| With the global climate change and population growth,and the demand for ecological civilization construction,improving the efficiency of forest cultivation and maximizing the multiple benefits and functions of forests is particularly important for us to meet human ecological needs and promoting sustainable economic and social development.Fast,real-time and accurate acquisition of forest structural and physiological parameters is the prerequisite for efficient precision forest cultivation and sustainable management,and it is also the basis for studying the interaction process between forests,society and the environment.Traditionally,the forest structural and physiological parameters are mainly obtained through field measurement and laboratory observation,but the realization of that with low cost,high precision and high efficiency is difficult.Therefore,in this study,four typical forests,including a secondary forest with Pinus massoniana and Sawtooth oak as the main tree species and the Metasequoia,Poplar,and Ginkgo plantations,were selected in the Jiangsu study area,combined with remote sensing data acquired by multiple platforms and multiple sensors,to precisely estimate the structural parameters(tree species,tree diameter and height,individual tree volume,number of trees,mean DBH and height,volume and biomass,etc.)and physiological parameters(chlorophyll and carotenoids content).Moreover,using a physiological process model to simulate the forest stand growth process and explore the influence of different cultivation methods on forest stand growth,for providing the necessary supports of technologies,means and methods for the rapid and efficient estimation of forest structural and physiological parameters,simulation and prediction of forest growth.The main conclusions are as follows:1.The multi-platform remote sensing data integration models were created and by optimizing the information extraction and models construction methods,the fast,accurate,and non-destructive individual tree species identification,DBH and volume,and physiological parameters estimation were realized.First,based on point cloud segmentation and sunlit canopy extraction,the airborne hyperspectral and Li DAR data acquired simultaneously were combined with a random forest classifier to identify individual tree species in a secondary forest;Moreover,high-density point cloud data which acquired through a lightweight,backpack-mounted Li DAR sensor(BLS),was combined with trunk detection(DBSCAN)and volume model construction methods,to conduct individual tree DBH and volume measurements in metasequoia and poplar plantations;In addition,the hyperspectral and high-density Li DAR point cloud data obtained by the UAV platform were used and combined with the proposed DSM-based data fusion method,to estimate the physiological parameters(chlorophyll and carotenoids contents)based on the regression method and radiative transfer model and quantitatively analyse the vertical distribution of physiological parameters in the forest canopy.The results indicated that:the tree delineation approach(point cloud segmentation algorithm)was suitable for detecting individual tree in this study(overall accuracy=82.9%).The classification approach provided a relatively high accuracy(overall accuracy>85.4%)for identifying five tree-species in the study site.The identification using both hyperspectral and Li DAR metrics resulted in higher accuracy than only hyperspectral metrics(the improvement of overall accuracy=0.4-5.6%).Comparing with the identification using whole crown metrics(overall accuracy=85.4-89.3%),using sunlit crown metrics(overall accuracy=87.1-91.5%)improved the overall accuracy of 2.3%;Moreover,the individual tree extraction used the density-based spatial clustering of applications with noise(DBSCAN)had high accuracy(overall accuracy=95.77%).In addition,the volume estimation based on the backpack Li DAR data and the volume equation method also performed well(metasequoia:R~2=0.76,r RMSE=20.02%;poplar:R~2=0.98,rRMSE=7.25%);In addition,a number of vegetation indices,derived from the UAS hyperspectral data,were strongly correlated with a number of physiological parameters(Adj-R~2=0.85-0.91;r RMSE=5.19-6.38%).For dawn redwood and poplar trees,the results were consistent in that the lower component of the canopy(least light impacted)had the highest chlorophyll and carotenoids contents.Moreover,the vertical distribution of physiological parameters on individual tree canopy surfaces changed with age likely due to the growth variation from the photosynthetic activity of the canopy.2.The multi-source remote sensing data fusion models were established.The stand-scale structural parameters estimation models combined empirical models and parameter optimization methods has been developed to realize accurate,efficient,and large-scale forest structural parameter acquisition.First,on the basis of correcting the pulse amplitude and waveform shape,the simultaneously acquired full-waveform(FWF)Li DAR and hyperspectral data were combined with multiple linear regression model to estimate the forest structural parameters(mean DBH,tree height,volume and aboveground biomass,etc.);Moreover,the multispectral and true color images from the UAV platform,and the digital aerial photogrammetry(DAP)method were used to generate a three-dimensional point cloud,then fitted a partial least squares(PLS)regression model to estimate the structural parameters of the Ginkgo plantation.The results indicated that:in the estimation of forest structural parameter combined with full-waveform Li DAR and hyperspectral data,the estimation of H_L had a relatively higher accuracy(Adjusted-R~2=0.88,relative RMSE=10.68%),followed by the estimation of AGB(Adjusted-R~2=0.84,relative RMSE=15.14%),and the estimation of V had a relatively lower accuracy(Adjusted-R~2=0.81,relative RMSE=16.37%);and the models including only DPC had the capability to estimate forest structural parameters with relatively high accuracy(Adjusted-R~2=0.52-0.81,relative RMSE=15.70-40.87%)whereas the integration of DPC,FW and HS can improve the accuracy of forest structural parameter estimation(Adjusted-R~2=0.68-0.88,relative RMSE=10.68-28.67%);Moreover,the models with spectral and structural metrics had the highest accuracy(R~2=0.82-0.93,r RMSE=4.60-14.17%).The combo models fitted with stratified sample plots had relatively higher accuracy than those fitted with all of the sample plots(ΔR~2=0-0.07,Δrelative RMSE=0.49-3.08%),and the accuracy increased with increasing stem density.3.The growth process models of Metasequoia and Poplar plantations were constructed and used to predict the growth of its structural parameters and productivity.At the same time,the influence mechanism of different cultivation measures on the growth of Metasequoia and Poplar was explored.First,the historical and future climate data which was simulated by the Climate AP,and the field measured and remotely sensed data were combined to adjust the parameters of the3-PG model that based on physiological processes to simulate the growth process of forest parameters(stand density,DBH,stem biomass and volume)in Metasequoia and Poplar plantations.At the same time,the initial planting density,thinning intensity and fertilization management measures of the forest stand were adjusted gradually to explore the influence of different cultivation measures on the growth of that two tree species.The results indicated that:through remote sensing estimated data verification,the forest structural parameter estimation using 3-PG model had high accuracy.The stand density had the highest accuracy(R~2=0.99,r RMSE=5.79%),followed by the volume(R~2=0.93,r RMSE=14.18%)and stem biomass(R~2=0.87,r RMSE=15.79%),the accuracy of DBH was the lowest(R~2=0.69,r RMSE=16.63%).For the Metasequoia plot without thinning,the DBH growth in the first 50 years was 23.22 cm,and the stand volume was 277.54 m~3/ha.For the poplar plot without thinning,the DBH growth was 47.63 cm and the stand volume in the first 50 years was 702.61 m~3/ha.For the Metasequoia plot with thinning,the DBH growth in the first 50 years was 31.76 cm,and the stand volume was269.30 m~3/ha;for the poplar plot with thinning,the DBH growth in the first 50 years was 72.77cm and the stand volume was 652.36 m~3/ha.Under different cultivation modes,with the increase of initial planting density,the average DBH growth in the first 50 years decreased gradually,and the growth of stand volume increased gradually.With the thinning intensity increases,the average DBH growth in the first 50 years increased gradually,the stand volume growth decreased gradually.With different fertilization modes,the fertilizer management in the middle of growth stage was beneficial to the growth of the average DBH and volume of the stand. |
PDF Full Text Request