Validity of extrapolating field CO2 experiments to predict carbon sequestration in natural ecosystems

One of the ultimate goals of ecosystem carbon dioxide (COL) experiments is to infer the capacity of terrestrial ecosystems to sequester carbon (C) in a CO2-enriched world. This modeling study examines C sequestration (C-seq) in natural ecosystems based on CO2 experiments. Most experiments are conduc ted by a step increase in CO2 concentration, whereas natural ecosystems are experiencing a gradual increase in atmospheric CO2 (C-a). To examine the e ffects of a step vs. gradual CO2 increase on ecosystem responses: we have d eveloped a terrestrial C sequestration (TCS) model that focuses on C and ni trogen (N) interactions in regulating C-seq. We used the model to: (1) comp are C-seq and N demand in response to the step vs, gradual increase in CO2; (2) identify mechanisms underlying different ecosystem responses to the st ep vs. gradual CO2 forcing; (3) examine key parameters in controlling C-seq and (4) explore three hypothesized N supply mechanisms in regulating photo synthetic acclimation and C-seq. Application of this model to simulate responses of a forest ecosystem with gross primary productivity of 1200 g C.m(-2).yr(-1) suggested that a step i ncrease in CO2 from 350 to 700 ppm resulted in C-seq of 263 g C.m(-2).yr(-1 ) in the first year. A gradual C-a increase led to the C-seq rates of 27 an d 58 g C.m(-2).yr(-1) in 1987 and 2085 when CO2 reached 350 and 700 ppm, re spectively. The model predicted that N demand required to balance the addit ional C influx was 4.1 g N.m(-2).yr(-1) in the step CO2 increase and only 0 .6 and 1.7 g N.m(-2).yr(-1) in 1987 and 2085, respectively, in the gradual C-a increase. The contrasting differences in C-seq and N demand between the two increase scenarios reflected the nature of C fluxes that were controll ed by the sizes of donor pools (i.e., donor-controlled system). Our modelin g analysis of four ecosystems (forest with high productivity [HP]; grasslan d with HP; forest with low productivity [LP]; and grassland with LP) indica ted that additional C influx and C relaxation time are the key parameters i n determining ecosystem C-seq. The additional C influx varied with ecosyste m productivity and N regulation, while C relaxation time differed between t he forests and grasslands due to woody tissues and litter in the forests. W e conclude that in spite of the fact that the step experiment is one of the most effective approaches in ecosystem studies, its results cannot be dire ctly extrapolated to predict terrestrial C-seq in natural ecosystems respon ding to a gradual C-a increase. In order to develop predictive understandin g from the: step experiments, we need not only to improve experimental desi gn and measurement plans, but also to develop new approaches, such as decon volution and inverse modeling, for data analysis and interpretation.