Skip to main content

Data from: Assessing spatial patterns of soil erosion in a high‐latitude rangeland

Cite this dataset

Streeter, Richard T.; Cutler, Nick A. (2020). Data from: Assessing spatial patterns of soil erosion in a high‐latitude rangeland [Dataset]. Dryad.


High‐latitude areas are experiencing rapid change: we therefore need a better understanding of the processes controlling soil erosion in these environments. We used a spatiotemporal approach to investigate soil erosion in Svalbarðstunga, Iceland (66° N, 15° W), a degraded rangeland. We used three complementary datasets: 1) high‐resolution UAV imagery collected from 12 sites (total area ~0.75 km2); 2) historical imagery of the same sites; and 3) a simple, spatially‐explicit cellular automata model. Sites were located along a gradient of increasing altitude and distance from the sea, and varied in erosion severity (5–47% eroded). We found that there was no simple relationship between location along the environmental gradient and the spatial characteristics of erosion. Patch‐size frequency distributions lacked a characteristic scale of variation, but followed a power‐law distribution on five of the 12 sites. Present total eroded area is poorly related to current, site‐scale levels of environmental stress, but the number of small erosion patches did reflect site‐level stress. Small (< 25 m2) erosion patches clustered near large patches. The model results suggested that the large‐scale patterns observed likely arise from strong, local interactions, which mean that erosion spreads from degraded areas. Our findings suggest that contemporary erosion patterns reflect historical stresses, as well as current environmental conditions. The importance of abiotic processes to the growth of large erosion patches and their relative insensitivity to current environmental conditions makes it likely that the total eroded area will continue to increase, despite a warming climate and reducing levels of grazing pressure.


See paper for details. All images collected in August 2017.

Usage notes

RG-NIR image

Images are 5000 x 5000 pixels in size, and each pixel is 0.05 m in size. (Site A is 4800 x 4800 pixels, at the same scale).

Filenames are in the format:  site_sitename_RGNIR.tif

RGB images

The RGB images are not necessarily aligned precisely with the RG-NIR images, but are cover a similar area. RGB images were not used in generating the thresholded images, and are for visual comparison only. Scale is 0.05 m pixel.

Filenames are in the format:  site_sitename_RGB.tif

Thresholded landcover images

These are based upon an NDVI calculated from the RG-NIR images. The NDVI values are then thresholded into vegetated/non-vegetated categories – see paper for details on this process. Pixels with values 255 represent eroded/non-vegetated areas, and pixels with value 0 are vegetated areas. Pixel size is 0.05 m.

Filenames are in the format:  site_sitename_thresholded.txt

Site Location data

The coordinates for the centroid of each site are given in site_locations.csv


University of Cambridge

Office of Polar Programs, Award: 1202692