Constraining Respiration Flux and Carbon Pools in a Simple Ecosystem Carbon Model
Abstract
Challenges in adequately capturing the balance of terrestrial carbon uptake (through photosynthesis) and release (through ecosystem respiration) limit our ability to model the response of the carbon cycle to global change. While many globally monitored proxies allow for the estimation of global plant productivity and photosynthesis, quantifying the global ecosystem respiration flux is impossible through direct measurements, and accordingly, there are no global constraints on the expenditure of the land carbon budget. Nonetheless, available data can be synthesized in process-based models to achieve reliable and interpretable carbon cycle representations. In this work, we investigate whether respiration can be constrained to model-data fusion instead. Specifically, we study the effects of assimilating multiple datasets in a simple ecosystem carbon model - CARDAMOM - and quantify their impact on constraining respiration fluxes. We modify the model such that solar induced fluorescence (SIF) and vegetation optical depth (VOD) data are assimilated in addition to the commonly used leaf area index (LAI), and explore different combinations of the assimilated data for five globally distributed FLUXNET sites. In all cases, adding SIF and VOD led to more accurate model predictions (measured as correlation coefficient (R) and root mean squared error (RMSE)) between the model outputs and FLUXNET data. For wet, light-limited sites, meteorological drivers are sufficient to achieve high R (0.91-0.98) by closely matching the seasonal cycle. Even in these cases, though, additional data yields reductions in RMSE for the respiration flux. For water-limited sites, the effect is even more pronounced, as in these climatic zones VOD and LAI are decoupled. Assimilating several diverse datasets reduces the chance of equifinality - i.e., when a strong correlation with observational data is achieved despite compensating errors in unobserved model states and suboptimal retrieved model parameters. Such evidence can be found through changes in the distribution of meaningful model parameters, e.g. a shift in canopy efficiency with SIF and VOD inclusion. With more high-quality global data available from remote sensing, even simple carbon models become a source of valuable inputs in carbon dynamic.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.B55A..07S