Stanford University

*** Ph.D. Thesis/ Oral Defense ***

Quantifying and mitigating the contribution of model uncertainty to predictions of the terrestrial carbon cycle.

Caroline Famiglietti

Friday, May 12, 11:30 am 

Green Earth Science 365

Department of Earth System Science

Advisor: Dr. Alexandra Konings

The terrestrial biosphere is an integral component of the Earth system, serving as a buffer against the impacts of anthropogenic climate change and providing a wealth of ecosystem services to communities across the globe. However, its behavior and responses to environmental perturbations are challenging to model, mainly due to uncertainties involving model structure (i.e., which physical processes to represent and in how much detail) and parameterization (i.e., how to accurately assign model coefficients in a global context). Despite a proliferation of targeted model development efforts, persistent disagreement between predictions from different state-of-the-art terrestrial biosphere models (TBMs) hinders the establishment and prioritization of robust management and restoration initiatives. 

In this dissertation, I develop methods for quantifying and mitigating the role of model uncertainty in predictions of the terrestrial carbon cycle. This was made possible via diverse, multi-scale Earth observations, which I integrate within a flexible ecosystem model–fusion framework. In Chapter 2, I map the relationship between TBM complexity and predictive skill, finding that robust model parameterization is a prerequisite for structural additions to improve performance. Building on this result, in Chapter 3, I develop and test an innovative parameterization scheme for implementation in TBMs by coupling satellite-derived parameter estimates with data describing local environmental characteristics. Lastly, in Chapter 4, I refine these parameter–environment relationships by investigating their stability in a changing climate. Together, this research provides new insight into the inter-relationship between structural and parametric uncertainties, and also highlights opportunities for using novel data to develop model parameterization strategies that yield realistic variability across the global land surface.

Zoom link: https://stanford.zoom.us/j/5477583499?pwd=dE92WVQvMFdnd1luOVdtZnFiR2xRQT09 (ID: 547 758 3499, password: 659546)