Summer School for Inverse Modeling of Greenhouse Gases (SSIM-GHG) 2024
Motivation
Rising concentrations of CO2 and CH4 are the driver of climate change and reflect a complex mix of natural and human sources and sinks. For example, plants and oceans absorb about half of human CO2 emissions every year, slowing the progression of climate change. Tracking progress to combat climate change requires good quality understanding of sources and sinks and underlying processes, which come from a variety of complex modeling tools that help scientists calculate the exchanges of CO2 and CH4 at the Earth’s surface from atmospheric data. This workshop, supported by the newly-established US Greenhouse Gas Center, aims to develop a future workforce skilled at using existing tools as well as building their own tools to understand sources and sinks of greenhouse gases.
Workshop goals
The goal of the workshop is to present and provide guidance and instruction of the state of the art in atmospheric data assimilation techniques needed to support current and future GHG observing systems. This includes different flux estimation techniques for GHGs, and retrieval techniques for estimating atmospheric GHGs from space-based and surface-based remote sensing platforms.
Sponsors





Presentations
Note: Some videos are concatenated across multiple speakers***
Week One:
Background/Intro – Tuesday, June 11
Instructors: Scott Denning, Sean Crowell, Gretchen Keppel-Aleks, Andy Jacobson
Speaker | Title | Video |
---|---|---|
Scott Denning | Carbon Cycle | Video |
Andy Jacobson, Gretchen Keppel-Aleks | Observations: CO2 through different types of observations | Video |
Scott Denning | Atmospheric transport at different scales | Video |
Bayesian Matrix Methods – Wednesday, June 12
Speaker | Title | Video |
---|---|---|
Michael Bertolacci | Bayesian Statistics Refresher | Video |
Andrew Schuh | Batch Inversion Method | Video |
4DVAR/Variational Methods – Thursday, June 13
Speaker | Title | Video |
---|---|---|
David Baker | 4DVAR/Variational Methods | Video |
Ensemble Kalman Filter Methods – Friday, June 14
Speaker | Title | Video |
---|---|---|
Andy Jacobson | Filtering Techniques | Video |
Week Two:
Introduction to CH4, plume methods, MMRV and Bayesian Hierarchical Models – Monday, June 17
Speaker | Title | Video |
---|---|---|
Alex Turner | Welcome to CH4 | Video |
Alana Ayesse | Plume Methods | Video |
Kim Mueller, Hannah Nesser | MMRV and Policy | Video |
Michael Bertolacci | Hierarchical Bayesian Models | Video |
Trace Gas Retrievals – Tuesday, June 18
Speaker | Title | Video |
---|---|---|
Chris O’Dell | Trace Gas Retrieval Theory | Video |
Field Trip to Niwot Ridge, CO (facilitated by NOAA-GML), Wednesday, June 19
LPDM Methods – Thursday, June 20
Speaker | Title | Video |
---|---|---|
Arlyn Andrews | Introduction to LPDM Methods | Video |
Hannah Nesser | How to handle boundary conditions | Video |
Vineet Yadav | Hierarchy of Steps in Regional Flux Inversion | Video |
Kim Mueller | Engineering a Regional Flux Inversion | Video |
State Reduction Techniques and Speed Talks – Friday, June 21
Speaker | Topic/Slides | Video |
---|---|---|
Hannah Nesser | State Reduction Techniques | Video |
Instructor Speed Talks | Instructor Speed Talk Slides | Video |
Student Speed Talks | Student Speed Talk Slides | Video |
Resources
References
Q&A
Q. Is there agreement in the community on what to set for the ‘sensitivity to surface’ height for LPDMs? In the literature I’ve seen 40m, 100m, boundary layer height, and in the slides here: half boundary layer height. – Daemon Kennett
In short, no. Historically I believe I maybe used 100-150 meters in Schuh et al 2010, 2013. Considering the PBL is reasonably well mixed, half the PBL height may not be terribly different than 100 meters. Plus it would track a collapsing boundary layer at night which might be appropriate. So, at least for STILT, I think it would probably be fine. However, it would be very interesting to see if 0.25 or 0.75 of the PBLH (as opposed to 0.5) would give you any different results.
Q. For surface in situ sites we typically assimilate afternoon observations when the boundary layer is well mixed. Any thoughts on conditions under which night-time observations could be assimilated, to help constrain the full diurnal cycle of fluxes? Maybe high wind-speed conditions? – Daemon Kennett
Yes, agree. I’d think like eddy covariance folks. I’d say set a threshold, wind speed for example, and make sure both model and obs see that level of winds. Turbulence at night should make the obs and model more robust.
Q. It was mentioned that LPDM should only be used for “long-lived” species. What are the limits on this and/or does it depend on your spatial resolution? – Betsy Farris
No, I wouldn’t say “long-lived”. I think the misconception is likely that folding multi-species complex chemistry could be tricky. However, species don’t have to be long lived. A Lot of this original work was based on radioactive species w/ varying half lives. The key is probably “simple” chemistry, decay, or passive.
Toy Example Code
Code for the toy examples presented in class can be found at the following link: https://github.com/US-GHG-Center/ssim-ghg
Please be aware the code and data are preliminary.