Using eddy covariance of CO2, 13CO2 and CH4, continuous soil respiration measurements, and PhenoCams to constrain a process-based biogeochemical model for carbon market-funded wetland restoration

Abstract:

We use multiple data streams in a model-data fusion approach to reduce uncertainty in predicting CO₂ and CH₄ exchange in drained and flooded peatlands. Drained peatlands in the Sacramento-San Joaquin River Delta, California are a strong source of CO₂ to the atmosphere and flooded peatlands or wetlands are a strong CO₂ sink. However, wetlands are also large sources of CH₄ that can offset the greenhouse gas mitigation potential of wetland restoration. Reducing uncertainty in model predictions of annual CO₂ and CH₄ budgets is critical for including wetland restoration in Cap-and-Trade programs. We have developed and parameterized the Peatland Ecosystem Photosynthesis, Respiration, and Methane Transport model (PEPRMT) in a drained agricultural peatland and a restored wetland. Both ecosystem respiration (R_eco) and CH₄ production are a function of 2 soil carbon (C) pools (i.e. recently-fixed C and soil organic C), temperature, and water table height. Photosynthesis is predicted using a light use efficiency model. To estimate parameters we use a Markov Chain Monte Carlo approach with an adaptive Metropolis-Hastings algorithm. Multiple data streams are used to constrain model parameters including eddy covariance of CO₂, ¹³CO₂ and CH₄, continuous soil respiration measurements and digital photography. Digital photography is used to estimate leaf area index, an important input variable for the photosynthesis model. Soil respiration and ¹³CO₂ fluxes allow partitioning of eddy covariance data between R_eco and photosynthesis. Partitioned fluxes of CO₂ with associated uncertainty are used to parametrize the R_eco and photosynthesis models within PEPRMT. Overall, PEPRMT model performance is high. For example, we observe high data-model agreement between modeled and observed partitioned R_eco (r² = 0.68; slope = 1; RMSE = 0.59 g C-CO₂ m^-2 d^-1). Model validation demonstrated the model’s ability to accurately predict annual budgets of CO₂ and CH₄ in a wetland system (within 14% and 1% of observed annual budgets of CO₂ and CH₄, respectively). The use of multiple data streams is critical for constraining parameters and reducing uncertainty in model predictions, thereby providing accurate simulation of greenhouse gas exchange in a wetland restoration project with implications for C market-funded wetland restoration worldwide.