Contribution And Propagation Of Remote Sensing Product Error In The Sea-Air CO₂Fluxes Estimation

Sea-air CO2fluxes estimation is significant for researches on carbon biogeochemical cycles and greenhouse effects. As the major methods, field measurements, however, are complicated, time-wasting and cannot meet the demand of wide range monitoring. Remote sensing monitoring techniques are therefore developed for synchronous observation of large area, which makes the researches on spatial and temporal variability of carbon fuxes on the sea surface available. Nevertheless, due to a great deal of uncertainties, the reliability of the Sea-air CO2fluxes remote sensing estimation products on the management is still limited. Based on the analysis of field measurement error sources, this study focused on the propagation and contribution of the remote sensing products error in the process of carbon dioxide exchange across air-sea interface. The major contents and main conclusion in this study were summarised as follows:(1) Field measurement error analysis. The field measurement methods and processes in the study of CO2fluxes between the interface of sea and air were described. Various of error sources generated in the in situ measurements were analysed, and summarized into three categories:sampling protocol error, man-made error and instrument platform error. Then the error framework map of field measurements was established.(2) Remote sensmg monitoring error analysis. The main source of data in remote sensing monitoring was introduced, and the data accuracy was evaluated. Based on the principle of remote sensing, the parameterize method of the direct control parameters of CO2flux (the gas transfer velocity-k; the sea surface CO2solubility-S; and the sea surface partial pressure of CO2-pCO2SW) were introduced. And the framework of various error sources emerged in the remote sensing routines of air-sea CO2fluxes estimation was built.(3) Instruments error propagation and contribution analysis in field measurements. In this study, the transfer principle and final error contribution analysis in the calculation of fluxes were carried out using Monte Carlo to simulate errors generated from the pCO2traveling monitoring system GO8050-Li-Cor NDIR(7000) and other instruments. Eight parameters [wind speed (Uiom), the actual temperature of the sea surface (SST), surface sea water salinity (SAL), mole fraction of CO2in surface seawater (xCO2air), the balance pressure of water vapor in the balancer (Peq), the temperature in the balancer (Teq), the mole fraction of CO2in the (xCO2air) and atomspheric pressure (Pair)] which were directly obtained by the sensors of instruments, were included. The results showed that errors of water vapor pressure (pH2O) and the uncorrected CO2partial pressure of sea surface temperature (pCO2sw’) were normally distributed, while the other primary parameter errors presented non-normal distribution, which were approximately exponential distributed after modeling. The instrument error was approximately in the range of±0.2mmol C·m-2-day-1, which accounting for1.3%of the actual flux results (in accordance with the calculation of the actual monitoring flux range of±15mmol C·m-2-day-1). Then the zero-error sensitivity of the single factor was carried out for the four direct controlling parameters of CO2flux (k/S/pCO2sw/pCO2air), which indicated that the error of the gas transfer velocity k was improved obviously, followed by pCO2SW, while the influence of the other two parameters (S/pCO2air) were not distinct flux calculations.(4) Remote sensing monitoring error propagation and contribution analysis. Based on the error-framework of the direct control parameters of CO2fluxes (k/S/pCO2sw), we set SST as the main flux influence factor and established the SST error as transfer flow chart. In order to understand the way SST error affects the fluxes interaction, the transfer law and the final error contribution analysis in the calculation of fluxes were carried out using Monte Carlo simulation. The results indicated that on the assumption that remote sensing SST error was±0.5℃and normally distributed, the SST error transfer law was exponential distributed in the k parameterization, and approximately exponential distributed in the S parameterization, while normally distributed in pCO2sw parameterization and exponential distributed in CO2fluxes, respectively; and when atmospheric CO2partial pressure was fixed at the value of370μatm, SST brought an error of±1.2mmol C·m-2-day-1to the final result of the flux, which accounted for about8%of the actual result.