| As the most widely distributed vegetation types in the world,grassland provides a series of important ecosystem products(meat milk,fur,various herbs)and services(wind and sand control,atmospheric regulation,water conservation,etc.)for human beings.However,due to climate change and human disturbance,grassland biodiversity and ecosystem functions have undergone tremendous changes.Plant functional diversity can explain the relationship between biodiversity and ecosystem functioning with respect to the aspects of plant growth,defense,resistance and other functional strategies,as an important indicator for evaluating ecosystem functions.Traditional means of measuring plant functional diversity are time-consuming and laborious,and also limited by time and spatial scales,making it difficult to expand to large scale studies.Recently,the development of remote sensing technology has facilitated the monitoring of plant functional traits and functional diversity over large areas,providing the possibility to further reveal the response of grassland ecosystems to climate change or anthropogenic disturbances at different spatial and temporal scales.However,compared with forest ecosystems,relatively few studies have been conducted using remote sensing technology to explore the relationships among plant functional traits,functional diversity and ecosystem functions in grassland,and need to be strengthened.This study was conducted in July-August 2020 in the meadow steppe of Ulagai management area of Xilinguole League in Inner Mongolia Autonomous Region as the research area.The leaf or canopy functional traits were spectrally estimated based on the linkage between spectral variation and plant functional traits using a combination of Sentinel-2 satellite images and field survey data.The retrieval accuracy of different functional strategies was estimated,and the distribution of functional traits and functional diversity in the experimental area were mapped.The main findings and conclusions include:(1)Leaf-scale functional trait spectral retrieval:Leaf trait was spectrally achieved using partial least squares regression based on the ASD filed spectrometer,especially for chlorophyll a(Chl-a),chlorophyll b(Chl-b),carotenoids(Car)with R2=0.81,0.79,0.69(p<0.001),respectivley.Non-structural carbohydrates(NSC),Lignin(Lig),Calcium(Ca),Leaf water content(LWC),Nitrogen(N),specific leaf area(SLA),cellulose(Cel)and other traits were also well predicted with R2 ranging from 0.52-0.67.At the leaf level,the correlations among the traits varied considerably(p<0.05).Leaf pigments(Chl-a,Chl-b and Car)showed significant positive correlations with each other,negative correlations with LWC and positive correlations with N.Ca showed positive correlations with LWC and weak correlations with SLA and NSC,negative correlations with all other traits,where Ca showed stronger correlations with leaf pigments.Lig was positively correlated with all other traits.SLA was positively correlated with all other traits except Cel,which showed negative correlation,but the correlation was poor.NSC was weakly correlated with all other traits.(2)Canopy-scale functional trait spectral retrieval:Leaf traits were extrapolated up to the canopy scale(10 m)based on species biomass,and canopy traits were inverted using partial least squares regression methods based on combined Sentinel-2 images.After scale conversion,all measured traits were elevated compared to leaf level except Chl-a(R2=0.71,p<0.001),Chl-b(R2=0.65,p<0.001),and Lig(R2=0.55,p<0.001);where Cel(R2=0.84,p<0.001)improved the most,by 32%,and LWC(R2=0.61,p<0.001)improved the least,by about 2%.At the canopy level,the differences in correlations between traits were small(p<0.05).lig was negatively correlated with N and SLA and NSC was weakly negatively correlated with SLA,but the rest of the traits were positively correlated with each other.In particular,there were strong positive correlations between canopy pigments,strong correlations between N,NSC and Cel and leaf pigments,strong correlations between Ca and LWC,SLA and N,and relatively weak correlations between the remaining canopy traits.Besides,we also found that community productivity was negatively correlated with SLA,N,and Chl-b,and positively correlated with Lig,NSC,Ca,and Cel,and was strongly correlated with SLA,Lig,and NSC.Species diversity was negatively correlated with all traits except SLA and N,while community productivity was negatively correlated with species diversity.(3)Remote sensing retrieval of functional diversity and its relationship with diversity and productivity:Functional traits were selected based on importance ranking combined with trait correlation,and NSC,Ca,and SLA were found to be important indicator traits for community productivity,while NSC,Ca,and Lig were important indicator traits for biodiversity.A comparative analysis of three regression methods,Partial Least Squares Regression(PLSR),Stepwise Regression and Random Forest Regression(RFR),revealed that PLSR could estimate functional diversity better.The prediction accuracies(R2)of functional richness(FRic),functional evenness(FEve)and functional dispersion(FDiv)based on species diversity were 0.30,0.29 and 0.40,respectively,while the accuracies of functional richness(FRic),functional evenness(FEve)and functional dispersion(FDiv)based on community productivity were 0.40,0.38 and 0.50,respectively.The remotely sensed FRic was negatively correlated with species diversity,while FEve and FDiv were positively correlated.but the grassland community productivity(above-ground biomass)was better predicted(R2=0.61,p<0.001),and FRic had the best relationship with productivity(R2=0.40),followed by FDiv(R2=0.28)and FEve(R2=0.27). |
