for one
of 130 m3/s, it would imply that there can be no reliable indication as to the
extent of the flood plain in the Castle Estate. It further implies that there
is no reliable estimate of appropriate flood levels. It follows that it is impossible
at present to judge the effectiveness of the proposed alleviation measures.
In the wider context it implies that current and proposed policies for the future
development in the Tolka catchment is also similarly flawed.
In contrast to the deplorable uncertainty regarding all the matters relating
to flooding in the Tolka catchment, there can be no doubt whatsoever as to the
inappropriateness of allowing further development in the Castle Estate, or anywhere
else in the vicinity of the Castle River, for the foreseeable future
I will now provide the evidence for this assertion.
The data available for analysis is taken from the appendices of the River Tolka
Flooding Study, Technical Report No.2 - River Modelling Report. The data refers
to the outflow from the (140 square kilometres) Tolka catchment measured at
Drumcondra, and consists of a total of 41 records of peak annual flows for a
period of 122 years from 1880. Of these only the 34 records from 1954 can be
regarded as reasonably reliable.
The consultants analysed the data in three sets (1) The Full Recorded Series
comprising the 34 most reliable Maximum Annual (MA) flows records, (2) the Peak
Over Threshold (POT) Series based on the 15 observations that exceeded a threshold
of 35m3/s and (3) an Expanded Series of 122 obtained by estimating the 76 missing
observations from the MA series. In each case the data were fitted to two commonly
used Extreme Values distribution functions, namely EV 1 also known as the Gumbel
and EV 2. The first has two parameters, the mean and standard deviation. The
second, EV 2 has an additional parameter (k) that modifies the shape.
The process of fitting involves estimating the parameters and this should be
done using an efficient statistical method. Two suitable methods are (1) by
using probability weighted moments (PWM) or (2) by maximum likelihood estimation
(MLE). Generally MLE is regarded as the "gold standard" and is normally
used by statisticians. An essential part of the process involves testing the
'goodness of fit' in order to measure the accuracy of the final probability
estimates. Here, the approved method involved determining the standard errors
and establishing confidence intervals.
There are several causes of concern regarding the analysis used in the Technical
Report
To begin, an ad hoc graphical method was used to estimated the parameters of
the EV 1 distribution instead of either of the approved methods. Also there
is no evidence to show that any form of significance testing was done.
The parameters for the EV 2 should be found using MLE, instead a fixed value
of k equal to -0.05 was used in addition to the parameters already estimated
for the EV 1 analysis. This method is quite incorrect because the mean and standard
deviation of the EV 2 distribution depend on the value of k and have to be estimated
accordingly. The results are obviously undependable and should not be used.
In the literature, POT data is analysed using a Generalised Pareto distribution
and the parameters are estimated using MLE. The consultants use a probability
plotting method to fit a Gumbel distribution. This gave a value of -2.82 for
the mean. This estimate is negative and therefore infeasible. Notwithstanding,
the consultants merely switch the sign and proceed regardless. It follows that
the results of the POT analysis are unusable.
The analysis of the Expanded Annual Maximum Series is also totally untrustworthy.
In addition to the inaccurate method used for parameter estimation, a disproportionately
large number (76 out of a total of 122) of missing observations were estimated
in order to expand the series. This entailed using a number of random samples
based on assumptions regarding the distribution of the missing data. Consequently
the results of the analysis are likely to be confounded by the assumed distribution
of the missing data and are therefore also of dubious value.
This leaves the Full Recorded Series. Before discussing the analysis of this
I will read the abstract of a paper entitled " ON THE APPROPRIATENESS OF
THE GUMBEL DISTRIBUTION FOR MODELLING EXTREME RAINFALL " presented at the
proceedings of the European Science Foundation's Life and Environmental Science
Committee (ESF LESC) Exploratory Workshop held at Bologna, Italy, October 24-25
2003., by Demetris Koutsoyiannis of the Department of Water Resources, Faculty
of Civil Engineering, National Technical University of Athens. It reads
For half a century, the Gumbel distribution has been the prevailing model for
quantifying risk associated with extreme rainfall. Several arguments including
theoretical reasons and empirical evidence are supposed to support the appropriateness
of the Gumbel distribution. These arguments are examined thoroughly in this
work and are put into question. Moreover, it is shown that the Gumbel distribution
may misjudge the hydrological risk as it underestimates seriously the largest
extreme rainfall amounts. Besides, it is shown that the three-parameter extreme
value distribution of type II is a more consistent alternative and it is discussed
how this distribution can be applied even with short hydrological records.
The first few lines from the introduction reads:-
Almost a century after the empirical foundation of hydrological frequency curves
…..the estimation of hydrological extremes continues to be highly uncertain.
This has been vividly expressed by Klemeš (2000) who argues that
"... the increased mathematicisation of hydrological frequency analysis
over the past 50 years has not increased the validity of the estimates of frequencies
of high extremes and thus has not improved our ability to assess the safety
of structures whose design characteristics are based on them. The distribution
models used now, though disguised in rigorous mathematical garb, are no more,
and quite likely less, valid for estimating the probabilities of rare events
than were the extensions 'by eye' of duration curves employed 50 years ago."
A similar critique was made by Professor Jim Dooge in 1986 (Sometime an Irish
Minister for Foreign Affairs and recognised world wide for his pioneering work
on River Basin Modelling)
In the light of the above I made analyses of the 34 MA observations in Full
Recorded Series using approved statistical methods to fit (1) the Gumbel and
(2) the GEV i.e. the three parameter EV 2. The 100 year return flow were as
follows:-
For the Gumbel (i) using MLE 80 m3/sec
DISPLAY
For the GEV (i) using PWM 129 m3/sec
(iii) using MLE 133 m3/sec
DISPLAY
I will now quote a few lines from the conclusion Kutsoyiannis's paper
Interestingly, EV1 has been the prevailing model for rainfall extremes despite
the fact that it results in the highest possible risk for engineering structures,
i.e. it yields the smallest possible design rainfall values in comparison to
those of the three-parameter EV2 for any value of the shape parameter. The empirical
investigation of this study demonstrates that the underestimation of design
rainfall by the EV1 distribution is quite substantial (e.g. 1:2 that is 100%)
for large return periods and this fact must be considered as a warning against
the adoption of the EV1 distribution for rainfall extremes.
On the contrary, the three-parameter EV2 distribution (the GEV distribution
bounded from the left) does not have the theoretical and empirical disadvantages
of the EV1 distribution.
I now wish to discuss the matter of Flood Mitigation
There has been a radical change of River Basin Management and Flood Mitigation
policy throughout Europe, in recent times, from the unsustainable traditional
one, that results in 'flushing' the flood water down the river, to an alternative
one of retaining flood waters in the sparsely populated upland parts of a river
catchment area where it can do relatively little damage. Reference to this can
be found in
There is sufficient evidence to show that the proper completion of the Tolka
Study was disrupted almost immediately after the flooding on 14th/15th November
2002. One can deduce that a decision was made, for political expediency, to
embark on a program of "quick fix" remedial measures in spite of the
contrary advice of the consultants doing the study. This was expressed by Minister
Parlon in the Dail on 19th November 2002, when he stated that in the absence
of the full report of the study any interim works could exacerbate flooding
problems upstream or downstream and moreover could have disastrous consequences
in exceptional circumstances.
The Flooding in 2002 presented he consultants with an unexpected opportunity
to establish the maximum levels reached by the flood, by direct observation
without the need for any complex mathematical modelling. Also, being the largest
flood for the River Tolka on record, it was presumably regarded as a suitable
basis for determining the level of protection required in the areas most at
risk. It is easy to appreciate that, without being able to assess the implications
that would otherwise have been provided by a detailed analysis of the problems,
the decision by those concerned led to the automatic adoption of the traditional
policy of flood mitigation.
Consequently, the 58 remedial measures proposed by the consultants almost entirely
involve raising or building embankments and walls or making modifications to
the river channel. Such measures reduce the normal role of the flood plains
and confine the flood water to the river channel. This inevitably results in
accelerating it downstream towards the most vulnerable and most densely populated
areas.
The emergency measures for Dunboyne, including the river and flood plains in
the Castle Estate involved massive changes to the river profile over a distance
of more than two kilometers. These measures result in accelerating the floodwaters
downstream where the impact on the critical conditions at Clonee could not have
been properly assessed without the aid of catchment modelling.
There are no automatic rain gauges in the Tolka catchment. There are no river
gauges at the outlets of the three main tributaries: the Tolka River upstream
of Clonee, the Castle River at Dunboyne and the twin Pilkeens at Mulhuddard.
Without measured flows in these tributaries, reliable mathematical modelling
(based on the St Venant wave equations) of the river from Dunboyne to Glasnevin
is impossible. It follows that there are no reliable means of assessing the
impact of any of the changes that have been made. Moreover, until such time
as proper measuring devices are installed and calibrated the affect of the measures
already adopted must remain unknown.
What is more, the interim measures were implemented without a proper Environmental
Impact Study and as a consequence there is no reliable information in the public
domain as to the impact of these works downstream. It follows that a reliable
assessment of any further change cannot be assessed at present.
In the wider context. the policy of retention, recommended throughout Europe,
was almost a mandatory option in the case of the Tolka because the scope for
providing adequate protection in the most vulnerable highly populated city area
downstream of Glasnevin is severely limited. The peak flow recorded at Glasnevin
for the 2002 event was 98 m3/sec, that is some 25% above the design capacity
of the river. However, the construction of the N3 and associated ribbon development
in recent years has radically altered the character of the river from Dunboyne
to Finglas. This results in a drastic reduction of the amplitude of the flood
wave as it travels downstream. Under previous conditions the peak flow at Glasnevin
in 2002 might well have been as high as 110 m3/sec, that is about 33% above
the capacity of the river (before recent modifications). Obviously, substantial
damage to property and perhaps loss of life might easily result from adopting
the wrong policy of "flushing" in the upstream part of the river.
The recent flooding also clearly revealed a critical problem in the Clonee/Dunboyne
area resulting from the construction of the N3 at Clonee which acts as a dam
across the river downstream of the village. This obstruction causes the floodwaters,
in the upstream area, to rise well above the levels that would have pertained
before the building of the road. Consequently, the risk of flooding at Dunboyne
as well as at Clonee has been a greatly increased. In fact there is no doubt
that N3 was responsible for a substantial part of the damage caused in both
villages by the last two floods.
This serious hazard was clearly recognised by the consultants and the initial
remedial measures proposed for the area, included suggestions for diversion
channels designed to redirect a substantial part of the floodwater away from
the areas at risk. However this option has now been abandoned, presumably, because
it would have caused dangerous problems downstream.
The remaining measures do not correct the chronic underlying problem caused
by the road works. Instead they act purely as defenses comprising the building
of a wall to protect the village of Clonee and a series of embankments. These
are designed to protect property and to contain the flood waters in an extensive
area comprising a large part of the original flood plains between Dunboyne and
Clonee. It is evidently now intended to treat this area as an alleviation basin.
However, this option was dictated by the initial failure to adopt a proper policy
of retention and provide for temporary storage of the flood waters in suitable
locations in the catchment further upstream.
It cannot be denied that, in the present climate of uncertainty regarding flooding,
the consequences of ill-advised action in the Dunboyne area could have disastrous
consequences downstream.
To summarize what I have said:-
The risk of flooding has been seriously underestimated and is currently used
as a basis for the implementation of an unsustainable flood mitigation policy
This has created an alarming state of uncertainty, ignoring the unpredictable
affect of future climate change, regarding all crucial aspects of flooding needed
for sound decision taking .
In relation to the proposed development:-
· There is uncertainty regarding the actual the extent of the existing
floodplains.
· There is uncertainty regarding potential flood levels that are essential
for the design of alleviation measures
· There is uncertainty regarding the potential impact of the proposed
development downstream under unpredictable circumstances and the additional
avoidable risk imposed.