科研之友
微信
新浪微博
Facebook
分享链接
摘要
The flux gradient eddy covariance approach is commonly used to measure fluxes of trace species at a field scale, and is particularly useful when fluxes must be measured at several locations using a single instrument, or where instrumentation is not capable of measuring concentration fluctuations at a high frequency. In addition to the need to resolve very small vertical concentration gradient, the technique also requires the estimation of eddy diffusivity. Here we perform an analysis to determine the sensitivity of the technique to errors in the estimation of this term. The eddy diffusivities, also termed as transfer coefficients, were estimated using a thermodynamic approach (denoted here as k(h)(t)) and a parameterization approach (denoted here as d(h)(p)). The transfer coefficients in both approaches were calculated from several other surface layer variables. We derived algebraic expressions describing the relationships between the error terms of the transfer coefficients and their component variables. Based on these relationships, a Monte-Carlo type analysis was performed to explore the dependency of the transfer coefficients on their contributing factors over a range of stability conditions in the atmospheric surface layer. Finally, the total relative uncertainty in the flux values due to errors in transfer coefficient and concentration differences were estimated. We found that the mean relative error in k(h)(t) is higher (15%) than the mean relative error in d(h)(p) (approximate to 8%), irrespective of stability. However, depending on the initial uncertainty among the surface layer variables, the uncertainty can range up to 80% and 49% for both k(h)(t) and 4, irrespective of stability. Under most conditions, total error contribution in k(h)(t) was found to be associated more with the temperature gradient term than the heat flux term, whereas the error contribution for (p)(h) was found to be associated more with the u(star) term than the L term. Errors in the concentration differences were estimated based on the minimum resolvable estimates from the gas analyzer and the associated random errors were found to be 6% and 8% for unstable and stable conditions. Finally, the total mean error in the N2O flux values was found to be approximately of the order of 9% and 12% for the parameterization method for unstable and stable conditions, respectively, and 16.5% for the thermal method, irrespective of stability.
- 出版日期2014-2
