Seabed mapping data from remote sensing and groundtruth samples taken at East of Haig Fras Marine Conservation Zone from 2012 to 2015

As described in the manuscript entitled "Limitations of predicting substrate classes on a sedimentary complex but morphologically simple seabed" remotely sensed variables and ground truth samples were distributed as part of a seabed mapping exercise. This holding contains the four raster layers provided to all participants for the East of Haig Fras MCZ and the groundtruth samples. These four rasters are: Multibeam echosounder bathymetry, multibeam echosounder backscatter, modelled maximum current velocity at the seabed and modelled mean current velocity at the seabed. All rasters are in UTM29N projection at a 2 m x 2 m resolution. The groundtruth samples were derived from drop camera still images of rock substrate and sediment substrate samples were derived from grabs classified using Particle Size Analysis. Groundtruth samples contain the field "Test". Test values of 0 were used to train the models and values of 1 were withheld from all analysts as an independent test set. Multibeam echosounder data were acquired during the UK MPA programme in 2012. Currents velocity data were modelled in 2019 using a Telemac 2D model. Groundtruth samples were acquired over three separate surveys by Cefas aboard the RV Cefas Endeavour. Samples were subsequently filtered for the purposes of examining different seabed mapping approaches. This dataset includes only those samples relevant to the different substrate classes having also applied a minimum distance between samples. Data is as presented to the participants in the mapping exercise described in "Limitations of predicting substrate classes on a sedimentary complex but morphologically simple seabed".