Dr Inez Fung: Carbon Dioxide Sources and Sinks

Data Assimilation Workshop

July 19, 2006

Notes by: Jing Lin and Kim Mueller

Motivation

The location and magnitude of carbon dioxide natural sources and sinks is not well known with reasonable certainty. This leads to:

  • Incomplete understanding of the carbon cycle
  • Questions about where to locate new measurement sites
  • Problems in predicting future evolution of CO2 and climate change
  • Challenges for protecting current sinks and reducing sources

Data Availability

Current available data and models that are used to infer sources and sinks are either limited or have significant amounts of uncertainty. Data includes:

  • Direct measurements: Coverage is not uniform. North America has the most coverage while other large regions are poorly constrained such as South America and Africa. Also, the oceans only have measurements along designated ship tracks. Areas that are difficult to measure tend to be not sampled.
  • Inventory data: Fossil fuel carbon dioxide sources are assumed to be “well-known” but there are incentives for some countries to report inaccurate emissions especially if they are concerned with their compliance with the Kyoto treaty.
  • Transport models: Basis functions that are used to transform sources and sinks to measurement space are based on global transport models. These models are difficult to construct especially in regards to turbulent mixing.

Problem Statement

How to estimate carbon dioxide sources and sinks with reasonable certainty using limited data and inaccuracies in transport models?

Initial Approach

Box models are good ‘back of the envelope’ calculations that can be used to validate or highlight discrepancies in the model.

To illustrate how observations and transport model can be combined to invert the sources and sinks, a simple box model was explored. In this case, north hemisphere and south hemisphere are treated as two boxes forced by surface fluxes (sources/sinks) and connected by inter-hemispheric transport with a known time scale. Given observed surface CO2 gradient and total atmospheric CO2 growth rate, the sources/sinks terms can be backed out for each of the hemispheres using back-of-envelope calculation from two first-order differential equations.

Questions

(1)How can limited ship track measurements be used to extrapolate to the entire oceans?

Scott Dolney presented Takahashi dataset and explained that this dataset is compiled from ~2 million samples (although the ocean hasn’t been completely sampled) was the result from extrapolation. This dataset does not have associated uncertainty.

(2)How are transport errors, representation errors, and other errors estimated and accounted in R?

R includes should not only include measurement errors but also transport, representative and aggregation errors.

(3)To what level of uncertainty and at what scale (both temporally and spatially) do we need to know sources and sinks?

This depends upon the problem statement. It was not explained to what level of uncertainty and scale needs to be achieved to satisfy policy requirements.

(4)How is the uncertainty associated with the priors determined?

Doug Nychka gave an example in his lecture.

(5)How could basis functions be constructed and used to in air pollution models for species that are reactive?

This will be covered in next week’s applications.