• A POSTnote on machine learning for environmental remote sensing will focus on setting out the ways in which machine learning analysis adds value to remote sensing data.
  • This POSTnote will examine the environmental applications of this interface.
  • It will look at the challenges facing the effective use of machine learning and the effective policies for realising opportunities.
  • In production. To contribute expertise, literature or an external reviewer please contact Berivan Esen.

Machine learning efficiently processes data while learning from it, leading to a reduction in data collection and processing times. Advances in this area are being applied to the monitoring and management of environmental issues. This POSTnote will examine how machine learning is improving environmental data collection, the challenges involved, and the effective policies for taking advantage of opportunities.

Remotely sensed data is collected by satellites and other air-borne instruments. This big data area (POSTnote 468) is suited for machine learning algorithms which can manage and learn from large data sets. For example, instruments like the Sentinel-1 satellite, (which is part of the European Space Agency’s and European Commission’s Earth observation programme) produces more than 10TB of Earth observation data per day, which is ideal for machine learning applications.

As a result of machine learning, image classification and instrument normalisation (sensors adapting to changes in light, weather, lenses) is becoming more automated.  Applications include a wide range of environmental issues, including water resource management, flooding and urban planning. With this technology, the need for human analysts, the supervision of data collection and labour-intensive ground work is reduced, but pre-processing of data remains a significant requirement.

Machine learning and remotely sensed data can be applied to measuring progress towards forest management and precision agriculture targets. For example, the Committee on Climate Change (CCC) report in January 2020 set out agricultural and land use policies to meet the Governments goal of net zero greenhouse gas emissions for the UK by 2050. There are also internationally relevant policy targets such as the UNFCCC, Reducing Emissions from Deforestation and forest Degradation (REDD+) program (POSTnote 466), to which the UK is a contributor.

The aims of this POSTnote are to provide MPs and Peers with an overview setting out the ways in which machine learning analysis adds value to remote sensing data. This POSTnote will also examine the environmental applications of this interface, the challenges facing the effective use of machine learning and effective policies for realising opportunities.