Computerized classification of Mediterranean vegetation using panchromatic aerial photographs

Abstract. Historical aerial photographs are an important source for data on medium- to long-term (10 - 50 yr) vegetation changes. Older photographs are panchromatic, and manual interpretation has traditionally been used to derive vegetation data from such photographs. We present a method for computerized analysis of panchromatic aerial photographs, which enables one to create high resolution, accurate vegetation maps. Our approach is exemplified using two aerial photographs (from 1964 and 1992) of a test area on Mt. Meron, Israel. Spatial resolution (pixel size) of the geo-rectified photos was 0.30 m and spatial accuracy (RMS error) ca. 1 m. An illumination adjustment prior to classification was found to be essential in reducing misclassification error rates. Two classification approaches were employed: a standard maximum-likelihood supervised classifier, and a modification of a supervised classification, which takes into account spectral properties of individual pixels as well as their neighbourhood characteristics. Accuracy of the maximum likelihood classification was 81 % in the 1992 image and 54 % in the 1964 image. The neighbour classifier increased accuracy to 89 % and 82 % respectively. The overall results suggest that computerized analysis of sequences of panchromatic aerial photographs may serve as a valuable tool for the quantification of medium-term vegetation changes.     

Predicting vegetation types at treeline using topography and biophysical disturbance variables

Abstract. The relationships between four vegetation types and variables representing topography and biophysical disturbance gradients were modeled for a study area in east-central Glacier National Park, Montana. Four treeline transition vegetation types including closed-canopy forest, open-canopy forest, meadow, and unvegetated surfaces (e.g. rock, snow, and ice) were identified and mapped through classification of satellite data and subsequent field verification. Topographic characteristics were represented using a digital elevation model and three variables derived from topoclimatic potential models (solar radiation potential, snow accumulation potential, and soil saturation potential). A combination of generalized additive and generalized linear modeling (GAM and GLM, respectively) techniques was used to construct logistic regression models representing the distributions of the four vegetation types. The variables explained significant amounts of variation in the vegetation types, but high levels of variation remained unexplained. A comparison of ‘expected’ and ‘observed’ vegetation patterns suggested that some unexplained variation may have occurred at the basin scale. A suite of tools and techniques is presented that facilitates predicting landscape-scale vegetation patterns and testing hypotheses about the spatial controls on those patterns.     

Implementation of a hierarchical global vegetation classification in ecosystem function models

Abstract. We propose an alternative approach for the currently used biogeographic global vegetation classifications. A hierarchical vegetation classification system is proposed for consistent and routine monitoring of global vegetation. Global vegetation is first defined into six classes based on plant canopy structure and dynamics observable by remote sensing from satellites. Additional biome variability is then represented through a remote sensing derived leaf area index map, and direct climate data sets driving an ecosystem model to compute and map net primary production and evapotranspiration. Simulation results from an ecosystem function model suggest that the six canopy structure-based classes are sufficient to represent global variability in these parameters, provided the spatio-temporal variations in Leaf Area Index and climate are characterized accurately. If a bioclimatically based classification is needed for other purposes, our six class approach can be expanded to a possible 21 classes using archived climatic zones. For example, tropical, subtropical, temperate and boreal labels are defined by absolute minimum temperature. Further separation in each class is possible through changes in water availability defined by precipitation and/or soils. The resulting vegetation classes correspond to many of the existing, conventional global vegetation schemes, yet retain the measure of actual vegetation possible because remote sensing first defines the six biome classes in our classification. Vegetation classifications are no longer an end product but a source of initializing data for global ecosystem function models. Remote sensing with biosphere models directly calculates the ecological functions previously inferred from vegetation classifications, but with higher spatial and temporal accuracy.     

Gene selection and classification of microarray data using random forest

Background  

Selection of relevant genes for sample classification is a common task in most gene expression studies, where researchers try to identify the smallest possible set of genes that can still achieve good predictive performance (for instance, for future use with diagnostic purposes in clinical practice). Many gene selection approaches use univariate (gene-by-gene) rankings of gene relevance and arbitrary thresholds to select the number of genes, can only be applied to two-class problems, and use gene selection ranking criteria unrelated to the classification algorithm. In contrast, random forest is a classification algorithm well suited for microarray data: it shows excellent performance even when most predictive variables are noise, can be used when the number of variables is much larger than the number of observations and in problems involving more than two classes, and returns measures of variable importance. Thus, it is important to understand the performance of random forest with microarray data and its possible use for gene selection.     

中国热带森林植被类型研究历史和划分探讨

热带森林是我国森林植被的重要组成部分,明确其森林植被类型分类对于生物多样性维持机制研究和保护管理等都具有重要意义。该文以中国热带森林植被分类研究中存在的问题为出发点,通过阐述我国各省区植被类型分类的研究历史,在综合考虑多种生物和非生物影响因素的基础上,提出一个新的热带森林植被类型分类框架以供探讨。结果表明:(1)尽管针对我国热带森林的分布范围和群落特征等都已开展了诸多研究,但对我国热带森林植被类型的划分依据和分类体系仍存在争议。(2)尽管我国的热带森林都处于季风气候区带内,但许多热带地区的森林植被类型并不只是受季风影响,而是气候带、关键气候因子、地形、土壤反馈和物种适应等多种因素共同作用的结果。(3)我国的热带森林植被包括5个植被型,即非典型性热带雨林、热带季雨林、热带山地雨林、热带山顶苔藓矮林(热带云雾林)和热带针叶林,其中热带季雨林植被型包含4种植被亚型[热带落叶季雨林、热带半落叶(半常绿)季雨林、热带常绿季雨林和热带石灰岩(石山)季雨林]。(4)阐明了上述热带森林植被型和植被亚型在我国各省区的分布情况,并提出未来有必要对人工恢复后的热带森林进行评估和植被类型划分。综上所述,该文提出一个新的热带森林植被类型分类框架,以期为今后基于不同地区开展热带森林比较研究提供参考。     

Remote sensing approach to detect post-fire vegetation regrowth in Siberian boreal larch forest

Remote sensing with time series data offers considerable potential in the trajectory of post forest fire dynamics beyond the current monitoring of structural attributes that are displayed in the post-fire area. Many studies have addressed this topic by using time series remote sensing indices; however, this approach has sometimes been demonstrated as an unrealistic and biased representation of the post-fire forest patterns due to the saturation issues of vegetation indices. These saturation issues then lead to an underestimation of the forest successional stages and an overestimation of the forest recovery rate. This paper aims to develop a framework for trajectory of the post-fire forest patterns in the Siberian boreal larch forest (Larix sibirica) with the synergistic use of different remote sensing based vegetation-cover indicators derived from the Landsat time series and the WorldView-2 images. A time-series of the forest recovery index (FRI) and fractional vegetation cover (FVC) has been analyzed to estimate the rates of forest regeneration and vegetation recovery across different burn severity levels in the Siberian larch forest. The results showed that the FRI method can be used to observe the regrowth of the larch forest from the tenth year after the fire overlapping with the period of significant increase in the sapling stem volume. The post-fire larch forest canopy can fully recover to the pre-fire condition with respect to the magnitude of the FRI values after 30–47 years where the highest regeneration rate was observed in the moderate burn severity areas followed by the low and high burn severity. On the other hand, the FVC method was positively correlated with burn severity and more sensitive for evaluating the early stages of the forest succession in which the FVC dramatically increases after 5–6 years after the fire. The significant growth of FVC was accentuated by the maximum emergence of the sapling density as well as the rapid growth of herbaceous plants, grasses, shrubs, and shade-intolerant trees immediately after the fire, which could not be evaluated using the FRI. Both time series of the FRI and the FVC are valuable tools for determining the dominant stages of the post-fire larch forest succession in order to understand the relationships between fire disturbance and natural cycles of the boreal larch forest.     

Extant and potential vegetation of an old-growth maritime Ilex opaca forest

The extant and potential (seed bank) vegetation of a rare maritime holly forest on Fire Island, New York was described to assess whether treefall gaps act as a mechanism for the persistence of the species composition of this plant community over time. The Sunken Forest overstory is dominated by Ilex opaca, Amelanchier canadensis and Sassafras albidum. A survey of canopy gaps indicated canopy openings compose 11.3% of the land within the Sunken Forest (16 ha). The composition and density of the seed bank were described using the emergence method. Germination from soil samples placed in the greenhouse was monitored over 2 years. Sixteen species germinated with an average propagule density of 215±41 germinants per square metre. An early successional species (Rhus copallinum) dominated the seed bank, but the late-successional, shade-tolerant I. opaca was also present. Though only one species in the seed bank did not appear in the current vegetation, species abundance differed between vegetation strata. The mean cover and density of the ground-layer flora were higher beneath treefall gaps than closed canopy. Sapling density did not differ between the two canopy conditions, but the dominant species differed with A. canadensis occupying several closed canopy plots and P. serotina saplings appearing more often in gap plots. Most of the dominant canopy species are present in the seed bank and ground layer but are not present in the shrub and sapling layer, with the exception of A. canadensis. Current (2002) sapling density is lower than three decades ago for all species except P. serotina, which is now the dominant woody species in the Sunken Forest understory. The results of this study indicate that if the cause of the sapling reduction is lessened or removed, the characteristic species of the overstory of this unusual plant community may rebound and redevelop a sapling and shrub layer akin to that present before the increase in Odocoileus virginianus on the island.     

Mapping large-scale distribution of submerged aquatic vegetation coverage using remote sensing

This study investigated the potential of using remote sensing for mapping and monitoring of submerged aquatic vegetation (SAV) on a large scale. The spectral characteristics of SAV with varied coverage were measured using a portable spectroradiometer on a man-made lake at the Chongming International Wetland Park, Shanghai, China. A good relationship between the coverage of SAV and their field spectral characteristics was established and the reflectance of SAV increased with its increasing coverage. A regression analysis was then carried out between the coverage and the reflectance at the wavelengths of four QuickBird bands. After making an atmospheric correction from a synchronous QuickBird image for the study site, the image digital number (DN) was converted into the ground reflectance. The reflectance image was then deduced into a distribution map of SAV coverage by using the results of the regression functions between the coverage of SAV and the reflectance rate measured in situ. An accuracy assessment indicated that this approach could be used to quickly monitor the distribution and growth situation of SAV. The implications of this observation, in terms of the ability of remote sensing to estimate and monitor the distribution and dynamics of SAV on a large scale are discussed.     

Carbon and nitrogen isotope composition of vegetation on King George Island,maritime Antarctic

We report abundance of ¹³C and ¹⁵N contents in terrestrial plants (mosses, lichens, liverworts, algae and grasses) from the area of Barton Peninsula (King George Island, maritime Antarctic). The investigated plants show a wide range of δ¹³C and δ¹⁵N values between −29.0 and −20.0‰ and between −15.3 and 22.8‰, respectively. The King George Island terrestrial plants show species specificity of both carbon and nitrogen isotope compositions, probably due to differences in plant physiology and biochemistry, related to their sources and in part to water availability. Carbon isotope compositions of Antarctic terrestrial plants are typical of the C₃ photosynthetic pathway. Lichens are characterized by the widest carbon isotope range, from −29.0 to −20.0‰. However, the average δ¹³C value of lichens is the highest (−23.6 ± 2.8‰) among King George Island plants, followed by grasses (−25.6 ± 1.7‰), mosses (−25.9 ± 1.6‰), liverworts (−26.3 ± 0.5‰) and algae (−26.3 ± 1.2‰), partly related to habitats controlled by water availability. The δ¹⁵N values of moss samples range widest (−9.0 to 22.8‰, with an average of 4.6 ± 6.6‰). Lichens are on the average most depleted in ¹⁵N (mean = −7.4 ± 6.4‰), whereas algae are most enriched in ¹⁵N (10.0 ± 3.3‰). The broad range of nitrogen isotope compositions suggest that the N source for these Antarctic terrestrial plants is spatially much variable, with the local presence of seabird colonies being particularly significant.     

《中国植被志》的植被分类系统、植被类型划分及编排体系

植被志(vegegraphy)是基于植被调查资料, 全面记叙植被的群落外貌、种类组成、结构和功能、生境条件, 以及地理分布等特征, 并对同类植被进行归纳和总结的志书。“植被志”的英文译为“vegegraphy”, 是本文的新造词, 它是由“vegetation”的词头“vege-”和英文后缀“-graphy” (记叙之意)组成的合成词。《中国植被志》的研编是一项时间紧迫、内容复杂、工作量浩繁的重大科学工程。它的完成将极大提升中国植被科学和生态学的研究水平, 并为中国植被资源的合理利用、生物多样性保护及生态环境治理等提供不可或缺的基础资料。本文首先简述了植被的基本特征(主要包括外貌特征、种类组成、群落结构以及动态变化等)和国内外植被调查的进展情况, 简要回顾了中国植被分类系统的研究历史, 并对以往的分类系统进行了若干修订。在此基础上, 着重讨论并提出了《中国植被志》卷册编排体系和用于《中国植被志》研编的植被类型划分方案。在对植被分类系统的修订方面, 主要对高级分类单位及相对应的英文名称进行了讨论和修订。按照本文修订的分类系统, 中国植被的分类单位及其对应的英文名称分别是: 植被型组(Vegetation Formation Group)、植被型(Vegetation Formation)和植被亚型(Vegetation Subformation)、群系组(Alliance Group)、群系(Alliance)和亚群系(Suballiance)、群丛组(Association Group)以及群丛(Association)。在植被型组中, 划分为9类: 森林、灌丛、草本植被、荒漠、高山冻原与稀疏植被、沼泽与水生植被、农业植被、城市植被及无植被地段。关于《中国植被志》的卷册编排和“植被类型”划分, 首先按高级分类单位——植被型划分相应的“卷”; 在此框架下, 模糊“植被亚型”、“群系组”和“群系”的概念, 确定“植被类型” (Vegetation type), 并将其作为植被志各卷中的“册”。这样处理不仅保证了研编工作的可操作性, 也保持了同一卷册中特定植被类型的完整性。《中国植被志》编排体系中的“植被类型”的划分很重要, 它是指具有相同建群种及相同优势类群(如种、属)的植被组合, 但它不是严格意义上的植被分类单位。“植被类型”的划分遵循“优势类群及生活型的同一性, 生境条件的相对重要性, 植被特征及用途的差异性, 以及突出植被志的应用性”等原则。按该编排体系, 《中国植被志》将由48卷约110册组成。     

西双版纳地区主要森林植被乔木多样性的时间变化

为了评估云南省西双版纳森林植被乔木多样性的时间变化,该研究通过样方调查收集了该地区4种主要森林植被(热带雨林、热带季节性湿润林、热带山地常绿阔叶林和暖热性针叶林)乔木多样性数据,结合遥感影像提取了该地区4种森林植被在1992年、2000年、2009年和2016年4个时期的分布,用Simpson、Shannon-Wiener和Scaling物种多样性指数对比4种森林植被乔木均匀度差异,并利用Scaling生态多样性指数和灰色关联评价模型,评估该地区在4个时期的森林乔木多样性的时间变化。结果表明:(1)森林面积比例变化有先减少后增加的趋势,表现为由1992年的65.5%减少至2000年的53.42%,减少到2009年的52.49%,再增至2016年的54.73%,但热带雨林呈现持续减少的趋势。(2) 4种森林植被对乔木多样性的贡献有明显差异,均匀度排序是热带雨林>热带山地(低山)常绿阔叶林>暖热性针叶林>热带季节性湿润林,丰富度排序是热带雨林>热带山地(低山)常绿阔叶林>热带季节性湿润林>暖热性针叶林,对乔木多样性贡献的排序是热带雨林>热带山地(低山)常绿阔叶林>热带季节性湿润林>暖热性针叶林。(3)热带雨林和热带季节性湿润林乔木多样性呈现持续减少趋势,4个时期西双版纳森林植被乔木多样性排序为1992年>2009年> 2016年> 2000年。以上结果表明经济活动是影响西双版纳生物多样性的重要原因,保护热带雨林对维持该地区生物多样性具有重要意义。     

Modeling grassland aboveground biomass using a pure vegetation index

Remote sensing can be the most effective means of scaling up grassland aboveground biomass (AGB) from the sample scale to the regional scale. Among the remote-sensing approaches, statistical models based on the vegetation index (VI) are frequently used to retrieve grassland AGB because of their simplicity and reliability. However, these types of models have never been comprehensively optimized to overcome VI insensitivity and soil effects. Because grassland AGB is related to grassland type, in our research the integrated orderly classification system for grassland (IOCSG) was used to differentiate grassland types. The study area, located in Inner Mongolia, China, included desert steppe, typical steppe and meadow steppe. A pure VI (PVI) was extracted from the normal VI using spectral mixture analysis (SMA). Using a proportional relationship, PVI models were then constructed based on grassland type. The results demonstrated that the PVI models can have clear advantages over the more commonly used VI models. They simplify the parameterization of VI models and thus enhance models constructed for different regions with different remote sensing data sources. Notably, detailed differentiation of grassland types can improve the accuracy of AGB estimates. The methodology proposed in this study is particularly beneficial for AGB estimates at a national scale, especially for countries such as China with many grassland types.     

Comparison of an intuitive mapping classification and numerical classifications of vegetation in south-east Queensland,Australia

Hierarchical agglomerative polythetic clustering of vegetation data for 51 sites from Brian Pastures Research Station, south-east Queensland, Australia, produced site and species groups that supported those formed during a previous intuitive mapping survey. On the basis of floristic composition, these analyses suggested the possible amalgamation of some map units. However, their segregation into separate units was justified on the basis of having distinct photopatterns and physiographic positions in the landscape. The classifications of the trees only (55 species) and woody plants only (115 species) data sets produced site groupings of high similarity to those constructed by the mapping survey. The addition of the herbaceous plants (266 additional species, of which 70% were annual species) tended to dominate the analyses, and produce site groupings that were less similar to the mapping groups. The distribution of many annual species appeared to be independent of the perennial plant species and underlying substrate. In wooded communities, herbaceous plants have only a minor influence on photopattern, which is the primary determinant of the mapping classification. Binary data retained a large percentage of the information contained in the quantitative data. The extra effort of gathering herbaceous data may not be justified for a broad scale mapping project, but is required for comprehensive nature conservation surveys, flora inventory, and vegetation monitoring purposes.     

随机森林模型在分类与回归分析中的应用

随机森林(random forest)模型是由Breiman和Cutler在2001年提出的一种基于分类树的算法。它通过对大量分类树的汇总提高了模型的预测精度,是取代神经网络等传统机器学习方法的新的模型。随机森林的运算速度很快,在处理大数据时表现优异。随机森林不需要顾虑一般回归分析面临的多元共线性的问题,不用做变量选择。现有的随机森林软件包给出了所有变量的重要性。另外,随机森林便于计算变量的非线性作用,而且可以体现变量间的交互作用(interaction)。它对离群值也不敏感。本文通过3个案例,分别介绍了随机森林在昆虫种类的判别分析、有无数据的分析(取代逻辑斯蒂回归)和回归分析上的应用。案例的数据格式和R语言代码可为研究随机森林在分类与回归分析中的应用提供参考。     

Bathymetry,water optical properties,and benthic classification of coral reefs using hyperspectral remote sensing imagery

The complexity and heterogeneity of shallow coastal waters over small spatial scales provides a challenging environment for mapping and monitoring benthic habitats using remote sensing imagery. Additionally, changes in coral reef community structure are occurring on unprecedented temporal scales that require large-scale synoptic coverage and monitoring of coral reefs. A variety of sensors and analyses have been employed for monitoring coral reefs: this study applied a spectrum-matching and look-up-table methodology to the analysis of hyperspectral imagery of a shallow coral reef in the Bahamas. In unconstrained retrievals the retrieved bathymetry was on average within 5% of that measured acoustically, and 92% of pixels had retrieved depths within 25% of the acoustic depth. Retrieved absorption coefficients had less than 20% errors observed at blue wavelengths. The reef scale benthic classification derived by analysis of the imagery was consistent with the percent cover of specific coral reef habitat classes obtained by conventional line transects over the reef, and the inversions were robust as the results were similar when the benthic classification retrieval was constrained by measurements of bathymetry or water column optical properties. These results support the use of calibrated hyperspectral imagery for the rapid determination of bathymetry, water optical properties, and the classification of important habitat classes common to coral reefs.     

Past and present forest vegetation in NE Slovenia derived from old maps

Abstract. Using old military and cadastral maps and modern vegetation maps, the changes in land use over the past 230 yr were followed. The following maps were used: the military map from the second half of the 18th century, the cadastral map of the ‘Economic cadastral survey for regulation of land taxes’ from the first half of the 19th century, and a field map made in the 1980s. A vegetation map of the area was made on the basis of satellite images. We used basic classification techniques combined with extensive field inspections and aerial photographs. The output of this procedure was verified in the field. Additionally, a comparison of statistical data about land use categories is presented. It was established that the last remaining areas of inundated riverine forest disappeared 200 yr ago, and since then only minor changes in land use have occurred.     

National forest inventories and biodiversity monitoring in Australia

Abstract

Forests currently cover over 20% of the Australian continent and are an important resource, subject to a wide range of economic and environmental pressures. These lands support substantial numbers of forest-dependent species with national forest inventories providing important information on biodiversity. National scale information on these forests has been collected or collated since 1988 under the National Forest Inventory (NFI) programme, but substantial problems with the ‘snap shot’ approach have been recognized, particularly with respect to monitoring change and a consequent move towards a permanent and sample-based continental forest monitoring framework (CFMF) has been proposed. CFMF is proposed to consist of three Tiers: (1) satellite imagery of the continent to identify forest and change in forest cover; (2) systematic high-resolution remotely sensed data and (3) permanent ground points at 20×20 km grid interception points. The CFMF approach is in line with the international trend of national forest inventories in developed countries although the Tier 2 approach offers a useful extension. An alternative inventory approach is provided by the National Carbon Accounting System (NCAS) which models the mass of carbon and nitrogen in seven separate living and dead biomass pools for any point under forest or agriculture land use since 1970. The NCAS approach allows fine spatial and temporal monitoring of changes in these carbon and nitrogen biomass pools, and predictions of changes that result from policy or management decisions. This paper briefly reviews NFI, NCAS and the proposed CFMF, with particular emphasis on issues of use and potential for monitoring biodiversity in this biologically very diverse country.