This project is currently Active
Vegetation maps provide an important tool to guide management actions for the restoration of terrestrial ecosystems in the Galápagos Islands by addressing questions, such as: 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 there plant species compositions that promote or limit the habitat range of threatened animals?
We use drone and satellite imagery to generate very high resolution maps that show the distribution and abundance of invasive plant species, like blackberry (Rubus niveus), guava (Psidium guajava), Cuban cedar (Cedrela odorata) and quinine (Cinchona pubescens), and key endemic plant species, like Scalesia pedunculata or Miconia robinsoniana.
Principal Investigator
Investigator
Investigator
Mateo Reyes (Technishe Universität Berlin), Jorge Luis Rentería (University of Utah), Daniel Orellana (Universidad de Cuenca)
We mixed a bit of fun flying drones in the fresh air of the Galapagos highlands with many cups of coffee, as we worked on machine learning algorithms behind the computer to arrive at the perfect recipe for mapping several plant species.
Distribution of endemic, native and invasive plant species at visitor site “Los Gemelos” on Santa Cruz, based on a WorldView-2 multispectral satellite image (left) and a photo captured with a DJI Inspire-1 drone (right). Maps: C. Carrión, CDF.For the mapping we are using cutting-edge technology. Our main resources are aerial photos captured with drones (DJI Inspire 1 and Mavic Pro) and satellite images with very high resolution (WorldView-2), donated by the DigitalGlobe Foundation, through Brown University. The photos captured by drones are very detailed, which allow us to visually identify and delineate the plant species present in our sampling area. While the WorldView-2 satellite images have a lower resolution than the drone photos, their potential lies in the fact that they cover the entire study area and are multispectral. This means that they have eight color channels (instead of four as the Google Earth satellite images). Among those eight spectral channels, there are some which cover the infrared range, which is very useful for characterizing plant species with different concentrations of chlorophyll. Combining the information captured with the drone with the color information in the satellite image, we train a statistical model called “Random Forest”. This model identifies a species based on its color characteristics and then extrapolates this information to the entire satellite image. This way, an approximation of the spatial distribution and abundance of each species is being attained.
Diagram of the applied methodology: Based on the photos captured with the drone, individual species are identified and delineated. The delineated shapes are then applied to the part of the multispectral satellite image shown on the right. A classification model based on the color characteristics of each species is created. This model will then be applied to the entire satellite image. Maps: C. Carrión, CDF.We have created maps of the distribution and abundance of the most dominant invasive and native plant species and key endemic species present in the humid zone of Santa Cruz, Floreana, Isabela and Santiago. We are also documenting changes in the plant communities over time by comparing maps generated from current images to those obtained from satellite images from previous years.
Our main goal is to produce detailed vegetation maps, which help guide management actions carried out by the Galapagos National Park Directorate (GNPD) and thus contribute to the conservation of the unique species of Galapagos.
A brief overview of some the maps we have produced within the context of this project for the protected areas.
Primary source (in parenthesis):
Reflects the final resolution of the map produced
Map elaborated by: C. Carrión, CDFKeywords: Invasive plant species, Endemic plant species, Maps, Drone, Multispectral Satellite Imagery
• Alvarez-Taboada, F., Paredes, C. y Julián-Pelaz, J. (2017). Mapping of the invasive species Hakea sericea using Unmanned Aerial Vehicle (UAV) and WorldView-2 imagery and an object-oriented approach. Remote Sens., 9(9), 913. doi:10.3390/rs9090913
• Asner, G.P., Knapp, D.E., Kennedy-Bowdoin, T. et al. (2008). Invasive species detection in Hawaiian rainforests using airborne imaging spectroscopy and LiDAR. Remote Sens Environ.;112(5):1942-1955. doi:10.1016/j.rse.2007.11.016
• Immitzer, M., Atzberger, C. y Koukal, T. (2012). Tree species classification with Random forest using very high spatial resolution 8-band worldView-2 satellite data. Remote Sens., 4(9), 2661-2693. doi:10.3390/rs4092661
• Laliberte, A.S., Goforth, M.A., Steele, C.M. y Rango, A. (2011). Multispectral remote sensing from unmanned aircraft: Image processing workflows and applications for rangeland environments. Remote Sens.;3(11):2529-2551. doi:10.3390/rs3112529
• Liaw, A. y Wiener, M. (2002). Classification and regression by randomForest. R news 2002, 2, 18–22.
• Toral-Granda, M.V., Causton, C.E., Jäger, H. et al. (2017). Alien species pathways to the Galapagos Islands, Ecuador. PLoS One, 12(9), e0184379. doi:10.1371/journal.pone.0184379
• Vilà, M., Espinar, J.L., Hejda, M. et al. (2011). Ecological impacts of invasive alien plants: A meta-analysis of their effects on species, communities and ecosystems. Ecol Lett., 14(7), 702-708. doi:10.1111/j.1461-0248.2011.01628.x