Examination of Witnesses (Questions 560
WEDNESDAY 12 DECEMBER 2007
Let us move on to the question of probability versus intensity.
One of the phenomena, again looking at it as a layman, is that
all of our flood protection work is done on a probabilistic basis,
preparing for 1:10, 1:30, 1:75 and so on but what this summer
seemed to suggest was that we had the amount of rain in these
probability possibilities visited in a very short space of time.
I think what I am trying to understand is: do we have to adapt
our thinking about the way that rainfall arrives? Are we going
to see more intense events? Should that, therefore, be built in
to the probability calculation, which in our earlier session for
example has an impact on the ranking order of flood protection
schemes? I am going to guess now. I do not know what 1:10 is but
if a 1:10 arrived in five minutes instead of 24 hours, at that
particular point there might be an awful lot of water to shift.
It is just trying to understand the relationship between the two
and do we need to change our predictive models to take this into
account. Perhaps you could give me a little synopsis or education
on that subject?
Mr Noyes: I will deal with how
we use probabilities in weather forecasts and I will hand over
to John about what that means in a climate change scenario. I
am sure colleagues on my left will also chip in. When we issue
a forecast for the coming days, we use probabilities increasingly.
That is particularly useful for exceptional weather events because
it is helpful to convey to the people who have to respondthe
emergency services, Environment Agency and so fortha degree
of confidence about what is going to happen. That could be snowfall,
high temperatures, severe rainfall events. That is slightly different
I think to where you are coming from in regard to return periods
and probabilities of whether this thing will be more frequent
in the future. I think it is useful just to clarify the difference.
They are both used in the context of these events.
Professor Mitchell: Just to clarify,
I think the question you are asking is slightly different from
the one I am going to answer, so please forgive me, but it is
relevant. In terms of weather forecasting, the way we deal with
uncertainty is to run a set of forecasts. The reason we need to
do that is that we know both our models are imperfect and our
initial data is imperfect. What we do is to run a set of forecasts
or we can do a set of forecasts for slightly different nascent
conditions, and that gives you a spread of outcomes. The usefulness
of that is, I think, two-fold: it identifies if there is a small
possibility of a very extreme event which will allow advance warning;
and also whether the spread is narrow or wide gives you an indication
of how confident you are in the forecast. Those are close to being
related to the actual probability of those outcomes happening.
It is a probabilities forecast in a very real sense. When it comes
to climate change, we have uncertainties in models and we have
no way of actually ranking those. What we tend to do is look,
as we have for example done in UKCIP, and Dr West will say more
about that, to using a variety of models and look at the spread
those models give. That is really just a frequency distribution
because we have not been able to assess which is a better model
and which is not a better model. The way we use probability in
climate change is different to the way we use probability in weather
prediction. I think the question you were asking is whether in
a 24-hour period it makes a difference if a given amount of rainfall
falls over 24 hours or it ends up in 20 minutes. In terms of the
model's ability to forecast in very short timescales, that will
certainly stretch the model. The time-step in the model is typically
20 minutes to an hour and so you could actually evaluate rainfall
totals, but I think you would have to question whether on that
timescale the model is accurate enough. We do routinely keep six-hour
totals, so we can look at a finer timescale. As you go to the
finer timescale, I suspect the accuracy becomes more suspect just
because of the shortcomings in the model and the fact that the
model resolution is limited. If you can indulge me just a bit
more, one of the issues that we have, I think, with the Environment
Agency is in looking at rivers, and some of those are in quite
small catchments. The forecast models at the moment are run at
12 kilometres and 4 kilometre resolution; 12 kilometres is much
too coarse for a lot of the catchments and for some catchments
a 4 kilometre scale is still not fine enough. We plan in the next
two years to run a 1.5 kilometre model over the UK, which would
allow for directly coupling in rainfall forecasts to the sorts
of models that Dr Marsh's unit produces. You could actually have
those combined rainfall and riverflow forecasts running in real
time. There is still a question of scale but one could then, hopefully
in the five years or so and after that is we get the enhancement
to computing, run ensembles at this very fine resolution which
would allow the Environment Agency to look at the risk of very
high rainfall in small catchments. In the future, there are things
that we can do better, both through better modelling and coupling
that modelling into the river routing models and the flow models
where the expertise is not in the Met Office but probably it would
have to be done on the operational side. There is the potential
in the future to bring those two things together. That would allow
us to look at shorter timescales because I guess the timescale
of Dr Marsh's model is hours. You could look in real time into
the future at what the potential river flow was.
Dr West: I think there is an issue
with return periods and probability in terms of the threshold
or the description of the event you are interested in. For a Boscastle-type
event, it is a matter of perhaps in less than an hour how much
rain falls in a small catchment. What I think we looked at last
summer was the combination of that heavy rainfall over a long
period, some months, and very intense rain. It is that combined
probability that we need, if you like, to quantify. Different
scales of event require different combinations of thresholds.
For temperature, a threshold may be important if the temperature
goes over a threshold for 20 minutes, or in terms of heat build-up
it may be over several days. One of the ways we anticipate that
users will interact with the next set of climate scenarios is
that we want them to be able to describe what is an important
threshold. For a seaside town, it may be sea level. For a flood
management, it may be the amount of rain over four days. When
they can define that threshold, and that requires them to look
back in time probably at what has happened in the past to consider
where they are vulnerable, then they can interact with these probabilistic
scenarios and interrogate them in a way that gives them an answer
to a question that is relevant to them. While not downplaying
the value of increasing the accuracy and the amount of information
in the climate scenarios, we feel it is very important that somehow
we train decision-makers to interrogate these more intelligently
than they perhaps do at the moment.
Do I interpolate from what you have said that you could have,
if you like, a double probability: you could put two bits of probability
togetherthe probability of a one in 75 year event or a
one in 100 year eventand say that that event occurs with,
if you like, a Boscastle intensity? In other words, you can say
that we will have one of these every 100 years but there is a
probability number that we can give to indicate whether it is
going to be intense or not. Hopefully, I am not muddling the two
things up too much there.
Dr West: I think the answer is
"yes". I think the probability is of your defined event,
and in defining the event you will have to talk about intensity.
If you are interested in how heavy the rainfall is at this instant,
we can probably tell you that, but for any flood situation, it
is how long it takes to get there.
When you define these one in some many year events, over what
time period is the rainfall defined or measured?
Dr West: I am going to ask the
Hadley people to help with that. If we look at last summer's event,
in terms of rainfall it is how much fell over a couple of months.
The thing that is still niggling at my mind is: if the events
measured out over a 24-hour period but what we actually had was
an event over two hours, it is still a 1 in 100 year event but
it occurred over two hours and that is the bit that has put so
much pressure on the existing systems to deal with the flood.
What I am trying to understand is whether that compaction of the
timescale of the 1 in 100 year event occurring is something which
can also be modelled with any degree of accuracy?
Dr West: I would suggest that
so much rain falling in 24 hours may be a one in five year event.
If it all falls in an hour, that is much more unlikely and therefore
has a much longer return period.
True, but if we come back to the fact that the reality was that
in these summer rains we had this very heavy concentration, when
you look at Sheffield, an area of low flood risk probability,
it was overwhelmed because a very unusual event occurred. Everybody
is now casting round saying, "Can we predict when such events
occur?" At the moment we have something that says a one in
so many year event. I am not quite clear over what time period
the rain is measured in terms of these. Professor Mitchell, you
are nodding like a man who could give me an answer.
Professor Mitchell: In terms of
statistics, I think one can look at shorter-term rainfalls in
observation. There are records of shorter-term variations than
a day. A day happens to be a convenient unit of time which people
use. I think the observational record exists for us to check on
that. In terms of weather prediction models and whether they can
predict on those shorter timescales, that is an interesting question.
High resolution models of the sort that we have used on occasions
recently, are certainly in a better position to do that than present-day
weather forecast models. Where I would question it is when one
starts looking at the models that we use for future climate, which
are at 25 kilometre resolution at best, is whether you are smoothing
out intense rainfall over the 25 kilometres squared, and so it
may tend to underestimate those totals.
Dr Calver, did you have a contribution to help me?
Dr Calver: With regard to recurrence
intervals in hydrology, we associate that with the time period
and so we would talk about the peak flow having a particular recurrence
interval or we would say that it is a rarity event based on a
30-day period, or something like that. The time is an integral
part of the number we come up with. There are one or two other
comments that may be helpful to add on to the meteorology. Hydrology
follows from meteorology but it also brings in other things like
the ground surface, the type of surface, the antecedent conditions
there, the topographic configuration and of course not least the
flood defences. So the recurrence interval of a rainfall event
when it becomes a hydrological event can be enhanced or reduced
by the nature of the surface, the condition of it. If it is very
wet, it may well be a higher recurrence interval event and vice
versa. What do we do if we do not do real-time forecasting? Some
of our research does go into the Environment Agency real-time
forecasting, but we do a lot of hydrology based on Met Office
and other output longer ahead into the future. We take the meteorology
of climate models and they can, if you like, be used to "drive"
hydrological models to derive possible future outcomes. In this
case, there is that difficulty that John mentioned earlier, the
difficulty of rainfall extremes. There is more accuracy expected
of models in the future than now. For example, at the moment Defra's
plus 20% allowance on peak flows for schemes over a 50-year scheme
lifespan is based on that sort of hydrological modelling, following
on from meteorological modelling. We can do it transiently (i.e.
our model gives a time series through time) or we can jump to
a future scenario for a certain date and we can do that in an
ensemble set of model runs as well. I think in this context what
is going to be a good analysis shortly to tell you about is river
flood frequencies under climate change when we have the downscaling
of UKCIP 08 scenarios
I am sorry; downscaling of what?
Dr Calver: It is to make the scale
of the outputs for the hydrological models so that we can take
into account some of the surface variability. Hydrology works
in a very variable domain in space and in time. The advantage
over what Defra is working from at the moment for its project
appraisal guidance will include that we maywhen I say "we"
I mean the hydrological communityand I stress may, be able
to make distinctions in different parts of the country. We will
include some sort of quantitative analysis errors. They may arise
from emission scenarios, from climate models, from hydrological
models, et cetera. We would probably call them uncertainties,
if you understand what I mean. What we also hope to do is compare
those flood frequencies in the future with what one would expect
anyhow from natural variability. An important point to make here
is that many errors in assigning recurrence intervals, flood frequencies,
et cetera, are not that different in percentage terms from some
which are predicted in climate change
Just to interpret this flow of future hydrological information
in terms of outputs, in other words I am a decision-maker and
I am salivating at the thought of the interrelationship between
hydrological models and meteorological information, and I am sensing
that I am going to get some better information about what might
happen on the ground in the future, what kinds of things can I
expect from this improvement?
Dr Calver: May I just say that
it is better in the sense that more things will have been taken
into account, but we often find when we quantify uncertainties
that that does not automatically mean that they are narrowed.
The sort of outcome you get is a time series of flow in various
rivers across the country. One could interpret that as what we
call a flood frequency curve and say the magnitude of the peak
flow in the river against how often on average you are likely
to get that flow. Hydrology and meteorological models are intimately
linked. They are done in different ways in a real time sense,
but please ask the Environment Agency for details in the longer
planning sense. There is also feedback from the hydrology of the
surface into meteorological models.
Dr West: I think it is helpful
to recognise that climate models deal with average values over
perhaps 25 kilometres squared and very often the decision-maker
is interested in that point, that time of day. What Ann referred
to as downscaling is using the current relationship between the
current climate and current weather on very short timescales,
on very local timescales, and saying, "We assume that that
relationship will be the same in the future". So you can
produce a simulation in the future from which you can say, "Yes,
if this river rises at this rate, this piece of land will be flooded",
but is an abstraction based on assuming that the relationship
between the weather and climate stays the same as it is now. So
that will be available, but it requires somebody with the sense
to take that information to hydrologists and convert climate information
on a grid square basis into real flow information.
Who is going to do that?
Dr Calver: There are various potential
people who can do that, whether in the public sector research
establishments or universities.
Is it something that should be done?
Dr Calver: I understand it will
be done, as far as I know.
Dr West: It will not be done as
part of what is described as UKCIP 08, which we will launch in
October, but it allows people who want that doneDefra flood
management, the Environment Agency, any consultantto take
that information and apply it to a particular situation. For example,
Leeds City Council might want to know about the local flow. Somebody
has to do that; it is not going to be provided for free as part
of UKCIP 08 but the tools to do that will be the best available.
What does UKCIP 08 stand for?
Dr West: Our programme is the
UK Climate Impacts Programme. Working with the Hadley Centre and
funded by Defra, we are producing in October a new set of future
climate scenarios for the UK which, as John described, for the
first time will have probabilistic type information. Instead of
saying, "That is the number at that time period in that place",
here is a range of numbers. What you do with those numbers requires
you to have some knowledge of what is important. If you just want
to report what the climate is at that time, you might pick a mid-value.
If you are very risk-averse, you might want to know what is the
level that might be exceeded at a 90% probability. Your attitude
to your own threshold will determine how you look at the data.
If you do not have that attitude to risk and knowledge about a
threshold, the information is still there but you will not be
making maximum use of it.
Q571 Patrick Hall:
If I understood what Dr Calver was talking about, it was about
river flood risk.
Dr Calver: It is, and I am totally
aware that there are other types of flood risk.
Q572 Patrick Hall:
I just want to ask about another type of flooding which we have
called earlier today flash flooding. Because of the likelihood
of greater and more often intense rainfall episodes due to climate
change, does it matter what the condition of the ground is in
terms of flash flood risk? I have heard people say that the ground
is so dry, it just ran off; and I have heard people say in other
circumstances that the ground was absolutely soaked so that it
had nowhere else to go and it just ran off. Does it matter what
the ground is like in terms of intensity and flash flood conditions?
Dr Calver: Yes. It is probably
the rainfall intensity coupled with the duration that is more
important but, yes, because otherwise a potential flood will not
happen. Things like the ground conditions are modifiers beyond
Q573 Patrick Hall:
But both have been cited as explanations as to why a particular
Dr Calver: I am sorry, the wet
and the dry. I think it is a bit like dry blotting paper that
is very difficult to blot but, once it gets a little bit wet,
it sucks up ink better. I am going to ignore the very dry conditions
and say that if it is very wet, any rain will flow off; if there
is capacity to take the water in, some of it will go into the
soil. I believe there is an analogous position to the very dry
blotting paper case, to do with soil pore water pressures.
Q574 Patrick Hall:
I think I can follow that.
Mr Noyes: If you happen to live
in a clay area, in the summer months if you have had a dry summer,
the clay will form a very hard surface, which is almost like concrete,
and so the water will run off quickly. Equally, if the clay is
saturated in the winter, then the same thing is going to occur.
There is an intermediate state between the two.
Q575 Patrick Hall:
That shows how complex these matters are.
Dr Calver: The science is the
spatial variability of what the rainfall is falling on and therefore
the hydrological response. Of course, when you come to urban areas,
there are other complexities and perhapsothers may wish
to comment differentlythese very intense rainfalls might
be most danger in urban environments.
Ms Yeates: It very much depends
on the urban environment: the height of the gutters, the shape
of them, the size of the sewers, how much has been paved over,
how much green space there is. There are people that do this sort
of modelling and this work looking at both what happens in the
current situation, particularly focusing on areas that have already
been affected where they have data, and then taking the climate
information and looking forward to what might happen in the future,
but it is very much dependent on the site and what the area is
actually like in the first place., They have examples where what
currently is a road or somebody's backyard suddenly becomes a
very fast-flowing river during the rainfall period. If the rain
stops a couple of hours later, it just becomes a pile of mud which
is what was left after that rainfall.
Q576 Paddy Tipping:
What are people doing about it?
Ms Yeates: The Pennine Water Group
has I think moved to Sheffield University now.
They are one of the key researchers in this area. I believe Richard
Ashley is still leader of that group.
Q577 Paddy Tipping:
So they are doing work on the natural environment and the interaction
with the built environment?
Ms Yeates: They are very much
focused on the built environment and looking at what happens when
the rain falls on to that built environment, recognising that
if you have a road, obviously you are not going to get rainfall
soaking into soil but you are going to get it running off into
gutters, and looking at those sorts of aspects of it. Instead
of saying, "We have a river basin and we have these areas
to soak up water", they have different types of areas to
soak up water and they are looking at how to manage that instead.
Mr Drew: I am sorry I missed the early
part of your comments. Can I look into the world of science fiction?
There is something that I wanted to ask you. Is it complete science
fiction at the moment actually to believe that one day if we were
to get prognosis of a real threat to an area then we could either
move that weather front in some way or possibly dilute the water?
There would have been no problem if the rainfall that we had in
Gloucestershire had either been slightly moved off over areas
where there was not the same level of population or, more particularly,
diluted so that the rainfall came down over a greater area.
Chairman: That is climate modification.
Q578 Mr Drew:
That is a good term.
Professor Mitchell: Climate modification
or weather modification: I am quite sceptical about our ability
to do that. I have just come back from China where I was told
that there is something like 20,000 people employed to modify
weather. They shoot something like silver iodide into clouds to
modify the cloud structure. I think, even if we could do it, it
is a very risky procedure because if you get it wrong
Q579 Mr Drew:
If you get it wrong, you are sacked.
Professor Mitchell: If you move
it from Gloucestershire and it lands in Warwickshire, you can
imagine all sorts of lawsuits occurring. Scientifically, I think
the jury is still out. Why the Chinese invest so much in it, I
am not sure. They obviously believe it works, but I think other
countries are much more sceptical.
13 Note by witness: The Pennine Water Group (PWG)
is an EPSRC funded Platform Grant centre dedicated to research
into water and wastewater. It is based at the Universities of
Sheffield and Bradford and is headed by Professor Richard Ashley
as Managing Director and Professor Adrian Saul as Research Director.
The PWG aims to advance engineering and scientific knowledge across
all aspects of potable water, stormwater and wastewater service
provision and management of associated assets. See http://pwg.group.shef.ac.uk/
for further information. Back