Mapping Ecosystem Services in the tropics using UAVs

Integration of high resolution imagery from UAVs for Mapping of Provisioning Ecosystem Services in the tropics

Ximena Tagle provides a summary of her Ph.D. research topic

I’m halfway of my Ph.D. research at the Laboratory of Geo-information Science and Remote Sensing at  Wageningen University and Research (WUR), in collaboration with the Research Institute of the Peruvian Amazon (IIAP in Spanish acronyms) and the University of Leeds thanks to the WWF Russell E. Train Fellowship (https://www.worldwildlife.org/stories/wwf-welcomes-the-2018-class-of-conservation-leaders#).

The main goal of my research is to combine ground data (field measurements), Unmanned Aerial Vehicles – UAV   RGB images (drone pictures) and satellite images to map the hotspots of palm fruit production (Mauritia flexuosa, Euterpe precatoria and Oenocarpus bataua) in the Loreto region. In order to show that the forests’ economic value can come from non-timber products and encourage conservation of the Peruvian Amazon.

Ximena flying the Phantom 4 RTK UAV in the Peruvian Amazon

For this, I have 3 research objectives:

With 4 research questions:

1. Is it possible to semi-automatically identify and quantify palm tree species using UAV imagery in a tropical rainforest?

We answered this question in the paper published last year: https://www.mdpi.com/2072-4292/12/1/9

2. How feasible is the Automatic detection and quantification of economically important palm species in a “larger scale” using UAVs?

I’m using Deep Learning to detect economically important palm species in 100 – 200 hectares per mosaic, in different areas of Loreto. This publication will come soon!

In addition, one of my MSc. Students developed a QGIS plug-in incorporating this model to detect Mauritia flexuosa, and soon also Euterpe precatoria and Oenocarpus bataua. This interface is user friendly and it is been used by the Protected National Areas Service in Peru (SERNANP acronym in Spanish).

Ernesto Fernandez from SERNANP showing the results obtained in a mosaic from the Madre de Dios region in Peru using the QGIS plug-in.

3. How accurate can UAV imagery determine palm fruit production semi-automatically?

I am trying to automatize this detection by using deep learning as well.

4. What is the spatial variability of the palm tree fruit production in the Loreto region? Is it possible to identify hotspots of palm tree fruit production?

By combining the UAV, ground data and satellite imagery, the idea is to generate a map of hotspots of palm fruit production for the Loreto region.

This research is a joined effort and would have not been possible without the support from my supervisors Martin Herold, Harm Bartholomeus, Tim Baker; IIAP collaborators Dennis Del Castillo and Euridice Honorio, and especially without the help of all the field team, including the local communities and the Protected Natural Areas Service (SERNANP) staff.

Leave a comment