Research On The Calculation Method Of Turbulent Flux In The Near-surface Layer Under Non-stationary Condition

Turbulent transport is important in the exchange of matter and energy between earth and air,and accurate turbulent flux results are of great help in understanding information about the exchange of matter and energy between various types of ecosystems and the atmosphere.In this paper,the accuracy of calculation results by eddy covariance method(EC),conditional sampling method(CS),multi-resolution flux decomposition method(MFD)and wavelet analysis method(WA)is compared.For turbulence in stationary state,the calculated results of eddy covariance were taken as the benchmark,and the calculated results of various methods were consistent with them.Mexhat wavelet method showed a tendency to underestimate,and Morlet wavelet method showed a tendency to overestimate.For turbulence in non-stationary state,the results calculated by eddy covariance method can no longer be used as benchmark.By adding periodic sinusoidal waves and random disturbances to the observed stationary turbulence data set,a non-stationary turbulence data set is concocted to simulate the non-stationary characteristics of large-scale vortices,turbulence intermittency and turbulence asymmetry in the real world.The theoretical"true value"obtained from this data set is used as the benchmark of turbulence flux in non-stationary state to evaluate the accuracy of various algorithms quantitatively.The results show that the deviation range of eddy covariance method and Morlet wavelet method is 50%-100%compared with the base value under different non-stationarity classes,especially when the non-stationarity grades are larger than 6,the accuracy of the results is greatly reduced,while the deviation range of Mexhat wavelet method is only half of these two methods.In addition,due to the high correlation between the Mexhat wavelet result and the benchmark,with the correlation R~2 exceeding 0.885 in different non-stationary grades,the regression slope between the Mexhat wavelet result and the benchmark can be improved to close to 1 by increasing the correction coefficient,then the results were verified by spectral analysis.