The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data

摘要: We describe a model-data fusion (MDF) inter-comparison project (REFLEX), which compared various algorithms for estimating carbon (C) model parameters consistent with both measured fluxes and states simple C model. Participants were provided the synthetic net ecosystem exchange (NEE) of CO2 leaf area index (LAI) data, generated from added noise, observed NEE LAI data two eddy covariance sites. endeavoured to estimate all cases over years provided, generate predictions one additional year without observations. Nine participants contributed results using Metropolis algorithms, Kalman filters genetic algorithm. For case, parameter estimates well true values. The analyses indicated that linked directly gross primary production (GPP) respiration, such as those related foliage allocation turnover, or temperature sensitivity heterotrophic best constrained characterised. Poorly estimated turnover fine root/wood pools. Estimates confidence intervals varied among but several successfully located values annual experiments within relatively narrow 90% intervals, achieving >80% success rate mean <110 gC m−2 year−1 case. Annual flux by generally agreed gap-filling approaches half-hourly data. estimation respiration GPP through MDF outputs partitioning studies Confidence limits on increased an average 88% in prediction previous year, when available. 30% used instead reflecting quantifying addition error. Finally, our incorporating constraints, pools (wood, soil roots) would help reduce uncertainties poorly served