Using Gaussian Process Emulation to Quantify the Global Methane Budget
Abstract
Methane is the second most important greenhouse gas, but its budget is poorly understood: emissions inventories and bottom-up models are inconsistent with atmospheric trends. The concentration of methane in the atmosphere has been growing since the industrial revolution, until 2000, when the growth stopped abruptly, before continuing in 2007. The reason for these changes is the subject of much debate.
Recently, there have been a range of studies using measurements and models, providing competing theories for the dominant drivers of recent methane trends. Since global three-dimensional atmospheric models are computationally expensive to run, these studies have used simplified atmospheric transport (e.g. box models) or have explored a limited part of the uncertain parameter space (e.g. by imposing fixed hydroxyl radical concentrations or ignoring uncertainty in isotopic source signatures). In this work, Gaussian Process emulation of a three-dimensional atmospheric transport model is used to explore the relationship between methane mole fraction and δ13C-CH4 observations, and a wide range of uncertain model inputs (e.g. wetland emission and source isotope ratio, fossil fuel emissions, atmospheric losses). The emulator allows millions of input parameter combinations to be tested against observations, based on only a relatively small set of forward model runs. This method has many advantages: three-dimensional, inter-annually varying transport, the Gaussian Process can deal with the non-linearities, and a wide range of input types can be sampled without any modification of the forward model. However, technical and computational challenges remain. In this work, we will present a pilot study in which hemispheric-mean methane mole fraction and isotopic observations are used to constrain trends in methane sources and sinks between 2000 and 2012 (Figure 1). We will demonstrate how uncertainties in the methane budget are influenced by factors not normally considered in three-dimensional model studies (e.g. isotopic source signature uncertainties) and outline which regions of the parameter space are consistent with the hemispheric-mean observations. This method could be further developed to model individual measurement sites, allowing more information to be extracted from the underdetermined system.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.B13O2514S
- Keywords:
-
- 0426 Biosphere/atmosphere interactions;
- BIOGEOSCIENCES;
- 0475 Permafrost;
- cryosphere;
- and high-latitude processes;
- BIOGEOSCIENCES;
- 0497 Wetlands;
- BIOGEOSCIENCES;
- 1615 Biogeochemical cycles;
- processes;
- and modeling;
- GLOBAL CHANGE