植被叶面积指数遥感监测模型

叶面积指数是植被定量遥感的重要参数,区域的时序列叶面积指数揭示了区域生态的演化过程,反演方法上主要是通过植被指数建立相关模型实现的,对于不同地区或不同气候带而言,模型的通用性以及各种植被指数在模型中的灵敏度都需做进一步的探讨。以江苏省宜兴市作为研究区,采用2002年8月22日获得的Landsat-5TM图像数据和2003年8月23～26日采用LAI-2000进行的野外实测植被叶面积指数(LAI)数据,分别探讨了植被指数(VI)与LAI的一元、多元线性回归模型和非线性回归模型,其中的非线性回归模型包括对数、指数、乘幂和多项式回归模型。结果表明,VI与LAI之间的最佳回归模型为多元线性回归模型,R2达0.864;采用逐步选择剔除法,遴选出了用于回归模型的植被指数为RVI、PVI、SAVIL=0.35、MSAVI、ARVIγ=1、ARVIγ=0.5和SARVI。经模型LAI=-ln((VI-VI∞)/(VIg-VI∞))/KVI检验,预测值(y)与实测值(x)的拟合度较好y=0.5345x 1.3304,R2为0.7379。RVI与LAI的三次多项式回归模型也较好,R2为0.7806。再次为RVI与LAI的一元线性回归模型,R2为0.7726,比值植被指数RVI在反演叶面积指数模型中具有较高的灵敏度。     

华南地区亚热带树木叶面积指数的高光谱反演研究

为构建树种叶面积指数的估算模型,以NDVI、RVI、FREP、CIGreen、CIRed-edge、MSAVI2为高光谱特征变量,通过统计分析,确定反演树种叶面积指数的最佳光谱特征变量,构建华南农业大学校园内50种亚热带树木的叶片反射率和叶面积指数(LAI)模型。结果表明,6种高光谱特征变量与树种叶面积指数间都具有极显著相关性,其中红边位置反射率(FREP)和比值植被指数(RVI)与LAI的拟合方程的R2都大于0.8,决定系数分别为0.820和0.811。经过精度验证,FREP估算的均方根误差(RMSE)只有0.13,该回归模型为估测亚热带典型树种的叶片LAI最佳模型。从高光谱遥感的角度结合亚热带植被的群落结构特点来看,建立的红边位置光谱反射率与叶面积指数的回归模型普遍具有较高的拟合度,所以利用高光谱特征变量反演亚热带树木叶片的叶面积指数等植被参数的应用前景较好。     

基于小波分析的大豆叶面积高光谱反演

实测了不同水肥耦合、经营制度及有效营养面积条件下的大豆(Glycinemax)冠层高光谱反射率与叶面积指数(LAI),并对光谱反射率、微分光谱与LAI的关系进行了分析;采用比值植被指数(RVI)与归一化植被指数(NDVI)建立了大豆LAI反演模型;采用小波分析对采集的光谱反射率数据进行了能量系数提取,并以小波能量系数作为自变量进行了单变量与多变量回归分析,对大豆LAI进行估算。结果表明:大豆LAI与光谱反射率在可见光波段呈负相关;在近红外波段呈正相关;微分光谱在红边处与大豆LAI密切相关(R2=0.92);RVI与NDVI可以提高大豆LAI的估算精度(R2分别达0.79、0.84);各植被指数各有优缺点,应根据需要进行选择;小波能量系数回归模型可以进一步提高大豆叶面积的估算水平,以一个特定小波能量系数作为自变量的回归模型,大豆LAI回归确定系数R2高达0.884;以4个和6个小波能量系数建立LAI回归分析模型(R2分别达0.92、0.93),2个模型LAI预测值与大豆LAI实测值线性回归确定性系数R2分别为0.90、0.92。比较可知,小波分析可以对高光谱进行特征变量提取,进而反演大豆生理参数,并且反演的LAI精度较光谱反射率、微分光谱及植被指数都有明显提高,小波分析在植被生理参数的高光谱提取方面有着广阔的应用前景。     

基于多角度植被指数的马尾松林LAI反演方法

CHRIS/PROBA是目前具有最高空间分辨率（17 m×17 m）的星载多角度高光谱数据,该款数据在反演植被垂直结构参数,如树高、叶面积指数（leaf area index,LAI）等方面具有重要的应用前景。基于四尺度几何光学模型得到马尾松（Pinus massoniana Lamb.）冠层的归一化差分植被指数（normalized difference vegetation index,NDVI）各向异性分布规律,利用CHRIS红光特征波段和近红外特征波段构建一种新型多角度植被指数（normalized hotspot-dark-spot difference vegetation index,NHDVI）,并将其应用于CHRIS数据对马尾松林的LAI遥感估算上。结果显示：（1）相比归一化差分植被指数（NDVI）与土壤调节植被指数（soil adjusted vegetation index,SAVI）而言,NHDVI能很好地融合光谱信息与角度信息,与地面实测LAI的决定系数达到0.7278;（2）利用NHDVI-LAI统计回归模型方法来反演LAI值,将得到的LAI值与地面实测值进行相关性分析,结果拟合优度达到0.8272,均方根误差RMSE为0.1232。与传统植被指数相比,包含角度信息的多角度植被指数对LAI的反演在精度上有较大提升,同时比基于辐射传输模型的反演方法更简易、实用。     

高光谱与拟合多光谱植被指数反演武夷山亚高山草甸LAI的对比研究

植被叶面积指数(Leaf Area Index, LAI)是重要的生态学参数, 被广泛用于指示植被密度、生物量、碳、氮物质循环以及气候变化对生态系统的影响, 也作为生态过程模型的重要输入参数。地面实测高光谱遥感数据能以更高的空间分辨率及更高的光谱分辨率监测植物的光谱特征, 为精准反演LAI提供了基础。本项研究以武夷山国家公园黄岗山顶的亚高山草甸为研究对象, 通过建立多种高光谱植被指数和拟合多光谱植被指数反演叶面积指数的统计模型, 并比较高光谱与多光谱对叶面积指数反演的效果, 阐明用于反演高覆盖率亚高山草甸的最适高光谱和拟合多光谱植被指数。结果表明: 高光谱新植被指数(NVI)对于反演LAI有最好的效果, R2 = 0.85, P < 0.01; 依据高光谱NVI拟合而成的多光谱NVI反演结果次之, R2 = 0.82, P < 0.01。几种常用比值植被指数NDVI、MSR、RVI和GNDVI在高光谱和拟合多光谱反演结果中相差不大, 表现较好, R2都在0.65以上。通过对比高光谱和拟合Sentinel-2A和Landsat-8两种多光谱卫星波段的反演结果发现, 光谱响应函数中具有更窄波段范围的近红外、红、绿、蓝波段构成的植被指数可以得到更好的反演结果, 而固定波段的高光谱植被指数未必在每种植被指数中都具有最好的反演效果。同时, 发现当某种植被指数反演LAI的线性回归方程的斜率越大, 说明这种植被指数越有可能随LAI的增大而出现饱和现象, 相反的, 斜率越小则说明该种植被指数没有出现饱和现象。此外, 在研究区内使用高光谱和拟合多光谱波段植被指数法反演LAI, NDVI都获得了较好的效果, 存在很好的线性关系, 之前的很多研究和判断都认为NDVI不适用于反演高覆盖植被的LAI, 这个发现是具有意义的, 表明高覆盖植被的叶面积指数在一定范围内是能够被NDVI(应用最广泛的植被指数)较好的反演, 进一步扩展了NDVI反演LAI的适用性和可能性。     

植冠下土壤类型差异对遥感估算冬小麦叶面积指数的影响

遥感是从田块到区域乃至全球范围无损探测叶面积指数(LAI)的有效方法。土壤背景是LAI遥感研究的重要制约因素之一,而土壤类型是组成土壤背景的主要部分,对植被冠层-土壤的光学性质有重要影响,但目前植冠下土壤类型背景对遥感LAI估算的影响尚不明确。该文通过分析归一化差异植被指数、修正型土壤调节植被指数、修正的叶绿素吸收比率指数、红边拐点、红边振幅、红边面积、红边对数指数和归一化差异光谱指数在不同土壤类型下对LAI的敏感性,挖掘最不敏感的光谱参数;通过比较两种回归模型(偏最小二乘回归和随机森林回归)在单一土壤类型和多种土壤类型区对LAI的预测精度,探究将单一土壤类型下发展的LAI估算模型应用到复杂土壤类型地区时可能出现的问题。结果表明:(1)虽然8种光谱指数对LAI的敏感性因土壤类型不同而差异明显,但红边拐点受植冠下土壤类型影响最小;"lambda-by-lambda"波段优选算法不仅可以提供对LAI最敏感的光谱区间,而且可在一定程度上为抵抗植冠下土壤类型差异影响的光谱指数构建提供可行思路;(2)回归模型的LAI预测精度因是否考虑土壤类型而不同,但在小区域尤其是田块尺度研究时,对变量的解释能力是选择模型的第一考虑,而偏最小二乘回归在此方面优于随机森林回归;在未知地表先验知识的前提下,随机森林回归对大区域LAI估算比偏最小二乘回归适合,但地表先验知识的获取对LAI遥感估算仍然十分必要。     

三江平原湿地植被叶面积指数遥感估算模型

利用中巴资源卫星CBERS-02影像提取的归一化植被指数(NDVI)和同期野外实测的叶面积指数(LAI)数据,分析了三江平原洪河自然保护区草甸、沼泽植被、灌丛和岛状林4种湿地植被及样本总体的NDVI与LAI之间的相关关系,建立了NDVI与不同湿地植被类型叶面积指数间的线性和非线性回归模型,并制作完成洪河自然保护区LAI空间分布图.结果表明,整个研究区样本总体的LAI估算效果不太理想,其NDVI与LAI的相关性仅为0.523;将研究区分为草甸、沼泽、灌丛和岛状林4种湿地植被类型,NDVI与各植被型LAI的相关性和估算效果均有很大程度的提高,所建立的LAI遥感反演模型以三次曲线回归方程拟合精度最高,R2分别达到0.723、0.588、0.837、0.720.以上结果表明,结合地面实测数据并基于遥感植被分类的基础上,CBERS-02遥感影像可用于较大区域内湿地植被生理参数的反演研究.     

光谱植被指数与水稻叶面积指数相关性的研究

 综合分析比较了几种常见光谱植被指数与水稻（Oryza sativa）叶面积指数的相关性及其预测力。结果表明,植被指数的预测力在水稻营养生长旺盛期间最好。植被指数的预测力主要依赖于叶面积指数(LAI)的整体变化范围。因此,综合不同生育时期和氮肥处理的试验资料,光谱植被指数能准确地预测LAI的变化。LAI与各植被指数均呈曲线相关,与比值植被指数(RVI)、再归一化植被指数(RDVI)和R810/R560显著幂相关,与归一化植被指数(NDVI)、垂直植被指数(PVI)、差值植被指数(DVI)、土壤调整植被指数(SAVI)和转换型土壤调整指数(TSAVI)显著指数相关。其中,近红外与绿光波段的比值R810/R560的预测力最佳。用不同移栽秧龄、不同密度、不同水分和氮肥处理的数据对R810/R560的表现进行了检验,结果表明估算精度平均为91.22%,估计的均方差根(RMSE)平均为0.480　5,平均相对误差为-0.013。表明宽波段光谱植被指数可以准确地用来监测水稻叶面积指数。     

基于多源遥感数据的大豆叶面积指数估测精度对比

近年来遥感技术的革新促使遥感源越来越丰富.为分析多源遥感数据的叶面积指数(LAI)估测精度,本文以大豆为研究对象,利用比值植被指数(RVI)、归一化植被指数(NDVI)、土壤调整植被指数(SAVI)、差值植被指数(DVI)、三角植被指数(TVI)5种植被指数,结合地面实测LAI构建经验回归模型,比较3类遥感数据(地面高光谱数据、无人机多光谱影像以及高分一号WFV影像)对大豆LAI的估测能力,并从传感器几何位置和光谱响应特性以及像元空间分辨率三方面分析讨论了3类遥感数据的LAI反演差异.结果表明： 地面高光谱数据模型和无人机多光谱数据模型都可以准确预测大豆LAI(在α=0.01显著水平下,R²均>0.69,RMSE均<0.40);地面高光谱RVI对数模型的LAI预测能力优于无人机多光谱NDVI线性模型,但两者差异不大(E_A相差0.3%,R²相差0.04,RMSE相差0.006);高分一号WFV数据模型对研究区内大豆LAI的预测效果不理想(R²<0.30,RMSE>0.70).针对星、机、地三类遥感信息源,地面高光谱数据在反演LAI方面较传统多光谱数据有优势但不突出;16 m空间分辨率的高分一号WFV影像无法满足田块尺度作物长势监测的需求;在保证获得高精度大豆LAI预测值和高工作效率的前提条件下,基于无人机遥感的农情信息获取技术不失为一种最佳试验方案.在当今可用遥感信息源越来越多的情况下,农业无人机遥感信息可成为指导田块精细尺度作物管理的重要依据,为精准农业研究提供更科学准确的信息.     

森林冠层叶面积指数遥感反演——以小兴安岭五营林区为例

以中国东北小兴安岭五营林区为研究区,基于MODIS BRDF遥感模型参数产品数据,首先利用4-Scale模型建立查找表计算像元尺度上各组分比例,估算研究区森林乔木冠层反射率,然后利用冠层反射率数据,获取研究区3种常用森林冠层植被指数,最后基于植被指数与实测叶面积指数构建研究区冠层叶面积指数反演模型,并选取最优模型实现研究区森林冠层叶面积指数反演。结果表明:研究区冠层LAI遥感反演模型中,基于比值植被指数SR(simple ratio,SR)构建的二次多项式反演模型精度最高,且反演精度比未考虑背景反射影响的SR反演模型精度有较大幅度提高,模型决定系数由0.38提高至0.54;反演获取的研究区冠层LAI在2.38~12.67,平均值6.52,LAI值在阔叶林区域相对较高。     

Estimation of leaf area index and canopy openness in broad-leaved forest using an airborne laser scanner in comparison with high-resolution near-infrared digital photography

We estimated leaf area index (LAI) and canopy openness of broad-leaved forest using discrete return and small-footprint airborne laser scanner (ALS) data. We tested four ALS variables, including two newly proposed ones, using three echo types (first, last, and only) and three classes (ground, vegetation, and upper vegetation), and compared the accuracy by means of correlation and regression analysis with seven conventional vegetation indices derived from simultaneously acquired high-resolution near-infrared digital photographs. Among the ALS variables, the ratio of the “only-and-ground” pulse to “only” pulse (OGF) was the best estimator of both LAI (adjusted R ² = 0.797) and canopy openness (adjusted R ² = 0.832), followed by the ratio of the pulses that reached the ground to projected lasers (GF). Among the vegetation indices, the normalized differential vegetation index (NDVI) was the best estimator of both LAI (adjusted R ² = 0.791) and canopy openness (adjusted R ² = 0.764). Resampling analysis on ALS data to examine whether the estimation of LAI and canopy openness was possible with lower point densities revealed that GF maintained a high adjusted R ² until a fairly low density of about 0.226 points/m², while OGF performed marginally when the point density was reduced to about 1 point/m², the standard density of high-density products on the market as of February 2008. Consequently, the ALS variables proposed in the present study, GF and OGF, seemed to have great potential to estimate LAI and canopy openness of broad-leaved forest, with accuracy comparable to NDVI, from high-resolution near-infrared imagery.     

Estimation of wetland vegetation height and leaf area index using airborne laser scanning data

Wetland vegetation is a core component of wetland ecosystems. Wetland vegetation structural parameters, such as height and leaf area index (LAI) are important variables required by earth-system and ecosystem models. Therefore, rapid, accurate, objective and quantitative estimations of wetland vegetation structural parameters are essential. The airborne laser scanning (also called LiDAR) is an active remote sensing technology and can provide accurate vertical vegetation structural parameters, but its accuracy is limited by short, dense vegetation canopies that are typical of wetland environments. The objective of this research is to explore the potential of estimating height and LAI for short wetland vegetation using airborne discrete-return LiDAR data.The accuracies of raw laser points and LiDAR-derived digital elevation models (DEM) data were assessed using field GPS measured ground elevations. The results demonstrated very high accuracy of 0.09 m in raw laser points and the root mean squared error (RMSE) of the LiDAR-derived DEM was 0.15 m.Vegetation canopy height was estimated from LiDAR data using a canopy height model (CHM) and regression analysis between field-measured vegetation heights and the standard deviation (σ) of detrended LiDAR heights. The results showed that the actual height of short wetland vegetation could not be accurately estimated using the raster CHM vegetation height. However, a strong relationship was observed between the σ and the field-measured height of short wetland vegetation and the highest correlation occurred (R² = 0.85, RMSE = 0.14 m) when sample radius was 1.50 m. The accuracy assessment of the best-constructed vegetation height prediction model was conducted using 25 samples that were not used in the regression analysis and the results indicated that the model was reliable and accurate (R² = 0.84, RMSE = 0.14 m).Wetland vegetation LAI was estimated using laser penetration index (LPI) and LiDAR-predicted vegetation height. The results showed that the vegetation height-based predictive model (R² = 0.79) was more accurate than the LPI-based model (the highest R² was 0.70). Moreover, the LAI predictive model based on vegetation height was validated using the leave-one-out cross-validation method and the results showed that the LAI predictive model had a good generalization capability. Overall, the results from this study indicate that LiDAR has a great potential to estimate plant height and LAI for short wetland vegetation.     

Global remote sensing of water–chlorophyll ratio in terrestrial plant leaves

I evaluated the use of global remote sensing techniques for estimating plant leaf chlorophyll a + b (C_ab; μg cm⁻²) and water (C_w; mg cm⁻²) concentrations as well as the ratio of C_w/C_ab with the PROSAIL model under possible distributions for leaf and soil spectra, leaf area index (LAI), canopy geometric structure, and leaf size. First, I estimated LAI from the normalized difference vegetation index. I found that, at LAI values <2, C_ab, C_w, and C_w/C_ab could not be reliably estimated. At LAI values >2, C_ab and C_w could be estimated for only restricted ranges of the canopy structure; however, the ratio of C_w/C_ab could be reliably estimated for a variety of possible canopy structures with coefficients of determination (R²) ranging from 0.56 to 0.90. The remote estimation of the C_w/C_ab ratio from satellites offers information on plant condition at a global scale.     

Rates of apparent photosynthesis,respiration and dry matter accumulation in maize canopies

The rates of canopy apparent photosynthesis (P_C) and canopy respiration (R_c) were studied during vegetation season in two erectophile and two planophile hybrids of maize (Zea mays L.) grown at two canopy densities [7.5 plants m^-2 (HD) and 4.5 plants m^-2 (LD)]. Large differences in P_C, R_c, R_C/P_C, leaf area index (LAI), dry matter accumulation and grain yield were found among hybrids and plant densities. Variations in P_C and R_C were associated mainly with changes in LAI. There was also found change in P_C per unit LAI with time. The average R_C/P_C was 28.9 % for all treatments throughout the vegetation season. P_C and R_C per unit dry matter were higher in LD than in HD and decreased throughout the measurement period. The HD stand had higher P_C and yield in hybrids with erectophile foliage, whereas LD stand had higher P_C after male tetrad and got higher yield in hybrids with planophile foliage. Only R_C in hybrids of the two foliage types was higher under HD than under LD throughout the vegetation period.     

一种估测小麦冠层氮含量的新高光谱指数

提出了一种估测小麦冠层氮含量的新高光谱指数--微分归一化氮指数(FD-NDNI)。以FieldSpec Pro FR地物光谱仪采集拔节后至孕穗前小麦的冠层光谱190份,随机抽取142份作为训练集,其余48份作为预测集。将光谱以小波阈值去噪法去噪后,利用其525、570 与730 nm处的一阶导数值,采用差值、比值以及归一化的方法构建了12种光谱指数以实现小麦冠层氮含量的估测,并与mNDVI705、mSR以及NDVI705等22种常用指数进行了比较分析。发现指数FD-NDNI对小麦冠层氮含量的估测结果最佳,其估测模型(指数形式)校正集决定系数(C-R²)与预测集决定系数(P-R²)分别达0.818与0.811,优于mNDVI705等常用指数。进一步分析表明,在各指数中,FD-NDNI对叶面积系数最不敏感,可最有效地避免冠层郁闭度等因素对氮含量估测的影响。为优化结果,采用最小二乘支持向量回归算法(LS-SVR)对模型进行了改进,当模型惩罚系数C与RBF核函数参数g取得最优解6.4与1.6时,其C-R²与P-R²分别提高至0.846与0.838,具有比指数模型更高的精度。结果表明:FD-NDNI是小麦冠层氮含量估测的优选指数,LS-SVR为建模的优选算法。     

祁连山区青海云杉林冠层叶面积指数的反演方法

叶面积指数(Leaf area index, LAI)是陆地生态系统的一个十分重要的结构参数。随着空间精细化模型的发展和基于过程的分布式模拟技术的应用, 对LAI的区域估算显得越来越重要, 但目前尚缺乏有效的估算手段。该项研究以青海云杉(Picea crassifolia)林为研究对象, 利用LAI-2000冠层分析仪、鱼眼镜头法和经验公式法对林冠层LAI进行了测定, 观测值分别为1.03~3.70、0.48~2.26和2.27~8.20, 显然, 仪器测定值偏低。针对针叶的集聚效应导致仪器测定值偏低的现象, 利用跟踪辐射与冠层结构测量仪(TRAC)测定的青海云杉林聚集系数计算调整系数, 对鱼眼镜头法获取的LAI值进行订正。根据高分辨率的遥感数据反演青海云杉林的植被指数与LAI的关系, 最后获得了较合理的该地区林冠层LAI的空间分布图。     

Root phenotyping by root tip detection and classification through statistical learning

Background and aims

Variations in the water and soil background in the signal path can cause variations in canopy spectral reflectance, which leads to uncertainty in estimating the canopy nitrogen (N) status. The primary objective of this study was to explore the optimum vegetation indices that were highly correlated with canopy leaf N concentration (LNC) but less influenced by the canopy leaf area index (LAI) and vegetation coverage (VC) in rice.

Methods

A systematic analysis of the quantitative relationships between various hyperspectral vegetation indices and LNC, VC and LAI was conducted based on 4-year rice field experiments using different rice varieties, N rates and planting densities. New spectral indices were derived to estimate LNC in rice under variable vegetation coverage.

Results

Although the newly developed simple green ratio indices, SR (R₅₅₃, R₅₃₇) and SR (R₅₄₅, R₅₃₈), and the three-band index (R₆₀₅-R₅₂₁-R₆₈₂)/(R₆₀₅+R₅₂₁+R₆₈₂) correlated well with the LNC. Only SR (R₅₅₃, R₅₃₇) was less influenced by VC/LAI and showed a stable performance in both the independent calibration and validation datasets. For the published indices tested in the present study, NDVIg-b and ND (R₅₀₃, R₄₈₃) showed a good predictive ability for the LNC. However, both of these indices and other published indices were found to be significantly dominated by the VC/LAI.

Conclusion

SR (R₅₅₃, R₅₃₇) was the best index to reliably estimate the LNC in rice under various cultivation conditions, and is recommended for this use. However, other spectral indices need to be examined to determine if they influenced by factors such as VC/LAI. Such studies will improve the applicability of these indices to different types of rice cultivars and production systems.     

林地叶面积指数遥感估算方法适用分析

叶面积指数是与森林冠层能量和CO2交换密切相关的一个重要植被结构参数,为了探讨估算林地叶面积指数LAI的遥感适用方法和提高精度的途径,利用TRAC仪器测定北京城区森林样地的LAI,从Landsat TM遥感图像计算NDVI、SR、RSR、SAVI植被指数,分别建立估算LAI的单植被指数统计模型、多植被指数组合的改进BP神经网络,获取最有效描述LAI与植被指数非线性关系的方法并应用到TM图像估算北京城区LAI。结果表明,单植被指数非线性统计模型估算LAI的精度高于线性统计模型;多植被指数组合神经网络中,以NDVI、RSR、SAVI组合估算LAI的精度最高,估算值与观测值线性回归方程的R2最高,为0.827,而RMSE最低,为0.189,神经网络解决了多植被指数组合统计模型非线性回归方程的系数较多、较难确定的问题,可较为有效的应用于遥感图像林地LAI的估算。     

Analysis of the relationship between the spectral characteristics of maize canopy and leaf area index under drought stress

Maize is one of the most widespread grain crops in the world; however, more than 70% of corn in China suffers some degree of drought disaster every year. Leaf area index (LAI) is an important biophysical parameter of the vegetation canopy and has important significance for crop yield estimation. Using the data of canopy spectral reflectance and leaf area index (LAI) for maize plants experiencing different levels of soil moisture from 2011 to 2012, the characteristics of the canopy reflective spectrum and its first derivative, and their relationships to leaf area index, were analyzed. Soil moisture of the control group was about 75% while that of the drought stress treatment was about 45%. In addition, LAI retrieval models for maize were established using vegetation indices (VIs) and principal component analysis (PCA) and the models were tested using independent datasets representing different soil water contents and different developmental stages of maize. The results showed that canopy spectral reflectances were in accordance with the characteristics of green plants, under both drought stress and at different developmental stages. In the visible band, canopy reflectance for both healthy and damaged vegetation had a green-wavelength peak and a red-wavelength valley; reflectance under drought stress, especially in the green peak (about 550 nm) and the red valley (about 676 nm) was higher than in the control group. In the near-infrared band, the canopy spectral reflectance decreased substantially between 780 and 1350 nm under drought stress. Moreover, the red edge of the spectrum was shifted toward blue wavelengths. The first derivative spectrum showed a double peak phenomenon at the edge of the red band at different developmental stages: the main peak appeared between 728 and 732 nm and the minor peak at about 718 nm. The double peaks become more obvious through the growth and development of the maize, with the most notable effect during the silking and milk stages, after which it gradually decreased. During maize growth, the LAI of all plants, regardless of soil moisture conditions, increased, and the largest LAI also occurred during the silking and milk stages. During those stages, the LAI of plants under different drought stress levels was significantly lower (by 20% or more) than in normal plants with sufficient water supplies. The LAI was highly significantly correlated with canopy spectral reflectance in the bands from 350 nm to 510 nm, from 571 nm to 716 nm, and from 1450 nm to 1575 nm. Also, the LAI was significantly correlated with red edge parameters and several VIs. The Perpendicular Vegetation Index (PVI) had the best correlation with LAI, with a coefficient of determination (R²) of 0.726 for the exponential correlation. Using dependent data, a LAI monitoring model for the maize canopy was constructed using PCA and VI methods. The test results showed that both the VI and PCA methods of monitoring maize LAI could provide robust estimates, with the predicted values of LAI being significantly correlated with the measured values. The model based on PVI showed higher precision under the drought stresses, with a correlation coefficient of 0.893 (n = 27), while the model based on PCA was more precise under conditions of adequate soil moisture, with a correlation coefficient of 0.877 (n = 32). Therefore, a synthesis of the models based on both VI and PCA could be more reliable for precisely predicting LAI under different levels of drought stresses in maize.