Multi-scale Analysis Of Time-lag Between Net Ecosystem Carbon Exchange And Environmental Factors In A Populus × Euramericana Plantation

There is a general time-lag phenomenon in the response between net ecosystem carbon exchange(NEE)and environmental factors such as water,light and heat in forest ecosystems.However,the in-fluence of time-lag was not considered in the analysis of environmental factors and the interpolation of eddy covariance data.Therefore,it 1s necessary to quantitatively study the effect of time-lag on the response between NEE and impact factors,to provide scientific basis for accurately evaluating the carbon sink of the ecosystem and simulating the dynamic changes of NEE.The NEE and microclimate of Populus x euramericana plantation were observed by using eddy covariance technology and mi-cro-meteorological observation system in Shunyi District,Beijing.The daily and annual data of NEE,air temperature(Ta),soil temperature(Ts)photosynthetically active radiation(PAR),vapor pressure deficiency(VPD)and soil volumetric water content at 25-cm depth(VWC25)during the growing sea-son in 2014 were selected.The time-lag relationship between NEE and these impact factors was ob-tained by combining regression analysis with wavelet analysis.Partial correlation and principal com-ponent analysis were used to evaluate the accuracy of fitting model with the determinant coefficient(R2),Lynn's harmonic coefficient(LCCC)and root mean square error(RMSE).The effect of ignoring time-lag on the accuracy of mipact factors and the fitting equation was discussed.The results showed that(1)PAR was the dominant factor affecting NEE at both daily and annual scales.Because of the high groundwater level and abundant water in the study area,the driving effect of VWC25 on NEE in any scales could be neglected.(2)There was no time lag relationship between NEE and PAR in all seasons.However,NEE reached its peak value earlier than Ta,Ts and VPD for 4.0,3.5 and 3.5 hours in spring,3.5,3.0 and 3.0 hours in summer,and 3.5,3.0 and 3.0 hours in autumn,re-spectively.At the annual scale,NEE changed synchronously with Ta,Ts and PAR,while it lagged be-hind VPD with a time-lag of 21 days.(3)Neglecting the time-lag on the daily scale would affect the real relationship between NEE and Ta in spring,while it would affect the real relationship between NEE and Ts in summer and autumn.However,the absolute values of partial correlation coefficients of NEE and PAR were the largest in each season,which indicated that ignoring the time-lag phenomenon would not affect the dominant factor of NEE.(4)The optimal logarithmic regression model could be obtained in every season before or after eliminating the time-lag phenomenon.The determinant coef-ficients R2 of the model increased by 5.23%,6.61%and 2.11%in spring,summer and autumn,LCCC increased by 2.41%,4.03%and 1.45%,and RMSE decreased by 0.473,4.459 and 0.159 ?mol·m-·s-1 respectively.For the annual scale,the optimal hyperbolic regression model could be obtained before or after eliminating the time-lag phenomenon.After eliminating the time-lag phenomenon,The R2 and LCCC of the model only increased by 1.11%and 0.47%,and RMSE decreased by only 0.006 umol·m-2·s-1.This result showed that the fitting accuracy of daily NEE data could be significantly improved by considering the time-lag phenomenon.Although the effect of the annual scale was not obvious,the accuracy was also improved.On this basis,the daily mean data obtained from the daily scale optimal regression model after eliminating the time-lag phenomenon were compared with those obtained from the simulation of the annual scale optimal regression model.It was found that the ac-curacy of daily value data obtained by using the daily scale model was higher than that obtained by using the annual scale model.R2 and LCCC increased by 6.51%and 7.14%respectively,while RMSE decreased by 24.245 ?mol·m-2·s-1.It showed that for the problem of time scale conversion of data,the simulation results at the short-time scale were more accurate than long-time scale.Therefore,the daily and annual dynamic changes of poplar plantation ecosystem were controlled by Ta,Ts,VPD and PAR.There was time-lag in both daily and annual scale.After eliminating the time-lag,the environmental impact factors of NEE could be accurately judged at daily scale,and the fitting ac-curacy of NEE could be improved at any time scale.