| The study was undertaken in three stands(hereafter refer as to stand 1,stand 2 and stand 3,respectively)of Larix principis-rupprechtii forest from 2007 ~ 2015,with different elevations in Pangquangou National Nature Reserve,to get the data sets of diurnal variation,seasonal variation and the spatial variation of soil respiration(Rs)and environmental factors.At the same time we also got the surface temperature(LST)and vegetation index(VIs)by using MODIS products of the three sampling sites.The purpose of the study was to 1)explore the spatial variation characteristics of Rs and determine the reasonable sampling number;2)understand the diurnal and seasonal variation of Rs and its controlling factors,and build the fitting models of Rs with time variation;3)try to use MODIS products to fit seasonal variation of Rs.Main results are as follows:(1)At spatial scale,the variation coefficients of Rs,total carbon(C),total nitrogen(N)and soil organic carbon(SOC)increased gradually with the decease of sampling scale,indicating that the greater the sample distance of these factors was,the better of their spatial representation was.Generally,the spatial variation of Rs was mainly controlled by physicochemical factors,which influence the soil microbial respiration and root respiration,such as C,N,SOC etc,and root biomass inside the PVC collar directly promoting root respiration intensity.The soil temperature and moisture had little effect on Rs spatial variation.(2)Necessary sample size of each factor was mainly decided by its spatial variability,and the value orders from high to low were: Nroot-weight > NRs > NC > Nlitter-weight > NSOC > NN > NSWC > NTs.Under 95% confidence level and 80% estimation precision,necessary sample size for Rs at 5 m,2.5 m and 1.25 m sampling scales was 23,27 and 28,respectively.(3)The diurnal variation of Rs and Ts all showed a single concave curve trend from 18:00 to 18:00 next day.The maximum Rs occurred at the period 14:00 to 18:00,and the minimum Rs was in the morning.The Rs diurnal variation was mainly controlled by Ts.And the determination coefficient of the fitted exponential function based on T10 was up to 0.9.(4)The seasonal variation of Rs,Ts,LST,and VIs during the nine-year measuring time basically showed a symmetrical ?bell? type,with the lower values in December and January at that time the air temperature was the coldest,and higher value in June and July at that time the temperature was highest over the season.The mean Rs across nine years measurement at the stand 1,stand 2 and stand 3 was 3.57 ± 1.94,4.95 ± 2.39 and 6.11 ± 2.94 ?mol CO2 m-2 s-1,respectively.Nevertheless the seasonal variation of SWC showed a fluctuating pattern,accompanied by a trend of high and low alternation due to the precipitation events.The SWC in three stands over the season was all about 30% ~ 40% or more.(5)The fitting efficiency of exponential and Arrhenius functions based on Ts and LSTnight was approximately equal,and the R~2 was more than 0.70.The latter function fitted with the same temperature has a slightly higher R~2 than the former,and also with a lower RMSE and AIC,but for both equations no significant difference existed.Among all kinds of temperatures,T5 and LSTan could be best used for fitting the seasonal variation of Rs.Nevertheless the SWC had little impact on Rs,and there was no significant relationship between Rs and SWC.From the fitted models of Rs over the season with 5 kinds of vegetation index,we found that the R~2 of the exponential function was higher than that of the linear.The function based on NDVI had a better fitting effect than other VIs had.(6)Results from the fitted two double-variable exponential models,one including both Ts and NDVI,and another including both LSTan and NDVI,showed that for each year data there was no significant difference of R~2 between the two models in each stand.This further indicated that the pure remote sensing model based on LSTan and NDVI could be used for predicting Rs seasonal variation in the study site. |
PDF Full Text Request