机载激光雷达森林垂直结构剖面参数的沿海平原人工林林分特征反演

中国是世界上人工林面积最大的国家,实时、定量、精确地获取人工林林分特征对于人工林资源监测、管理以及全球碳循环具有重要意义。以北亚热带沿海平原人工林为研究对象,借助机载激光雷达LiDAR(Light Detection And Ranging)点云数据并结合地面实测的55个样地来反演人工林林分特征。首先,构建冠层高度分布剖面CHD(Canopy Height Distribution)和枝叶剖面FP(Foliage Profile);然后,通过Weibull函数分别对CHD和FP进行拟合并提取Weibull参数作为特征变量(第1组);同时,还直接基于点云提取了LiDAR高度变量HRM(HeightRelated Metrics)和冠层密度变量DRM(Density-Related Metrics)(第2组);最后,结合地面实测数据和两组特征变量构建了多元回归模型用于预测各林分特征(即林分密度、平均胸径、胸高断面积、Lorey’s树高、蓄积量和地上生物量)。结果表明:(1)与只使用基于点云的特征变量(即第2组)相比,结合点云特征变量(第2组)和冠层垂直结构剖面特征变量(第1组)的各林分特征预测精度均有所提升(ΔAdjusted R2=0—0.13,ΔrRMSE=0.08—3.65%);(2)对各林分特征预测的结果中,Lorey’s树高(Adjusted R2=0.85, rRMSE=7.66%)和蓄积量(Adjusted R2=0.84,rRMSE=14.27%)的预测精度最高,地上生物量(Adjusted R2=0.78, rRMSE=14.15%)、胸高断面积(Adjusted R2=0.73, rRMSE=14.70%)和平均胸径(Adjusted R2=0.64, rRMSE=15.05%)次之,林分密度(Adjusted R2=0.58,rRMSE=26.16%)的预测精度最低;(3)Weibull函数较准确地反映了亚热带人工林垂直冠层结构,可以有效提高林分特征反演精度。     

层析SAR反演森林垂直结构参数现状及发展趋势

森林垂直结构参数反演是进行森林资源管理、森林蓄积量估算及全球碳循环研究的基础。层析合成孔径雷达TomoSAR(Tomography Synthetic Aperture Radar)是随着InSAR/Pol-InSAR技术的日益发展而产生的,更适用于森林垂直结构参数反演。本文首先介绍了TomoSAR的概念与实现方式:PCT(Polarization Coherence Tomography)、多基线干涉层析SAR MB-InTomoSAR(Multi-baseline Interferometric Tomographic SAR)、多基线极化层析SAR MBPolTomoSAR(Multi-baseline Polarization Tomographic SAR);概括了目前应用TomoSAR技术反演森林垂直结构参数的技术方法与信号模型等;论述了应用TomoSAR技术提取森林垂直结构参数的现状,最后分析了应用TomoSAR技术提取森林垂直结构参数可能的发展方向。     

地基激光雷达的玉兰林冠层叶面积密度反演

叶面积密度LAD(Leaf Area Density)是表征冠层内部叶面积垂直分布的重要参数,其分布廓线的准确反演对研究植被碳氮循环、初级生产力和生物量估算等具有重要意义。本文在电子科技大学校内建立实验样区,利用地基激光雷达Leica Scan Station C10和数码相机获取玉兰林高分辨率3维激光点云数据和真彩色影像。利用监督分类将真彩色影像中枝干等非光合组织与叶片分离,再将像素分类信息映射给点云数据,从而提取叶片点云。通过点云数据体元化,并引入2维凸包算法确定垂直方向分层树冠边界,获取激光接触冠层的频率;随机选择不同高度的多个叶片,利用特征值法进行叶片平面拟合,估算出叶倾角,并结合天顶角估算叶倾角校正因子;最后基于体元的冠层分析VCP(Voxel-based Canopy Profiling)方法实现树林冠层LAD反演。结果表明体元化的叶片点云数据能准确确定树林冠层边界和统计接触频率实现LAD反演;反演的LAD变化走势与区域林木冠层叶片垂直分布相吻合,在冠层中下部随着高度的增加叶面积密度也随之增加,在4 m高度处达到最大值1 m2/m3,之后随着高度的增加叶面积密度逐渐降低。根据LAD计算得到的累积叶面积指数LAI为3.20 m2/m2,与LAI-2200实测的叶面积指数相比,相对误差为1.26%。     

激光雷达森林参数反演研究进展

激光雷达通过发射激光能量和接收返回信号的方式,来获取高精度的森林空间结构和林下地形信息。全波形激光雷达通过记录返回信号的全部能量,得到亚米级植被垂直剖面;离散回波激光雷达记录的单个或多个回波,表示来自不同冠层的回波信号。星载激光雷达一般采用全波形或光子计数激光剖面系统,仅能获取卫星轨道下方的单波束或多波束数据,用于区域/全球范围的森林垂直结构及变化观测。机载激光雷达多采用离散回波或全波形激光扫描系统,能够获取飞行轨迹下方特定视场范围内的扫描数据,用于林分/区域范围的森林结构观测。地基激光雷达多采用离散回波激光扫描系统,获取以测站为中心的球形空间内扫描数据,用于单木/样地范围的森林结构观测。激光雷达单木因子估测方法可分为CHM单木法、NPC单木法和体元单木法3类。CHM单木法通过局部最大值识别树冠顶点,采用区域生长或图像分割算法识别树冠边界或树冠主方向,NPC单木法一般通过空间聚类或形态学算法识别单木,体元单木法在3维体元空间采用区域生长或空间聚类算法识别树冠。根据激光雷达冠层高度分布可以估测林分因子,冠层高度分布特征来自于离散点云或全波形。多时相激光雷达可用于森林生长量、生物量变化等监测,以及森林采伐、灾害等引起的结构变化监测。随着激光雷达技术的发展,它将在森林调查、生态环境建模等生产与科学研究领域中得到更为广泛的应用。     

激光雷达在森林参数反演中的应用

激光雷达是近年来国际上发展十分迅速的主动遥感技术,在森林参数的定量测量和反演上取得了成功的应用。在林业上,高采样密度激光雷达能够获取单株木3维结构特征,采用不同的数据处理方法,可以得到不同精度的单株木参数。利用激光雷达测量森林参数不仅节省了人力,还提高了工作效率,现在已经成为快速获取树木几何参数的一种有效方法。文中主要介绍了LiDAR工作原理、类型及特点、影响LiDAR数据质量的因素、国内外LiDAR的发展状况及应用领域,重点介绍了国内外利用LiDAR数据反演森林参数(树高、郁闭度、冠幅、林分密度、断面积和蓄积量等)的方法和研究进展,同时对今后LiDAR在森林参数反演方面的研究作了展望。     

基于激光雷达卫星(GEDI)的广东省森林冠层高度和生物量估算

森林冠层高度和生物量估算对估算森林碳收支方面起到重要作用。本文以广东省森林为研究对象,以全球生态系统动态调查(GEDI)激光雷达卫星为数据源,分别采用回归树和克里金插值算法,对广东省的森林冠层高度和生物量进行反演。研究结果表明,广东省的树木高度普遍在10～20 m,占比超过50%。树高高值出现在粤北的韶关、肇庆等市,树高普遍在15～20 m;而湛江市的平均树高最低,普遍不足10 m。广东省森林生物量最大值为335.85 t/hm²,最小值为5.25 t/hm²,平均值为98.27 t/hm²。森林生物量高值区域主要分布在粤东山区和粤西山区,而广东省平原和城市化地区森林生物量则较低。本文结果为估算广东省森林生态系统碳吸收提供科学依据。     

芦苇地叶面积指数的遥感反演

 叶面积指数（LAI）是植被冠层结构的一个重要参数，它的改变标志着植被发生了生物物理变化。本文提出了一种利用混和模型 从TM图像上获取叶面积指数的方法。首先，利用冠层反射率（FCR）模型计算并得到查找表； 然后，利用从查找表得到的统计关系进行 LAI制图。试验表明，该方法简单易行，并可较精确地用来反演芦苇地的叶面积指数。     

森林地上生物量遥感反演方法综述

森林地上生物量反演对理解和监测生态系统及评估人类生产生活的影响有着重要作用,日益发展的遥感技术使全球及大区域的生物量估算成为可能。近年来,不同的遥感技术和反演方法被广泛用于估算森林生物量。本文首先总结了现有的全球及区域生物量产品及其不确定性,然后综述了3类方法在森林地上生物量遥感反演中的应用,即基于单源数据的参数化方法、基于多源数据的非参数化方法和基于机理模型的反演方法,阐述了各类反演方法的特点、优势及局限性。最后从机理模型研究、多源遥感数据协同、生物量季节变化研究和遥感数据源不断丰富4个方面对今后的生物量遥感反演研究进行了展望。     

基于极化相干层析算法反演森林垂直结构剖面研究

森林的高度和垂直结构剖面都是重要的森林空间结构信息,极化干涉SAR的出现使定量获取森林结构参数成为可能。在分析比较三阶段法、DEM差分法和相干幅度法估计森林高度的基础上,采用相干幅度算法提取的树高,通过勒让德多项式展开以及单基线极化干涉数据提取森林的垂直结构剖面。利用仿真SAR数据进行实验分析,研究表明,通过极化相干层析技术,能够获取准确的森林垂直结构函数。     

利用激光雷达和多角度频谱成像仪数据估测森林垂直参数

植被的结构参数如植被高度、生物量、水平和垂直分布等，是影响陆地与大气能量交换乃至生物圈多样性的重要因素。多数遥感系统虽然可以提供植被水平结构的图像，但是不能提供植被成分垂直分布的信息。大尺度激光雷达仪器如LVIS产生的激光雷达信号，已成功地用于估计树高和森林生物量，然而大多数激光雷达仪器不具备图像能力，只能提供一个区域内的采样数据。其他的遥感数据如多角度高光谱、多频率多时相辐射计或雷达数据，可根据GLAS（Geoscience Laser Altimeter System）采样的测量用来推断出连续的森林结构区域覆盖参数。 MISR（Multi-angle Imaging Spectrometer）对陆表多角度的成像能力，可以通过BRDF的各向异性提供植被的结构信息。结合激光雷达的垂直采样和MISR的图像，区域内乃至全球性的森林空间参数的成像是可能的。ICESat卫星上的GLAS数据、Terra卫星上的MISR数据为区域或全球性森林结构参数提供了可能。本文的研究目的是评估GLAS数据，分析类似于MISR的数据对森林结构参数的估计能力。本文中使用了LVIS、AirMISR和GLAS数据。通过对GLAS树高的测量与GLAS像元内来自LVIS的平均树高对比，发现它们是高度相关的。同时还探讨了多角度频谱成像仪数据预测树高信息的能力，这将在今后区域内森林结构参数映射加以研究。     

基于被动微波遥感技术的玉米冠层叶面积指数反演

基于玉米冠层结构参数实测数据和Matrix-Doubling(MD)模型构建了玉米出苗期至抽穗期的冠层多波段、双极化微波辐射特性模拟数据库;通过对模拟数据的回归分析得到了玉米冠层在各波段的微波发射率及其与透过率之间的经验关系,并将经验关系应用于0阶微波辐射传输模型;结合土壤发射率模型构建了玉米冠层覆盖地表的微波辐射亮温参数化计算模型,并基于该参数化模型、利用玉米样地微波亮温观测试验数据,采用迭代方法进行了玉米叶面积指数(LAI)的反演.研究表明,LAI反演值与实测值的相关系数r＞0.9,说明多波段被动微波遥感数据在植被冠层LAI反演方面具有较大的应用潜力.     

Characterization of the horizontal structure of the tropical forest canopy using object-based LiDAR and multispectral image analysis

This article's goal is to explore the benefits of using Digital Surface Model (DSM) and Digital Terrain Model (DTM) derived from LiDAR acquisitions for characterizing the horizontal structure of different facies in forested areas (primary forests vs. secondary forests) within the framework of an object-oriented classification. The area under study is the island of Mayotte in the western Indian Ocean. The LiDAR data were the data originally acquired by an airborne small-footprint discrete-return LiDAR for the “Litto3D” coastline mapping project. They were used to create a Digital Elevation Model (DEM) at a spatial resolution of 1 m and a Digital Canopy Model (DCM) using median filtering. The use of two successive segmentations at different scales allowed us to adjust the segmentation parameters to the local structure of the landscape and of the cover. Working in object-oriented mode with LiDAR allowed us to discriminate six vegetation classes based on canopy height and horizontal heterogeneity. This heterogeneity was assessed using a texture index calculated from the height-transition co-occurrence matrix. Overall accuracy exceeds 90%. The resulting product is the first vegetation map of Mayotte which emphasizes the structure over the composition.     

Modeling BVOC isoprene emissions based on a GIS and remote sensing database

This paper presents a geographic information systems (GIS) model to relate biogenic volatile organic compounds (BVOCs) isoprene emissions to ecosystem type, as well as environmental drivers such as light intensity, temperature, landscape factor and foliar density. Data and techniques have recently become available which can permit new improved estimates of isoprene emissions over Hong Kong. The techniques are based on Guenther et al., 1993, Guenther et al., 1999 model. The spatially detailed mapping of isoprene emissions over Hong Kong at a resolution of 100 m and a database has been constructed for retrieval of the isoprene maps from February 2007 to January 2008. This approach assigns emission rates directly to ecosystem types not to individual species, since unlike in temperate regions where one or two single species may dominate over large regions, Hong Kong's vegetation is extremely diverse with up to 300 different species in 1 ha. Field measurements of emissions by canister sampling obtained a range of ambient emissions according to different climatic conditions for Hong Kong's main ecosystem types in both urban and rural areas, and these were used for model validation. Results show the model-derived isoprene flux to have high to moderate correlations with field observations (i.e. r² = 0.77, r² = 0.63, r² = 0.37 for all 24 field measurements, subset for summer, and winter data, respectively) which indicate the robustness of the approach when applied to tropical forests at detailed level, as well as the promising role of remote sensing in isoprene mapping. The GIS model and raster database provide a simple and low cost estimation of the BVOC isoprene in Hong Kong at detailed level. City planners and environmental authorities may use the derived models for estimating isoprene transportation, and its interaction with anthropogenic pollutants in urban areas.     

Moving from plot-based to hillslope-scale assessments of savanna vegetation structure with long-range terrestrial laser scanning (LR-TLS)

Reliable quantification of savanna vegetation structure is critical for accurate carbon accounting and biodiversity assessment under changing climate and land-use conditions. Inventories of fine-scale vegetation structural attributes are typically conducted from field-based plots or transects, while large-area monitoring relies on a combination of airborne and satellite remote sensing. Both of these approaches have their strengths and limitations, but terrestrial laser scanning (TLS) has emerged as the benchmark for vegetation structural parameterization – recording and quantifying 3D structural detail that is not possible from manual field-based or airborne/spaceborne methods. However, traditional TLS approaches suffer from similar spatial constraints as field-based inventories. Given their small areal coverage, standard TLS plots may fail to capture the heterogeneity of landscapes in which they are embedded. Here we test the potential of long-range (>2000 m) terrestrial laser scanning (LR-TLS) to provide rapid and robust assessment of savanna vegetation 3D structure at hillslope scales. We used LR-TLS to sample entire savanna hillslopes from topographic vantage points and collected coincident plot-scale (1 ha) TLS scans at increasing distances from the LR-TLS station. We merged multiple TLS scans at the plot scale to provide the reference structure, and evaluated how 3D metrics derived from LR-TLS deviated from this baseline with increasing distance. Our results show that despite diluted point density and increased beam divergence with distance, LR-TLS can reliably characterize tree height (RMSE = 0.25–1.45 m) and canopy cover (RMSE = 5.67–15.91%) at distances of up to 500 m in open savanna woodlands. When aggregated to the same sampling grain as leading spaceborne vegetation products (10–30 m), our findings show potential for LR-TLS to play a key role in constraining satellite-based structural estimates in savannas over larger areas than traditional TLS sampling can provide.     

Predicting grain protein content of winter wheat using remote sensing data based on nitrogen status and water stress

Advanced site-specific knowledge of grain protein content of winter wheat from remote sensing data would provide opportunities to manage grain harvest differently, and to maximize output by adjusting input in fields. In this study, remote sensing data were utilized to predict grain protein content. Firstly, the leaf nitrogen content at winter wheat anthesis stage was proved to be significantly correlated with grain protein content (R² = 0.36), and spectral indices significantly correlated to leaf nitrogen content at anthesis stage were potential indicators for grain protein content. The vegetation index, VI_green, derived from the canopy spectral reflectance at green and red bands, was significantly correlated to the leaf nitrogen content at anthesis stage, and also highly significantly correlated to the final grain protein content (R² = 0.46). Secondly, the external conditions, such as irrigation, fertilization and temperature, had important influence on grain quality. Water stress at grain filling stage can increase grain protein content, and leaf water content is closely related to irrigation levels, therefore, the spectral indices correlated to leaf water content can be potential indicators for grain protein content. The spectral reflectance of TM channel 5 derived from canopy spectra or image data at grain filling stage was all significantly correlated to grain protein content (R² = 0.31 and 0.37, respectively). Finally, not only this study proved the feasibility of using remote sensing data to predict grain protein content, but it also provided a tentative prediction of the grain protein content in Beijing area using the reflectance image of TM channel 5.     

Mapping beech (Fagus sylvatica L.) forest structure with airborne hyperspectral imagery

Estimating forest structural attributes using multispectral remote sensing is challenging because of the saturation of multispectral indices at high canopy cover. The objective of this study was to assess the utility of hyperspectral data in estimating and mapping forest structural parameters including mean diameter-at-breast height (DBH), mean tree height and tree density of a closed canopy beech forest (Fagus sylvatica L.). Airborne HyMap images and data on forest structural attributes were collected from the Majella National Park, Italy in July 2004. The predictive performances of vegetation indices (VI) derived from all possible two-band combinations (VI_(i,j) = (R_i − R_j)/(R_i + R_j), where R_i and R_j = reflectance in any two bands) were evaluated using calibration (n = 33) and test (n = 20) data sets. The potential of partial least squares (PLS) regression, a multivariate technique involving several bands was also assessed. New VIs based on the contrast between reflectance in the red-edge shoulder (756–820 nm) and the water absorption feature centred at 1200 nm (1172–1320 nm) were found to show higher correlations with the forest structural parameters than standard VIs derived from NIR and visible reflectance (i.e. the normalised difference vegetation index, NDVI). PLS regression showed a slight improvement in estimating the beech forest structural attributes (prediction errors of 27.6%, 32.6% and 46.4% for mean DBH, height and tree density, respectively) compared to VIs using linear regression models (prediction errors of 27.8%, 35.8% and 48.3% for mean DBH, height and tree density, respectively). Mean DBH was the best predicted variable among the stand parameters (calibration R² = 0.62 for an exponential model fit and standard error of prediction = 5.12 cm, i.e. 25% of the mean). The predicted map of mean DBH revealed high heterogeneity in the beech forest structure in the study area. The spatial variability of mean DBH occurs at less than 450 m. The DBH map could be useful to forest management in many ways, e.g. thinning of coppice to promote diameter growth, to assess the effects of management on forest structure or to detect changes in the forest structure caused by anthropogenic and natural factors.     

Application of 3D triangulations of airborne laser scanning data to estimate boreal forest leaf area index

We propose 3D triangulations of airborne Laser Scanning (ALS) point clouds as a new approach to derive 3D canopy structures and to estimate forest canopy effective LAI (LAI_e). Computational geometry and topological connectivity were employed to filter the triangulations to yield a quasi-optimal relationship with the field measured LAI_e. The optimal filtering parameters were predicted based on ALS height metrics, emulating the production of maps of LAI_e and canopy volume for large areas. The LAI_e from triangulations was validated with field measured LAI_e and compared with a reference LAI_e calculated from ALS data using logarithmic model based on Beer’s law. Canopy transmittance was estimated using All Echo Cover Index (ACI), and the mean projection of unit foliage area (β) was obtained using no-intercept regression with field measured LAI_e. We investigated the influence species and season on the triangulated LAI_e and demonstrated the relationship between triangulated LAI_e and canopy volume. Our data is from 115 forest plots located at the southern boreal forest area in Finland and for each plot three different ALS datasets were available to apply the triangulations. The triangulation approach was found applicable for both leaf-on and leaf-off datasets after initial calibration. Results showed the Root Mean Square Errors (RMSEs) between LAI_e from triangulations and field measured values agreed the most using the highest pulse density data (RMSE = 0.63, the coefficient of determination (R²) = 0.53). Yet, the LAI_e calculated using ACI-index agreed better with the field measured LAI_e (RMSE = 0.53 and R² = 0.70). The best models to predict the optimal alpha value contained the ACI-index, which indicates that within-crown transmittance is accounted by the triangulation approach. The cover indices may be recommended for retrieving LAI_e only, but for applications which require more sophisticated information on canopy shape and volume, such as radiative transfer models, the triangulation approach may be preferred.