Data assimilation of eddy flux measurements coupled with an ecosystem process model to generate optimized predictions of carbon and water exchange in a high elevation coniferous forest

Recent studies in forest ecosystem modeling have employed data assimilation techniques to generate optimal parameter sets, providing highly accurate model estimates of net carbon uptake. Although the carbon and water cycles are linked there has been little focus on applying similar techniques to ecosystem water exchange.  We coupled SIPNET (Simple Photosynthesis EvapoTranspiration), a simplified model of ecosystem function, with a data assimilation system to estimate parameters leading to model predictions most closely matching the net carbon and water fluxes measured by eddy covariance in a sub-alpine forest at Niwot Ridge, CO. When optimized using measurements of carbon exchange, the model matched observed net ecosystem exchange (RME = 0.495 g C m^–2) but underestimated transpiration calculated from sapflow measurements by a factor of 4. Consequently, the carbon only optimization was not sensitive to imposed changes in water availability. Adding both carbon and water exchange data to the optimization reduced the overall model fit to the observed fluxes only slightly (RME = 0.528 g C m^–2), however this parameterization also reproduced direct sapflow measurements of transpiration (R² =0.81, slope =0.96). Optimizing parameters using water fluxes alone led to spurious estimates of carbon exchange. In ecosystem models where the link between the carbon and water cycles is explicitly modeled we conclude that a significant amount of information can be extracted from simultaneous measurements of carbon and water exchange and that failure to include both data streams can generate misleading ‘optimal’ results.