Understanding and managing forests is crucial to understanding and potentially mitigating the effects of climate change, invasive species, and shifting land use on natural systems and human society. However, collecting data on individual trees in the field is expensive and time consuming, which limits the scales at which this crucial data is collected. Remotely sensed imagery from satellites, airplanes, and drones provide the potential to observe ecosystems at much larger scales than is possible using field data collection methods alone. This data science competition focuses on using remote sensing data to quantify the locations, sizes and species identities of millions of trees and on determining how these methods generalize to other forests.

Enter your email to be notified of updates and keep reading below for more details.

What is it?

  • Data science competition to develop and/or test algorithms on two tasks (1) delineation of tree crowns and (2) classifcation their species identity on airborne remote sensing images
  • Participants can engage in one or both of the competition tasks
  • Uses 0.25 – 1 m resolution remote sensing data (RGB, lidar and hyperspectral) from three forest sites of the National Ecological Observatory Network (NEON; www.neonscience.org/data-collection/airborne-remote-sensing)
  • Cross-site testing:  how well do algorithms developed at one site work at other sites?

Why?

Understanding and managing forests is crucial to understanding and potentially mitigating the effects of climate change, invasive species, and shifting land use on natural systems and human society. However, collecting data on individual trees in the field is expensive and time consuming, which limits the scales at which this crucial data is collected. Remotely sensed imagery from satellites, airplanes, and drones provide the potential to observe ecosystems at much larger scales than is possible using field data collection methods alone. This data science competition focuses on using remote sensing data to quantify the locations, sizes and species identities of millions of trees and on determining how these methods generalize to other forests.  The competition capitalizes on the idea that people from various fields and backgrounds (data and computer scientists, ecological remote sensing specialists, computational biologists) can advance this field more quickly than groups working on their own.

What do I have to do?

  • Overall: based on training data from 2 sites, develop algorithms that can map tree crown boundaries and species identities from airborne image data at 3 NEON sites.
  • Crown delineation task: training and output data are boundary boxes that delineate the boundaries of individual tree crowns
  • Species classification task: training data are bounding boxes with species already identified.  Output data will be classification of species on bounding boxes of unknown species identity
  • Cross-site comparison:  performance will be tested on the same sites for which the training data are provided as well as one additional site from which none of the training data was provided
  • Opportunity to publish results in a collection of papers from the competition (see link to previous competition)

Timeline

  • 1 Feb 2020: competition announced and participants solicited
  • 1 Mar 2020: detailed instructions, training and image data provided to participants
  • 1 Apr 2020: submission process open for live evaluations
  • 1 May 2020: deadline for participants to submit data for evaluation
  • 8 May 2020: announcement of results/winner and organization of manuscript collection
  • 15 May 2020: deadline for participants to sign up for manuscript submission

Find out more

Here are the results from the 2017 competition

Enter your email to hear more