Science challenge

In order to manage UK land for a carbon neutral future and net biodiversity gain, detailed, accurate knowledge of the land surface is essential.

Project summary

UKCEH's Land Cover Map (LCM) is an uninterrupted national dataset of land cover classes from grassland, woodland and fresh water to urban and suburban built-up areas. The maps provide essential information for government agencies, water companies, land managers, NGOs and researchers about the state of the British landscape, locally or nationally, and how it is changing. Knowing what habitats there are, and where, is crucial for planning developments and environmental improvements.

To create the Land Cover Map, we use time series of satellite data, mostly obtained from the Google Earth Engine. This spatial data is integrated with training data, automatically derived from our historical series of Land Cover Maps, and a machine learning algorithm, to produce a new map. Therefore each new map is produced from the previous one in a process known as bootstrap training. To ensure good quality land cover products we produce a new map every year. 

Validation of the Land Cover Map product is carried out by systematic visual inspection using online image resources and harvesting data from a wide range of other field activities. The UK-SCAPE Land Cover team actively engages with field-based surveying communities through the UK Environmental Observation Framework to advise and encourage data collection suitable for satellite-based vegetation assessments.

Objectives

• To provide a high quality, comparable, satellite-derived time series of UK land cover maps
• To produce satellite-derived metrics of habitat condition and function
• To integrate time series maps of habitat distribution, stock, condition and function to support understanding of UK land cover dynamics.

Resources

Papers and reports

• LCM2017, LCM2018 and LCM2019 dataset documentation (zip 1.34Mb)
• Land Cover Change 1990 to 2015 dataset documentation (zip 938kb)
• Revised Land Cover for 1990 dataset documentation (pdf 1.34Mb).

Data

The following products and associated data sets have been produced so far. The project will produce a new land cover map every year. In addition we are developing advanced methods for creating our land cover maps:

• Latest Land Cover Map: 2021
• Land Cover Maps for previous years: 2017 to 2020
• An assessment of land cover change from 1990 to 2015
• A revised Land Cover Map for 1990 designed to be comparable to recent land cover products.

Note: To obtain Land Cover Maps, please contact the UKCEH Data Licensing Team using our Data Request Form or by emailing Spatial Data for more information. The way we license our Land Cover Maps depends upon how the dataset will be used.

Models

• We have developed an in-house machine-learning suite for supervised classification of land cover. Software is written in Java and uses a machine learning toolkit from the University of Waikato, New Zealand. We are able to use a range of classification tools but have found the Random Forest classifier to be superior
• We are looking to translate our classification functionality into a cloud-processing environment. In the long-term this will lead to near real-time satellite-derived assessment of the land.

Research facilities

• This work is carried out on a high powered modelling PC, with satellite data derived and downloaded from Google Earth Engine. However, to realise long-term objectives we are investigating the use of high powered, cloud processing.

Interactions

• UKCEH national capability: Land cover information is a fundamental and generic resource. These data are used across the UKCEH National Capability programme
• UKCEH co-aligned research: LCM was incorporated into work undertaken in the UKCEH-Defra Memorandum of Agreement (MoA) project.