Mapping Invasive Plants to Inform Restoration Efforts in Galapagos

The Galapagos Islands are a living laboratory and a place to observe evolution in real time. Most of the unique species that inspired Darwin’s “On the Origin of Species” are still present in the different ecosystems. However, almost 60% of the Galapagos flora is introduced (about 810 species vs. 550 native species). These species were brought in by humans, either intentionally or accidentally, and many of them are beneficial, like agricultural and ornamental plants. But 6% have become seriously invasive, like blackberry (Rubus niveus), quinine (Cinchona pubescens), guava (Psidium guajava) and Cuban cedar (Cedrela odorata). These species are threatening the native and endemic plant species, since they compete for space, light and nutrients. To evaluate the threat that these invasive species pose to the local flora, we have to know their distribution and abundance.

Some of the invasive plants in Galapagos. Left to right: Blackberry (Rubus niveus), quinine (Cinchona pubescens), guava (Psidium guajava) and Cuban cedar (Cedrela odorata).

The Charles Darwin Foundation (CDF), in collaboration with the Galapagos National Park Directorate (GNPD), is mapping the invasive plant species in the humid zone of the Galapagos Islands. The area to map within the National Park is large and often difficult to access and some of these invasive species occur in small patches. Therefore, the team needed satellite imagery of very high spatial and spectral resolution to assess those areas, which was generously provided through a grant from Maxar.

Maxar’s optical satellite images have the natural color spectral bands (red, green, and blue) but also include infrared and short-wave infrared spectral bands, which are very useful for recognizing different plant species due to differences in their spectral characteristics. To gather data for the machine learning model that would ultimately create maps of where invasive species are located, the CDF team flew a drone to obtain footage that allowed for the identification of plant species visible at the top of the canopy. The species identified were delineated onto the satellite imagery and served as training data for the model.

Diagram of the modelling process using drone and satellite imagery

The use of imagery from the WorldView constellation and machine learning allowed for the detection of invasive plant species, not only in large monospecific stands but also within mixed forests. This is particularly useful for the detection of individual crowns of the Cuban cedar tree and blackberry, which often occur in small patches and would not be detectable in satellite images of a lower resolution. The end result were maps of the invasive species based on the satellite imagery, covering an area of 55 square miles on Santa Cruz and 15 square miles on Floreana.

The maps created by CDF help address conservation priority questions like:

  • What is the approximate area covered by an invasive plant species and how much effort would it take to control it?
  • How are plant species compositions changing over time?
  • Are the conservation management activities successful?
  • Are there plant species compositions that are promoting or limiting the habitat range of endangered animals?

These maps allow for a more accurate estimation of the actual distribution and abundance of plant species, which in turn can guide restoration efforts and preventions of new invasions.