基于多源遥感的森林地上生物量KNN-FIFS估测

【目的】针对多源遥感数据及其派生特征因子数据维度高、信息冗余、易造成估测模型过拟合等问题,从高维度遥感特征因子中高效优化特征组合,优化区域森林地上生物量(AGB)的k最近邻(k-NN)估测模型。【方法】提出基于快速迭代特征选择的k最近邻法(KNN-FIFS),以森林资源样地调查数据计算的森林AGB为参考,以留一法交叉验证(LOO)相应的k-NN模型反演的森林AGB均方根误差(RMSE)最小为原则,依次迭代选取遥感特征,优化区域森林AGB的k-NN估测模型。以大兴安岭根河森林保护区为研究区,结合Landsat-8 OLI各波段光谱信息、植被指数、纹理、地形因子、机载合成孔径雷达(SAR)P-波段HV极化后向散射强度信息(PHV)以及森林资源样地调查数据,利用KNN-FIFS方法估测研究区森林AGB,并与多元线性逐步回归法(SMLR)进行对比分析。【结果】利用KNN-FIFS方法,得到当k为3,特征组合为PHV、短波红外波段一均一性(H6)、短波红外波段一二阶矩(S6)、短波红外波段二二阶矩(S7)、海蓝波段相关性(Cr1)、近红外波段相关性(Cr5)、海蓝波段相异性(D1)、增强型植被指数(EVI)时,研究区森林AGB估测结果最优,其精度(R~2=0.77,RMSE=22.74 t·hm~(-2))显著优于SMLR估测精度(R~2=0.53,RMSE=32.37 t·hm~(-2))。【结论】KNN-FIFS方法相比SMLR更适用于森林AGB多源遥感估测;KNN-FIFS方法可以从高维度遥感特征因子中高效选取相关特征进行森林AGB估测。     

联合GF-3 PolSAR数据与Landsat-8 OLI数据的森林地上生物量估测

【目的】森林是陆地生态系统的重要组成部分,精确估测森林地上生物量对森林资源的经营管理具有指示作用,对研究全球碳循环具有重要意义。为了改善单一来源遥感数据估测森林地上生物量的不足,探讨了联合高分三号(Gaofen-3,GF-3)全极化(Polarimetric synthetic aperture radar,PolSAR)数据极化分解参数和Landsat-8 OLI数据估测森林地上生物量的可行性,并针对多源遥感数据的冗余问题优化特征组合。【方法】以广西南宁市高峰林场为研究区,结合森林样地调查数据,提取GF-3 PolSAR数据的后向散射系数、极化分解参数和Landsat-8 OLI数据的光谱信息、植被指数、纹理,使用基于序列前向特征选择的K最近邻法(K-nearest neighbor based on sequence forward feature selection,KNN-SFS)估测研究区的森林地上生物量,以留一法交叉验证得到的森林地上生物量预测值和实测值之间的均方根误差(Root mean square error,RMSE)最小为原则,对比验证使用多源遥感数据和单一来源遥感数据时的估测结果,寻求估测森林地上生物量的最优特征组合,基于最优特征组合绘制研究区的森林地上生物量空间分布图。【结果】结合GF-3 PolSAR数据和Landsat-8 OLI数据估测研究区森林地上生物量的精度为RMSE=21.05 t·hm~(-2),R~2=0.75,优于仅使用GF-3 PolSAR数据估测的精度(RMSE=28.38 t·hm~(-2),R~2=0.47)和仅使用Landsat-8 OLI数据估测的精度(RMSE=29.52 t·hm~(-2),R~2=0.42)。【结论】多源数据协同反演森林地上生物量可以提高估测的精度,基于KNN-SFS方法联合GF-3 PolSAR数据与Landsat-8 OLI数据可以较好地估测森林地上生物量。     

高分一号遥感影像下崂山林场林分生物量反演估算研究

以高分一号遥感影像为数据源,结合外业实测数据,通过提取植被指数因子、波段组合因子以及地形因子等19个因子,建立崂山林场生物量多元线性反演模型,进而分析该反演模型的精度和优缺点,估测崂山林场森林生物量。结果表明:应用多元线性回归法建立的生物量模型平均反演精度达到80.75%;根据反演出的林分生物量估测模型,估测崂山林场2013年的林分生物量为402 485.44 t。     

森林地上生物量的多基线InSAR层析估测方法

【目的】发展一种森林地上生物量(AGB)的多基线干涉合成孔径雷达(InSAR)层析估测方法,解决热带雨林森林AGB遥感估测常规方法的信号"饱和"问题,为区域及全球森林生物量估测和碳储量研究提供关键技术支撑。【方法】以法属圭亚那巴拉库(Paracou)热带雨林为研究对象,以Tropi SAR 2009 P-波段多基线机载SAR数据和85块样地调查数据为主要数据源。首先,根据HH极化层析相对反射率的三维分布信息提取林下地表高度,对HV极化多基线InSAR数据进行地形相位去除;然后,对HV极化多基线InSAR数据进行三维成像,并对其进行地理编码,得到地理坐标空间层析相对反射率的三维分布信息;最后,利用样地调查数据,分析不同高度处层析相对反射率与森林AGB的相关性,进而建立以层析相对反射率为输入特征的森林AGB估测模型,同时采用留一交叉验证法(LOOCV)对其估测模型进行精度评价。【结果】20 m以下各高度处层析相对反射率与森林AGB呈不同程度的负相关关系,以5 m高度处层析相对反射率与森林AGB的负相关性最强(相关系数达到-0.58);20 m以上各高度处层析相对反射率与森林AGB呈不同程度的正相关关系,以25 m高度处层析相对反射率与森林AGB的正相关性最强(相关系数达到0.63)。采用5 m高度处层析相对反射率构建模型的估测精度为88.44%,均方根误差为49.85 t·hm-2(相对均方根误差为13.56%);采用25 m高度处层析相对反射率构建模型的估测精度为88.82%,均方根误差为47.30 t·hm-2(相对均方根误差为12.87%);同时采用5 m和25 m高度处层析相对反射率联合构建模型的估测结果最优,估测精度为89.17%,均方根误差为46.45 t·hm-2(相对均方根误差为12.63%)。【结论】通过多基线InSAR层析技术得到的层析相对反射率信息有效解决了热带雨林森林AGB遥感估测常规方法的信号"饱和"问题。采用5 m和25 m高度处层析相对反射率可反演得到高精度的森林AGB,表明多基线InSAR层析技术得到的特定高度处层析相对反射率对热带雨林森林AGB具有良好的指示作用;同时利用5 m和25 m高度处层析相对反射率进行联合估测可进一步提高森林AGB的估测精度,说明充分利用不同层次的森林垂直结构信息可进一步提高复杂森林空间结构条件下的森林AGB估测精度。     

机载高分辨率遥感影像的傅氏纹理因子估测温带森林地上生物量

【目的】从反映森林冠层大小的树冠纹理结构出发,利用高空间分辨率遥感影像中树冠纹理的周期性信息,提取基于傅里叶变换纹理序列的纹理指数(FOTO,Fourier-based textural ordination)估测森林地上生物量,探究FOTO纹理因子在温带森林生物量估测上的潜力,为提取新型纹理参数估算森林生物量提供新的参考途径。【方法】以2009年9月获取的小兴安岭地区凉水国家自然保护区(47°11'N,128°53'E)高分辨率机载航空影像(空间分辨率0.5 m)为例,通过提取CCD影像的FOTO纹理参数,采用多元逐步回归方法对森林地上生物量进行参数反演,并对CCD 3个波段影像提取的9个FOTO纹理因子以及波段平均影像提取的3个FOTO纹理因子2种方法的生物量估测结果进行比较。同时,在研究中尝试采用5种不同尺寸(60 m×60 m,80 m×80 m,100 m×100 m,120 m×120 m和150 m×150 m)的窗口,产生不同尺寸的FOTO因子与生物量进行回归建模。最后,将FOTO纹理因子作为自变量与激光雷达反演的参考生物量进行拟合,利用多元逐步回归方法建立生物量模型,并采用十折交叉验证评估预测模型的泛化能力。【结果】FOTO纹理因子与森林生物量的相关性较高,CCD影像3个波段的9个FOTO纹理因子与生物量的R~2均高于0.67,窗口60 m×60 m,80 m×80 m,100 m×100 m,120 m×120 m和150 m×150 m的估测精度分别为67.3%,73.4%,74.4%,78.3%和80.9%。CCD影像波段平均影像的3个FOTO纹理因子与生物量的R~2均高于0.57,5种窗口尺寸的估测精度分别为58.2%,62.1%,64.3%,67.4%和70.9%。根据最优预测模型获得分辨率100 m的凉水试验区全覆盖生物量结果图,精度为74.41%,RMSE为50.55 t·hm~(-2)。【结论】基于FOTO算法提取的纹理因子与森林地上生物量密切相关且无明显饱和现象,对我国北方温带混交林区的生物量反演有极大潜力。FOTO纹理因子与森林地上生物量的多元线性逐步回归模型R~2达0.81,RMSE为46.78 t·hm~(-2)。     

基于仿真大光斑激光雷达和多层感知器的森林地上生物量估算模型构建

森林是全球重要的陆地生态系统,各国普遍采用地面样地调查的方法评估其资源量和生物量。随着激光雷达技术的发展,采用星载大光斑激光雷达估算大区域森林地上生物量将成为另一种选择。为探索利用大光斑激光雷达估算森林地上生物量的方法,提出了一种基于仿真大光斑激光雷达和多层感知器的森林地上生物量估算模型。比较仿真大光斑激光雷达波形参数13种组合拟合森林地上生物量的效果后,认为多层感知器的估测精度高于多元线性回归。与样地实测地上生物量相比,多元线性回归估测结果的偏差范围为-34.96~23.28t/hm²,多层感知器估测结果的偏差范围更小,为-19.09~20.19t/hm²。因此,多层感知器估测森林地上生物量的效果优于多元线性回归。     

基于Landsat TM的香格里拉市高山松生物量估测重建

基于香格里拉市2006年TM影像、森林资源二类调查数据、外业调查数据,利用随机点所在小班的遥感因子平均值建立数据集,通过数据筛选和相关性分析,选出123个样地数据及14个遥感因子,建立了基于遥感因子的高山松生物量估测的非线性和线性模型,并讨论了模型精度及其预测精度。研究结果表明,线性模型(R2adj=0.406、RMSE=34.18 t/hm~2、rRMSE=38.54%)比非线性模型的精度(R_(adj)~2=0.286、RMSE=37.79 t/hm~2、rRMSE=42.60%)高;运用交叉验证法得到的线性模型的预测精度(RMSE’=35.12 t/hm~2、rRMSE’=39.59%)也要高于非线性模型的预测精度(RMSE’=38.44 t/hm~2、rRMSE’=43.34%)。与其它2个同类研究相比,重建的高山松生物量模型虽然在模型精度上还略有欠缺,但是建模数据更为随机、合理,对因子提取方法进行了一定改进,建立的模型为高海拔地区利用遥感数据和森林资源二类调查数据估测典型乔木森林生物量提供参考,为地形起伏较大区域进行森林生物量估测提供了较为完整的技术方法。     

旺业甸林场人工林生物量遥感反演研究

森林生物量能直接反映森林质量,遥感技术结合地面样地能实现林分或区域范围森林生物量的反演,反演结果对制定森林资源合理利用、生态环境改善等方面的政策具有重要的指导意义。论文以旺业甸林场Landsat8 OLI影像为数据源,从影像中提取161个植被指数,对比Pearson相关系数法和随机森林法进行特征变量选择,分别筛选出合适的因子作为模型自变量,结合实地调查数据,建立多元线性逐步回归、地理加权回归、kNN回归和随机森林等4种生物量反演模型,并对模型结果进行精度验证。研究结果表明:1)利用Pearson相关系数法进行特征变量选择要优于随机森林法。2)短波红外光和近红外区间波段组合得到的植被指数与生物量的相关性显著,相关性系数最高的前五个因子为SR627、SR637、SR647、SR64、SR213,分别达到了0.776、0.761、0.730、0.702和0.657;3)4种生物量反演模型中,随机森林模型效果最好,决定系数R2为0.72,RMSE=8.12,EA=76.54%;线性逐步回归模型次之,R2为0.65,RMSE=9.01,EA=72.88%;其次是kNN回归模型,R2为0.59,RMSE=9.75,EA=74.89%;地理加权回归模型效果最差,R2为0.58,RMSE=13.75,EA=53.95%;4)利用随机森林模型对研究区进行生物量反演,反演结果生物量空间分布与实际情况基本一致,反演效果较好。     

基于机载P-波段全极化SAR数据的复杂地形森林地上生物量估测方法

【目的】利用国产合成孔径雷达(SAR)系统(CASMSAR)获取的机载P-波段全极化SAR(PolSAR)数据,分析SAR对森林地上生物量(AGB)的响应与地形的关系,建立融合地形因子的高精度多项式模型,以提高森林AGB的估测精度。【方法】首先以基于机载激光雷达(Li DAR)数据得到的研究区坡度分布图与结合实测样地AGB数据得到的森林AGB分布图作为参考数据进行系统抽样,分析森林AGB与P-波段PolSAR后向散射强度的关系以及不同坡度下二者的相关性变化;然后利用Li DAR得到的高精度数字高程模型(DEM)结合机载P-波段的轨道数据计算当地入射角,进而建立以后向散射强度、当地入射角以及雷达视角为输入特征的多项式统计模型,同时将以上系统抽样得到的样本一部分作为模型训练样本,一部分作为精度检验样本。为避免样本尺度引起的偶然性,检验了20 m×20 m至100 m×100 m不同样地尺度下的估测精度。【结果】以90 m×90 m样本为例,当坡度为0°~5°时,引入当地入射角(第2组特征)的估测精度与未引入当地入射角(第1组特征)的估测精度分别为:决定系数(R~2)为0.634和0.634,均方根误差(RMSE)为12.07和12.08 t·hm~(-2),总精度(Acc.)为78.91%和78.89%;当坡度为5°~10°时,第2组特征与第1组特征的估测精度分别为:R2为0.524和0.523,RMSE为13.52和13.97 t·hm~(-2),Acc.为80.57%和80.52%;当坡度大于10°时,第2组特征与第1组特征的估测精度分别为:R~2为0.628和0.519,RMSE为13.16和15.70 t·hm~(-2),Acc.为81.05%和78.55%。随着样地尺度增大,2组特征的估测精度均增大,且第2组特征的估测精度大于第1组。【结论】当坡度小于10°时,地形对森林的后向散射强度几乎无影响;当坡度大于10°时,地形的影响显著,在不同尺度下,引入当地入射角的估测模型均可以有效提高估测精度,充分说明模型的有效性和稳定性。此外,随着尺度增大,无论采用的模型是否考虑了地形影响,其估测精度都逐渐提高并趋于稳定,揭示出对复杂地形下森林AGB估测模型效果的评价必须考虑尺度的影响,且参考样地要足够大,否则难以得到客观的结论。     

崂山林场林分生物量遥感反演估算研究

以Landsat TM影像和高分一号影像为数据源,结合外业实测数据,利用遥感影像和实测数据建立崂山林场生物量多元线性反演模型,比较分析不同数据源下反演出的模型精度,估测了崂山林场森林生物量。研究发现,利TM遥感影像作为数据源的崂山林场森林生物量反演模型平均精度为77.12%。高分一号遥感数据反演的生物量模型平均反演精度达到80.75%,高于TM数据源下的生物量反演模型精度。分别根据TM遥感影像和高分一号遥感影像林分生物量估测模型,估测的崂山林场2009年的林分生物量为401185.62t,2013年的林分生物量为402485.44t。     

Browsing Damage on Pine (Pinus sylvestris and P. contorta) by a migrating moose (Alces alces) Population in Winter: Relation to Habitat Composition and Road Barriers

In Scandinavia, moose (Alces alces L.) sometimes cause severe browsing damage to economically-important pine. Moose-vehicle accidents have spurred construction of fences along roads, and these may interfere with moose migration between summer and winter ranges, or the road alone may be a barrier. If this happens and moose build up along roads, landowners may suffer economically. Therefore, this study investigated whether roads, fences or other factors influence the use of young pine stands by moose. Eighty stands along roads in northern Sweden were evaluated in which individually-browsed branches were counted on 9972 pines. Moose browsing was not significantly related to birch (Betula pendula Roth, B. pubescens Ehrh.) density, nor did it differ between pines (Pinus contorta Douglas or P. sylvestris L.). However, increased pine density, site productivity and proximity to a highway were associated with increased browsing. Further large-scale studies are needed to understand moose habitat selection and the effects of roads.     

Effects of Root Freezing on the Physiology and Growth of Picea glauca,Picea mariana and Pinus banksiana Seedlings Under Different Soil Moisture Regimes

The root systems of 2-yr-old Picea glauca, Picea mariana and Pinus banksiana seedlings were submitted to various frost temperatures during an artificial frost to induce different levels of root damage. Frost-damaged and control seedlings were placed in a greenhouse under high and low soil moisture regimes. Seedling growth and physiology were evaluated periodically. Seedling survival was reduced when root damage reached levels of 60-80%. Root systems of all three species showed partial to total recovery by the end of the experiment. In general, root freezing damage caused reductions in seedling growth, with these reductions becoming less significant over time. Root damage had little to no effect on black spruce and jack pine seedling physiology, while white spruce CO 2 uptake decreased with increasing root damage. Shoot nitrogen content of all three species decreased slightly with increasing root damage.     

Stand Structure in Pine,Spruce and Deciduous Forests: A Comparison Between Finnish and Russian Karelia

This study investigated the stand structure in pine, spruce and deciduous forests in the border district of Finland and Russia. A total of 46 mature forest stands was selected as pairs, the members of each pair being as similar as possible with respect to their forest site type, age, moisture and topography. The stands were then compared between the two countries by means of basal areas and number of stems. The proportions of dominating tree species were 2-12% lower, and correspondingly the proportions of secondary tree species higher, in Russian forests. The density of the forest stock was also higher in each forest type in Russia. The forests in the two countries differed most radically in terms of the abundance of dead trees. The amount was two to four times higher in Russian deciduous and spruce forests, and in pine forests the difference was 10-fold. The stand structures indicated that Russian coniferous stands, in particular, were more heterogeneous than intensively managed pine and spruce stands in Finland.     

Growth and Terminal Bud Formation in Picea abies Seedlings Grown with Alternating Diurnal Temperature and Different Light Qualities

The root collar diameter and the height:diameter ratio are of particular importance in container-grown seedlings where a high density in the containers may produce spindly seedlings. Temperature regimes and light quality are known to affect plant growth. The aim of this study was to identify responses in Picea abies (L.) Karst. seedlings grown with light providing different red:far-red ratios and under temperature regimes with alternating day (DT) and night temperature (NT) from negative (DT < NT) to positive (DT > NT) difference (DIF) between DT and NT. Experiments were conducted in controlled environment chambers and in a daylight phytotron. Only limited thermoperiodic responses appeared in P . abies seedlings with respect to seedling height and dry weight accumulation. The formation of terminal buds, however, was clearly delayed in seedlings grown at negative DIF. The results indicate a requirement for day extension light that is high in far-red, to prevent terminal bud formation under natural short-day conditions. An extended study should be conducted to clarify the minimum level of light intensity and the optimal light quality needed to prevent terminal bud formation under natural short-day conditions.     

CAD与建筑设计的关系

本文分析了CAD在设计中引起的正面和负面影响,并进行了系统的阐述,从而使设计者在应用中能够保持客观的态度。     

Risk Perception Among Non-industrial Private Forest Owners

Based on an enquiry, risk perception among non-industrial private forest owners is described in relation to climate change and forestry hazards. Of the respondents, 11% took action to remedy the effects of climate change. Out of a given set, hazards were ranked according to each respondent's experience of recent substantial financial loss to the estate and in relation to his or her willingness to make investments aimed at risk reduction. For each hazard, the respondent assessed the risk in four classes ranging from very high to negligible risk. Six hazards were considered most problematic in all three aspects: browsing damage, falling timber prices, damage by wind, spruce bark beetle, root rot and pine weevil. A majority of the respondents claimed to take action to reduce the risk associated with at least one hazard, while 35% did not know whether they did. Excluding climate change, the need for decision support was the largest in relation to damage by wind owing to a combination of perceived high risk and a high level of ignorance in relation to whether risk-reducing measures were taken.     

Interactions in decomposition and N mineralization between tropical legume residue components

In situ produced plant residues contain a mixture of different plant components of varying quality. To assess synergistic or antagonistic interactions occurring during the decomposition and mineralization of such mixtures, individual plant parts (stems, leaves, leaf litter and roots) or the mixture of stems, leaves and leaf litter of the agroforestry species pigeonpea (Cajanus cajan) or of crop residues of peanut (Arachis hypogaea) or of the weed hairy indigo (Indigofera hirsuta) were incubated in pots for 19 weeks. Periodically, remaining plant residues were sieved out (>2 mm), weighed and N content as well as soil mineral N determined. Above- and below-ground residues of peanut decomposed fastest and showed the largest N release in agreement with their high N concentration and low-acid detergent fibre (ADF) : N ratio. Hairy indigo was hypothesized to be of lower quality than pigeonpea because of its high-polyphenol content. However, it decomposed faster than pigeonpea, largely because of the higher N and lower lignin concentration of its components. Ranking of individual plant components for N mineralization resulted in the following pattern, leaves > leaf litter > roots > stems. In mixtures of the different plant components a similar species order in decomposition was obtained, e.g. peanut > hairy indigo > pigeonpea. The amount of N released from the mixture was dominated by stem material that comprised 46–61% of the mixture. The interactions in mixtures were relatively small for peanut (generally high-quality components) as well as for pigeonpea (low proportion of high-quality components, i.e. N rich leaf material). However, a positive interaction occurred during later stages of N mineralization in the mixture of hairy indigo as it had a significant proportion of N rich components and absence of highly reactive polyphenols. Thus, for plants with low to intermediate chemical quality attributes, manipulating plant composition (e.g. by varying harvest age, affecting stem and leaf proportions) will be important to obtain significant interactions during decomposition when its components are mixed.     

Pasture production and tree growth in a young pine plantation fertilized with inorganic fertilizers and milk sewage in northwestern Spain

The use of organic waste materials such as milk sewage as an organic fertilizer could have the dual advantages of organic-waste disposal and reduced dependence on inorganic fertilizers. The effects of fertilization with (1) conventional mineral fertilization, (2) milk sewage sludge at 40 kg N ha⁻¹ target rate and (3) no fertilization on pasture production and tree growth were examined in an experiment consisting of two pasture mixtures under a one-year-old Pinus radiata plantation with a density of 2500 trees ha⁻¹. The two pasture mixtures were: (1) Dactylis glomerata L. var. saborto (25 kg ha⁻¹) + Trifolium repens L. group Ladino (4 kg ha⁻¹) + Trifolium pratense L. var. Marino (1 kg ha⁻¹); (2) Lolium perenne L. var. Tove (25 kg ha⁻¹) + Trifolium repens L. group Ladino (4 kg ha⁻¹) + Trifolium pratense L. var. Marino (1 kg ha⁻¹). The experiment began in the spring of 1995 using a randomized block design with three replicates in Castro Riberas de Lea (Lugo, Galicia, north-western Spain). Plot size was 12 × 8 m², with a 1 m buffer strip between plots. Two-year data showed that fertilization with either material had a positive effect on pasture production, with no significant difference between the two fertilization treatments. Tree growth in the milk sewage sludge plot was significantly higher than in the control plots. Inorganic fertilization increased pasture production, but affected tree growth negatively. The results show that milk sewage sludge could be used as a fertilizer in silvo-pastoral systems. This revised version was published online in June 2006 with corrections to the Cover Date.     

Key Habitats in the Norwegian Production Forest: A Case Study

This study investigated the forest history, structural characteristics, the presence of fungal species suggested to indicate ecological continuity in old-growth forest, and the species composition of saproxylic beetles in 30 old forest ''woodland key habitats'' (WKHs), and compared them with production forest in the same age-class in south-eastern Norway. No statistically significant differences in forest characteristics, community of saproxylic beetles or number of red-listed beetles were found between the WKHs and the production forest, probably owing to a combination of profound long-term logging and a lack of will to delineate sufficiently large WKHs in the one area with significantly less forestry impact. The study indicates the advantage of including forest history information in WKH selection where such data exist, and the importance of restoring habitats in heavily exploited forest landscapes.