Abstract
Carbon sequestration is one of the key ecosystem services provided by tropical forests. Its accurate assessment is an important task, as the amount of carbon a forest can sequestrate is an indicator of ecosystem health as well as a contribution to maintaining of environmental sustainability. Nowadays, remote sensing has become a widely used technique to acquire data for ecosystem service valuation (ESV). However, in part due to its abundance, it is impossible to get insights from the data without the use of strong computational algorithms, such as Machine Learning (ML). Accordingly, the present study considers the following three questions: (1) what are the specific methods necessary to preprocess EOS data, making it suitable for ESV? (2) what ML algorithm performs best in forest health assessment when using EOS data? and (3) what ESV model can be recommended based on the outcome of the ML experiments? In the presentation, we discuss challenges of preprocessing EOS data that would make it appropriate for the assessment of specific ecosystem services. We investigate different ML algorithms for classifying the forest stands and experiment with tuning of the selected algorithm to fit the problem. Finally, we provide recommendations on an ESV method most suitable for assessing carbon sequestration, based on the output that the algorithm generates.
Presenters
Polina SysoevaGraduate Student, School of Information Studies, York University, Ontario, Canada Peter A. Khaiter
Associate Professor, School of Information Technology, York University, Ontario, Canada
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Sustainable Management, Remote Sensing, Earth Observing System, Machine Learning
