Prediction and assessment of bark beetle-induced mortality of lodgepole pine using estimates of stand vigor derived from remotely sensed data

The current outbreak of mountain pine beetle (Dendroctonus ponderosae Hopkins) in British Columbia (BC), Canada, has led forest managers to consider thinning as a means of decreasing residual tree susceptibility to attack and subsequent mortality. Previous research indicates that susceptibility to mountain pine beetle is a function of a tree's physiological vigor and the intensity of attack. Trees able to produce ≥ 80 g (g) of wood per m² of projected leaf area annually are highly resistant, because they are able to shift resource allocation locally from wood to resin production to isolate blue-stain fungi introduced by attacking beetles. Typically, the leaf area of susceptible stands must be reduced by two-thirds to permit most residual trees to increase their vigor to a safe level. We evaluate whether Landsat Thematic Mapper (TM) imagery (30 × 30 m) provides a means to assess the maximum leaf area index (LAI) of unthinned stands and the extent that thinning reduces LAI. The extent that residual trees in thinned stands may have increased their resistance to attack from mountain pine beetle is predicted from a non-linear relationship between % maximum LAI and mean tree vigor.We investigated the merits of this approach in the vicinity of Parson, British Columbia using four stands of lodgepole pine (Pinus contorta Dougl.), two of which were heavily thinned (stands were spaced to 4 and 5 m, approximately 70% reduction in stand density). An analysis of archived Landsat TM imagery indicated that prior to thinning in 1993, all four stands had full canopy, which, for mature stands, would translate to mean tree vigor between 40 and 70 g of annual wood production per m² of foliage. By 1995, based on estimated changes in LAI derived from a second data of Landsat TM imagery, stand vigor in the unthinned stands had not changed; however, in the thinned stands, a nearly two third reduction in LAI resulted in a predicted increase in vigor to between 100 and 160 g wood m^− 2 of leaf area. A subsequent assessment in 2001 indicated that stand vigor remained higher in the thinned stands relative to the control stands. Following an infestation of mountain pine beetle in the study area in 2002, mortality data indicated that the thinned stands experienced no mortality relative to the unthinned stands which experienced 5.5% mortality in the initial years of the attack. In the larger area surrounding the study site, a general relationship was found between predicted stand vigor and mountain pine beetle-induced mortality as estimated from aerial overview survey data (r² = 0.43, p < 0.01).     

利用风云卫星MERSI数据反演叶面积指数的研究

风云三号A星上搭载的中分辨率成像光谱仪(Medium Resolution Imaging Spectrometer)MERSI从2008年5月底开始对地球观测,其中5个波段250m分辨率的数据包含了丰富的植被信息,在全球同类传感器数据中独具特色,在其基础上反演的陆表植被数据产品目前还不多见。利用2013年生长季在河北固城观测获取的冬小麦光谱数据,结合MERSI 250m数据计算的NDVI值,建立二者NDVI之间的线性转换模型Y=1.1458 X+0.1916;同时利用地物光谱NDVI与实测叶面积指数构建了NDVI-LAI指数模型Y=0.0899e4.459 X;然后,利用MERSI 250m数据反演出华北太行山前平原区冬小麦的叶面积指数,经与大田观测的叶面积指数以及同期MODIS的叶面积指数产品对比验证,结果表明:反演的MERSI-LAI与实际观测叶面积指数接近且具有很好的线性关系,其空间分布与MODIS的叶面积指数相近,但MODIS-LAI数值明显偏小。     

基于BP神经网络的夏玉米多生育期叶面积指数反演研究

叶面积指数（Leaf Area Index,LAI）是生物地球化学循环中重要的植被结构参数。针对目前基于我国GF-1 WFV卫星影像的夏玉米多生育期LAI反演研究较少的问题,基于不同隐含层构建BP神经网络模型（BP1模型和BP2模型）,对比分析BP1模型、BP2模型和6种统计模型（NDVI、RVI、DVI、EVI、SAVI、ARVI）反演之间的精度差异,并根据实测数据绘制BP1模型和BP2模型的夏玉米多生育期LAI动态变化图。结果表明:LAI与6种常用的统计模型均有良好相关性,其中NDVI指数方程式回归模型拟合度最优;BP神经网络模型整体R ²略小于统计模型,而RMSE则小于统计模型,取得了与实测值差异更小的结果,统计模型与BP神经网络模型各有优劣之处;BP2模型在R ²和RMSE均优于BP1模型,能获得更为精确的反演值,BP2整体预测精度更高;基于BP神经网络模拟夏玉米生育期反演,LAI值呈现缓慢升高—快速增长—逐渐减小的S型变化过程,基本符合作物生长规律。该研究结合不同隐含层建立的BP神经网络模型,为GF-1卫星在作物叶面积指数多生育期反演的应用推广提供了方法支撑。     

不同叶面积指数遥感数据模拟中国总初级生产力的时空差异

陆地生态系统总初级生产力（GPP）反映了植物吸收固定大气中CO₂的能力,是碳循环过程中的重要环节。光能利用率（LUE）模型被广泛应用于GPP模拟。叶面积指数（LAI）数据是LUE模型的重要输入数据,不同的LAI数据差异较大,从而导致GPP模拟存在很大差异。利用3种常用的卫星遥感LAI数据（MCD15、GLASS和GlobMap）和气象数据模拟中国2003~2017年的GPP,比较了3种LAI数据在中国区域的时空差异,分析不同LAI数据模拟的中国GPP的时空差异。研究结果表明：3种LAI数据在中国区域的年平均值和LAI变化趋势的空间分布格局存在明显差异,森林区域的差异较大;2003~2017年间,中国区域3种LAI年平均值均呈显著增加趋势（p<0.01）,但不同LAI数据年平均值的年际变化差异明显;站点尺度GLASS LAI模拟的GPP与观测值相关性较好;不同LAI数据模拟的中国GPP总量多年平均值差异明显,最大值为7.46 Pg C a^-1 （GLASS）,最小值为6.39 Pg C a^-1 （GlobMap）;3种LAI数据模拟的中国GPP总量在2003~2017年呈显著增加趋势（p<0.05）,但不同的LAI数据模拟的中国GPP年总量的年际变化差异明显;不同LAI数据模拟的年均GPP和GPP变化趋势的空间分布格局存在明显差异,森林和农田区域的差异较大。研究结果有助于评估由于LAI数据造成的区域GPP模拟结果的不确定性。     

面向遥感叶面积指数产品的地形校正研究

地形校正是提高复杂地形区地表参数遥感定量化反演精度的重要手段。当前广泛应用的遥感叶面积指数产品（Leaf Area Index, LAI）多具有一定的地形误差,减少地形影响、提升其产品精度有着非常重要的意义。以我国江西省千烟洲地区为研究区域,利用地面实测LAI数据、LandsatTM数据和高程数据等,基于高程标准差和GLOBMAP LAI产品值的关系,建立面向叶面积指数产品的地形校正模型,利用这一模型对GLOBMAP LAI产品进行地形校正。结果表明：校正后的LAI与地面实测数据更为接近,LAI产品与地面测量值的RMSE由2.11下降到2.04;校正后LAI产品的标准差由2.08下降至1.69,LAI产品的地形误差得到了较好的改正。该方法较好地完成了LAI产品的地形校正,进一步提高了产品精度,具有一定的实用价值。     

我国叶面积指数卫星遥感产品生产及验证

利用卫星遥感生产叶面积指数（Leaf area index： LAI）产品并进行真实性检验是植被定量遥感的一项重要研究内容。过去10 a,我国研究人员利用MODIS或AVHRR观测数据生产了GLOBALBNU, GLASS, GLOBMAP,MuSyQ和FSGOM等数套全球和全国LAI产品,受到了国内外的广泛关注和应用。在产品生产的同时,我国学者也广泛开展了LAI产品在全球和区域尺度的真实性检验研究工作。本文总结了我国LAI卫星产品生产和验证研究工作的现状和趋势。近年来,我国在本领域相关的理论、技术和方法研究方面取得了全面进展,LAI产品精度和连续性已与国外先进水平相当,但仍然存在数据源单一且依赖国外、算法不确定性不清、生产不连续以及缺乏充分验证等问题,客观上影响了LAI产品应用的广度及深度。未来应充分利用新的卫星数据特别是国产数据,加强遥感机理模型、反演算法以及应用的创新研究,生产具有特色的高质量LAI产品,满足地球系统科学的研究需求。同时,应加强LAI验证基础设施建设,发掘利用更广泛的验证站点,同时增进国际合作,加强产品的推广使用,在与用户的互动交流反馈中进一步提高产品的市场。随着我国对相关研究投入的增加,可以预期未来20 a将是我国LAI遥感产品及验证研究由“跟跑”国际先进水平向“并跑”乃至“领跑”过渡的机遇期。     

Monitoring canopy biophysical and biochemical parameters in ecosystem scale using satellite hyperspectral imagery: An application on a Phlomis fruticosa Mediterranean ecosystem using multiangular CHRIS/PROBA observations

This study focuses on the potential of satellite hyperspectral imagery to monitor vegetation biophysical and biochemical characteristics through narrow-band indices and different viewing angles. Hyperspectral images of the CHRIS/PROBA sensor in imaging mode 1 (5 observation angles, 62 bands, 410-1005 nm) were acquired throughout a two-year period for a Mediterranean ecosystem fully covered by the semi-deciduous shrub Phlomis fruticosa. During each acquisition, coincident ecophysiological field measurements were conducted. Leaf area index (LAI), leaf biochemical content (chlorophyll a, chlorophyll b, carotenoids) and leaf water potential were measured. The hyperspectral images were corrected for coherent noises, cloud and atmosphere, in order to produce ground reflectance images. The reflectance spectrum of each image was used to calculate a variety of vegetation indices (VIs) that are already published in relevant literature. Additionally, all combinations of the 62 bands were used in order to calculate Normalized Difference Spectral Indices (NDSI_(x,y)) and Simple Subtraction Indices (SSI_(x,y)). The above indices along with raw reflectance and reflectance derivatives were examined for linear relationship with the ground-measured variables and the strongest relationships were determined. It is concluded that higher observation angles are better for the extraction of biochemical indices. The first derivative of the reflectance spectra proved to be very useful in the prediction of all measured variables. In many cases, complex and improved spectral indices that are proposed in the literature do not seem to be more accurate than simple NDSIs such as NDVI. Even traditional broadband NDVI is proved to be adequate in LAI prediction, while green bands seem also very useful. However, in biochemical estimation narrow bands are necessary. Indices that incorporate red, blue and IR bands, such as PSRI, SIPI and mNDVI presented good performance in chlorophyll estimation, while CRI did not show any relevance to carotenoids and WI was poorly correlated to water potential. Moreover, analyses indicated that it is very important to use a near red-edge band (701 nm) for effective chlorophyll index design. SSIs that incorporate 701 nm with 511 or 605 nm showed best performance in chlorophyll determination. For carotenoid estimation, a band on the edge of carotenoid absorption (511 nm) combined with a red band performed best, while a normalized index of two water absorption bands (945, 971 nm) proved to be an effective water index. Finally, the attempt to investigate stress conditions through pigment ratios resulted in the use of the band centred at 701 nm.     

Filtering and Reconstruction of LAI Time Series Data by S-G Filter based on Pixel Quality Analysis and Outlier Detection

The Leaf Area Index(LAI) is an important parameter of the canopy structure of the vegetation.There are problems such as missing data and low quality in the MODIS LAI data product because of the influence of atmospheric conditions and other factors,which seriously affects its the application.As a case study of Jiangxi Province,MODIS LAI time series products data set in 2009-2013 were reconstructed by integrating pixel quality analysis,S-G filter and annual sequence abnormal value detection and filtering technology.The results show that the high quality of the broad-leaved forest was the lowest,only 51.76%,and proportion of low quality and retrieval failed pixels reached to 20% to 30%.The integrated filtering method proposed for low quality data set reconstruction.Compared with S-G filtering,the mean value of the reconstructed LAI by the integrated filtering method is more consistent with the original mean for high quality pixel.And its correlation coefficient between the high quality pixel and the original data reaches 0.97.So it has better fidelity.The outliers of the low- and medium-quality pixel reconstruction are filtered,and the area of null value is filled.Meanwhile,the standard deviation is reduced,and the low-value area or abnormal point is better identified and repaired.The overall stability becomes better and it can fit the time series change curve well.     

Influence of woody elements of a Norway spruce canopy on nadir reflectance simulated by the DART model at very high spatial resolution

A detailed sensitivity analysis investigating the effect of woody elements introduced into the Discrete Anisotropic Radiative Transfer (DART) model on the nadir bidirectional reflectance factor (BRF) for a simulated Norway spruce canopy was performed at a very high spatial resolution (modelling resolution 0.2 m, output pixel size 0.4 m). We used such a high resolution to be able to parameterize DART in an appropriate way and subsequently to gain detailed understanding of the influence of woody elements contributing to the radiative transfer within heterogeneous canopies. Three scenarios were studied by modelling the Norway spruce canopy as being composed of i) leaves, ii) leaves, trunks and first order branches, and finally iii) leaves, trunks, first order branches and small woody twigs simulated using mixed cells (i.e. cells approximated as composition of leaves and/or twigs turbid medium, and large woody constituents). The simulation of each scenario was performed for 10 different canopy closures (CC = 50-95%, in steps of 5%), 25 leaf area index (LAI = 3.0-15.0 m² m^− 2, in steps of 0.5 m² m^− 2), and in four spectral bands (centred at 559, 671, 727, and 783 nm, with a FWHM of 10 nm). The influence of woody elements was evaluated separately for both, sunlit and shaded parts of the simulated forest canopy, respectively. The DART results were verified by quantifying the simulated nadir BRF of each scenario with measured Airborne Imaging Spectroradiometer (AISA) Eagle data (pixel size of 0.4 m). These imaging spectrometer data were acquired over the same Norway spruce stand that was used to parameterise the DART model.The Norway spruce canopy modelled using the DART model consisted of foliage as well as foliage including robust woody constituents (i.e. trunks and branches). All results showed similar nadir BRF for the simulated wavelengths. The incorporation of small woody parts in DART caused the canopy reflectance to decrease about 4% in the near-infrared (NIR), 2% in the red edge (RE) and less than 1% in the green band. The canopy BRF of the red band increased by about 2%. Subsequently, the sensitivity on accounting for woody elements for two spectral vegetation indices, the normalized difference vegetation index (NDVI) and the angular vegetation index (AVI), was evaluated. Finally, we conclude on the importance of including woody elements in radiative transfer based approaches and discuss the applicability of the vegetation indices as well as the physically based inversion approaches to retrieve the forest canopy LAI at very high spatial resolution.     

Performances of neural networks for deriving LAI estimates from existing CYCLOPES and MODIS products

This paper evaluates the performances of a neural network approach to estimate LAI from CYCLOPES and MODIS nadir normalized reflectance and LAI products. A data base was generated from these products over the BELMANIP sites during the 2001-2003 period. Data were aggregated at 3 km × 3 km, resampled at 1/16 days temporal frequency and filtered to reject outliers. VEGETATION and MODIS reflectances show very consistent values in the red, near infrared and short wave infrared bands. Neural networks were trained over part of this data base for each of the 6 MODIS biome classes to retrieve both MODIS and CYCLOPES LAI products.Results show very good performances of neural networks to estimate the original LAI products with an overall root mean square error (RMSE) around 0.5 for MODIS LAI from both MODIS and CYCLOPES normalized reflectances and a RMSE ranging between 0.12 (CYCLOPES reflectances) and 0.29 (MODIS reflectances) for CYCLOPES LAI. A drop of 15% of performance was found by training MODIS biome dependant algorithm by a single network over all the classes at the same time. More detailed analyses show that CYCLOPES and MODIS LAI values are very consistent for grasses and crops. Conversely, other biomes including shrubs, savanna, needleleaf and broadleaf forests show significant discrepancies, mainly due to differences between LAI definitions used between CYCLOPES (closer to effective LAI) and MODIS (closer to true LAI). However, products derived from the original CYCLOPES LAI products show a better agreement with both effective and true LAI ground measurements values. MODIS LAI products show more instability, partly because of the slightly shorter temporal resolution as compared to CYCLOPES.These results confirm the interest and versatility of neural networks for operational algorithms. This approach could be extended to other products or sensors, and may constitute a step forward for the fusion of data from several sensors, hence contributing to develop ‘virtual constellations’.