Proceedings of the 19th IAHR-APD Congress 2014, Hanoi, Vietnam

ISBN 978604821338-1

CLIMATE CHANGE, RAINFALL EXTREMES, AND CHANGES TO DESIGN PRINCIPLES IN

ENGINEERING HYDROLOGY

ASHISH SHARMA(1)
(1) School of Civil and Environmental Engineering, The University of New South Wales, Sydney, NSW 2052, AUSTRALIA
a.sharma@unsw.edu.au

ABSTRACT

This paper reviews recent advances in our understanding of the implications changes in climate will have on design flood
estimation. The paper stems from a need to reflect on the changes that are supported by observations currently available,
instead of overly relying on a suite of complex models whose reliability for practical applications is often found wanting.
It is suggested that if observations are supporting theoretical hypotheses of changes in extremes, then strategies be put in
place to modify design sequences to reflect the changes that are likely. If climate model simulations are to be used, they
should ideally be restricted to variables that have been characterised as reliably simulated for future (warmer) climates.
The review is formulated using a series of questions that are addressed using observational records on a range of variables
that impact design flood estimation.

Keywords:Climate Change, Extremes, Design Flood Estimation, Antecedent Rainfall

1. INTRODUCTION
Conditi Condition
Climate change is now undeniable. Questions still remain, Number Uncon Conditio oned ed MK on
however, as to what exactly are the impacts it will have on Study of ditione ned MK MK on Filtered
hydrology. The only agreement we have till date is that Period Stations d MK on SAM Nio 3.4 IPO
the wet will get wetter, and the dry drier. But is this the
30-yr 330 7 (75) 3 (45) 7(49) 5 (49)
only change an altered climate will bring? What about its
implication on the biggest problem engineering hydrology 40-yr 77 2 (16) 0 (1) 2 (11) 2 (13)
focusses on design flood estimation? Will future design 50-yr 21 1 (6) 0 (0) 1 (6) 0 (3)
floods be greater, or smaller? How would the design
The key conclusions that were drawn from this study
rainfall change to force such change in floods? What about
were:
vegetation and catchment changes? What about overall
wetness in the catchment, and the structure of the design 1. There exist a statistically significant number of stations
storm being used? These are some of the questions exhibiting trends in annual maximum flows in
researchers are grappling with all over the world. This Australia. This point is all the more noteworthy when
paper offers a reflection on these questions based on the we take into consideration the fact that all the above
research the author has been engaged in over the years. stations are checked and found free from
anthropogenic influences.
This paper is broken up into a number of questions that
are reflected on one by one. So let us start with the first 2. Of these stations, a majority exhibit a decrease in the
question that needs to be thought about are floods really annual maximum flood.
changing?
3. The number of such significant trend stations
2. ARE FLOOD CHANGING? decreases when the effect of climatic covariates (such
as the El Nino Southern Oscillation (ENSO) or the
In one of the most extensive studies to date on assessing
Interdecadal Pacific Oscillation (IPO)) is taken into
climate change impacts on annual maximum floods, Ishak
consideration. As the above mentioned covariates are
et al. [2013] asked the above question using data from 491
known to impact extremes, any asymmetric sampling
small to medium sized catchments that were classified to
of records towards one phase of the climatic anomaly
be anthropogenically unaffected. The study used
(such as a negative IPO phase in contrast to a positive
catchments with at least 30 year long streamflow records,
phase) can lead to an artificial trend in the overall
and assessed changes using standard trend tests but also
record.
by segregating flows as a function of low-frequency
modulations of the climate such as the El Nino Southern 4. The fact that the number of stations exhibiting
Oscillation (ENSO) or the Inter-decadal Pacific Oscillation significant trends is more than 10% after the climatic
(IPO) or the Southern Annular Mode (SAM). Results from covariates are taken into consideration, forces one to
this exercise are summarized in Table 1 below. consider that this may be a result of climate change.
Table 1. A summary of the Mann Kendall (MK) test results on 5. The above consideration becomes even stronger when
study catchments with and without conditioning on a range of we acknowledge that the climatic covariates are
climatic co-variates. Source: Ishak et al. [2013]. themselves significantly correlated to the global

1
 warming trend, suggesting the true number of
stations that exhibit significant trends is closer to that
reported under the unconditional case (column 2).
Consequently, one can conclude that annual maximum
floods are changing. What could possibly be the reason
behind these changes remains elusive, but given all other
significant causative factors have been addressed in
deriving the above results, the most likely reason behind
the change is changes to the climate. Which brings us to
our second question should we expect floods to change
due to global warming?
3. WHY SHOULD FLOODS CHANGE DUE TO
GLOBAL WARMING?
Global warming is a result of an increase in (mostly
anthropogenic) greenhouse gas emissions and
concentrations, which lead to an intensification of the
energy cycle, specifically an increased entrapment of the
longwave radiation fluxes in the earths atmosphere. This
continued entrapment leads to an increase in the overall
temperature of the planet, that then goes on to result in a
range of changes, including an increase in the amount of
water that can be stored in an atmospheric column at any
instant of time.
The question this then raises is should floods be
changing as a result of the changes noted above? For
floods to change in a future climate, one or more of the
following conditions need to be created:
1. Extreme precipitation leading to the flood changes.
2. Catchment wetness preceding the extreme (flood
causing) rain changes.
3. The catchment response to a rainfall event changes
through a change in transpiration characteristics
associated with given vegetation types.
Each of these is discussed in more detail in the discussion
that follows.
4. IS EXTREME (FLOOD CAUSING) RAIN
CHANGING?
The keyquestion we need to ask is whether there is likely
to be a change in extreme rainfall and its characteristics
(temporal and spatial patterns) due to global warming?
General Circulation Model (GCM) simulations of likely
future climates, in general, predict that the wetter
latitudes will get wetter, and the drier latitudes drier. But
there general changes may not have much relevance when
the rainfall intensity associated with extreme events is
under question.
The key physical change that is likely in a future warmer
climate is related to the atmosphere being capable of
holding a greater volume of moisture that what is the case
now. This extra holding capacity of the atmosphere is
explained through the Clausius-Clapeyron law, and has
been used to argue that an atmospheric column will have
a greater holding capacity of moisture, and consequently,
can lead to greater extreme events if the moisture source is
constrained to fall within this column [Lenderink and Van
Meijgaard, 2008]. In related work using extensive sub-daily
rainfall data across Australia, Hardwick-Jones et al. [2010]
identified clear intensification of sub-daily rainfall with
increased temperatures, with the increase becoming less
significant as the rainfall duration became larger. Selected
results from this study are presented in Figure 1.
Figure 1. Variations in the 99th percentile 60-minute extreme
rainfall across Australia. The red dots represent locations which
indicate increases, and the blue dots decreases. The increase in the
extreme rainfalls is clear from the above figure, whereas the
decreases are most likely a result of additional moisture sources
contributing to the extreme rainfall. Source: [Jones et al., 2010].
Given that these increases are more likely to be associated
with shorter duration extreme rainfall events, the
implications of this result are likely to be felt in the smaller
urban catchments most. While the same arguments hold
for the larger catchments too, the assumption that the
moisture source is constrained to the atmospheric column
is more likely to be violated.
One issue that has not been addressed yet is whether it is
only the change in the extreme rainfall amount that is
important for resulting floods to change. The follow-up
question to this is whether a change in spatial or temporal
characteristics associated with the design rainfall burst can
lead to consistent changes in the flood. Design flood
estimation studies require the specification of a temporal
pattern for the design storm, and studies over larger
catchments require a similar specification for a design
spatial pattern too. An intensification of the atmospheric
circulation suggests that storms corresponding to warmer
climate may be intensifying (or, getting narrower and
more peaked). This is a hypothesis that has received
limited attention in the literature, but one likely to lead to
increases in the resulting flood.
5. ARE ANTECEDENT CONDITIONS
CHANGING?
In a recent study assessing the reasons for changes in
annual maximum floods between opposing phases on the
Interdecadal Pacific Oscillation (IPO), Pui et al. [2011]
found that while floods were remarkably different
between the two IPO phases, the causative rainfall leading
to the flood was not. Further investigation revealed that
the differences in floods were a result of markedly
different antecedent conditions preceding the flood
causing rainfall. These antecedent conditions were clearly
linked with the IPO, even when the extreme rainfall was
not. Similarly, Pathiraja et al.[2012] showed the
importance of antecedent conditions on the resulting
flood, with changes of the order of 10 times being noted
when wet antecedent conditions were used instead of dry
ones for selected Australian catchments.
2
In the context of the present discussion, the question that
arises from the above finding is whether antecedent
conditions corresponding to the extreme floods recorded
on the catchment, exhibit any trend that would indicate
that these are affected by global warming. To our
knowledge, this is a question that has not been
investigated in significant detail as of yet. However, for
antecedent conditions to be different, one would need to
see a difference in the incident rainfall, and also in the
evaporation associated with this rainfall. Alternately, one
would be able to spot such changes in soil moisture, a
variable that global records for exist through advances in
satellite remote sensing.
Dorigo et al. [2012] recently investigated trends in satellite
derived soil moisture across the world and found 27% of
the area sampled indicated significant trends, with
roughly 3/4th of these locations exhibiting negative
trends. Figure 2 reproduces results from the mentioned
paper, indicating the locations where significant trends
were identified.
Figure 2. Mann Kendall test results on global reconstructed
satellite derived soil moisture estimates. These estimates cover a
time period from 1988-2010. Locations a-h indicated on the plot
are explored further in the supplementary material included in
the paper. Source: Dorigo et al. [2012].
The results in figure 2 are possible only if there has been
an overall increase in evaporation for the continent,
something that is to be expected with increase in global air
temperature. Readers are referred to [Johnson and
Sharma, 2010] for an assessment of the changed in pan
evaporation in our instrumented record, an assessment of
whether pan evaporation decreases (associated with the so
called pan evaporation paradox) can be reconciled with
the overall increases one would expect in a warming
climate, and an assessment of the evaporation that is
simulated across a range of General Circulation Model
(GCM) simulations representing likely future climates.
The study shows that even if there is a question about pan
evaporation decreases in the current climate, caused due
to the win-forced (or aerodynamic) component of the
evaporation decreasing over certain segments of our
instrumented record, the radiation forced component (or
the radiative component) of evaporation will overshadow
any such decreases through the added energy that will be
available in the atmosphere as the full impact of global
warming unfolds.
6. IS THE CATCHMENT RESPONSE
MECHANISM CHANGING?
The last part of the puzzle lies in pondering on whether a
warmer climate may lead to a change in the rainfall-runoff
transformation mechanism. The key reason this
transformation may change is through the change in
transpiration properties associated with the catchment.
Transpiration, on its own, should change for two reasons.
The first reason is the change in the overall atmospheric
temperature, leading to an increased humidity, causing
possibly a tendency to reduce the transpiration that would
occur. The second reason is more related to the change in
vegetative attributes, brought on because of a change in
climate (rainfall and evaporation). It is this latter change
that may make the biggest impact when it comes to the
rainfall-runoff transformation.
Has this change already been noted to be occurring? To
the best of knowledge, statistically conclusive evidence to
this effect are not clearly available, especially if the region
being considered is large and remains beyond the
influence of other changes such as anthropogenic impacts.
However, this is an area that needs further investigation,
along with an assessment how this change could be
modelled for warmer, more humid climates.
7. SO HOW SHOULD DESIGN FLOODS FOR A
WARMER CLIMATE BE ESTIMATED?
The above sections have concluded several points. Firstly,
there is evidence that extreme flood events are changing
on a regional basis. Secondly, there is empirical as well as
theoretical evidence that extreme (flood causing) rainfall is
increasing for shorter time scales when associated with
convection. Thirdly, regional changes in pre-extreme
rainfall antecedent wetness conditions are being observed
around the world, suggesting that the flow response to the
same extreme rainfall will be different in the future.
Lastly, it is possible that changes in transpiration
properties for a future vegetated catchment will cause
changes in flood response, although these changes are
likely to affect larger catchments more than smaller
catchments around the world. The question this leaves us
with is just how should we approach the flood estimation
problem as a result?
Before we address this key question, thought must be
given to the approach being adopted to assess changes to
flood characteristics in a warmer climate. Typically,
estimating a future design flood is attempted by
formulating a changed future design rain which is then
translated through a rainfall-runoff model to ascertain the
resulting flood value. The future design rain is usually
ascertained by either performing some type of bias
correction on directly simulated rainfall from a GCM or a
GCM+RCM, or statistically downscaled using bias
corrected atmospheric variable simulations from the GCM
or GCM+RCM. If direct GCM simulations (or
corresponding RCM simulations that use lateral boundary
conditions from a GCM simulation) are used, case much
be taken to ensure that the variable of interest is likely to
be simulated with some level of accuracy for use in
practical applications such as design flood estimation. A
study by Johnson and Sharma [2009] suggests that the
consistency between simulations across multiple GCMs
representing a future climate is considerably lower for the
case of rainfall compared to other atmospheric variables
such as atmospheric pressure or temperature. Hence,
derivations of rainfall for future climates stemming from
the use of the relatively stable climate predictors from
GCMs is a more prudent choice to undertake.
Thought must also be given to whether it is sensible to
correct simulation biases in the variables that are used for
downscaling, in ways that involve disconnecting the
temporal dependence that leads to variables evolving over
3
time. Approaches that involve transforming GCM
simulations to match observed quantiles fall into this
category, their use becoming even more dangerous when
multiple variables are transformed in this manner thereby
ensuring their inter-dependence attributes are also
distorted along with their temporal dependence. While
bias correction will lead to an alteration of the physical
consistency associated with these variables, approaches
that use the entire sequence of simulations suitably
transformed (ideally in a multivariate manner) to maintain
persistence attributes in resulting sequences, should be
preferred. The Nested Bias Correction (NBC) suite of bias
correction alternatives [Johnson and Sharma, 2011; 2012;
Mehrotra and Sharma, 2012], specifically designed to ensure
an appropriate representation of persistence across
multiple time scales, is one such approach to consider.
Even if a design rainfall storm corresponding to a future
climate can be ascertained through sensibly constructed
downscaling and bias correction models, the question still
remains as to how catchment antecedent conditions and
their changes can be included in the flood estimation
exercise. An option that has not been explored enough is
the use of continuous simulation alternatives, suitably
devised to ensure representation of antecedent conditions
which can be altered to take into account the effect of
global warming. Examples of such approaches for use in
both gauged and ungauged locations around Australia are
given in [Mehrotra et al., 2012; S. Westra et al., 2013; Seth
Westra et al., 2012].
Regarding changes to catchment response, this is an issue
that needs significantly more research, following which
stress functions can be developed that allow
transformation of fluxes that are likely in a future climatic
setting. Incorporation of these functions into existing
rainfall-runoff models may offer the best alternative to
include their effect on future design floods.
8. CONCLUSIONS
This paper reflected on some of the challenges facing
design flood estimation in a warming climate. It
recognised that traditionally followed approaches such as
flood frequency analysis or the use of a design storm to
estimate the flood, become especially flawed when used in
a knowingly nonstationary setting. It suggested that any
approach that be used, needs to recognise three key
changes associated with future extremes that antecedent
rainfall preceding the extreme rainfall event will change,
that the design storm itself is going to intensify especially
for the shorter design durations, and that catchment
response to the rainfall event may be different in the
future because of changes in humidity and vegetation.
The paper went on to propose that continuous rainfall
simulation for a warming climate may offer a means to
address many of the difficulties noted in the traditional
alternatives above. Continuous sequences could be
formulated based on daily rainfall sequences, with the
daily sequences being designed to accommodate the
changes altered antecedent conditions might bring. They
may then be further disaggregated in a manner that
ensured that the extreme subdaily rainfall exhibited
greater extremes corresponding to change in the climate,
again possible to accommodate as long as the
disaggregation model is structured to do so. While this
approach will still use an existing rainfall-runoff
transformation to generate flows, this issue can be
addressed too through carefully designed studies. The
framework described, on the whole, will be capable of
simulating design floods for future warmer climates,
something that is highly needed across the world as we
protect our existing and planned infrastructure against the
extremes of tomorrow.
ACKNOWLEDGMENTS
This paper is a result of funding and support from a
number of agencies over the years, primary amongst
which are the Australian Research Council. Also, the work
has been done in collaboration with a number of research
students and colleagues over the years. These include Raj
Mehrotra, Fiona Johnson, Seth Westra, Alex Pui, Conrad
Wasko, Jingwan Li and FitsumWoldemeskel. Their
contribution in getting this work to the level it is now is
gratefully appreciated.
REFERENCES
Dorigo, W., R. de Jeu, D. Chung, R. Parinussa, Y. Liu, W.
Wagner, and D. Fernandez-Prieto (2012), Evaluating
global trends (1988-2010) in harmonized multi-satellite
surface soil moisture, Geophysical Research Letters, 39,
doi:Artn L18405, Doi 10.1029/2012gl052988.
Ishak, E. H., A. Rahman, S. Westra, A. Sharma, and G.
Kuczera (2013), Evaluating the non-stationarity of
australian annual maximum flood, Journal of Hydrology,
494, 134-145.
Johnson, F., and A. Sharma (2009), Measurement of GCM
Skill in Predicting Variables Relevant for
Hydroclimatological Assessments, Journal of Climate,
22(16), 4373-4382, doi:10.1175/2009JCLI2681.1.
Johnson, F., and A. Sharma (2011), Accounting for
interannual variability: A comparison of options for water
resources climate change impact assessments, Water
Resources Research, 47(4), doi:10.1029/2010wr009272.
Johnson, F., and A. Sharma (2012), A nesting model for
bias correction of variability at multiple time scales in
general circulation model precipitation simulations, Water
Resources Research, 48(W01504), 16,
doi:10.1029/2011WR010464.
Lenderink, G., and E. Van Meijgaard (2008), Increase in
hourly precipitation extremes beyond expectations from
temperature changes, Nat Geosci, 1(8), 511-514,
doi:10.1038/ngeo262.
Mehrotra, R., and A. Sharma (2012), An improved
standardization procedure to remove systematic low
frequency variability biases in GCM simulations, Water
Resources Research, 48(12), doi:10.1029/2012WR012446.
Mehrotra, R., S. Westra, A. Sharma, and R. Srikanthan
(2012), Continuous rainfall simulation: 2 - A regionalised
daily rainfall generation approach, Water Resources
Research, 48(1), W01536 doi:10.1029/2011WR010490.
Pathiraja, S., S. Westra, and A. Sharma (2012), Why
Continuous Simulation? The Role of Antecedent Moisture
in Design Flood Estimation, Water Resources Research,
48(W06534), doi:10.1029/2011WR010997.
Pui, A., A. Lal, and A. Sharma (2011), How does the
Interdecadal Pacific Oscillation affect design floods in
Australia?, Water Resources Research, 47(5),
doi:10.1029/2010wr009420.
4
Westra, S., J. P. Evans, R. Mehrotra, and A. Sharma (2013),
A conditional disaggregation algorithm for generating fine
time-scale rainfall data in a warmer climate, Journal of
Hydrology, 479, 86-99.
Westra, S., R. Mehrotra, A. Sharma, and R. Srikanthan
(2012), Continuous rainfall simulation: 1 - A regionalised
sub-daily disaggregation approach, Water Resources
Research, 48(1), W01535, doi:10.1029/2011WR010489.

Copyrights
Paper(s) submittedto the IAHR-APD2014 are interpreted
as declaration that the authors obtained the necessary
authorization for publication.