Land cover and land use information to support UNFCCC emissions reporting

This service will create annual land cover maps in order to identify land use change and subsequent emissions changes.

The following presentation on this service was given at the MELODIES: Exploiting Open Data conference in October 2016. Slides are available here.

Service Q&A

Who are the target users of this service?

Government departments (and their contractors) who have a responsibility to compile information to support UNFCCC emissions reporting.

How will these users access this service?

Through a working relationship with project partners. The dataset produced during MELODIES will also be available through the CEH Environmental Information Platform

What products does this service provide?

A map of consistent annual land cover - on a 500m resolution grid covering the whole of UK. Feasibility for scaling up to include other countries will be investigated in the latter stages of the MELODIES project.

How will these products benefit users?

One of the critical uncertainties in greenhouse gas emissions is annual land cover and land-use change. At present, the highest quality and resolution land cover product for the UK is produced only once every 7-10 years. This service will produce annual land-cover maps. This will provide a step-change in inventory methodology and substantially reduce current uncertainties in greenhouse gas emissions.

Which Open Data sources drive this service?

What processing is performed on this data?

Firstly, the satellite input data are screened for clouds and a time-series of surface 'colour' is compiled.

Then we use a machine-learning algorithm called a support vector machine to analyse this time-series and intelligently assign land cover categories. In order to train the support vector machine we have to be certain of a few pixels and know the specific class to which they belong (e.g. grassland, cropland, urban area etc), for this we use the high-resolution CEH land cover map.

A whole time period (2000 - 2010 initially) is classified and output as gridded land cover products.

We are currently working on the algorithm in order to determine plausible land use changes from the satellite input data.

The processing chain that we're developing is designed to be generic and transferable. In principle we can extend this to other geographic areas or time periods (given the required ancillary data), or use sensors other than MODIS.

How Open Data has improved this service

The Open data used in this service is provided from large data-centers, which makes accessing the data very straightforward. The openness of the data also means that others have also used and validated the data, giving usĀ  greater awareness of any problems and also more confidence in its quality.

How the Shared Platform has improved this service

We are developing in-house. In time, we will use the shared platform to determine the computational cost of scaling the processing up to larger geographic areas, without having to over-size our in-house computing resources, simply to do the appropriate tests.