Study On CO₂flux And Carbon Stable Isotope Of Terrestrial Ecosystem

Carbon problem is a hot topic recently which seems to hit the scientific creationists thehardest, also has interesting implications for today’s earth. Research on carbon sink/source ofterrestrial ecosystem is one of the main issues of global carbon cycle. The rapid developmentof remote sensing, eddy covariance technology flux measurement and stable isotopetechnology have greatly improved the progress of research on terrestrial ecosystem carboncycle. Currently, modeling is one of the tools to study regional/global terrestrial ecosystemcarbon cycle. Developing of fusion approach （flux data, remote sensing data and model）,increasing the precision of measurement and modeling could greatly improve ourunderstanding/modeling/predicting of terrestrial ecosystem carbon cycle.This study carefully selected8different terrestrial ecosystems （Howland Forest, HarvardForest, Wind River Forest, Rannells Prairie, Freeman Ranch, Chestnut Ridge, Metolius andMarys River）. By measure and model the ecosystem CO₂flux and its Carbon stable isotope,we suggest that the equilibrium boundary layer budget method can serve as a routine,diagnostic analysis to interpret long-term NEE observations in flux networks, providing anintermediate-level analysis to complement aircraft/MODIS based integration efforts forestimates of continental carbon budget. We find that the carbon isotope composition ofecosystem respired CO₂is positive to PPF, TA and VPD, negative e to SWC. We noted theimportance of the variation of annual ecosystem（13）^C discrimination when using inversionmodel. The important results list as follows:1Measured and compared different ecosystem （Howland Forest, Harvard Forest, WindRiver Forest, Rannells Prairie） CO₂flux, interannual and seasonal variation. By understoodthe factors which affect the variation, we found out that the growing season precipitation wasthe main factor which could affect ecosystem interannual NEE variation.2Introduced and compared the approaches of study ecosystem CO₂flux. Also weanalyzed the factors that contribute to the uncertainty of each approach. The uncertainty intower eddy CO₂fluxes include but are not limited to: instrument deficiency （e.g. loss ofspectrum energy under low turbulence）, oversimplification in theory （e.g. zero divergence incomplex terrain） and representative errors （e.g. tower locations not representative of dominant vegetation types）. Uncertainties associated with the inverse modeling approach employed hereinclude: errors in the CO₂mixing ratio measurements （both near the surface and in the freetroposphere）, unreliable estimates of boundary layer heights and vertical transport across thetop of the atmospheric boundary layer, and the assumption of an equilibrium boundary layerover time periods longer than synoptic scales. There are several sources of uncertaintyassociated with EC-MOD flux estimates, including uncertainties in the: eddy fluxmeasurements, land cover, model structural, representativeness of the AmeriFlux sitelocations （Xiao et al.2008,2001）, and finally, the negligence of fossil CO₂fluxes.3Introduced the CBL model. We found the seasonal variation of BLH, highest at July,lower at winter. We conducted a series uncertainty analysis of CBL model and discussedtoward equilibrium H2O and CO₂exchange in the ABL at monthly time scales. We evaluatedthe contribution and limitation of CBL model.4We evaluated the EC-MOD approach, and discussed the uncertainty of estimating NEE.5We analyzed the factors which attributes to the difference in the model and tower NEE.The first type of systematic differences was found in the wintertime fluxes, in which CBLmodel fluxes were consistently larger than tower NEE at all but Wind River site. The EC-MOD approach consistently predicted wintertime fluxes closer to the tower measurementsthan the CBL model, irrespective of EC-MOD spatial scale. The factor influence wintertimefluxes is the effects of fossil fuel emission which could proved by the regional fossil fuelfluxes of CO₂estimated from CO data measured and δ（13）^C values associated with net CO₂fluxes （δ（13）^Cnet）. The second type of systematic differences results from a scale mismatchbetween tower and model footprints. Inversion models relying on mixing ratio observationshave often reported smaller fluxes when compared to tower-based NEE measurements.6We compared the Midday CO₂mixing ratios and δ（13）^C of atmospheric CO₂measuredabove the canopy at8ecosystem（Howland Forest, Harvard Forest, Wind River Forest,Rannells Prairie, Freeman Ranch, Chestnut Ridge, Metolius and Marys River fir）. We foundthe seasonal interannual variation. We analyzed the factors which affect the variation. Wefound the reverse variation of CO₂mixing ratios and δ（13）^C of canopy atmospheric CO₂whichcaused by plant photosynthetic discrimination of（13）^C. Also we compared the CO₂mixingratios and its δ（13）^C between canopy and BL.7We estimated the ecosystem （13）^C values associated with NEE （δ（13）^Cbio）, and comparedδ（13）^Cbioof different8ecosystems. Rannells prairie showed the lowest δ（13）^Cbiovalues. Generallyspeaking, δ（13）^Cbiokept consistent （better consistence appeared at Harvard Forest，the greaterinterannual variation appeared at Howland Forest）. Monthly δ（13）^Cbioshowed also consistent（δ（13）^Cbioof Wind River almost the same from May to September. the greater monthly variation appeared at Freeman Ranch and Rannells prairie）.8We measured and compared the carbon isotope ratios of ecosystem respired CO₂（δ（13）^CR）. Higher δ（13）^CRvalues （-24‰å'Œ-30‰） appeared at forest ecosystem and lower δ（13）^CRvalues （-12‰and-18‰） appeared at grassland. Seasonal fluctuations of δ（13）^CRwererepeatedly observed in many of our study sites. Generally speaking, values of δ（13）^CRare moredepleted in winter, with progressive enrichment from the onset of a growing season to the endof summer. This seasonal pattern of δ（13）^CRappeared to be more consistent in coniferousforests。Freeman Ranch showed a bigger variation （-16‰and-30‰）.9We analyzed the relationship between δ（13）^CRand environmental variables on weeklyand monthly time scales and found out that δ（13）^CRwere positive to PPF TA and VPD,negative to SWC （higher correlation coefficient at weekly scale）.10we used a flux-weighted approach to estimate canopy-level（13）^C of plantdiscrimination （ΔA） for each of the study sites and compared the ΔA value of differentecosystem. These observation-based estimates of ΔA fall in the range between17.3‰and20.0‰with an average ca.18.5‰in forest ecosystems. Our estimates are very close to thevalue expected from wood cellulose samples that was thought to be most appropriate foratmospheric inversion analysis The Latewood cellulose was extracted from tree ring samplesand analyzed for（13）^C ratios from1996to2009. These ΔA estimates from woody tissues werecompared with those derived from bulk leaves, Keeling-plots and the flux-weighted approach.Our results suggest3important patterns:（1） ΔA values derived from nocturnal air samples（Keeling-plots） were consistently2-4‰higher than those derived from wood cellulose （15-16.7‰）,（2） ΔA values derived from Keeling-plots were similar to those derived from theflux-weighted approach, with an average value¹9‰between2002-2007at the Wind Riversite, and （3） bulk-leaf ΔA values were close but somewhat smaller than Keeling-plot ΔA. Weanalyzed the relationships between annual-mean ΔA and the growing-season water availability.ΔA positive correlation was observed between ΔA and water availability within a site, likelyas a result of a general stomata response to drought stress during dry years. AΔ change from18‰to20‰approximately corresponds to a change from0.59to0.68in Ci/Ca.11We estimated and compared the ecosystem WUE at6ecosystems （Howland Forest,Harvard Forest, Wind River Forest, Freeman Ranch, Metolius and Marys River）. The WUE ofFreeman Ranch and Metolius showed the lower values. We found a strong negativecorrelation between WUE and VPD for all sites, with Harvard Forest site and MetoliusPonderosa Pine site showing the highest and lowest sensitivity, respectively. We comparedecosystem WUE calculated using eddy covariance flux （EC WUE） and carbon isotope （δ（13）^CR-WUE） measurements. There was good agreement between WUE values calculated using eddy covariance flux and carbon isotope measurements at all but the Freeman Ranch site. EC-WUEvalues underestimate long-term expected WUE at Freeman Ranch site （an open savannasystem） and Marys River Fir （a coastal forest with frequent fog formation） site. At these twosites, LE likely overestimate canopy transpiration, resulting in an underestimation ofecosystem WUE when compared to（13）^C-based WUE values.12We evaluated the main4ecosystems. Harvard Forest consistently showed the greatestNEE and GPP, while Howland forest showed the greatest R. The net annual C uptake in theHowland Forest is comparable to Harvard Forest, the longer growing season in HowlandForest compensates with the result that both sites have the similar annual rates of C storage.Although maximum rates of uptake are greater at Harvard Forest than at Howland Forest. Theuptake ability of the forest ecosystems （Harvard Forest site> Howland Forest> Wind RiverForest） were greater than grasslands. Also we analyzed the increasing tendency of CO₂uptakeability and regional fossil fuel emission （especially at Harvard Forest）.