本文選題：植被冠層反射率模型　切入點(diǎn)：模型弱敏感參數(shù)　出處：《電子科技大學(xué)》2017年博士論文　論文類型：學(xué)位論文

【摘要】：20世紀(jì)70年代以來,植被冠層反射率模型研究得以蓬勃發(fā)展,涌現(xiàn)出SAIL、GEOSAIL等眾多經(jīng)典的冠層反射率模型。基于這些經(jīng)典模型的植被關(guān)鍵參數(shù)反演研究也在近20年中如火如荼地開展著,并廣泛應(yīng)用于農(nóng)作物估產(chǎn)、生態(tài)環(huán)境監(jiān)測(cè)與保護(hù)、自然災(zāi)害(干旱、火災(zāi)等)評(píng)估及預(yù)警、水文水資源管理等關(guān)乎國(guó)家安全、全球氣候變化的重大需求中。但是,近年來植被冠層反射率建模研究進(jìn)展緩慢;植被關(guān)鍵參數(shù)反演方面也一直主要集中在幾個(gè)模型敏感參數(shù)(如葉面積指數(shù)、冠層含水量等)的反演,而重要但模型弱敏感的參數(shù)研究卻鮮有踏足,導(dǎo)致理論研究與實(shí)際應(yīng)用之間的落差越來越大。鑒于此,本論文站在前人對(duì)模型敏感參數(shù)成功反演的基礎(chǔ)上,對(duì)模型弱敏感參數(shù)進(jìn)行遙感定量反演研究,重點(diǎn)研究了模型弱敏感參數(shù)干物質(zhì)重量(DMC)及其衍生的草地地上生物量(AGB)和冠層可燃物含水率(FMC)的反演理論與方法。AGB是區(qū)域碳循環(huán)研究中的重要一環(huán),同時(shí)也是農(nóng)作物估產(chǎn)的研究重點(diǎn);FMC是描述植被點(diǎn)燃概率及火災(zāi)蔓延速率的關(guān)鍵指標(biāo)因子,因此是眾多火災(zāi)模型的關(guān)鍵輸入?yún)?shù)。論文通過對(duì)這二者參數(shù)反演方法及應(yīng)用研究,旨在構(gòu)建一套適用于模型弱敏感參數(shù)反演的方法體系,同時(shí)拓展植被冠層反射率模型的應(yīng)用范圍,服務(wù)于農(nóng)作物估產(chǎn)、區(qū)域生態(tài)安全、野火風(fēng)險(xiǎn)評(píng)估及預(yù)警、全球氣候變化等領(lǐng)域。論文主要研究工作和成果如下:(1)分析植被冠層反射率模型及參數(shù)敏感性的基礎(chǔ)上,提出了模型弱敏感參數(shù)遙感反演策略。論文指出模型弱敏感參數(shù)成功反演的關(guān)鍵在于增強(qiáng)弱敏感參數(shù)的敏感性,并提出五個(gè)弱敏感參數(shù)反演策略:基于高光譜數(shù)據(jù)的模型弱敏感參數(shù)反演;參數(shù)化模型敏感參數(shù),提高模型弱敏感參數(shù)的敏感性;基于面向?qū)ο蟮哪Ｐ腿趺舾袇?shù)反演;基于多時(shí)相遙感數(shù)據(jù)的模型弱敏感參數(shù)反演;基于真實(shí)實(shí)測(cè)數(shù)據(jù)的模型弱敏感參數(shù)反演。(2)研究了緩解植被冠層反射率模型參數(shù)反演的病態(tài)問題,提出了基于貝葉斯網(wǎng)絡(luò)算法的反演方法。病態(tài)反演問題會(huì)極大地降低模型敏感參數(shù)的反演精度,對(duì)模型弱敏感參數(shù)正確反演的影響更為強(qiáng)烈。為緩解這一問題,本論文以貝葉斯網(wǎng)絡(luò)算法為基礎(chǔ),通過考慮模型參數(shù)之間固有的關(guān)聯(lián)特性,構(gòu)建了更符合自然界實(shí)際情況的自由參數(shù)先驗(yàn)聯(lián)合概率分布,降低了錯(cuò)誤自由參數(shù)組合出現(xiàn)的概率,緩解了病態(tài)反演問題,提高了葉面積指數(shù)(LAI)及植被冠層含水量的反演精度,為后續(xù)基于植被冠層反射率模型的草地AGB及植被冠層FMC反演奠定基礎(chǔ)。(3)提出了一套新的反演草地AGB的方法,突破了傳統(tǒng)基于實(shí)測(cè)統(tǒng)計(jì)擬合估算草地AGB的思路。該方法假設(shè)草地AGB可通過LAI與葉片干物質(zhì)重量(DMC)的乘積近似表示。首先,基于PROSAIL模型對(duì)實(shí)驗(yàn)區(qū)草地LAI及DMC進(jìn)行反演;同時(shí)通過考慮模型參數(shù)間的相關(guān)特性及融入MODIS LAI產(chǎn)品等方法提高草地AGB的敏感性,反演獲得草地AGB。實(shí)驗(yàn)中以我國(guó)青海湖流域草地AGB反演為例,論證該方法的有效性;同時(shí),實(shí)驗(yàn)中使用了植被指數(shù)法、偏最小二乘法(PLSR)、人工神經(jīng)網(wǎng)絡(luò)法(ANN)等三種傳統(tǒng)經(jīng)驗(yàn)估算草地AGB的方法進(jìn)行對(duì)比分析。實(shí)驗(yàn)結(jié)果顯示,使用本論文方法反演的草地AGB精度高于植被指數(shù)法及ANN,與PLSR精度不相上下,但本論文方法不依賴地面實(shí)測(cè)數(shù)據(jù),因此比PLSR更具普適性,是具有前景的草地AGB反演方法。(4)構(gòu)建了植被冠層FMC反演方法體系,包括基于區(qū)間估計(jì)LAI的草地冠層FMC遙感反演方法和基于耦合輻射傳輸模型的森林冠層FMC遙感反演方法。FMC是一個(gè)與LAI無關(guān)的量,但在基于物理模型的FMC反演中,LAI的不確定性對(duì)FMC的正確反演具有較強(qiáng)的干擾作用。為緩解這一問題,提高FMC的敏感性,實(shí)驗(yàn)結(jié)合MODIS LAI產(chǎn)品,利用降尺度及區(qū)間估計(jì)的LAI參數(shù)化模型LAI輸入,以此降低LAI的不確定性,從而提高草地冠層FMC的反演精度。在森林冠層FMC的反演中,針對(duì)本研究區(qū)森林植被具上層喬木下層草本植被的特點(diǎn),即雙層冠層結(jié)構(gòu),實(shí)驗(yàn)通過耦合描述草地冠層反射率特征的SAIL模型及描述森林冠層反射率特征的GEOSAIL模型,近似模擬該雙層冠層結(jié)構(gòu)的森林植被反射率特征,降低模型系統(tǒng)誤差對(duì)弱敏感參數(shù)FMC的影響,從而正確反演了森林冠層FMC。(5)構(gòu)建了基于遙感技術(shù)的區(qū)域大尺度野火風(fēng)險(xiǎn)評(píng)估及預(yù)警雛形。論文以澳大利亞和四川西昌瀘山為例,首先應(yīng)用上述植被冠層FMC反演方法,進(jìn)行植被冠層FMC定量反演,生產(chǎn)了 2001-2015年的澳大利亞15年的植被FMC產(chǎn)品。其次,基于該套產(chǎn)品,結(jié)合MODIS歷史火災(zāi)產(chǎn)品(MCD64),應(yīng)用Logistic模型計(jì)算得出澳大利亞2001-2015年燃燒指數(shù)(FI)產(chǎn)品,以此對(duì)野火風(fēng)險(xiǎn)進(jìn)行評(píng)估及預(yù)警。通過澳大利亞歷史上三次重大火災(zāi)(2003年堪培拉火災(zāi)、2009年維多利亞州火災(zāi)及2013年新南威爾士州火災(zāi))爆發(fā)前植被冠層FMC及FI演變情況顯示,該兩套產(chǎn)品對(duì)野火風(fēng)險(xiǎn)具有較好指示作用,該工作對(duì)于今后全球植被冠層FMC產(chǎn)品化及野火風(fēng)險(xiǎn)評(píng)估應(yīng)用推廣具有重要的示范意義。
[Abstract]:Since 1970s, the research of canopy reflectance model of vegetation flourish, the emergence of SAIL, GEOSAIL and many other classical models of canopy reflectance. Study on the key parameters of the classical model based on inversion of vegetation also in the past 20 years to carry out like a raging fire, and widely used in crop yield estimation, monitoring and protection of the ecological environment, natural disasters (drought. Fire) assessment and early warning, hydrology and water resources management related to national security, great demand of global climate change. However, recent progress in modeling the vegetation canopy reflectance is slow; the key parameters inversion of vegetation has been mainly focused on the sensitive parameters of several models (such as leaf area index and canopy water content inversion, etc.) the important parameters of the model but rarely set foot weakly sensitive, leading to more and the gap between the theoretical research and practical application of the larger. In view of this, the The station based on the predecessor model sensitive parameters inversion of success, the quantitative remote sensing model of weak inversion of sensitive parameters, focusing on the model of weak sensitive parameters of dry weight (DMC) and its derivative on the grassland biomass (AGB) and canopy fuel moisture content (FMC).AGB inversion theory and method the study area is an important part of the carbon cycle at the same time, and also a research focus in crop yield estimation; FMC is the key factor of vegetation index and the rate of fire spread fire probability description, so it is a key input parameter of many fire model. Based on the parameter inversion method and application of the two, aims to establish a suitable model weak sensitive parameter inversion method system, and expand the scope of application of the vegetation canopy reflectance model, in crop yield estimation, regional ecological security, wildfire risk assessment and early warning, global climate Changes in other fields. The main research work and results are as follows: (1) analyzing the sensitivity of the model and the parameters of canopy reflectance, proposes a model of weak sensitive parameters of remote sensing inversion strategy. The paper pointed out that the key to success of the model of weak sensitive parameters inversion is to enhance the sensitivity of weak sensitive parameters, and puts forward five strategies of weak inversion sensitive parameters hyperspectral data model: weak sensitive parameter inversion based on parametric model; sensitive parameters, improve the sensitivity of model weak sensitive parameters; object oriented model of weak sensitive parameter inversion based on multitemporal remote sensing data model; weak sensitive parameter inversion based on real measured data model; weak inversion based on sensitive parameters (2) were studied. To alleviate the ill posed problem of vegetation canopy reflectance model inversion, the inversion algorithm based on Bayesian networks. The ill posed inversion problem will be greatly reduced The precision of inversion parameters of the low sensitive model, weak influence on the model of sensitive parameters inversion is more intense. In order to alleviate this problem, this thesis is based on the Bayesian network algorithm, by considering the inherent characteristics of correlation between model parameters, build more in line with the actual situation of the free parameters a priori nature of joint probability distribution, reducing the probability of error free parameter combinations, alleviate the ill posed inversion problem, improve the leaf area index (LAI) and vegetation canopy water content inversion precision, for subsequent AGB and grassland vegetation canopy FMC inversion of vegetation canopy reflectance model based on the foundation. (3) proposed a new method of the inversion of the grassland AGB, breakthrough the traditional statistical fitting estimation of grassland AGB based on the idea. The method assumes that the grassland AGB by LAI and leaf dry weight (DMC) of the product approximation. First, based on the PROSAIL Inversion of experimentation area grassland LAI and DMC model; at the same time by considering the correlation between model parameters and integration of MODIS LAI products and other methods to improve the sensitivity of grassland AGB, derived from the AGB. experiment in Qinghai Lake Valley meadow grassland in China AGB inversion as an example to demonstrate the effectiveness of the method; at the same time, the vegetation index method using experiment in the partial least squares (PLSR), artificial neural network (ANN) analyzed the grassland AGB estimation method of three kinds of traditional experience. The experimental results show that using the method of inversion accuracy in AGB grassland vegetation index method and the accuracy of PLSR and ANN, be roughly the same, but the method does not rely on the ground the data, which is more universal than PLSR, is the grassland AGB inversion method with great prospect. (4) constructed the system of vegetation canopy FMC inversion method, including interval estimation of LAI grass canopy FMC based on Remote Sensing Modeling method and.FMC forest canopy FMC remote sensing inversion method based on coupled radiative transfer model is an independent of LAI, but in FMC based on physical model inversion, inversion LAI uncertainty of FMC has the strong jamming. To alleviate this problem, improve the sensitivity of FMC, combined with MODIS LAI products, with model LAI input and the interval estimation of scale parameter of LAI, in order to reduce the uncertainty of LAI, so as to improve the inversion precision of grassland canopy FMC. In the inversion of the forest canopy FMC, according to the characteristics of forest vegetation in the study area with lower tree layers of herbaceous vegetation, namely double canopy structure, GEOSAIL model by describing the coupled reflectance characteristics of SAIL model and describe the forest canopy characteristics of canopy reflectance of grassland experiment, approximate simulation of forest vegetation reflectance characteristics of the double canopy structure, reduce the system model The error of weak sensitive parameters affecting FMC, and thus the correct inversion of the forest canopy FMC. (5) constructed the remote sensing technology of large scale regional wildfire risk assessment and early warning based on prototype. In order to Australia and Sichuan Xichang Lushan as an example, the first application of the FMC inversion of vegetation canopy, vegetation canopy FMC quantitative inversion, production 2001-2015 years of Australia 15 years of vegetation FMC. Secondly, the set of products based on the combination of MODIS (MCD64), history of fire products using Logistic model to calculate the Australian 2001-2015 year combustion index (FI) products, in order to carry out assessment and early warning of wildfire risk. Through the history of Australia three major fire (in Canberra in 2003 fire, fire in Vitoria in 2009 and 2013 the state of New South Wales) before the outbreak of the vegetation canopy fire display FMC and FI evolution, the two sets of products with wildfire risk This work has an important demonstration significance for the application and popularization of FMC production and wildfire risk assessment in the world.

【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP79

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 李宗南;陳仲新;王利民;姜志偉;;基于ACRM模型不同時(shí)期冬小麥LAI和葉綠素反演研究[J];中國(guó)農(nóng)業(yè)資源與區(qū)劃;2012年01期

2 姜志偉;陳仲新;任建強(qiáng);;基于ACRM輻射傳輸模型的植被葉面積指數(shù)遙感反演[J];中國(guó)農(nóng)業(yè)資源與區(qū)劃;2011年01期

3 楊飛;孫九林;張柏;姚作芳;王宗明;王卷樂;樂夏芳;;基于PROSAIL模型及TM與實(shí)測(cè)數(shù)據(jù)的MODISLAI精度評(píng)價(jià)[J];農(nóng)業(yè)工程學(xué)報(bào);2010年04期

4 孫焱鑫;王紀(jì)華;李保國(guó);劉良云;黃文江;趙春江;;基于BP和GRNN神經(jīng)網(wǎng)絡(luò)的冬小麥冠層葉綠素高光譜反演建模研究[J];遙感技術(shù)與應(yīng)用;2007年04期

5 蔡博峰;紹霞;;基于PROSPECT+SAIL模型的遙感葉面積指數(shù)反演[J];國(guó)土資源遙感;2007年02期

6 黃健熙;吳炳方;曾源;田亦陳;;基于蒙特卡羅方法的森林冠層BRDF模擬[J];系統(tǒng)仿真學(xué)報(bào);2006年06期

7 李小文;;定量遙感的發(fā)展與創(chuàng)新[J];河南大學(xué)學(xué)報(bào)(自然科學(xué)版);2005年04期

8 陳良富,徐希孺;植被遙感的Monte Carlo模擬研究[J];地理科學(xué)進(jìn)展;2000年01期

9 李小文,王錦地,胡寶新,A.H.Strahler;先驗(yàn)知識(shí)在遙感反演中的作用[J];中國(guó)科學(xué)(D輯:地球科學(xué));1998年01期

10 李小文,A.Strahler,朱啟疆,劉毅,劉廣成,張仁華,虞獻(xiàn)平,李繼泉;地物二向性反射幾何光學(xué)模型和觀測(cè)的進(jìn)展[J];國(guó)土資源遙感;1991年01期

相關(guān)博士學(xué)位論文 前2條

1 肖艷芳;植被理化參數(shù)反演的尺度效應(yīng)與敏感性分析[D];首都師范大學(xué);2013年

2 王東偉;遙感數(shù)據(jù)與作物生長(zhǎng)模型同化方法及其應(yīng)用研究[D];北京師范大學(xué);2008年

相關(guān)碩士學(xué)位論文 前1條

1 全興文;高原濕地植被參數(shù)遙感定量反演及同化技術(shù)研究[D];電子科技大學(xué);2013年

，

本文編號(hào)：1557395