Document Type


Publication Date



Dryland ecosystems have complex vegetation communities, including subtle transitions between communities and heterogeneous coverage of key functional groups. This complexity challenges the capacity of remote sensing to represent land cover in a meaningful way. Many remote sensing methods to map vegetation in drylands simplify fractional cover into a small number of functional groups that may overlook key ecological communities. Here, we investigate a remote sensing process that further advances our understanding of the link between remote sensing and ecologic community types in drylands. We propose a method using k-means clustering to establish soft classes of vegetation cover communities from detailed field observations. A time-series of Sentinel-2 satellite imagery and a random forest classification leverages the mixing of different phenologies over time to impute such soft community classes over the landscape. Next, we discuss the advantages of using a fuzzy confusion approach for soft classes in cases such as understanding subtle transitions in ecotones, identifying areas for targeted remediation or treatment, and in ascertaining the spatial distribution of non-dominant covers such as biological soil crusts and small native bunchgrasses which have typically been difficult to map with traditional remote sensing classifications. Our pixel-level analysis is relevant to the scale of management decisions and represents the complexity of the landscape. The combination of cloud computing with the spatial, temporal, and spectral observations from Sentinel-2 allow us to develop these ecologically-meaningful observations at large spatial extents, including complete coverage at landscape scales. Re-interpretation of large extent maps of soft classes may be helpful to land managers who need community-level information for fuel breaks, restoration, invasive plant suppression, or habitat identification.

Creative Commons License

Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License

Included in

Biology Commons