Parameter-induced uncertainty quantification of soil N2O, NO and CO2 emission from Höglwald spruce forest (Germany) using the LandscapeDNDC model

[摘要] Assessing the uncertainties of simulation results of ecologicalmodels is becoming increasingly important, specifically if thesemodels are used to estimate greenhouse gas emissions on site toregional/national levels. Four general sources of uncertaintyeffect the outcome of process-based models: (i) uncertainty ofinformation used to initialise and drive the model, (ii) uncertaintyof model parameters describing specific ecosystem processes, (iii)uncertainty of the model structure, and (iv) accurateness ofmeasurements (e.g., soil-atmosphere greenhouse gas exchange) whichare used for model testing and development.

The aim of our study was to assess the simulation uncertainty of theprocess-based biogeochemical model LandscapeDNDC. For this we set upa Bayesian framework using a Markov Chain Monte Carlo (MCMC) method,to estimate the joint model parameter distribution. Data for modeltesting, parameter estimation and uncertainty assessment were takenfrom observations of soil fluxes of nitrous oxide (N₂O),nitric oxide (NO) and carbon dioxide (CO₂) as observed overa 10 yr period at the spruce site of the Höglwald Forest,Germany. By running four independent Markov Chains in parallel withidentical properties (except for the parameter start values), anobjective criteria for chain convergence developed byGelman et al. (2003) could be used.

Our approach shows that by means of the joint parameterdistribution, we were able not only to limit the parameter space andspecify the probability of parameter values, but also to assess thecomplex dependencies among model parameters used for simulating soilC and N trace gas emissions. This helped to improve theunderstanding of the behaviour of the complex LandscapeDNDC modelwhile simulating soil C and N turnover processes and associated Cand N soil-atmosphere exchange.In a final step the parameter distribution of the most sensitiveparameters determining soil-atmosphere C and N exchange were used toobtain the parameter-induced uncertainty of simulated N₂O,NO and CO₂ emissions. These were compared to observationaldata of an calibration set (6 yr) and an independent validationset of 4 yr.The comparison showed that most of the annual observed trace gasemissions were in the range of simulated values and were predictedwith a high certainty (Root-mean-squared error (RMSE) NO: 2.4 to18.95 g N ha⁻¹ d⁻¹, N₂O: 0.14 to21.12 g N ha⁻¹ d⁻¹, CO₂: 5.4 to11.9 kg C ha⁻¹ d⁻¹).However, LandscapeDNDCsimulations were sometimes still limited to accurately predict observedseasonal variations in fluxes.

[学科分类] 地球化学与岩石