Spatiotemporal Variation Of Bamboo Forest Phenology Based On Remote Sensing Inversion And Its Influence Mechanism On Carbon Cycle

The impact of phenology and its spatiotemporal variation on forest ecosystem carbon cycle is a hot topic both at home and abroad.Bamboo forest is a special kind of subtropical forest.It has been shown in recent studies have shown that bamboo forests have high productivity and carbon sequestration capacity and plays an important role in maintaining global carbon balance in climate change.The bamboo forest exhibits the outstanding phenological phenomena of biennial on-and off-years.Bamboo shoots grow into new bamboo stalks in spring during the on-years,and old bamboo leaves turn yellow and fall off in spring during the off-years.The carbon budget of bamboo forest ecosystem in different phenological stages is extremely sensitive to climate and environmental factors such as temperature and precipitation.Therefore,it has important theoretical significance for further study on carbon formation mechanism of bamboo forest,and has important practical significance for management of bamboo forest resources.In this study,bamboo forest in Zhejiang Province was taken as the object,and MODIS LAI(leaf area index)time series products from 2001 to 2018 were used as the data source.Firstly,a bamboo forest phenology model is innovatively constructed based on LAI data assimilation technology,and then used to retrieve the bamboo forest start of the growing season(SOS),end of the growing season(EOS)and length of the growing season(LOS).Secondly,the response of phenology to environmental factors was analyzed by combining with meteorological data.Thirdly,Phenological coupling random forest(RF)model was used to estimate the aboveground biomass(AGB)of bamboo forest,and the impact of phenology on AGB was analyzed.Fourthly,bamboo forest phenology was used to drive the Integrated Terrestrial Ecosystem Carbon-budget(In TEC)model to simulate gross primary productivity(GPP),net primary productivity(NPP)and net ecosystem productivity(NEP)of bamboo forest ecosystem based on the data of bamboo forest distribution,LAI and meteorology,and then the response mechanism of phenology on the spatiotemporal variation of carbon cycle was analyzed by structural equation model under the background of climate change.And then the productivity evaluation of bamboo forest ecosystem related to phenology was also discussed.The main conclusions are as follows:1.Based on LAI data assimilation technology,the bamboo forest phenology remote sensing inversion model was established,and therefore accurate spatiotemporal inversion of bamboo forest phenology was realized.(1)The accuracy of LAI products obtained by data assimilation algorithm was high with R² of 0.91,and root mean square error(RMSE)was small with 0.52.(2)Predicting SOS and EOS using assimilated LAI time series were more reliable than that with MODIS LAI and EVI.The accuracy of SOS and EOS obtained by assimilating LAI was the highest with R² of 0.56 and0.54,RMSEs were the lowest with 6.34 days and 5.41 days,respectively.(3)From the perspective of spatial pattern,SOS had a decreasing trend from northwest to southeast,and EOS and LOS were increasing from northwest to southeast.From the perspective of time evolution,about 70%of the bamboo forest distribution areas in Zhejiang Province showed an early trend of SOS,about 60%of the regional EOS showed a delayed trend,and about 70%of the regional LOS showed a lengthen trend.The average SOS was significantly advanced at a rate of 0.83 days per year,while EOS and LOS were significantly delayed and lengthened by 0.28 and 1.12 days per year,respectively.2.Precipitation was the dominant factor affecting bamboo phenology.An increase of 100 mm in regional annual precipitation would advance the average SOS by 0.18 days and delay the average EOS by 0.12 days.And other environmental factors such as climate also had obvious effects on the temporal and spatial variation of bamboo forest phenology.3.The spatiotemporal estimation of AGB of bamboo forest in Zhejiang Province was achieved by coupling phenology and RF model.(1)The training and testing accuracies(R²)of AGB were higher with 0.56 and 0.52,and RMSEs were lower with 7.25 Mg ha^-1 and 7.62 Mg ha^-1.(2)The average annual value of bamboo forest AGB increased significantly at the rate of 0.27 Mg ha^-1 yr^-1(p<0.01)from 2001 to 2018.The AGB had obvious spatial heterogeneity,with the characteristics that northern was larger than southern and the central part was lower.(3)Bamboo phenology had a significant impact on the spatial and temporal distribution pattern of AGB.An advanced,delayed and lengthen of one day in SOS,EOS and LOS increased the annual average AGB by 0.23 Mg ha^-1,0.46Mg ha^-1 and 0.17 Mg ha^-1,respectively.4.The canopy scale photosynthesis simulation module of the In TEC was improved by integrating bamboo forest phenology,and then the spatial and temporal distribution of bamboo forest ecosystem carbon cycle in Zhejiang Province was simulated with high precision.The influence of phenology on the spatiotemporal variation of bamboo forest carbon cycle under the background of climate change was clarified.(1)The In TEC model with phenology had higher accuracy in carbon cycle simulation.Compared with temperature-based phenology of the In TEC model,the accuracies of GPP and NEP estimation based on assimilated-LAI phenology were improved by>1.24 time and RMSE decreased by>9.35%.(2)The inter-annual variation of GPP and NPP showed a significant increase at a rate of 2.0 g C m^-2 yr^-1 per year from 2001 to 2017,while NEP increased slowly.From the perspective of spatial pattern,the distribution pattern of GPP,NPP and NEP were relatively higher in northwest,southwest and northeast of Zhejiang Province.(3)The bamboo forest phenology had a significant influence on the interannual variation of carbon cycle.An advance in the SOS by 1 day would increase the GPP and NPP by 2.50 g C m^-2 yr^-1 and 2.52 g C m^-2 yr^-1 respectively;An delay in the EOS by 1 day would increase the GPP and NPP by 7.31 g C m^-2 yr^-1 and 8.19 g C m^-2 yr^-1respectively;An lengthen in the LOS by 1 day would increase the GPP and NPP by 2.07 g C m^-2 yr^-1and 2.15 g C m^-2 yr^-1 respectively.(4)The structural equation model based on phenology,climate and carbon cycle factors could explain 78%of the changes in GPP,83%in NPP and 73%in NEP.However,the absolute value of direct path coefficient(1.11)of phenology factor was greater than that of climate factor(0.52),which indicated that phenology factors played a more important role in explaining of bamboo forest ecosystem carbon cycle than climate factor.5.Based on 27 factors,such as phenology,carbon cycle,LAI,bamboo distribution,meteorology,soil and topography,a three-layer index system of bamboo forest productivity evaluation was established.And then the productivity level of bamboo forest ecosystem was evaluated using the partial least square path model.(1)The path analysis showed that 15 factors including SOS,EOS and LOS had great influence on bamboo forest ecosystem carbon cycle.(2)The productivity level of bamboo forest ecosystem can be divided into five classes.Approximately 16.53%of the bamboo forest area was classified as having a high productivity level(classes I and II),and its distribution was relatively clustered and mainly distributed in the northwest,southwest and central parts of Zhejiang Province.Accounting for 43.20%and 31.80%of the classes III and IV were medium level,which are mainly distributed in the southeast,northeast and western of Zhejiang Province.