The Ecosystem Monitoring and Management Application (EMMA) Workflow: Automating Change Detection and Reporting in the Hyperdiverse Fynbos of South Africa
Brian Salvin Maitner, University at Buffalo, bmaitner@gmail.com (Presenter)
Jasper Slingsby, University of Cape Town, South Africa; South African Environmental Observation Network, Cape Town, South Africa, jasper.slingsby@uct.ac.za
Glenn R Moncrieff, South African Environmental Observation Network, Cape Town, South Africa; University of Cape Town, South Africa, gr.moncrieff@saeon.nrf.ac.za
Yingjie Hu, University at Buffalo, yhu42@buffalo.edu
Yue Ma, University at Buffalo, yma28@buffalo.edu
Adam Wilson, University at Buffalo, adamw@buffalo.edu
Open ecosystems are ecologically and economically important and house more than 40% of total ecosystem carbon. However, they are also disproportionately at risk due to threats including land use change, fires, and alien plant invasions. There is thus a pressing need for methods to rapidly detect threats in near-real-time and inform relevant authorities and land managers. This continually-updated change detection has proven challenging in open ecosystems due to their dynamic and strongly seasonal nature. Further, there are informatics challenges with running a frequently-updated workflow that integrates multiple data products on a near-real-time basis (e.g. daily or weekly).
Here, we present the Ecosystem Monitoring and Management Application (EMMA) workflow which tackles these challenges in an Open Science framework. The goal of the EMMA workflow is to automate ecosystem change detection and reporting in one of the world’s biodiversity hotspots, the Fynbos of South Africa. The workflow has four main components: 1) a data collection, integration, and processing module; 2) a modeling and change detection module, which uses hierarchical Bayesian methods to model vegetation changes over time and detect deviations from expectations; 3) a change classification module which uses machine learning techniques to categorize changes according to likely causes (e.g. invasion, fire, land clearing); and 4) a module that generates protected area reports for land managers and other relevant authorities. The reports have been co-development with stakeholders, ensuring they are relevant to management needs. The workflow is publicly available via GitHub, and can be readily modified to focus on other regions, providing a generalizable solution to change detection in open ecosystems.
Poster: Poster_Maitner__56_35.pdf
Associated Project(s):
Poster Location ID: 3-46
Presentation Type: Poster
Session: Poster Session 3
Session Date: Thu (May 11) 3:00-5:00 PM
CCE Program: BDEC