UK Land Cover Map 2015

 

One of the main reasons why I moved to the UK and started working at the CEH was the opportunity of being involved in the creation of the UK Land Cover Map 2015 (LCM2015). This is a nation-wide habitat mapping project that will have a big impact in future UK research and it will be used by government departments, environmental management agencies, charities and many more. After spending all 2016 and part of 2017 working on it, the LCM2015 was finally released last April.  

                        

LCM2015 is derived from satellite images (mainly Landsat-8) and provides land-cover information of the whole UK. The main product is a 22 land-cover class vector map, based on a parcel-based spatial framework, from which a 25m-pixel raster product is derived. However, many other secondary products at different scales can be obtained. 

One of my favourite features about the LCM2015 is that the algorithm used to predict the land cover type (based on satellite spectral data and other ancillary information) is a Random Forest. As you probably know, this is a classificatory algorithm that I have been using for a long time in my research on animal distribution modelling. It is also becoming very popular to perform supervised classifications of satellite images, and we used it for the LCM2015, obtaining great results. Using this algorithm allowed us to provide a "probability map", giving the user an estimate of how accurate is the assignment of each parcel or pixel to a land-cover type. I think this is going to be a key feature of future LCM's, as it will allow us to add the uncertainty of the habitat identification, when using remote sensing derived data, into our species distribution models. 

The production of the LCM2015 meant many months of hard work, but it has been a great experience for me. Now, I can't wait to start exploring its potential for ecological research, especially for bird distribution research in the UK (stay tuned!). Here I link a short video of the LCM2015, created by the CEH (I did that 3D model!):