Volcanic (LIP) gas fluxes in geological history using geochemical models
PI: Isabel Fendley (Geosciences)
Plan for funding tuition for graduate students, or the remainder of the researcher’s salary for postdoc and research faculty: Partial TA position and/or start up funds; or summer project only (for 2 summers)
Large igneous province (LIP) eruptions are among the largest volcanic events in Earth’s history and are frequently linked with mass extinctions, ocean anoxic events, and severe climate change. LIPs affect the environment through the release of climate-changing gases such as SO2, whose effects occur on geologically short timescales (<5000 years), and CO2, which has effects on both short timescales (<100 years) and very long timescales (>100 thousand years). In our recent work, we have demonstrated that volcanic eruptions affect the Earth system, especially the ocean carbon cycle, on this entire timescale range, even if the individual eruptions last for only a few hundred years. Thus, to understand the relationship between LIPs and environmental change, it is necessary to constrain volcanic gas fluxes at high temporal resolution (10-100 yr resolution). This is challenging because LIPs erupted millions of years ago, and large portions of the LIP lavas are eroded, on the seafloor, or otherwise inaccessible.
One way to estimate LIP gas emissions is to measure the mercury concentration within sedimentary rocks during the time of LIP activity. Volcanoes emit mercury gas, and LIP eruptions have been found to coincide with global increases in mercury content in rocks. We use box models of the global mercury cycle to estimate the flux of LIP mercury required to produce observed mercury concentration changes in sedimentary records. Because volcanic mercury has a geologically short residence time in the surface environment, we can estimate LIP gas emissions at high temporal resolution with this proxy.
We have successfully applied this technique to individual and small clusters of eruptive events (Fendley et al., 2019; Fendley et al., 2024), showing the promise of this approach. As a next step, we have recently developed a Bayesian inversion framework to compare an actual mercury data timeseries (with non-uniform time sampling) with many synthetic mercury records with prescribed eruptive histories and statistically evaluate which are most similar. This results in an estimate of the volcanic mercury gas emissions over an entire record with many eruptions, rather than individual events.
However, a key scientific challenge with this approach is to analyze the results in a statistically robust manner and assess both the volcanic gas fluxes (using Hg) and the Earth system response (e.g., temperature, pH, carbon isotopes) in intermediate complexity Earth system models. Because we need to estimate gas fluxes at a scale of <100 years, the model runs need to have small time-steps (<1 year) but long durations (millions of years). Additionally, for robust inversion with uncertainty estimates, we need to construct millions of synthetic eruptive histories by varying the size and duration of eruptions, how frequently they occur, and allowing for the possibility of multiple long (>100 thousand years) hiatuses. Using the results of the Hg and Earth system models, we can generate synthetic multi-parameter sedimentary records sampled at the same chronological points as the actual/measured timeseries to account for sampling resolution and smoothing. A key requirement for doing this work for multiple records for different LIPs is to use high-performance computing, since each of thousands of model scenarios can take multiple hours to run on a single CPU. Hence, ICDS computational resources are needed for full inversions.
The key goals of this project are a) to finalize the framework for data-model comparison (e.g., evaluate parameter choices, test various metrics for statistically comparing records), b) optimize the Earth system and Hg cycle code for computational efficiency and the same for parameter sampling in the Bayesian framework, and c) set up and run the model inversion on the Roar Collab Cluster.
Level of effort: 1 year of graduate student funding @ 25% effort, or two summers
Specific areas of computational and/or data science expertise or skills needed: Experience with Python, coupled ordinary differential equation models, and data inversion techniques.
Requirements or expectations of ICDS Junior Researcher: Availability to meet regularly with the PI during the project. Expertise with cloud computing/parallel model runs and data inversion techniques. Limited expertise in geochemical modeling is ok for the project since the model frameworks can be learned by the Junior researcher during the project.
Specific objectives supported by this call: a) to finalize the Bayesian framework for datamodel comparison of Hg records, b) optimize the Earth system and Hg cycle code for computational efficiency and the same for parameter sampling in the Bayesian framework, and c) set up and run the model inversion on the Roar Collab Cluster. The final goal would be a scientific publication with the model results as well as preliminary results for future proposals.
Medium to long term goals: Use results from this study to submit a collaborative proposal to NSF EAR, in particular the P4Climate (Paleo Perspectives on Present and Projected Climate) call (annually in October), as LIPs are commonly used as analogues for anthropogenic climate change.
Statement explaining the connection of the project to ICDS’s mission: This project is led by a new (early career) Penn State faculty member. The proposed research is interdisciplinary and would lead to the development of advanced computational tools that can be used by geoscientists and volcanologists to understand the global effects of volcanic impacts in Earth history. The project also aligns with ICDS’s mission to combine topic expertise (in this case, volcanology, geochemistry, and Earth history) with advanced computational and data science approaches. The tools developed in this project would be of broad use for the paleoclimate community and serve as a basis for the development of a general geochemical data-model proxy inversion framework.
Team member’s recent and/or planned engagement with ICDS: Fendley is not currently associated with ICDS but is a new faculty member at PSU and is keen to be involved by attending seminars and contributing to ICDS wherever I can be most useful. My research involves using computational techniques to understand geochemical changes in Earth history. I look forward to collaborating with ICDS, especially using various resources for research (both active projects and future proposals) and teaching activities.