Documents to download

Environmental remote sensing involves the use of satellites and other air-borne instruments to collect data about the environment. Substantial quantities of data are produced in this way and environmental remote sensing is now considered an area of Big DataExperts are using artificial intelligence (AI) tools such as machine learning for more efficient data analysis of such data.  

Machine learning algorithms allow a system to learn and improve from data and experience without being specifically programmedreducing the level of human intervention. This data-driven approach means valuable information about a natural phenomenon can be extracted from the data aloneThis has benefits such as being able to manage more complex environmental data but has challenges such as data accessibility. 

Key Points: 

  • Machine learning with remote sensing can help to improve predictions about the behaviour of environmental systems, improve the automation of data analysis, lead to a better management of resources and the discovery of new insights from complex data sets.  
  • Applications at this interface include improved weather forecasting, flood and drought prediction, precision agriculturemanaging forests and in marine conservation and coastal clean-up projects. 
  • Wider implementation for remote sensing is limited by the availability of accessible, and representative datasets for training the machine learning algorithm. Specific challenges include: the availability of Analysis Ready Data (ARD) that requires expertise, time and computational power to prepare; the demands on storage, transfer and processing of large data sets; and, the demands on having an accurate and well-developed training data set. 
  • Inherent to all remotely sensed data used to train and test the models is a level of uncertainty and error including error introduced by a human analyst. Best practice guidelines vary and there is an overall lack of regulation by Government or relevant public bodies. Uncertainties may also be poorly understood by stakeholders 
  • Machine learning models are sometimes referred to as ‘black box’, implying that it is not easy to determine how an algorithm made a decision. For a public sector organisation, where any aspect of a decision is opaque, that raises issues of accountability and having the authority to make a decision. 
  • As machine learning becomes of greater value to the UK economy, challenges could be addressed by stakeholders such as NGO’s, academics and commercial EO companies working with Government to establish good practice codes, standards and regulatory frameworks.

Acknowledgements  

POSTnotes are based on literature reviews and interviews with a range of stakeholders and are externally peer reviewed. POST would like to thank interviewees and peer reviewers for kindly giving up their time during the preparation of this briefing, including: 

Alberto Arribas, Met Office 

Heiko Baltzer, University of Leicester 

John Bloomfield, British Geological Survey 

Julie Bremner, University of East Anglia & CEFAS* 

Rene Breton, University of Manchester 

Sue Chadwick, Pinsent Masons LLP* 

Michael Cross, Rezatec* 

Timothy Darlington, Met Office 

Tony Dolphin, CEFAS* 

Joseph Fennell, University of Manchester 

Karen Frake, Natural Scotland* 

Tanvir Islam, NASA Jet Propulsion Laboratory 

Tom Jones, Satellite Applications Catapult 

Gwawr Jones, JNCC* 

Alexandra Kilcoyne, Natural England 

Peter Kohler, CEFAS 

Stefan Leutenegger, Imperial College London 

Paula LightfootJNCC* 

Encarni Medina-Lopez, University of Edinburgh 

Paul Monks, University of Leicester 

Boguslaw Obara, Durham University* 

Rami Qahwaji, University of Bradford 

John Remedios, NCEO and University of Leicester 

Cristian Rossi, Satellite Applications Catapult 

Edward Salakpi, University of Sussex 

Anna Scaife, University of Manchester 

Ivan Tyukin, University of Leicester 

Hong Wei, University of Reading 

Daniel Wicks, Satellite Applications Catapult 

 *denotes people and organisations who acted as external reviewers of the briefing. 


Documents to download

Related posts

  • The use of technology to perpetrate domestic abuse, referred to as tech abuse, has become increasingly common. Domestic abuse charity Refuge reported that in 2019, 72% of women accessing its services said that they had been subjected to technology-facilitated abuse. Common devices such as smartphones and tablets can be misused to stalk, harass, impersonate and threaten victims. Some groups have raised concerns that the growing use of internet-connected home devices (such as smart speakers) may provide perpetrators with a wider and more sophisticated range of tools to harm victims. How is technology being used to perpetrate domestic abuse, how can this be prevented and what role can technology play in supporting victims?

  • People’s behaviour has a major role in the success of test, trace and isolate programmes. Uncertainty about whether to report symptoms, low perceived risk of COVID-19 disease and concerns about the consequences of self-isolation are among the barriers to adherence. Has the Scientific Advisory Group for Emergencies looked at adherence to TTI? What evidence is there on people’s understanding and willingness to be tested, provide contact details and self-isolate? Is there anything that can be done to improve this?