【作者】 王强；

【导师】 胡海清；

【作者基本信息】 东北林业大学 ， 森林保护学， 2012， 博士

【摘要】 林火是森林生态系统主要的因子,在大气的化学循环和碳循环中起着重要作用。森林火灾尤其是特大森林火灾的频繁发生不但破坏森林生态系统,而且造成了碳温室气体的大量释放。森林火灾释放的碳主要是以CO2或CO的形式排放,其余的以CH4、多碳烃等有机氧化物形式排放到大气中,对全球气候环境影响巨大。开展林火强度、森林可燃物载量消耗及林火气体释放量等参数的研究是开展林火对大气环境所产生的影响研究的基础,对于了解林火对全球碳循环的影响和改进林火管理策略也有着重要意义。本文将以森林火灾基本信息为研究问题,基于MODIS、TM遥感数据、历史火灾数据和森林调查数据,针对森林火灾森林可燃物载量的消耗及污染气体的释放、构建林木可燃物载量估测、消耗模型及林火排放气体计算模型等问题展开研究,初步完成特大森林火灾的释放污染物量化估算。基于这些问题本文开展了如下研究：完成了遥感数据及其他资料的搜集工作,对遥感数据进行了包括几何校正和辐射校正等预处理,对多时相的遥感数据的进行了平滑处理,完成了正确获取遥感信息的基础工作。完成了基于MODIS多时相遥感数据进行森林可燃物分类的工作,证明选定合适的分类方法和必要的他源数据参与分类,可以获得较为理想的分类精度。研究结果说明利用决策树方法对MODIS植被指数数据分类,可以获得83.2%的精度,要高于利用TM数据进行监督分类的精度,这样的精度基本满足生产需要；使用决策树方法的关键在于获取地物判别阈值并制定好分类规则,TM、MODIS图像在各关键期的可获取性及数量在一定程度上也决定了可燃物的分类精度。完成了基于MODIS数据利用混合像元分解技术进行森林火灾面积估算的方法研究,像元分解后像元精度可以达到92%、总量精度可以达到90%左右,基本上杜绝了以前估测过火面积时对非植被区和火场边缘像元提取精度不高的问题。与传统的飞机航拍圈划面积的方法相比在大空间尺度上操作上更方便、且时效性更强,相比利用高分辨率数据而言又具有较高的时效性,且精度也满足生产的需要。完成了对2006年松岭798高地森林火灾过火面积进行估测,估测过火面积为242030公顷,其中重度火烧面积为23781公顷、中度火烧145408公顷、轻度火烧72964公顷,其中重度火烧、中度火烧、轻度火烧所占比例分别为9.82%、60.05%、30.13%。建立了基于TM图像估测森林可燃物载量的模型,并且利用模型反演大兴安岭地区森林可燃物载量,反演的精度基本符合要求,但还需要进一步提高。在估测可燃物载量的过程中,本文对比应用偏最小二乘估计模型和二阶最小二乘估计模型,通过对模型精度、决定系数及均方根差的误差评估比较,选择二阶最小二乘估计模型反演的可燃物载量,取得了比较准确的反演结果。但由于样地设置不多以及大兴安岭地区的特殊性,估测模型还需要进一步的修正,局部反演精度还需提高。得到的乔灌层、地表可燃物估测模型分别为：Y=4.542-3.226×B2+2.126×B3-1.524B5+0.432×Prin2+0.469×ent2+39.706×sec2+70.628×corr4+0.179×DBH、Y=8.975+0.043×B3-0.008×B4-0.256×homo1+0.144×cont6+0.692×DBH。完成了大兴安岭松岭798高地火灾可燃物消耗的估算,火灾中各类型森林火灾乔灌层损失可燃物载量为2946805.78t,其中落叶松林乔灌层损失1503946.10t,占总损失可燃物载量的51.04%左右；针阔混交林乔灌层损失872662.73t,占总损失可燃物载量的29.73%左右；阔叶混交林中乔灌层损失566631.95t,占总损失可燃物载量的19.23%左右。结果显示不同类型的森林火灾造成的可燃物载量损失的差别比较显著的,如重度森林火灾造成的可燃物载量损失为667515.03t,占总损失可燃物载量的22.65%,而中度火烧和轻度火烧造成的可燃物载量损失分别为1962732.38t、316557.36t,分别占总损失可燃物载量的66.61%、10.74%。完成了大兴安岭松岭798高地森林火灾地表可燃物载量消耗量的估测,火灾消耗地表可燃物载量2083668.65t,其中重度火烧、中度火烧、轻度火烧消耗地表可燃物载量分别为226286.75t、1427932.29t、429450.04t。本文采用排放因子法估算了森林火灾气体释放总量,得到较好的效果。经过计算大兴安岭松岭798高地森林火灾中森林可燃物释放共释放CO2气体量约为149187.66t,其中乔灌层可燃物释放CO2气体量约为88110.43t,地表可燃物释放CO2气体量约为61077.23t；共释放CO气体量约为21187.70t,其中乔灌层可燃物释放CO气体量约为12010.07t,地表可燃物释放CO气体量约为9177.63t；共释放CH气体量约为1925.41t,其中乔灌层可燃物释放CH-气体量约为1512.15t,地表可燃物释放CH气体量约为413.26t；共释放NO气体量约为470.76t,其中乔灌层可燃物释放NO气体量约为272.19t,地表可燃物释放NO气体量约为198.57t；共释放SO2气体量约为658.77t,其中乔灌层可燃物释放SO2气体量约为301.07t,地表可燃物释放SO2气体量约为357.70t。更多还原

【Abstract】 Forest fire, as a main factor in forest ecosystem, played an important role in atmospheric chemical cycles and the carbon cycle. The high frequency of forest fire will not only detory the ecosystem but also result in massive release of carbon greenhouse gases. The forest fire estmissions is mainly in the form of CO2and CO, and others being emitted into atmosphere as CH4, multi-carbon hydrocarbons and volatile organic oxides. This will be breaking the balance of carbon in atmosphere, and then making a critical effect on the global climate environment. The research on the relationship between forest fire and carbon emission will contribute to understanding of how the impact that fire on global carbon cycle is and improvement of strategies of fire management. The estimation of composition and release of fire gas is a basic work for the research of atmosphere environment.There are four main research aspects (RA) in this study based on MODIS, TM remote sensing data, historical fire and forest survey data RA (1) estimating the consumption of forest fuel, and emission of polluting gas emission; RA (2) establishing the estimation and consumption model of forest fuel load; RA (3) developing the model of emission of forest fire; RA (4) evaluating the loss of forest fire and estimating the emission of polluting gases. The detailed work as follows:Preprocessed the remote sensing data,Which includes geometic and radiometric correction. Smooth processed the multi-temporal remote sensing data and completed the basic work of getting remote sensing information.Completed forestfuel classification based on MODIS and proved that appropriate classi-ficationmethodsand the participation of other necessary data can obtain more satisfact-ory resolution of classification. The results showed that the resolution based ondecision tree methodis83.2%, higher than supervised classification in terms of TM data.This resolution can meet the production needs. The keys ofdecision tree method are getting the threshold of surface features and developing a suitable rule of classification.However, accessibility and quantity of TM and MODIS images are able to affect the classification accuracyofthe com-bustibles.In this study, we estimated the burned area using mixed-pixel decomposition techni-que based on MODIS data. After pixel decomposition, the resolution of pixel can reach about92%, and the total precision can get90%, basically putting an end to the problem that extracting pixel accuracy of non-vegetated areas and the fire edge is low. Compared with conventionalaircraft, this method is more convenient in operation on large spatial scales and more time-sensitive also. Furthermore, it has high level of timeliness and can meet the production needs better than high-resolution data.Estimated the burned area of forest fire in798highland of Songling, which is242,030hectares totally, including severe fire (23,781ha), moderate fire (145,408ha) and low-intensity fire (72,964ha) and the percentage is9.82%.60.05%、30.13%correspondingly.Established a TM image-based estimation model of forest fuel load, then deducted the fuel load of entire Daxing’an mountains based on the model. We obtained a resasonable resolution; however, a further improvement is still needed. The Partial Least Squares and Second-order Least Squares models have been used to deduct the forest fuel load in this study.We made a comparison between these two models in terms of different statistics and the results indicated that Second-order Least Squares model is superior over Partial Least Squares and the function of estimation models of thearbor, shrub layer, surface fuel are as follows:Y=4.542-3.226×B2+2.126×B3-1.524×B5+0.432×Prin2+0.469×ent2+39.706×sec2+70.628×corr4+0.179×DBH、Y=8.975+0.043×B3-0.008×54-0.256×homo1+0.144×cont6+0.692×DBH.Completed the estimation the consumption of forest fuel in798high-land of Songling. The total loss of fuel load of canopy shrub is2946805.78t in different fire severity, in which Larch canopy shrub is1503946.10t accounting for51.04%;Mixed coniferous canopy shrub is872662.73t (29.73%);Broad-leaved mixed canopy shrub is566631.95t (19.23%).Besides, there is a significant distinction between different severity of fires, such as sever fire caused a667515.03t fuel load loss occupied22.65%; meanwhile,moderatefire and low intensity fire lead to1962732.38t and316557.36t respectively fuel load loss and accouting to66.61%.10.74%of total loss of fuel load. The consumption of surface fuel load in798high-land of Songling is2083668.65t, in which the number of burned fuel load by high intensity fire、 moderate fire、low intensity fire is226286.75t、1427932.29t、429450.04t respectively.This paper, the emission factor wasused to estimate the emission of gases from forest fires. The results revealed that there were149187.66tCO2emitted from forest fire, in which88110.43t come from canopy shrub fuel,61077.23tfrom surface fuel. The total amount of CO was12010.07t and canopy shrub fuel and surface fuel12010.07t and occupied9177.63trespectively.The total amount of CH was1925.41t, canopy shrub fuel1512.15t and surface fuel413.26t. The total amount of NO was470.76t, canopy shrub fuel272.19t and surface fuel198.57t. The total amount of SO2was658.77t, canopy shrub fuel301.07t and surface fuel357.70.更多还原