Data Assimilation for the REsilient City (DARE)

Data assimilation is an emerging mathematical technique for improving predictions from large and complex forecasting models, by combining uncertain model predictions with a diverse set of observational data in a dynamic feedback loop. The project will use advanced data assimilation to combine a range of advanced sensors with state-of-the-art computational models and produce a step-change in the skill of forecasts of urban natural hazards such as floods, snow, ice and heat stress.

The research will use synthetic aperture radar (SAR) data to develop a tool for real-time detection of flooded urban areas. SAR sensors take images from space over a wide area and can see through clouds. The sensors have resolutions as high as 1m, and are able to image flooded streets. However, substantial areas of urban ground surface may not be visible to the SAR due to shadows caused by buildings. Furthermore, shadowed areas may be misclassified as water even if dry. Our new approach is to use a SAR simulator in conjunction with lidar data. The SAR simulator estimates regions in the image in which water will not be visible due to shadow, and masks these out from the ground surface considered, resulting in a more accurate flood extent. This type of information could be used by first responders to monitor vital infrastructure and understand the extent and depth of the evolving flood.

SAR images can also be used to extract water level observations, which may be assimilated into a flood inundation model, to calibrate the system and keep predictions on track. Our recent ground-breaking work demonstrates the possibility of earth-observation-based flood inundation data assimilation and forecasting over a rural area. In this new project we aim to carry out scientific and mathematical studies to increase the flexibility of our flood data assimilation system, so that it can be straightforwardly applied at any location in the UK (including urban areas). For example, the behaviour of the system is expected to change for larger floods, steeper rivers, faster flow etc. In addition, we will develop techniques to derive new types of water level observations from smartphone photographs, traffic and river CCTV cameras, that can also be assimilated to improve predictive skill.

A number of environmental hazards are caused by the weather (e.g., heat stress, high winds, fog). The skill of numerical weather prediction is strongly constrained by the accuracy of the initial data, as estimated by assimilating expensive observations. There are burgeoning sources of inexpensive datasets of opportunity (citizen science, sensor networks etc.) that could be used, however lack of knowledge about natural variability in urban areas hinders uptake of these data. This proposal addresses uncertainty due to urban natural variability in observation-model comparisons, by considering numerical weather prediction models on a range of scales, and observational data with different "footprints". We will apply these results to citizen science automatic weather station data, car temperature sensors and commercial aircraft reports made to air traffic control (used to derive observations of winds and temperature).

The impact of this research will be guaranteed by working with operational providers of flood warnings and weather forecasts (the Environment Agency and Met Office). Commercialization of aspects of the research will be pursued in conjunction with the Institute for Environmental Analytics.

A network of researchers and industry working with digital technology at the "Living with Environmental Change" interface will be formed. This will have a programme of workshops, webinars, training and industry study groups to cross barriers between academic disciplines, creating bridges between academia and industry and providing space for junior and senior researchers to explore ideas. Funded pilot projects will kick-start activities and help define the future research agenda.

The project will provide new observations and research software tools for short term prediction and real-time detection of flooded urban areas that are expected to benefit the Environment Agency (EA) and EA/Met Office Flood Forecasting Centre (and their Scottish and Welsh equivalents): the UK public agencies responsible for flood risk guidance and management. The research could be developed into decision tools combining forecasts with geographical information to aid preparedness (e.g., protecting an electricity substation) and guide mitigating actions (e.g., deployment of flood alleviation schemes). Real-time information could be used by first responders to understand the extent of the evolving flood. Data assimilation reanalyses should improve estimates of past floods, allowing better design of new flood defences. Satellite information services such as Copernicus Emergency Management Services, and engineering consultants such as CH2M, JBA may also be interested in commercializing aspects of this research.

Our work on characterizing the uncertainty of urban observations will feed into Met Office operations. Current observing networks employed by the Met Office are very expensive (e.g. the satellite data used by the Met Office costs around £2 billion per year, although these costs are shared internationally). The research will provide the underpinning science that is needed for the use of new inexpensive urban observations and thus has potential for reduction in cost of the Met Office observing network in the future. Outcomes from the research will feed into improved public weather service (PWS) forecasts and are expected to promote better decisions along the existing and wide ranging forecast value chain. The 'Public Weather Service Value for Money Review 2015' indicates that the benefits of the PWS to the UK economy are around £1.5 billion per annum and it is impossible to describe all of the sectors affected in this short summary. However, market reports will be produced during the project to explore and define research questions for new applications of urban meteorology. Some examples relevant to the resilience of urban infrastructure are: better forecasts of fog and high winds would allow better air traffic management; better forecasts of snow and ice would allow transportation managers to target de-icing activities more effectively, and ensure that they are well supplied with materials, equipment and staff at critical times; better forecasts of storms and floods would enable a reduction in damage, reducing costs for the insurance industry; better forecasts of heat stress enable better preparedness for hospitals and public health agencies.

Outcomes from the research will also provide information for policy-makers and their networks such as the London Climate Change Partnership and the UK Natural Hazards Partnership programme. The research will provide evidence for future observing network design and natural hazard modelling strategies, including how to get the best value for money from observations. The project will also produce market reports detailing the needs and opportunities at the digital technology/environment sector interface and road-maps for future research strategy of use to EPSRC and other funders.

The proposed network will feature a range of commercial and public sector partners, and is expected to spin up new partnerships between industry and academia, increasing intellectual capital for UK business and competitiveness in the global economy. Early career scientists will receive training in computer modelling, data assimilation and multi-disciplinary working: skills at the top of the NERC/LWEC "Most Wanted II" list. Public engagement events will aid public understanding and enthusiasm for science, as well as improving personal understanding of what to do when extreme weather hits.