Societal and economic benefits from improved flood modelling based on pioneering Lancaster research in to risk and uncertainty in environmental modelling.
Submitting Institution
Lancaster UniversityUnit of Assessment
Earth Systems and Environmental SciencesSummary Impact Type
SocietalResearch Subject Area(s)
Mathematical Sciences: Statistics
Information and Computing Sciences: Artificial Intelligence and Image Processing, Computation Theory and Mathematics
Summary of the impact
Improved flood risk modelling based on the application of research led by
Keith Beven at Lancaster has had global impacts in improved flood defence
policies and planning by governments, and in assisting insurers with their
underwriting (for example in pricing and policy decisions). The benefits
are not just financial — they are human too: improved understanding of
flood risk and resilience protects life and assets, and has a positive
impact on the well-being of many of those at risk. These impacts are at
the centre of flood risk management across the UK, are being applied in
nine other European countries, and now becoming the methods of choice for
flood mapping in developing countries such as Thailand.
Underpinning research
The Lancaster Environment Centre (LEC) has been at the forefront of
modelling catchment hydrology for several decades, based on fundamental
research led by Prof. Keith Beven (1985-present). The impact described in
Section 4 is based on the application of Beven's pioneering research in
computer modelling for hydrological prediction. Impacts have been achieved
especially through the development of the Generalised Likelihood
Uncertainty Estimation (GLUE) approach, which required innovative methods
to implement the computationally demanding Monte Carlo simulation at its
core and, with Prof. Peter Young, Data Based Mechanistic modelling
techniques for adaptive probabilistic real-time forecasting. The
application of computationally intensive modelling has proved especially
valuable in river catchment management. It enables realistic modelling of
phenomena such as water quality and flooding, in combination with the
methods needed to assess risk and uncertainty that attempt to deal with
the sources of error in testing models more explicitly than traditional
statistical methods (e.g. see references 1 and 2, Section 3). Early
applications included novel analysis of uncertainty in flood-plain
inundation models (reference 3). Between 1993 and 2012, research in this
area has attracted research funding totalling approximately £4.5M and
supported c. 18 PhD projects.
Beven and his co-workers recognized that the effective application of
models for risk and uncertainty assessments in the environmental sciences
required the application of then state-of-the-art advances in computing
power. That insight led Beven's team to pursue the innovative use of
parallel computing techniques emerging from the domain of High Performance
Computing into hydrological modelling, an advance which underpinned
numerous academic publications (see for example references 4-5 in Section
3). This early work laid the foundation for the later realisation by one
of Beven's students that computing systems designed for video gaming could
be adapted cheaply to overcome computational constraints on
two-dimensional flood inundation models used for flood risk management
(see Section 4). There is on-going iteration between Beven's fundamental
research (e.g. reference 6) and its application by government and industry
stakeholders, which continues to enhance impact. Beven led the Risk and
Uncertainty component in Phase 1 of the Flood Risk Research Management
Research Consortium (FRMRC1) funded by EPSRC, Defra/EA and others
(2004-2012), and the GridStix forecasting project within the NERC FREE
Programme. Beven also contributed to the Environment Agency SC080030
Probabilistic Flood Forecasting project led by Atkins Global (2008-2010).
His role as leader of the Catchment Change Network (a NERC Knowledge
Exchange project: 2009-2012) secures an on-going (through ccmhub.net),
direct route between LEC's fundamental research and a wide range of
stake-holders.
References to the research
* indicates the three references that were most central to the impact
that has been achieved.
1*. Beven, K.J. (1993), Prophecy, reality and uncertainty in distributed
hydrological modelling, Adv. in Water Resourc., 16, 41-51. (464 citations
in Web of Science)
2. Beven, K J, 2006, A manifesto for the equifinality thesis, J.
Hydrology, 320, 18-36 (470 citations in Web of Science)
3. Aronica, G, Hankin, B.G., Beven, K.J., 1998, Uncertainty and
equifinality in calibrating distributed roughness coefficients in a flood
propagation model with limited data, Advances in Water Resources, 22(4),
349-365. (106 citations in Web of Science)
4*. Lamb, R., K. Beven, and S. Myrabo, (1998) Use of spatially
distributed water table observations to constrain uncertainty in a
rainfall-runoff model. Advances in Water Resources. 22:
305-317. (now 99 citations in Web of Science)
5*. Beven, K and Freer, J (2001) Equifinality, data assimilation, and
uncertainty estimation in mechanistic modelling of complex environmental
systems using the GLUE methodology Journal of Hydrology: 249:
11-29. (559 citations in Web of Science)
6. Pappenberger, F., and K. J. Beven (2006), Ignorance is bliss: Or seven
reasons not to use uncertainty analysis, Water Resour. Res., 42, W05302,
doi:10.1029/2005WR004820 (127 citations in Web of Science)
Details of the impact
LEC's research into catchment and flood modelling has resulted in
substantial improvements in modelling hydrological risks world-wide,
resulting in significant economic and societal benefits. One of the most
important hydrological risks is flooding, which causes huge damage
globally, leading to enormous financial and human costs. Flooding is
managed through a combination of investment in flood defences, development
planning, insurance and emergency or disaster responses. All these
activities depend on accurate and detailed flood mapping or real-time
forecasts, both based on computer modelling. Governments and insurers
require flood models at multiple scales, from national assessments down to
local modelling in urban environments. Delivering such models is a
significant science challenge because flooding involves many factors that
control the occurrence, magnitude and timing of any given event.
Lancaster's developments in environmental modelling have been at the
leading edge of applying uncertainty concepts to risk models, and have
been applied world-wide by a range of end users. However, for this case
study we will focus on one specific example in which well-documented
impacts delivered by a commercial research user can be directly attributed
to Beven's research.
Tools for operational flood forecasting typically predict the depth of
water at specific locations on a river, but for planning and insurance
purposes it is much more important to have two-dimensional (2D) maps of
potential flood risk. To produce the most realistic mapping possible
demands 2D dynamic hydraulic flow models, which are computationally very
intensive, something that constrained widespread application by industry.
However, these constraints were lifted through insights resulting from
Beven's research on uncertainty methods which gave a stimulus to apply
parallel computing techniques for hydrological modelling (see Section 2).
This is summarised in the following timeline of some of the key
developments.
1993-1996 Beven's fundamental research on uncertainty methods
includes pioneering use of parallel computing in environmental modelling
(references 3 & 4 in Section 2), including in PhD research by one of
Beven's students, Rob Lamb (1993-1996).
1996. Lamb graduates and maintains links with Lancaster research
on risk and uncertainty
2002. Lamb joins JBA, a consultancy working in flood risk
managementA which has its own in- house 2D flood inundation
model, JFlowB.
2005-2007. Drawing on his Lancaster research Lamb exploits
mass-market graphics processing units (GPUs, parallel processors designed
for computer graphics), allowing JBA to run flood models up to a thousand
times fasterC,D.
2006 onwards. JBA successfully apply JFlow GPU (see evidence of
impacts below)D.
2009-2011. Lancaster and JBA demonstrate use of JFlow within the
GLUE framework in the EPSRC-funded FRMRC project to produce the first
uncertainty-based flood maps, including Google earth visualisations.
Workshops run with EA, consultants, and town planning staff under
Catchment Change Network to disseminate the advances.
In summary, co-ordinated fundamental and technological advances, driven
by Lancaster's pioneering research and Lancaster-trained personnel, were
the enabling research that led JBA to develop the world's first
commercially successful GPU-based flood model, and Lancaster to produce
the first probabilistic flood risk maps. JBA's software system is now
being used world-wideB,C. While the major impacts of JFlow are
undoubtedly on those communities and individuals vulnerable to flooding
(see below) it is worth noting the commercial success of JBA as a UK
company. Since its formation in 1995, it has grown to become a major
employer of environment risk analysts, with more than 50 staff directly
employed in applying JFlow through JBA Risk management LtdD.
Revenues relating to JFlow are currently around £3 million annually, and
the company's annual turnover overall is now approximately £13 millionE.
In 2012, JBA was a finalist for the Royal Academy of Engineers (RAEng)
MacRobert AwardF for innovation, for its use of GPU processing
technology in JFlow. This award "seeks to demonstrate the importance of
engineering and the role of engineers and scientists in contributing to
national prosperity and international prestige". In selecting JBA as a
finalist, the RAEng recognised JFlow for "outstanding technical
innovation with benefits to the community and commercial success".
Know-how gained through Lancaster research was instrumental to this
success, which has enabled JBA to become a world leader, modelling more
rivers in more locations than any other company, for private clients and
government agencies in the UK and elsewhereD.
JFlow delivers maps and analytical products underpinning many assessments
of flood risk in the UK and beyondF. In England, the
Environment Agency's National Assessment of Flood Risks (NaFRA) 2009
estimates that one in six homes are at risk, with annual damage costing
more than £1billion. Following major floods in 2007, Sir Michael Pitt's
review included the urgent recommendation for improved flood mapping, in
particular for floods caused not by rivers bursting their banks but by
overland runoff, and the incorporation of prediction uncertainties into
flood forecasting and planning. This type of mapping had not been done
comprehensively, partially because of the high computational cost. However
JFlow-GPU allowed JBA to produce the first national surface water flood
map, which JBA licensed to the Environment AgencyG. "JBA
has subsequently delivered two major updates for the Environment Agency's
national surface flood mapsD,H. JBA also used JFlow for reservoir
inundation maps meeting Government's strategic needs under emergency
planning legislation, modelling potential impacts from breaching of 2100
reservoirs in the UKH - This provides information necessary for
public security at one tenth of the cost of alternative modelling
approaches also testedE.
The RAEng noted that "JBA Consulting's flood risk modelling system,
JFlow, has become ... an essential tool for the UK's insurance industry"I.
According to JBA, in 2013 "GPU-based JFlow software is now used by 70%
of UK and 80% of Irish Insurers, including the top five UK insurers by
market shareD. Maps produced using JFlow were described
by leading insurer Aviva as "revolutionary"J. At a time when
many householders face problems obtaining household insurance due to flood
risk, a major societal benefit arising from the improved precision and
confidence delivered by JBA's use of JFlow has been to allow insurers to
qualify more than 600,000 properties in flood risk areas for insuranceK.
With the productivity gains realised through GPU parallel technology, JBA
has also been able to develop national flood maps for Ireland, France,
Germany, Poland, the Czech Republic, Slovakia, Hungary, Belgium and
LuxembourgC,K. The lower cost of using JFlow instead of
more "traditional" approaches makes it ideal for modelling flood risk in
developing countries, where flood mapping has previously been uneconomic.
A prime example is the response to major flooding in Thailand in 2011.
This was one of the five costliest natural disasters in modern history
(with World Bank estimates of USD45.7 billion economic losses and some 1.5
million homes and other buildings affected). JBA "was able to respond
by using JFlow to produce flood maps of the whole country within four
months"E. This mapping helped to provide credible
information about flood risks that has been taken up by insurers and helps
to underpin the marketL
While JFlow's adoption of parallel computing technology forms a focussed
case study of the impact of developments driven by Lancaster's research,
there are other cases where our research has been at the leading edge of
applying risk and uncertainty concepts to floods and flood risk. One
example is the incorporation of Lancaster's flood forecasting methods into
the Deltares flood early-warning systemM that is "... applied
as the primary operational flood forecasting tool used by flood
management authorities in basins across the continental United States
and Alaska, in England and Wales, Scotland, Ireland, Netherlands,
Germany, Austria, Spain, Italy, Switzerland, Taiwan, Pakistan, the
Zambezi basin, Ghana, Canada, Colombia, Indonesia, Bolivia and by the
Mekong River Commission"N. Impact is still growing, for
example new knowledge and understanding is informing new industry user
guidance, such as the CIRIA (Construction Industry Research and
Information Association) Framework for assessing uncertainty in fluvial
flood risk mapping, Report C721 (in press, 2013) and directly to end-users
via the Catchment Change Network.O
Based on pioneering research, the appointment of LEC-trained individuals
to key posts with stakeholders working with the water industry, and
on-going collaborative research, Beven's research has delivered the
tangible benefits that we anticipate that continue to grow in the future.
Sources to corroborate the impact
A. http://www.jbaconsulting.co.uk/
B. http://www.jbaconsulting.com/products/JFlow
C. http://www.intermap.com/Portals/0/doc/Newsletters/RMA_Newsletter_Spring_2010.pdf
D. Letter from the Managing Director of JBA risk management
E. Company data on JBA Ltd.
F. http://www.raeng.org.uk/prizes/macrobert/pdf/MacRobertAwardBrochure_2013.pdf
G. Hunter, N, Waller, S, Balmbra, V, Hankin, B, Faulkner, D, Lamb, R,
Horritt, M, Wyse, P. (2010) Broad Scale Mapping of Surface Water Flooding
- Present Status and Future Improvements, Proceedings of the Environment
Agency FCRM>10 Conference, 29 June - 1 July 2010, Telford Conference
Centre, Paper O80.
H. Environment Agency (2013), Updated Flood Map for Surface Water,
National Scale Surface Water Flood Mapping Methodology, Final Report
version 1.0 (May 2013), Environment Agency Bristol.
I. Environment Agency Reservoir Flood Maps: External Guidance (October
2011) http://a0768b4a8a31e106d8b0-50dc802554eb38a24458b98ff72d550b.r19.cf3.rackcdn.com/flho0612bwnn-e-e.pdf.
J. http:/www.aviva.co.uk/media-centre/story/1684/norwich-unions-revolutionary-flood-map-begins-roll/
K. http://www.insurancejournal.com/news/international/2010/05/26/110194.htm
L. http://insurancenewsnet.com/oarticle/2012/03/19/jba-launches-first-flood-model-for-thailand-a-334973.html#.UlvA21DEMqh
M. Letter from Deltares Inland Water Systems - Operational Water
Management, Delft, The Netherlands
N. http://oss.deltares.nl/web/delft-fews/implementations
O. http://www.catchmentchange.net/