Using satellite data to respond to environmental disasters in Malaysia, Ethiopia and Kenya

The challenge of providing a rapid response to environmental disasters as varied as flooding, drought, illegal logging and oil spills is the focus of two new projects in which the University of Oxford is a key partner.

Using satellite imagery as a starting point, the aim of both projects is to create information for key decision makers in each country so that they are able to intervene as early as possible to protect people and the environment. The advantage of satellite data is that it can quickly identify small changes on the surface of the earth or sea that may be indicators of a larger problem in the making. A new 'hole' appearing in a forest can provide evidence of illegal logging, or a slight colour change in crops may show the early effects of drought. Combining data from these images with other data sources has the potential to create powerful information for governments and other actors.

In Malaysia the project consortium will be working with government agencies to tackle flooding, oil pollution and illegal logging, all of which pose serious social and economic threats to Malaysian people. Monsoon flooding is a major annual issue, and the project aims to enable evacuation plans and flood defences to be activated much faster. It will also generate data that will help the authorities to quickly identify and track oil leaks from shipping which are causing irreparable damage to Malaysia's mangrove swamps, and to locate areas where illegal logging is taking place.

In Ethiopia and Kenya the focus will be on creating an improved understanding of flood and drought risk, thus helping to build local people's resilience to these natural disasters and alleviate poverty. The intention is to use the same data to provide an emergency response where needed and to help develop longer-term strategies and solutions to drought and flood. In Kenya the project will also be generating tools to support the micro-insurance market, which is of key importance to farmers who have little or no access to insurance, by providing independent data about crop damage to verify farmers' claims.

Satellite imagery is very useful for quickly generating independent data from a wide variety of events on the earth as they unfold. The difficulty is how to organise and process this vast quantity of data and to combine it with other data from the earth's surface so that it can be used to inform decision-makers in the most effective way. There may also be gaps in the data, or some of it may be unreliable.

This is a field in which the Department of Engineering Science at Oxford University has considerable expertise. Researchers have already developed sophisticated machine learning tools that are capable of automating and processing large quantities of data from satellite images. The software can reconcile inconsistent data, filter out unreliable sources, and integrate information derived from other sources such as social media. It is even able to interpolate what may lie in the data 'black spots' between known observations, thus 'filling in the gaps' in the overall picture.

In collaboration with several other partners with different types of expertise, Oxford will be bringing these tools to bear on the real-world problems identified in Malaysia, Ethiopia and Kenya, and working out how they can be applied most effectively in these different contexts. In the drought-response work in Ethiopia and Kenya, for example, Oxford engineers will be working with colleagues from the School of Geography and the Environment who specialise in hydrology. Together they will be investigating, with partners in business how to use machine learning to integrate data from satellite imagery of crops with information of both surface and subterranean water resources. Combining views from above and below in this way is more powerful than looking at each one individually, and will create a much more accurate early warning of drought.

Dr Steven Reece, Oxford lead on machine learning, said: "Machine learning is having a positive impact on many walks of life, supporting evidence-based decision making across a wide range of vastly different application domains. Coupled with recent developments in data science and the availability of vast quantities of data, the time is right to deliver truly ground breaking and effective data centred solutions to key societal problems, including natural disasters. Our work will provide a more timely, accurate and detailed understanding of a disaster situation than is currently available."

The two projects, Earth and Sea Observation System (Malaysia) and Earth Observation for Flood and Drought Resilience in Ethiopia and Kenya, will have a total investment of £21 million and have attracted UK Space Agency funding of over £10 million. Both projects are directly relevant to many of the UN's Sustainable Development Goals. It is hoped that successful lessons from the projects can be applied in other areas of the world and to other environmental threats in the future.