Assessing Annual Net Ecosystem Carbon Balance Using a Data-Model Fusion Approach – Application of Data from National Ecological Observatory Network

Abstract:

Terrestrial ecosystems help mitigate climate change through removal of carbon (C) dioxide from the atmosphere and sequestering it in biomass, litter, and soil. It is important to understand the mechanisms of C sequestration as well as their responses to changing environmental conditions to predict whether ecosystems will continue sequestering atmospheric C in the future. However, gaining mechanistic understanding of C cycle is often challenging, because observational studies tend to focus on isolated ecosystem components or have short-term period of observation. To overcome these obstacles and facilitate understanding and forecasting of ecosystem dynamics, the National Ecological Observatory Network (NEON) set up data collection sites across 20 ecoclimatic domains in the USA. The volume and diversity of the collected data are increasing, and research is needed on ways to leverage these data to improve models and forecasts of C cycle dynamics. In this study, we combined observations from NEON and a data assimilation technique to calibrate parameters in the Terrestrial ECOsystem (TECO) model and evaluate whether an ecosystem is a net C sink or source. We evaluated data availability across the NEON field sites and chose the observations from Smithsonian Conservation Biology Institute (SCBI) for the 2016-2018 period. The observations helped constrain 18 out of 39 model parameters. We found that temperature sensitivity of litter and soil organic C (Q₁₀) was 3.8 and fell outside of the range of widely used values. After parameter calibration TECO performed well in reproducing LAI, wood biomass, fine root biomass, and soil organic carbon; however, simulations of net ecosystem exchange (NEE), leaf litterfall, and fine wood litterfall were not as aligned with the observed data. Data-constrained TECO model indicated that SCBI site was a much stronger C sink (-379.5 - -222.5 gCm^-2y^-1), compared to the estimate derived from the observations (-136.5 - -10.7 gCm^-2y^-1). Because the estimates from the data-constrained model represent the theoretical knowledge about C cycle fine-tuned to the given site by assimilating multiple types of observations, they are likely more reliable compared to estimates obtained by statistical extrapolation of observed NEE.