Probability, Uncertainty and Risk in the Natural Environment

Lead Research Organisation: Imperial College London
Department Name: Civil & Environmental Engineering

Abstract

Natural hazards pose serious problems to society and to the global economy. Recent examples in the UK include the cold winters of 2009 and 2010 and the eruptions of the Grimsvotn and Eyjafjallajökull volcanoes with the consequent disruption to air travel. Moving further afield, the first half of 2011 saw major disasters in Australia (flood), New Zealand (earthquake), Japan (earthquake and tsunami) and the US (hurricanes).

It would be nice if scientists could provide precise information to help with the management of such events. This is unrealistic, however, for several reasons: data are usually incomplete (e.g. not available at all required locations) and measured with error; predictions are made using computer models that can at best approximate reality; and our understanding of some phenomena is limited by lack of experience (for example, the historical tsunami record is relatively limited). Therefore, natural hazard scientists must acknowledge the uncertainty in the information they provide, and must communicate this uncertainty effectively to users of the science. However, neither of these tasks is easy. Moreover, scientists do not always understand what users want and need; and users themselves often are uncomfortable with uncertainty.

Despite these problems, modern statistical methods are available for handling uncertainty in complex systems using probability theory. In parallel, social science researchers are interested in understanding how people react to and understand uncertainty. By bringing these two developments together, and linking with scientists from several hazard areas along with a variety of users, we aim (a) to demonstrate a generic framework for handling uncertainty across hazards; and (b) to develop improved tools for communicating uncertain information.

The generic framework considered here has three core components. The first is the treatment of uncertainties arising from our imperfect models and imperfect understanding of any complexsystem. The second is the combination of information from various sources that are all judged to be relevant: this is particularly important in event management situations where decision-makers must take rapid action based on multiple strands of evidence that might be apparently contradictory. The third is the treatment of uncertainties that are deemed to be "unquantifiable" or too hard to handle:an example from the insurance industry involves how much money to set aside to cover the cost of an event that is known to be possible but for which no historical loss data are available (such as an Atlantic tsunami caused by the collapse of the Cumbre Vieja volcano in La Palma). Five case studies will be used to illustrate the framework: (1) flood risk management in the UK; (2) earthquake hazard in the UK (relevant to the nuclear power industry) and in Italy; (3) tsunami hazard and risk assessment, including the development of methods to improve real-time warning systems; (4) the interpretation of days-ahead weather forecasts (focusing on wind speeds and cold weather); (5) volcanic ash dispersal, again including real-time warning systems.

A final, and critical, component of the proposed research relates to the communication and use of the uncertainty information derived from the three previous components. Working with industrial partners, we will demonstrate how an improved understanding of uncertainty in the hazard itself can be translated through into risk assessments (which focus on the consequence of the hazard, for example the economic loss or damage to infrastructure). We will also carry out research to understand better how people perceive and use risk information. The results will be used to inform the development of novel methods for communicating natural hazard risk information to specialist and non-specialist users; and also (in collaboration with the PURE Network) to produce a handbook of risk communication for natural hazards.

Planned Impact

The proposed research will benefit all individuals and organisations with an interest in understanding, responding to and planning for natural hazards and their consequences. Excluding academic beneficiaries, these include:

- Business and industry, in particular the financial (notably insurance), energy, aviation and built environment sectors;
- Organisations such as DEFRA, the Environment Agency and SEPA, with responsibility for natural hazard risk management in the UK and elsewhere;
- Agencies responsible for the provision of risk and hazard management information, such as the UK Meteorological Office (UKMO);
- The general public, including schoolchildren.

For these non-academic beneficiaries, the primary impact of the research will arise from improved communication between the science and user communities, so that the science becomes more relevant to the users and the users are better able to understand the science. The requirements here work both ways. Our engagement with users and industrial partners, and research on communication under Work Package D, aims to foster better understanding of user needs by scientists. Simultaneously however, we will help users to develop a better understanding of what science can and cannot be expected to provide, and to make effective use of uncertain information in decision-making. Apart from the direct engagement with our industrial partners, much of this work will be carried out via dissemination, engagement and training events organised in collaboration with the PURE Network.

Further details of the research impact can be found in our "Pathways to Impact" statement.

Publications

10 25 50
publication icon
Almeida S (2015) Accounting for dependencies in regionalized signatures for predictions in ungauged catchments in Hydrology and Earth System Sciences Discussions

publication icon
Blair P (2016) Socio-hydrological modelling: a review asking "why, what and how?" in Hydrology and Earth System Sciences

publication icon
Blair P (2015) Modelling socio-hydrological systems: a review of concepts, approaches and applications in Hydrology and Earth System Sciences Discussions

publication icon
Vitolo C (2015) Web technologies for environmental Big Data in Environmental Modelling & Software

 
Description We have developed and published methodologies to better couple environmental models for improved predictions on flood hazard.
Exploitation Route All methods have been published and the computer code made available in the public domain to allow other research groups to use and improve the code.
Sectors Environment

 
Title High-resolution satellite-gauge merged precipitation climatologies of the tropical Andes 
Description A set of digital precipitation maps of the tropical Andes, covering Colombia, Ecuador and Peru at a 5km resolution. The maps represent different realizations of mean precipitation totals of the period 1981-2010 using different satellite-gauge merging methods. The work draws on a large database of 723 rain gauges and the full 5km Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (TPR) record from 1998 to 2014. 
Type Of Material Database/Collection of data 
Year Produced 2016 
Provided To Others? Yes  
Impact This dataset is being used by the National Service of Hydrology and Meteorology (SENAMHI), Peru 
URL https://catalogue.ceh.ac.uk/documents/74a588cc-723c-4a35-ac0c-223f5b92ee36
 
Title Time series and geospatial informations for the Pontbren catchments 
Description A QA workflow (see entry for the "pure" R-packages) was utilised to transform the raw time series data into a format compatible with the FUSE framework. Geospatial information is currently being used to validate modelling results. 
Type Of Material Data handling & control 
Provided To Others? No  
Impact The raw time series data is highly irregular and, as it is, could have not been used for modelling purposes. The derived time series are regular and can be used with the FUSE model as well as any other similar tool. 
 
Title Automatic Model Configuration Algorithm (AMCA R-package) 
Description This is a library developed for the R statistical environment and currently contains a data mining procedure based on unsupervised machine learning techniques to automatically configure hydrological conceptual rainfall-runoff models such as FUSE. 
Type Of Technology Software 
Year Produced 2014 
Open Source License? Yes  
Impact This library provides functions to perform an automatic selection of hydrological model configurations and visualisation tools to explore the multidimensional result space. 
URL http://cvitolo.github.io/r_amca/
 
Title Framework for Understanding Structural Errors (FUSE R-package) 
Description R-package containing the implementation of the framework for hydrological modelling FUSE, based on the Fortran version described in Clark et al. (2008). The package consists of two modules: Soil Moisture Accounting module and a Gamma routing module. It also contains default parameter ranges and data objects used for testing purposes. 
Type Of Technology Software 
Year Produced 2014 
Open Source License? Yes  
Impact This library allows to experiment on structural uncertainty of hydrological rainfall-runoff models. 
URL https://r-forge.r-project.org/R/?group_id=411
 
Title Hydrological data discovery tools (hddtools R-package) 
Description This is a library developed for the R statistical environment and currently contains reusable function to access and retrieve information from the following data sources: - Global Runoff Data Centre - Koppen Climate Classification map - Data60UK - NASA TRMM mission 
Type Of Technology Software 
Year Produced 2014 
Open Source License? Yes  
Impact This library facilitates hydrological data discovery and retrieval from data sources that do not provide programmatic access. 
URL http://cvitolo.github.io/r_hddtools/
 
Title R tools for the UK National River Flow Archive (R package) 
Description This is a library developed for the R statistical environment and currently contains tools to interact with the UK National River Flow Archive via an experimental API. 
Type Of Technology Software 
Year Produced 2014 
Open Source License? Yes  
Impact This library provides tools to facilitate data discovery, retrieval, parsing and mapping of data and metadata obtained from the UK National Flow Archive. 
URL http://cvitolo.github.io/r_rnrfa/
 
Title RHydro 
Description Package for hydrological modelling in the R environment 
Type Of Technology Software 
Year Produced 2009 
Open Source License? Yes  
Impact This software consists of hydrological models and tools to represent and analyze hydrological data. 
URL https://r-forge.r-project.org/projects/r-hydro/
 
Title Tool to facilitate pre-processing, hydrological modelling and flood frequency analysis under uncertainty (PURE R-package) 
Description This is a library developed for the R statistical environment and currently contains guidelines for data pre-processing and modelling using the FUSE framework. Work is in progress to add modular components to carry out flood frequency analysis under uncertainty. 
Type Of Technology Software 
Year Produced 2014 
Open Source License? Yes  
Impact This library providing guidelines to carry out modelling tasks in a more objective and less modeller dependent way, hence enhancing reproducible research. 
URL http://cvitolo.github.io/r_pure/
 
Description Short course: Introduction to environmental data analysis and simulation using R 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Professional Practitioners
Results and Impact 2 day workshop for 15 professionals, mostly from the insurance industry
Year(s) Of Engagement Activity 2015