A multisensor fusion approach to improve LAI time series

High-quality and gap-free satellite time series are required for reliable terrestrial monitoring. Moderate resolution sensors provide continuous observations at global scale for monitoring spatial and temporal variations of land surface characteristics. However, the full potential of remote sensing systems is often hampered by poor quality or missing data caused by clouds, aerosols, snow cover, algorithms and instrumentation problems. A multisensor fusion approach is here proposed to improve the spatio-temporal continuity, consistency and accuracy of current satellite products. It is based on the use of neural networks, gap filling and temporal smoothing techniques. It is applicable to any optical sensor and satellite product. In this study, the potential of this technique was demonstrated for leaf area index (LAI) product based on MODIS and VEGETATION reflectance data. The FUSION product showed an overall good agreement with the original MODIS LAI product but exhibited a reduction of 90% of the missing LAI values with an improved monitoring of vegetation dynamics, temporal smoothness, and better agreement with ground measurements.     

Parametric (modified least squares) and non-parametric (Theil-Sen) linear regressions for predicting biophysical parameters in the presence of measurement errors

Remote sensing often involves the estimation of in situ quantities from remote measurements. Linear regression, where there are no non-linear combinations of regressors, is a common approach to this prediction problem in the remote sensing community. A review of recent remote sensing articles using univariate linear regression indicates that in the majority of cases, ordinary least squares (OLS) linear regression has been applied, with approximately half the articles using the in situ observations as regressors and the other half using the inverse regression with remote measurements as regressors. OLS implicitly assume an underlying normal structural data model to arrive at unbiased estimates of the response. OLS regression can be a biased predictor in the presence of measurement errors when the regression problem is based on a functional rather than structural data model. Parametric (Modified Least Squares) and non-parametric (Theil-Sen) consistent predictors are given for linear regression in the presence of measurement errors together with analytical approximations of their prediction confidence intervals. Three case studies involving estimation of leaf area index from nadir reflectance estimates are used to compare these unbiased estimators with OLS linear regression. A comparison to Geometric Mean regression, a standardized version of Reduced Major Axis regression, is also performed. The Theil-Sen approach is suggested as a potential replacement of OLS for linear regression in remote sensing applications. It offers simplicity in computation, analytical estimates of confidence intervals, robustness to outliers, testable assumptions regarding residuals and requires limited a priori information regarding measurement errors.     

Seasonal variations of leaf area index of agricultural fields retrieved from Landsat data

The derivation of leaf area index (LAI) from satellite optical data has been the subject of a large amount of work. In contrast, few papers have addressed the effective model inversion of high resolution satellite images for a complete series of data for the various crop species in a given region. The present study is focused on the assessment of a LAI model inversion approach applied to multitemporal optical data, over an agricultural region having various crop types with different crop calendars. Both the inversion approach and data sources are chosen because of their wide use. Crops in the study region (Barrax, Castilla-La Mancha, Spain) include: cereal, corn, alfalfa, sugar beet, onion, garlic, papaver. Some of the crop types (onion, garlic, papaver) have not been addressed in previous studies. We use in-situ measurement sets and literature values as a priori data in the PROSPECT + SAIL models to produce Look Up Tables (LUTs). Those LUTs are subsequently used to invert Landsat-TM and Landsat-ETM+ image series (12 dates from March to September 2003). The Look Up Tables are adapted to different crop types, identified on the images by ground survey and by Landsat classification. The retrieved LAI values are compared to in-situ measurements available from the campaign conducted in mid July-2003. Very good agreement (a high linear correlation) is obtained for LAI values from 0.1 to 6.0. LAI maps are then produced for each of the 12 dates. The LAI temporal variation shows consistency with the crop phenological stages. The inversion method is favourably compared to a method relying on the empirical relationship between LAI and NDVI from Landsat data. This offers perspectives for future optical satellite data that will ensure high resolution and high temporal frequency.     

Spatially and temporally continuous LAI data sets based on an integrated filtering method: Examples from North America

Leaf Area Index (LAI) is an important biophysical variable for characterizing the land surface vegetation. Global LAI product has been routinely produced from the MODerate resolution Imaging Spectroradiometer (MODIS) aboard the Terra and Aqua satellite platforms. However, the MODIS standard LAI product is not continuous both spatially and temporally. To fill the gaps and improve the quality, we have developed a data filtering algorithm. This filter, called the temporal spatial filter (TSF), integrates both spatial and temporal characteristics for different plant functional types. The spatial gaps are first filled with the multi-year averages of the same day. If the values are missing over all years, the pixel is filled with a new estimate using the vegetation continuous field-ecosystem curve fitting method. The TSF integrates both the multi-seasonal average trend (background) and the seasonal observation. We implement this algorithm using the MODIS Collection 4 LAI product over North America. Comparison of the TSF results with the Savitzky-Golay filter indicates that the TSF performs much better in restoring the spatial and temporal distribution of seasonal LAI trends. The new LAI product has been validated by comparing with field measurements and the derived LAI maps from ETM+ data at a broadleaf forest site and an agricultural site. The validation results indicate that the new LAI product agrees better with both the field measurements and LAI values obtained from the ETM+ than does the MODIS LAI standard product, which usually shows higher LAI values.     

Integration of MODIS data into a simple model for the spatial distributed simulation of soil water content and evapotranspiration

A precise simulation of soil water content (SWC) and actual evapotranspiration (ET_a) in a region or a catchment depends on the accuracy of the spatial data inputs. In this study, we developed a simple grid-based soil water balance model. In this model, remotely sensed vegetation data are used to estimate spatial distributions of daily SWC and ET_a rates. The model was validated by comparing simulated SWC with the measured by gravimetric method and time domain reflectometry (TDR) at an experimental test site located in Northeastern Germany in the time period 1993-1998. The index of agreement IA and the root-mean-square error obtained from the comparison of the TDR measurements to the simulated values ranged from 0.45 to 0.80 and from 0.029 to 0.061 cm³/cm³, respectively. The comparison of simulated ET_a rates to those measured by four large-scale lysimeters at another test site showed IA values above 0.87 and R² values higher than 0.59. For the regional application of the model, a method was developed to integrate the Moderate Resolution Imaging Spectrometer (MODIS) vegetation data into the model. The MODIS data used in our study consist of 16-day normalized difference vegetation index and 8-day leaf area index products. Regarding the spatial application of the model, our approach was tested in a catchment located in Northeastern Germany in 2001-2003. A sufficient correlation between daily discharge rates measured at two observation gauges in the catchment and the corresponding simulated discharge rates and also good correlations between the simulated ET_a rates and the MODIS-leaf area index values indicate that the model is an appropriate simulation tool at regional scale if the corresponding additional spatial databases regarding surface and soil properties are available.     

水稻冠层光谱特征及其与LAI的关系研究

氮素营养是影响作物生长与产量的最主要限制因子之一。准确及时地监测或诊断出作物氮素营养状况,对提高氮素利用效率和作物管理水平、减少过度施氮造成的环境污染具有重要意义。本研究在不同施氮水平处理的水稻试验小区,对水稻整个生长期内冠层反射光谱进行了较系统、密集的测定,同时测定了几个重要生育期水稻的叶面积指数。研究结果表明:随着施氮量的增加,水稻冠层光谱在各生育期间呈现出一定的规律性,在近红外部分(710～1 220 nm),冠层光谱反射率随着施氮水平的提高而升高,而在可见光部分(460～680 nm),水稻冠层的光谱反射率反而逐渐降低。经冠层光谱差异显著性检验发现:水稻灌浆期以前,对施氮水平最为敏感的波段是绿光(560～610 nm)和近红外(710～760 nm)部分;转换为归一化植被指数(NDVI)以后,差异最显著的是(R760-R560)/(R760+R560)。不同氮肥处理的水稻LAI随时间变化曲线大致都呈抛物线型,中低水平施氮肥水稻LAI随时间的变化曲线比较平缓,而高水平施氮肥LAI曲线则变化比较剧烈。冠层光谱反射与叶面积的相关分析结果表明:在水稻抽穗前,叶面积与冠层光谱反射率相关性较差;而抽穗后,叶面积与冠层光谱有较高的相关性。     

森林叶面积指数遥感反演模型构建及区域估算

基于eCognition面向对象分类算法及校正后的TM遥感影像,获取研究区2010年土地利用/覆被数据。同时在ArcGIS平台下,提取遥感影像6个波段反射率及RVI、NDVI、SLAVI、EVI、VII、MSR、NDVIc、BI、GVI和WI等10个植被指数,并辅助于DEM、ASPECT、SLOPE等地形信息,在与植物冠层分析仪(TRAC)实测各森林类型叶面积指数相关性分析的基础上,研究表明:相对多元线性回归方法,偏最小二乘法能够更好地把握各森林类型LAI动态变化,而后结合研究区森林覆被信息进行区域估算。     

An interpolation procedure for generalizing a look-up table inversion method

The inversion of physically based reflectance models is increasingly efficient for extracting vegetation variables from remote sensing images. It requires a vegetation reflectance model and an inversion method that are accurate and efficient. Usually, the complexity of reflectance models implies to use specific inversion methods (e.g., look-up table and neural network). Unfortunately, these methods are valid only for the view-sun directions for which they are designed. A developed look-up table based inversion method avoids this limitation: it generalizes any look-up table for any view-sun direction, and more generally for any input parameter value. It uses a look-up table made of c_i coefficients of any analytical expression h that fits a set of reflectance values simulated by the Discrete Anisotropic Radiative Transfer (DART) model. Interpolation on coefficients c_i allows h to give reflectance values for any input parameter value. We settled some options of the inversion method with sensitivity studies: tree covers are simulated with 4-tree scenes, expression h has six coefficients c_i and the interpolation is the continuous first derivative interpolation method. Moreover, the robustness of the inversion method was validated. The ability to generalize a look-up table for any view-sun direction was successfully tested with the inversion of SPOT images of Fontainebleau (France) forest. LAI maps proved to be as accurate (i.e., RMSE≈1.3) as those obtained with classical relationships that are calibrated with in situ LAI measurements. Here, the advantage of our inversion method was to avoid this calibration.     

Airborne discrete-return LIDAR data in the estimation of vertical canopy cover, angular canopy closure and leaf area index

Remote sensing of forest canopy cover has been widely studied recently, but little attention has been paid to the quality of field validation data. Ecological literature has two different coverage metrics. Vertical canopy cover (VCC) is the vertical projection of tree crowns ignoring within-crown gaps. Angular canopy closure (ACC) is the proportion of covered sky at some angular range around the zenith, and can be measured with a field-of-view instrument, such as a camera. We compared field-measured VCC and ACC at 15° and 75° from the zenith to different LiDAR (Light Detection and Ranging) metrics, using several LiDAR data sets and comprehensive field data. The VCC was estimated to a high precision using a simple proportion of canopy points in first-return data. Confining to a maximum 15° scan zenith angle, the absolute root mean squared error (RMSE) was 3.7-7.0%, with an overestimation of 3.1-4.6%. We showed that grid-based methods are capable of reducing the inherent overestimation of VCC. The low scan angles and low power settings that are typically applied in topographic LiDARs are not suitable for ACC estimation as they measure in wrong geometry and cannot easily detect small within-crown gaps. However, ACC at 0-15° zenith angles could be estimated from LiDAR data with sufficient precision, using also the last returns (RMSE 8.1-11.3%, bias -6.1-+4.6%). The dependency of LiDAR metrics and ACC at 0-75° zenith angles was nonlinear and was modeled from laser pulse proportions with nonlinear regression with a best-case standard error of 4.1%. We also estimated leaf area index from the LiDAR metrics with linear regression with a standard error of 0.38. The results show that correlations between airborne laser metrics and different canopy field characteristics are very high if the field measurements are done with equivalent accuracy.