Informing Carbon Dynamics in the Community Land Model with Observations from Across Timescales

Abstract:

Correct simulation of carbon dynamics in Earth System Models is required to accurately predict both short and long-term land carbon-cycle climate and concentration feedbacks. As new model structures and parameterizations of increasing complexity are introduced there is an ever present need for data to inform these developments, either indirectly through benchmarking activities, or directly through model-data fusion techniques.

Here we briefly describe a very rich source of data that will come from the National Ecological Observatory Network (NEON), a continental-scale facility that will collect freely available biogeochemical and biophysical data from 60 sites representative of a full range of ecosystems across the USA over 30 years. Relevant data at each site include a full suite of micrometeorology measurements, profiles of CO₂ and H₂O vapor isotopes, soil temperature, moisture and CO₂ flux, fine root images, and plot-based NPP, leaf area and litterfall estimates. This is accompanied by Lidar and hyperspectral derived biomass, leaf area and canopy chemistry at < 1m resolution of 100s km². Critically, these observations are well calibrated and highly standardized across sites allowing comparisons, whilst plot and site selection has been designed to optimize representativeness and spatial scaling opportunities.

To illustrate the potential utility of these data in constraining models, we show the range of Community Land Model (CLM) output at NEON site locations, and in model-space look at a number of different functional responses that characterize the model in space and time and could be tested with data.

These observations can be used most directly through a data assimilation (DA) system and we demonstrate how we have developed support for CLM within the Data Assimilation Research Testbed (DART) that uses ensemble techniques for state estimation. Using an observing system experiment, we investigate how infrequent observations of carbon stocks constrain model dynamics and how these observations types can be used with more frequently available flux and leaf area index observations. We demonstrate the use of the latter with real Ameriflux and MODIS data.