FLUXNET-CH4 V2.0 Data Processing Workshop
On 21-23 October 2024, the hybrid FLUXNET-CH4 V2.0 Data Processing Workshop took place in Berkeley, California, as well as virtually on Zoom. This workshop focused on flux data processing, highlighting quality assurance/quality control (QA/QC), to facilitate the inclusion of new sites into FLUXNET-CH4 V2.0 and the new FLUXNET Data System Initiative. In total, approximately 180 scientists participated over the three days from various places across the globe; mostly the Americas, but also as far afield as Asia, Africa, and Europe
Workshop Objectives
The objectives of this gathering were as follows:
- To provide an overview of FLUXNET-CH4 V2.0 and the FLUXNET 2025 Data System Initiative;
2. To provide instruction and tutorials to facilitate the inclusion of new or existing sites into FLUXNET-CH4 V2.0 and the FLUXNET 2025 Data System Initiative;
3. To provide guidance on gap-filling, and uncertainty quantification for flux tower measurements, with an emphasis on CH4 fluxes;
3. To facilitate the inclusion of CH4 fluxes in the ONEFlux pipeline, thereby informing the next generation of the FLUXNET database;
4. To foster existing and new international community collaborations by bringing
together students/postdocs and PIs from across the globe.
Workshop Summary
All workshop sessions were recorded on Zoom, and these are now available in a playlist on the newly created EcoFlux Lab YouTube channel.
The morning session on the first day began with presentations introducing FLUXNET-CH4 V2.0 and the FLUXNET Data System from Sara Knox, Rob Jackson, Trevor Keenan, Housen Chu, and Dario Papale. This was followed by Sara Knox’s assessment of the past successes of FLUXNET-CH4 V1.0, and Qing Zhu along with Gavin McNicol covered regional to global upscaling, and multiscale analysis and modeling. To round out the session, high frequency data considerations and opportunities were presented, with a discussion on frequency response corrections led by Toprak Aslan, and the detection of ebullitive methane fluxes presented by Ben Runkle.
The Zoom chat was busy with discussion throughout this morning session; with ~120 scientists online, and about 25 in-person. The other workshop sessions had between 60–80 scientists online, with 10–20 in-person, many of whom were returning participants. There was a wide range of scientists including undergraduate, masters, and PhD students, as well as postdocs, research associates, professors, technical staff, and representatives from regional flux networks and FLUXNET.
Excerpt from Toprak Aslan’s presentation on Frequency Response Correction of Noisy Measurements
Leveraging High Frequency Data to Detect Ebullitive Methane Fluxes; excerpt from presentation by Ben Runkle
The day one afternoon session revisited high frequency data with a presentation on EddyPro automation and streamlining from June Skeeter. Following this, the focus turned to data formatting, specifically how to process flux data including QA/QC, to prepare datasets for submission to regional networks, such as AmeriFlux. Rosie Howard presented a summary of the flux data cleaning pipeline developed by the EcoFlux Lab located at McGill University in Montreal, Canada. Next, attendees were given demonstrations of two in-house data visualization tools, a Matlab graphical user interface used for data inspection and analysis during data cleaning, created and presented by Paul Moore, and an RShiny app intended for displaying real-time flux data, written and presented by Sara Knox.
Next in the afternoon session, Rosie Howard led participants through the data cleaning process step by step in a hands-on tutorial, following the lab’s online documentation. During this interactive portion of the workshop, extra support was provided via the Zoom chat and moreover in Zoom breakout rooms, as well as in-person, by EcoFlux Lab members Zoran Nesic, Paul Moore, June Skeeter, and Sara Knox. This allowed participants who were experiencing problems along the way to receive immediate assistance, while the remaining attendees could continue on following the main tutorial. The data cleaning pipeline primarily uses Matlab; during the third stage cleaning, Matlab interfaces with some R-code to run the REddyProc package for USTAR-filtering, gap-filling, and CO2 flux partitioning. Participants were also provided with access to the necessary open-source data pipeline library, developed by the EcoFlux Lab.
To extend accessibility, regarding time zones and scheduling in general, the data cleaning pipeline presentation, visualization tools demonstrations, and interactive tutorial were repeated twice on day two of the workshop, once during the morning and then again in the afternoon. This also allowed earlier participants to return and solve issues during the later sessions, adding the helpful element of repetition learning.
On the morning of day three, which was the final session of the workshop, the topic shifted to CH4 flux gap-filling and uncertainty quantification. Current approaches were presented by Gavin McNicol, then Fa Li led a tutorial on implementing these approaches; Matt Fortier then provided details of new AI approaches for CH4 flux gap-filling. Finally, the workshop concluded with a lively and open discussion on next steps for implementing CH4 fluxes in ONEFlux, including identifying the best CH4 flux gap-filling algorithm and implementing the full uncertainty quantification for CH4. This was moderated by Sara Knox, Gavin McNicol, Dario Papale, and Gilberto Pastorello.
Outcomes and Next Steps
Following the workshop, the EcoFlux lab received a lot of positive and constructive feedback, mostly via a Google form. The lab has provided an email address to all workshop participants in case of follow-up questions or if any issues arise when using the pipeline to clean their own data. Requests are being actively tracked and follow-up assistance provided as far as possible. Additionally, the EcoFlux lab YouTube channel now has nine subscribers, and the playlist has almost 250 views.
Development of the data cleaning pipeline is ongoing; the pipeline documentation, initially presented in a “quick start” style, is now being expanded to help answer many of the requests received, as well as incorporating constructive suggestions, more details in general, and adding in advanced topics. Additionally, as a long-term goal, the EcoFlux lab is planning to convert the data pipeline code from Matlab to open-source Python code for wider accessibility.
Overall, the scale of outreach as a result of the workshop has been large, spanning multiple countries, continents, and climate regions. Many connections have been made helping to foster international community collaborations, which hopefully will result in new and updated data submissions to regional networks, to ultimately lead to a more complete global understanding of flux measurements and related uncertainties.
Sponsorship
Thanks to the FLUXNET Cooperation Project (funded by NSF), the Global Carbon Project, the Moore Foundation, Environment and Climate Change Canada, and WWF-Canada for supporting this workshop.
Workshop leaders and presenters:
Thanks to the workshop co-organizers: Sara Knox, Rob Jackson, Alison Hoyt, Kyle Delwiche, and Gavin McNicol.
Workshop presenters and Zoom breakout room support:
Sara Knox, Rob Jackson, Trevor Keenan, Dario Papale, Housen Chu, Qing Zhu, Gavin McNicol, Toprak Aslan, Ben Runkle, June Skeeter, Rosie Howard, Paul Moore, Zoran Nesic, Fa Li, Matt Fortier, and Gilberto Pastorello.