【作者】 王雪军；

【导师】 孙玉军；

【作者基本信息】 北京林业大学 ， 森林经理， 2013， 博士

【摘要】 森林资源监测是林业重要的基础性工作,是国情国力调查的重要组成部分。传统森林资源监测的间隔周期一般为5—10年,信息产出的时效性相对滞后,但是随着遥感RS技术、GIS技术和GPS技术在森林资源监测中的综合研究深入和不断进展,基于遥感技术的森林资源的动态监测正在受到世界各国越来越多的重视。本研究旨在提高森林资源动态监测能力,增强森林资源监督管理的针对性和时效性,选取辽宁省鞍山市为示范研究区,采用多源、多级分辨率、多时相的遥感数据,结合地面调查数据,通过遥感反演技术来获得森林相关参数(例如：森林面积、林分郁闭度、森林蓄积量等),利用实地调查数据对模型计算得到的森林参数进行检验,通过多时相的遥感数据,对所获得的森林参数进行动态分析,完成对示范监测区森林资源动态变化的监测工作。研究结果表明：(1)在现行森林资源调查方法的基础上,以PPS抽样为基础,综合利用多级遥感数据,提出基于抽样技术的多级遥感监测技术方法,并以鞍山市为例,利用2006年TM、MODIS和航片等遥感数据,进行多级遥感监测技术试验,试验结果表明,与传统遥感方法相比,可以大幅提高MODIS低分辨率影像提取大区域森林分布信息的精度,分类精度达到86.77%。(2)以TM为主要数据源,利用线性光谱解混模型对研究区三期影像的郁闭度进行估测,预测值与野外实测值比较正确率达到80%以上,生成研究区三期郁闭度图层,并进行了森林郁闭度动态变化分析,结果显示：在1995至2006年期间,鞍山市林分郁闭度呈现先下降后上升的趋势,但林分整体郁闭度水平1995年为最好。(3)采用分水岭算法,利用航片影像资料,通过一类调查样地法和目视样地法进行单木森林结构参数提取实验,提取了单木的株树、树冠的平均面积,树冠的平均冠幅,以及单木的位置等信息,并对实验结果进行了分析和讨论。结果显示,利用高分辨影像资料,采用分水岭算法,可以有效提取单木森林结构参数(株数、位置、冠幅等),这种方法可应用于少林地区或城市区域的森林资源调查。(4)以三期TM为主要数据源,结合一类调查样地数据,采用逐步回归方法进行森林蓄积量遥感估测模型的建模实验。利用所建立的遥感估测模型,分别对鞍山市1995年、2000年、2006年3期森林蓄积量进行了估测,并分析了其时空动态变化特征。结果表明,三期模型得到的森林蓄积量预测值分别为688.11×105m3、755.86×105m3和1042.67×105m3；说明鞍山市森林蓄积量从1995至2006年呈现增长的趋势；并利用2007年森林资源二类调查小班统计蓄积进行验证,2006年模型估测值与2007小班蓄积统计值相比,两者相差13.79%,模型精度达到85%以上,较为理想。(5)采用一类调查数据资料,利用立地分级方法建立平均树高、平均胸径、公顷株数和公顷蓄积的生长模型,通过所建立的生长模型,可应用于森林资源调查中相关因子的更新。同时,通过研究森林蓄积、林分生长量和消耗量与年龄、平均树高、平均胸径和郁闭度等等林分因子之间的关系,建立了森林蓄积、生长量和消耗量的估测模型。结果表明,基于生长模型来更新森林资源档案数据,实现森林资源的年度监测的方法是可行的。(6)研究发展了基于多源数据源的森林资源遥感技术方法,建立了市级森林资源年度动态遥感监测体系,并成功应用于鞍山市的森林资源年度动态监测中,取得了很好的效果。通过鞍山市示范应用,能及时掌握鞍山市2007-2009年森林资源动态变化,为鞍山市制定林业发展、生态环境宏观规划提供可靠的依据。本研究首次综合利用多源、多级遥感数据,结合地面调查样地资料,采用基于PPS抽样技术的多级遥感监测方法,对研究区森林面积进行定量提取研究。在森林资源动态监测中,采用遥感技术和空间信息技术为采集数据的主要手段,遥感宏观监测与典型地面调查相结合,定量分析与定性评价相结合,模型分析与实地测量相结合的方法,解决了目前森林资源监测迫切需要的关键技术,建立了鞍山市森林资源年度监测技术体系,并成功应用于鞍山市森林资源的年度动态监测中,大大提高了鞍山市森林资源监测能力,增强森林资源监督管理的针对性和时效性,降低野外调查费用,满足市级森林资源日常调查业务、行政性监管与执法职能的要求,填补市级森林资源年度监测技术的空白。更多还原

【Abstract】 Forest resource monitoring is an important basic work of forestry, also an important part of our national situation. Due to the long monitoring period for5to10years and lagging information output of traditional forest resource, it is hard to meet the needs of the forestry development and ecological construction. With the deeper research and application of RS, GIS and GPS technology in forest resources monitoring, dynamic monitoring based on remote sensing technology is brought to the attention of the countries all over the world more and more. This study aims to improve the dynamic monitoring ability of forest resource, which helps strengthen the pertinence and efficiency of forest resources supervision and management. We selected Anshan (in Liaoning province) as demonstration research area, used multi-source, multi-resolution, multi-temporal remote sensing data, combined with field inventory data, to obtain some indexes (such as forest area, canopy density, volume, etc.) through remote sensing inversion technique. Then, the field survey data were used for model testing. Meanwhile, the dynamic of forest resource indexes were analysised by multi-temporal remote sensing data. Finally, we completed the monitoring work of dynamic changes for forest resources at the demonstration zone of Anshan. The results showed that:(1) Based on the current investigation method for forest resources and-PPS sampling, with comprehensive utilization of multiple remote sensing data, multi-level remote sensing method based on sampling technology was put forward. Taking Anshan as an example, TM, MODIS and aerial photographs of2006were used for the multi-level remote sensing experiment. Results showed that, compared with the traditional remote sensing method, the new method can significantly improve the classification accuracy of large area forest distribution by low resolution MODIS images for, which reached to86.77%.(2) Taking TM image as main source of data, and using linear spectral mixing model, the crown density indexes in the images at3stages for the studied area were evaluated. Comparing with field measured values, the prediction accuracy was up to80%. Crown density layers of3stages were generated then, for the analysis on the dynamic changes of forest canopy density. The results indicated that forest crown density in Aanshan showed a trend of "U-shaped" during1995to2006, with a best level in1995.(3) Based on the data of aerial photography, with field inventory and visual plot materials, forest structural indexes were extracted by watershed algorithm. The extraction of tree number, canopy coverage, crown width, and location for single tree were analyzed and discussed then. Results showed the forest structural indexes can be extracted effectively, and it could be applied in forest survey for open forestland or urban area. (4) Taking TM images at3stages in1995,2000and2006,as main data source, combined with the national forest inventory data, forest stock volume was estimated by stepwise regression method. The space-time dynamic progress of stock estimations for3stages were analysed respectively. Results showed a growth trend for forest stock volume from1995to2006, with estimations followed by688.11×105m3、755.86×105m3and1042.67×105m3. Comparing estimated volume in2006with sub-comparment statistic result of that in2007from forest resource inventory planning and design, we found no significant difference between them (13.79%), which suggested the model can ideally estimate volume, with high accuracy (above85%).(5) Based on the national forest inventory data, growth models for height, DBH (diameter at breast height) and volume were finished by site classification method. All of that growth models were used for the updates of tree measured factors in forest resources survey. Meanwhile, prediction models for volume, growth and consumption at stand level were bulited with some related factors as age, height, DBH crown density, etc. Results showed that archival data of forest resources can be updated based on growth model, which would helped for the annual monitoring of forest resources.(6). Our research improved the technology of forest resources monitoring based on Multi-data Source, and built a city level system for the annual monitoring of forest resources, which had been applicated successfully in Anshan. With the demonstration application, we learned the dynamic of forest resource from2007to2009, and also provided reliable basis and construction suggestions for forestry development and eco-environment planning.(7) Our study promoted to comprehensive utilization of multi-source, multi-level remote sensing data. We studied quantitative extraction of forest area by multilevel remote sensing method based on PPS sampling, and combined with field survey data. Focused on dynamic monitoring of forest resource, we took remote sensing technology and spatial information technology as main methods for gathering data. And then, a system for the annual monitoring of forest resources in Anshan was built by series methods as remote sensing macroscopic monitoring&field investigation, quantitative analysis&qualitative analysis, model estimation&field measurements. This study improved the urgently need technology to monitor forest resource dynamic, and successfully applied to Anshan annual monitoring of forest resources dynamic. In conclusion, our research filled the blank of the municipal annual monitoring technology of forest resources. The techonology could greatly improve the ability of forest resources monitoring in Anshan, reduce field investigation cost, and met the need of municipal survey, supervision and enforcement of forest resources.更多还原