Background

Many endangered plants worldwide are specialists on unusual soils that support limited plant cover and species diversity. An example is the dwarf bear-poppy (Arctomecon humilis), which grows only on barren gypsum soils close to a rapidly expanding urban area, St. George, Utah. A clear understanding of its distribution and habitat requirements is key to managing for its survival in the face of urban development. Gypsum soils are so fragile that even foot traffic is damaging, so that dwarf bear-poppy populations have never been adequately censused. With the ‘magic carpet’ of drone technology, we can carry out a census in two days that would take two botanists a month to complete on the ground, with virtually no impact to fragile soils and biological crusts.

 

 

 

Research

Our first step was to obtain imagery at different altitudes to determine how high we could fly our Phantom 3 Advanced drone and still be able to identify poppies in the imagery. We chose 50-m altitude for the census flights, but also obtained imagery at 15-m altitude over smaller areas to check the accuracy of our identifications. Once the complex procedure of processing the imagery was complete, we used ARCGIS mapping software to scan systematically and mark poppies. When we checked the 15-m validation imagery, we found that some plants marked as poppies at 50 m were actually not poppies, and many more poppies were too small to see at 50 m. We corrected our estimates of total numbers based on the validation imagery. We were also able to use the census maps to examine poppy density on different surface types within the gypsum environment. We discovered that poppies prefer eroding slopes where gypsum is directly exposed, and are much less common on soils with dark lichen crusts. Our next steps will address two limitations of the current method: acquisition of higher-resolution imagery to increase accuracy and development of a machine learning approach to automate poppy detection in the imagery.

 

 

Key Findings

• Dwarf bear-poppy, and likely other rare plant species with distinctive morphology and color that are found in simple plant communities of open habitats, can successfully be censused using drone imagery.
• Census maps can be combined with habitat classification to examine fine-scale habitat requirements.
• Drone-based census will provide managers with distribution data over the entire species range, permitting them to better prioritize management activities.
• Drone imagery at higher resolution can also be used to carry out yearly monitoring with an image acquisition protocol that will be economical and user-friendly, so that managers will potentially be able to carry out these activities without expert assistance.