Skip to main content
U.S. flag

An official website of the United States government

Initializing carbon cycle predictions from the Community Land Model by assimilating global biomass observations (Invited)

Presentation Date
Thursday, December 14, 2017 at 2:10pm
New Orleans Ernest N. Morial Convention Center - 356-357



The locations and longevity of terrestrial carbon sinks remain uncertain, however it is clear that in order to predict long-term climate changes the role of the biosphere in surface energy and carbon balance must be understood and incorporated into earth system models (ESMs). Aboveground biomass, the amount of carbon stored in vegetation, is a key component of the terrestrial carbon cycle, representing the balance of uptake through gross primary productivity (GPP), losses from respiration, senescence and mortality over hundreds of years.

The best predictions of current and future land-atmosphere fluxes are likely from the integration of process-based knowledge contained in models and information from observations of changes in carbon stocks using data assimilation (DA). By exploiting long times series, it is possible to accurately detect variability and change in carbon cycle dynamics through monitoring ecosystem states, for example biomass derived from vegetation optical depth (VOD), and use this information to initialize models before making predictions.

To make maximum use of information about the current state of global ecosystems when using models we have developed a system that combines the Community Land Model (CLM) with the Data Assimilation Research Testbed (DART), a community tool for ensemble DA. This DA system is highly innovative in its complexity, completeness and capabilities.

Here we described a series of activities, using both Observation System Simulation Experiments (OSSEs) and real observations, that have allowed us to quantify the potential impact of assimilating VOD data into CLM-DART on future land-atmosphere fluxes. VOD data are particularly suitable to use in this activity due to their long temporal coverage and appropriate scale when combined with CLM, but their absolute values rely on many assumptions. Therefore, we have had to assess the implications of the VOD retrieval algorithms, with an emphasis on detecting uncertainty due to assumptions and inputs in the algorithms that are incompatible with those encoded within CLM. It is probable that VOD describes changes in biomass more accurately than absolute values, so in additional to sequential assimilation of observations, we have tested alternative filter algorithms, and assimilating VOD anomalies.

Funding Program Area(s)