Mapping biological soil crusts in a Hawaiian dryland

ABSTRACT

Historical and ongoing land use patterns in the Hawaiian Islands have degraded the Islands’ drylands, causing erosion and detrimentally affecting adjacent coastal marine ecosystems. Biological soil crust (biocrust) communities have been shown to increase soil stability in drylands worldwide, but their efficacy in mitigating soil erosion in Hawaiian drylands is largely unknown. Using a combination of field data and imagery collected by small unmanned aerial systems (sUAS), we mapped biocrusts and examined their influence on soil stability in the Kawaihae watershed, an erosion-prone dryland on leeward Hawai`i Island. We created classified maps of biocrust cover from imagery collected at three spatial resolutions (1.2, 2.1 and 2.8 cm/pixel) using the pixel-based Support Vector Machine (SVM) classifier and investigated the impacts of spatial resolution and biocrust level of development on classification accuracy. Our medium (2.1 cm) resolution image produced the highest overall classification accuracy when biocrust was treated as a single class (82.1%). We explored the spatial impacts of biocrusts on soil loss via sUAS-derived measurements of elevation change over a four-year time span. We found differences in soil loss among land cover types, but robustly quantifying these was a challenge, as much of the change fell below statistically significant limits of detection. We investigated the relationship between biocrust development and soil stability by conducting soil aggregate stability testing at the three biocrust levels of development (LODs) present at the study site. We found a significant increase in soil stability from soils without surface biocrusts (LOD score of 0) to those with biocrusts at any development level (LOD 1–3). Our research adds to the body of biocrust knowledge by presenting new information about biocrust distribution and soil stabilization capabilities in Hawaiian drylands. We also provide insights into the trade-offs between spatial resolution and classification accuracy for biocrust classification and land cover analysis.

Acknowledgements

We’d like to thank the Parker Ranch for providing study site access, the Hawai`i Geographic Information Coordinating Council and the Society for Conservation Biology for providing financial support, and James Melcher and Aloha Kapono for providing project support. We are grateful to M. Villareal for comments on a previous draft of the manuscript. S.R. acknowledges support from the U.S. Geological Survey Ecosystems and Core Science Systems Mission Areas. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available via email from the corresponding author, E. Collier, upon reasonable request.