Assimilation of flux measurements into the NEON-NCAR land surface model

Background/Question/Methods

Water, carbon and climate dynamics influence many aspects of ecological function. Eddy covariance flux towers provide direct measurements of the ecosystem exchange of water, carbon and energy at small spatial scales but estimates of these fluxes at the regional and continental scale are required to diagnose, understand and predict the response of the global water and carbon cycles to a changing climate. The National Ecological Observatory Network (NEON) is a continental scale facility that will collect ecological data, including eddy covariance flux observations, from 60 sites in the continental US, Alaska, Hawaii, and Puerto Rico over 30 years. In order to perform both spatial extrapolation from flux tower sites and temporal forecasting on decadal timescales we are developing a model-data fusion framework in which NEON data can be combined with the Community Land Model (CLM) using a Bayesian approach to produce optimal solutions for model states and fluxes and parameter values, with their associated uncertainties, at a regional scale. Our initial approach has been to use flux data from a number of existing FLUXNET sites to assess the performance of CLM across a number of biomes and formally assess the sensitivity of CLM outputs to parameter uncertainty for the first time. We have then used this work to identify a set of parameters that can potentially be constrained by covariance between model outputs and flux observations. In tandem, we have developed the Community Earth System Model infrastructure to allow multiple instances of CLM to work simultaneously, which has allowed us to implement a coupling of CLM with the National Center for Atmospheric Research’s Data Assimilation Research Testbed (DART). Together these investigations have allowed us to investigate CLM parameter optimization using the Ensemble Kalman Filter (EnKF).

Results/Conclusions

Here we present some early results from these investigations focusing on: (i) the effects of parameter uncertainty on modeled fluxes; (ii) the identification of parameters that can be constrained with flux data; and (iii) the effectiveness of the EnKF for parameter estimation in a complex land surface model using both synthetic, model generated data streams and existing observational data. We identify a number of remaining challenges and discuss how other ecological data streams available from NEON in the future may be incorporated into this model-data fusion framework.