Hyperspectral Reflectance Data Research Articles

Hyperspectral field reflectance measurements to estimate wheat grain yield and plant height

Hyperspectral crop reflectance data are useful for several remote sensing applications in agriculture, but there is still a need for studies to define optimal wavebands to estimate crop biophysical parameters. The objective of this work is to analyze the use of narrow and broad band vegetation indices (VI) derived from hyperspectral field reflectance measurements to estimate wheat (Triticum aestivum L.) grain yield and plant height. A field study was conducted during the winter growing season of 2003 in Campinas, São Paulo State, Brazil. Field canopy reflectance measurements were acquired at six wheat growth stages over 80 plots with four wheat cultivars (IAC-362, IAC-364, IAC-370, and IAC-373), five levels of nitrogen fertilizer (0, 30, 60, 90, and 120 kg of N ha-1) and four replicates. The following VI were analyzed: a) hyperspectral or narrow-band VI (1. optimum multiple narrow-band reflectance, OMNBR; 2. narrow-band normalized difference vegetation index, NB_NDVI; 3. first- and second-order derivative of reflectance; and 4. four derivative green vegetation index); and b) broad band VI (simple ratio, SR; normalized difference vegetation index, NDVI; and soil-adjusted vegetation index, SAVI). Hyperspectral indices provided an overall better estimate of biophysical variables when compared to broad band VI. The OMNBR with four bands presented the highest R² values to estimate both grain yield (R² = 0.74; Booting and Heading stages) and plant height (R² = 0.68; Heading stage). Best results to estimate biophysical variables were observed for spectral measurements acquired between Tillering II and Heading stages.

Differentiation of turfgrass and common weed species using hyperspectral radiometry

Hand-held hyperspectral reflectance data were collected in the summers of 2002, 2003, and 2004 to differentiate unique spectral characteristics of common turfgrass and weed species. Turfgrass species evaluated were: bermudagrass, ‘Tifway 419’; zoysiagrass, ‘Meyer’; St. Augustinegrass, ‘Raleigh’; common centipedegrass; and creeping bentgrass, ‘Crenshaw’. Weed species evaluated were: dallisgrass, southern crabgrass, eclipta, and Virginia buttonweed. Reflectance data were collected from greenhouse and field locations. An overall classification accuracy of 85% was achieved for all species in the field. A total of 21 spectral bands between 378 and 1,000 nm that were consistent over the three data collection periods were used for analysis. Only centipedegrass, zoysiagrass, and dallisgrass were correctly classified less than 80% of the time. An overall classification accuracy of 69% was achieved for the greenhouse species. Spectral bands used in this analysis ranged from 353 to 799 nm. Creeping bentgrass and Virginia buttonweed were classified correctly at 96 and 92%, respectively.

Assessment of rice leaf growth and nitrogen status by hyperspectral canopy reflectance and partial least square regression

Diagnosis of rice growth and nutrient status is critical for prediction of rice yield and grain quality and prescription of nitrogen topdressing at panicle initiation stage. Two experiments, one in 2000 and one in 2003 were conducted to construct a partial least square model for assessing rice leaf growth and nitrogen status non-destructively at the Experimental Farm (37°16′N, 126°59′E) of Seoul National University, Suwon, Korea. The experiment included two cultivars (Hwasungbyeo and Dasanbyeo) and four levels of nitrogen (N) application in year 2000 and four rice cultivars (Hwasungbyeo, SNU-SG1, Juanbyeo, and Surabyeo) and 10 N treatments in year 2003. Hyperspectral canopy reflectance (300–1100 nm) data recorded at various growth stages before heading were used for partial least square regression (PLS) model to predict four crop variables: leaf area index, leaf dry weight, leaf N concentration, and leaf N density. Three hundred and forty two observations from two experiments were randomly split for model calibration (75%) and validation (25%). Coefficient of determination ( R 2), root mean square error in prediction (RMSEP) and relative error of prediction (REP) of model calibration and validation were calculated for the model quality evaluation. The results revealed that PLS model using hyperspectral canopy reflectance data to predict four plant variables produced an acceptable model precision and accuracy. The model R 2 and REP ranged from 0.84 to 0.87 and 10.0 to 23.8% for calibration and 0.79 to 0.84 and 11.1 to 25.6% for validation, respectively. The most important reflectance as judged by factor loading in PLS model for rice leaf characterization was at various bands such as near-infrared (>760 nm) and visible (355, 420, 524–534, 583 and 687 nm) and red edge (707 nm) region.

Novel Analysis of Hyperspectral Reflectance Data for Detecting Onset of Pollen Shed in Maize

Knowledge of pollen shed dynamics in and around seed production fields is critical for ensuring a high yield of genetically pure corn seed. Recently, changes in canopy reflectance using hyperspectral reflectance have been associated with tassel emergence, which is known to precede pollen shed in a predictable manner. Practical application of this remote sensing technology, however, requires a simple and reliable method to evaluate changes in spectral images associated with the onset of tassel emergence and pollen shed. In this study, several numerical methods were investigated for estimating percentage of plants with visible tassels (VT) and percentage of plants that initiated pollen shed (IPS) from remotely sensed hyperspectral reflectance data (397 to 902 nm). Correlation analysis identified regions of the spectra that were associated with tassel emergence and anthesis (i.e., 50% of plants shedding pollen). No single band, however, generated correlations greater than 0.40 for either VT or IPS. Classification using an artificial neural network (ANN) was predictive, correctly classifying 83.5% and 88.3% of the VT and IPS data, respectively. The extensive preprocessing necessary and the black box nature of ANNs, however, rendered analysis of spectral regions difficult using this method. Partial least squares (PLS) analysis yielded models with high predictive capability (R2 of 0.80 for VT and 0.79 for IPS). The PLS coefficients, however, did not exhibit a spectrally consistent pattern. A novel range operator-enabled genetic algorithm (ROE-GA), designed to consider the shape of the spectra, had similar predictive capabilities to the ANN and PLS, but provided the added advantage of allowing information transfer for increased domain knowledge. The ROE-GA analysis is the preferred method to evaluate hyperspectral reflectance data and associate spectral changes to tassel emergence and the onset of pollen shed in corn on a field scale.

3D modelling of forest canopy structure for remote sensing simulations in the optical and microwave domains

A detailed 3D structural model of a conifer forest canopy was developed in order to simulate the reflectance (optical) and backscatter (microwave) signals measured remotely. We show it is feasible to model forest canopy scattering using detailed 3D models of tree structure including the location and orientation of individual needles. An existing structural growth model of Scots pine ( Pinus sylvestris L.), Treegrow, was modified to simulate observed growth stages of a Scots pine canopy from age 5 to 50 years. The 3D tree models showed close structural agreement with in situ measurements. Needles were added to the structural models according to observed phyllotaxy (distribution). Individual trees were used to generate model canopies, which in turn were used to drive optical and microwave models of canopy scattering. Simulated canopy radiometric response was compared with airborne hyperspectral reflectance data (HyMAP) and airborne synthetic aperture RADAR (ASAR) backscatter data. Model simulations agreed well in general with observations, particularly at optical wavelengths where model simulations of low and high density canopy stands were shown to bracket observations. Relatively small sensitivity of observed reflectance to canopy age was captured reasonably well by the simulations. The choice of needle shape and phyllotaxy was shown to have a significant impact on multiple scattering behaviour at the branch scale. In the microwave domain, simulated backscatter values agreed reasonably well with observations at L-band, less so at X-band. L-band simulated backscatter significantly underestimated observed backscatter at younger canopy ages, probably as a result of inappropriate modelling of soil/understory. It is demonstrated that a combined structural and radiometric modelling approach provides a flexible and powerful method for simulating the remotely sensed signal of a forest canopy in the optical and microwave domains. This is particularly useful for exploring the impact of canopy structure on the resulting signal and also for combined retrievals of forest structural parameters from optical and microwave data.

Hyperspectral Crop Reflectance Data for characterising and estimating Fungal Disease Severity in Wheat

Many studies have shown the usefulness of hyperspectral crop reflectance data for detecting plant pathological stress. However, there is still a need to identify unique signatures for specific stresses amidst the constantly changing background associated with normal crop growth and development. Comparing spatial and temporal patterns in crop spectra can provide such signatures. This work was concerned with characterising and estimating fungal disease severity in a spring wheat crop. This goal can be accomplished by using a reference data set consisting of hyperspectral crop reflectance data vectors and the corresponding disease severity field assessments. The hyperspectral vectors were first normalised into zero-mean and unit-variance vectors by performing various combinations of spectral- and band-wise normalisations. Then, after applying the same normalisation procedures to the new hyperspectral data, a nearest-neighbour classifier was used to classify the new data against the reference data. Finally, the corresponding stress signatures were computed using a linear transformation model. High correlation was obtained between the classification results and the corresponding field assessments of fungal disease severity, confirming the usefulness and efficiency of this approach. The effects of increased disease severity could be characterised by analysing the resulting disease signatures obtained when applying the different normalisation procedures. The low computational load of this approach makes it suitable for real-time on-vehicle applications.

Prediction of topsoil organic matter and clay content from measurements of spectral reflectance and electrical conductivity

Soil organic matter (SOM) and clay content of a soil characterized as a coarse sandy loam were modelled using hyperspectral reflectance data acquired with a spectrometer and soil electrical conductivity (SEC) data acquired with an EM38 instrument manufactured by Geonics Ltd. The partial least squares (PLS) regression method was applied and the results validated using cross validation. First, the models were calibrated using only spectral reflectance data; then EM38 data were included in the X-matrix of predictors. Although SEC is significantly correlated with clay content, the results showed that EM38 data did not improve model performance for the estimation of soil organic matter content and clay content, despite the fact that EM38 showed significant correlation with clay content.

Remote Sensing to Detect Herbicide Drift on Crops

Glyphosate and paraquat herbicide drift injury to crops may substantially reduce growth or yield. Determining the type and degree of injury is of importance to a producer. This research was conducted to determine whether remote sensing could be used to identify and quantify herbicide injury to crops. Soybean and corn plants were grown in 3.8-L pots to the five- to seven-leaf stage, at which time, applications of nonselective herbicides were made. Visual injury estimates were made, and hyperspectral reflectance data were recorded 1, 4, and 7 d after application (DAA). Several analysis techniques including multiple indices, signature amplitude (SA) with spectral bands as features, and wavelet analysis were used to distinguish between herbicide-treated and nontreated plants. Classification accuracy using SA analysis of paraquat injury on soybean was better than 75% for both 1/2- and 1/8× rates at 1, 4, and 7 DAA. Classification accuracy of paraquat injury on corn was better than 72% for the 1/2× rate at 1, 4, and 7 DAA. These data suggest that hyperspectral reflectance may be used to distinguish between healthy plants and injured plants to which herbicides have been applied; however, the classification accuracies remained at 75% or higher only when the higher rates of herbicide were applied. Applications of a 1/2× rate of glyphosate produced 55 to 81% soybean injury and 20 to 50% corn injury 4 and 7 DAA, respectively. However, using SA analysis, the moderately injured plants were indistinguishable from the uninjured controls, as represented by the low classification accuracies at the 1/8-, 1/32-, and 1/64× rates. The most promising technique for identifying drift injury was wavelet analysis, which successfully distinguished between corn plants treated with either the 1/8- or the 1/2× rates of paraquat compared with the nontreated corn plants better than 92% 1, 4, and 7 DAA. These analysis techniques, once tested and validated on field scale data, may help determine the extent and the degree of herbicide drift for making appropriate and, more importantly, timely management decisions.

Detection of pitted morningglory (Ipomoea lacunosa) by hyperspectral remote sensing. I. Effects of tillage and cover crop residue

Field experiments were conducted to evaluate the potential of hyperspectral reflectance data collected with a hand-held spectroradiometer to discriminate soybean intermixed with pitted morningglory and weed-free soybean in conventional till and no-till plots containing rye, hairy vetch, or no cover crop residue. Pitted morningglory was in the cotyledon to six-leaf growth stage. Seven 50-nm spectral bands (one ultraviolet, two visible, four near-infrared) derived from each hyperspectral reflectance measurement were used as discrimination variables. Pitted morningglory plant size had more influence on discriminant capabilities than tillage or cover crop residue systems. Across all tillage and residue systems, discrimination accuracy was 71 to 95%, depending on the size of pitted morningglory plants at the time of data acquisition. The versatility of the seven 50-nm bands was tested by using a discriminant model developed for one experiment location to test discriminant capabilities for the other experiment, with discrimination accuracy across all tillage and residue systems of 55 to 73%, depending on pitted morningglory plant size.

Reflectance measurement of canopy biomass and nitrogen status in wheat crops using normalized difference vegetation indices and partial least squares regression

Hyperspectral reflectance (438 to 884 nm) data were recorded at five different growth stages of winter wheat in a field experiment including two cultivars, three plant densities, and four levels of N application. All two-band combinations in the normalized difference vegetation index ( λ1− λ2)/( λ1+ λ2) were subsequently used in a linear regression analysis against green biomass (GBM, g fresh weight m −2 soil), leaf area index (LAI, m 2 green leaf m −2 soil), leaf chlorophyll concentration (Chl conc, mg chlorophyll g −1 leaf fresh weight), leaf chlorophyll density (Chl density, mg chlorophyll m −2 soil), leaf nitrogen concentration (N conc, mg nitrogen g −1 leaf dry weight), and leaf nitrogen density (N density, g nitrogen m −2 soil). A number of grouped wavebands with high correlation ( R 2>95%) were revealed. For the crop variables based on quantity per unit surface area, i.e. GBM, LAI, Chl density, and N density, these wavebands had in the majority (87%) of the cases a center wavelength in the red edge spectral region from 680 to 750 nm and the band combinations were often paired so that both bands were closely spaced in the steep linear shift between R red and R nir. The red edge region was almost absent for bands related to Chl conc and N conc, where the visible spectral range, mainly in the blue region, proved to be better. The selected narrow-band indices improved the description of the influence of all six-crop variables compared to the traditional broad- and short-band indices normally applied on data from satellite, aerial photos, and field spectroradiometers. For variables expressed on the basis of soil or canopy surface area, the relationship was further improved when exponential curve fitting was used instead of linear regression. The best of the selected narrow-band indices was compared to the results of a partial least square regression (PLS). This comparison showed that the narrow-band indices related to LAI and Chl conc, and to some extent also Chl density and N density, were optimal and could not be significantly improved by PLS using the information from all wavelengths in the hyperspectral region. However, PLS improved the prediction of GBM and N conc by lowering the RMSE with 22% and 24%, respectively, compared to the best narrow-band indices. It is concluded that PLS regression analysis may provide a useful exploratory and predictive tool when applied on hyperspectral reflectance data.

Relationship Between Hyperspectral Reflectance, Soil Nitrate-Nitrogen, Cotton Leaf Chlorophyll, and Cotton Yield: A Step Toward Precision Agriculture

ABSTRACT Modern agriculture uses large amounts of organic and inorganic nutrients to optimize productivity. Excessive nutrient applications sometime lead to adverse effects on the environment and human health. Precision agriculture is evolving with the objectives of minimizing these adverse effects by enabling farmers to manage nutrient applications more efficiently while sustaining precious environmental resources. To develop a method that uses nutrients more efficiently on cotton, a field experiment involving three sources and three rates of nitrogen with and without nitrification inhibitor was carried out in four replications at Belle Mina, AL during the 1994-97 crop seasons. In 1997, these plots were used to determine if there was a relationship between remotely sensed hyperspectral reflectance data and three field measurements that included cotton leaf chlorophyll (defined as measurements of five leaves using a Minolta Chlorophyll SPAD Meter to represent cotton canopy), soil nitrate-nitrogen, and cotton yield. Our results showed that hyperspectral reflectance in the 807.6 nm region had the highest significant correlation with cotton leaf chlorophyll. Cotton leaf chlorophyll correlated significantly with soil nitrate-nitrogen and cotton yield. Because leaf chlorophyll is an indicator of nitrogen deficiency, our results suggest that hyperspectral reflectance may be used as a tool to help farmers determine nitrogen deficiency, which may subsequently lead to increased crop productivity and reduced environmental pollution.

Feature Vector Based Analysis of Hyperspectral Crop Reflectance Data for Discrimination and Quantification of Fungal Disease Severity in Wheat

The impact of plant pathological stress on crop reflectance can be measured both in broad-band vegetation indices and in narrow or local characteristics of the reflectance spectra. This work is concerned with using the whole spectra in the objective examination of how different parts of the spectrum contribute in describing disease severity in wheat. A hyperspectral reflectance spectrum was considered as a mixed signal, i.e. the integration of the effects of all active objects in the investigated area. Independent component analysis (ICA) was used to blindly separate mixed statistically independent signals. Principal component analysis (PCA) was also used to extract interesting components. The ICA or PCA results had then to be interpreted efficiently. This was achieved by using a technique called feature-vector-based analysis (FVBA), which produces a number of ‘component–feature vector’ pairs, which represent the spectral signatures and the corresponding weighting coefficients of the different constituting source signals. These weighting coefficients were proportional to field assessments of fungal disease severity in a spring wheat crop, in percentage necrosis of leaf area, and high correlations were shown. Two effects of increased disease severity were observed: (1) a flattening of the green reflectance peak together with a general decrease in reflectance in the near-infrared region and (2) a decrease of the shoulder of the near-infrared reflectance plateau together with a general increase in the visible region between 550 and 750 nm.

Multiplatform optical monitoring of eutrophication in temporally and spatially variable lakes

Representative spatial patterns of eutrophication variables cannot be produced using traditional in situ sampling techniques. Spatial heterogeneity complicates the study of seasonal and long-term trends and the evaluation of water management policies. Remote sensing, however, with its broad view has the potential to deliver the relevant information. This paper will address the added value of synoptic eutrophication maps to the standard monitoring program of two large, spatially and temporally variable lakes in the Netherlands, Lakes IJssel and Marken. Remote sensing images were obtained from SeaWiFS; and combined with hyperspectral reflectance data from the airborne EPS-a sensor and the shipboard PR-650 spectroradiometer. The PR-650 data were used in selecting the most appropriate algorithms for SeaWiFS and EPS-a. A special algorithm for case II waters with high chlorophyll content was applied to SeaWiFS data to obtain chlorophyll concentrations. Synoptic maps of suspended matter were retrieved using inversion of a model for irradiance reflectance. For the airborne sensor inversion of reflectance was used for both suspended matter and chlorophyll. Satellite and airborne sensors clearly are complementary to each other. Comparison of satellite data with the airborne data and the (scarcely available) in situ data reveal underlying problems with: (i) validation of remote sensing images; and (ii) comparing data at different spatial and temporal scales. In our study, we found a reasonable agreement between different data sources at seasonal time scales, but at shorter time scales the differences can be (much) larger. In situ data suffer from poor reproducibility, related to the natural variability at small spatial scales (patchiness), combined with a significant temporal variability. The standard in situ monitoring program in Lakes IJssel and Marken lacks both the necessary spatial coverage as well as an appropriate sampling frequency. This indicates that for reliable monitoring, a synoptic data set, sampled at a high frequency is required. Remote sensing can partially fulfil this demand but still lacks the demanded frequency, mainly due to regular cloud cover. The answer may be in a multiplatform monitoring approach, as used in our study (combining in situ data with shipboard, airborne and satellite optical data) and in combining monitoring data with models. Satellite remote sensing is most powerful in determining properties that are inherent to the whole lake system, like the overall mean chlorophyll- a concentration. Computational models may meet the demand for a sufficiently high sampling frequency by deterministic interpolation of the data in time.

Hyperspectral identification of grassland vegetation in Xilinhot, Inner Mongolia, China

Derivative spectroscopy of hyperspectral reflectance data of grassland vegetation was used to extract the wavelength peaks of different grass species of Xilinhot, Inner Mongolia, China. Caragana microphylla, which is characteristic of dry areas, was successfully detected among the four dominant grass species found in the typical Xilinhot grassland, using fourth-derivative peaks around 670 nm and 720 nm. A radiative transfer simulation showed that the identification scheme was not distorted by the atmosphere during airborne remote sensing. This suggests that high-order derivative spectroscopy of hyperspectral reflectance data from grassland vegetation can be an effective tool for advanced mapping and for the monitoring of grasslands.

Quantifying the Optical Properties and Chromophore Concentrations of Turbid Media by Chemometric Analysis of Hyperspectral Diffuse Reflectance Data Collected Using a Fourier Interferometric Imaging System

A non-contact Fourier transform interferometric imaging system was used to collect hyperspectral images of the steady-state diffuse reflectance from a point source in turbid media for the spectral range of 550–850 nm. Steady-state diffuse reflectance profiles were generated from the hyperspectral images, and partial least-squares (PLS) regression was performed on the diffuse reflectance profiles to quantify absorption (μ a) and reduced scattering (μ' s) properties of turbid media. The feasibility of using PLS regression to predict optical properties was examined for two different sets of spatially-resolved diffuse reflectance data. One set of data was collected from 40 turbid phantoms, while the second set was generated by convolving Monte Carlo simulations with the instrument response of the imaging system. Study results show that PLS prediction of μ a and μ' s was accurate to within ±8% and ±5%, respectively, when the model was trained on turbid phantom data. Moreover, PLS prediction of optical properties was considerably faster and more efficient than direct least-squares fitting of spatially-resolved profiles. When the PLS model was trained on Monte Carlo simulated data and subsequently used to predict μ a and μ' s from the diffuse reflectance of turbid phantom, the percent accuracies degraded to ± 12% and ± 5%, respectively. These accuracy values are applicable to homogenous, semi-infinite turbid phantoms with optical property ranges comparable to tissues.

Retrieval of sea water optically active parameters from hyperspectral data by means of generalized radial basis function neural networks

The authors present a new methodology for estimating the concentration of sea water optically active constituents from remotely sensed hyperspectral data, based on generalized radial basis function neural networks (GRBF-NNs). This family of NNs is particularly suited to approximate relationships like those between hyperspectral reflectance data and the concentrations of optically active constituents of the water body, which are highly nonlinear, especially in case II waters. Three main water constituents are taken into account: phytoplankton, nonchlorophyllous particles, and yellow substance. Each parameter is estimated by means of a specific multi-input single-output GRBF-NN. The authors adopt a recently proposed network learning strategy based on the combined use of the regression tree procedure and forward selection. The effectiveness of this approach, which is completely general and can be easily applied to any hyperspectral sensor, is proved using data simulated with an ocean color model over the channels of the medium resolution imaging spectrometer (MERIS), the new generation ESA sensor to be launched in 2001. The authors define the estimation algorithms over waters of cases I, II, and I+II and compare their performance with that of classical band-ratio, single-band, and multilinear algorithms. Generally, the GRBF-NN algorithms outperform the classical ones, except for the multilinear over case I waters. A particular improvement Is over case II waters, where the mean square error (MSE) can be reduced by one or two orders of magnitude over the error of multilinear and band-ratio algorithms, respectively.

Comparing prediction power and stability of broadband and hyperspectral vegetation indices for estimation of green leaf area index and canopy chlorophyll density

Hyperspectral reflectance data representing a wide range of canopies were simulated using the combined PROSPECT+SAIL model. The simulations were used to study the stability of recently proposed vegetation indices (VIs) derived from adjacent narrowband spectral reflectance data across the visible (VIS) and near infrared (NIR) region of the electromagnetic spectrum. The prediction power of these indices with respect to green leaf area index (LAI) and canopy chlorophyll density (CCD) was compared, and their sensitivity to canopy architecture, illumination geometry, soil background reflectance, and atmospheric conditions were analyzed. The second soil-adjusted vegetation index (SAVI2) proved to be the best overall choice as a greenness measure. However, it is also shown that the dynamics of the VIs are very different in terms of their sensitivity to the different external factors that affects the spectral reflectance signatures of the various modeled canopies. It is concluded that hyperspectral indices are not necessarily better at predicting LAI and CCD, but that selection of a VI should depend upon (1) which parameter that needs to be estimated (LAI or CCD), (2) the expected range of this parameter, and (3) a priori knowledge of the variation of external parameters affecting the spectral reflectance of the canopy.

Conifer species recognition: Effects of data transformation

In situ hyperspectral data obtained with a high spectral resolution radiometer were analysed for identification of six conifer species. Hyperspectral data were measured in the summer and late fall seasons at 15-20 cm above portions of tree canopies from both the sunlit and shaded sides. An artificial neural network algorithm was applied for identification purposes. Six types of transformation were applied to the hyperspectral reflectance data ( R ), preprocessed with a simple smoothing, followed by band aggregation. These include log( R ), first derivative of R, first derivative of log( R ), normalized R, first derivative of normalized R, and log(normalized R ). First derivative of log( R ) and first derivative of normalized R resulted in best species recognition accuracies with greater than 90% average accuracies, more than 20% greater than the average accuracy obtained from the pre-processed hyperspectral data. The effect of hyperspectral data taken from the shade sides of tree canopies can be minimized by applying normalization or by taking the derivatives after applying a logarithm to the pre-processed data. We found that a big difference in solar angle did not cause a noticeable difference in accuracies of species recognition.

Deriving Water Content of Chaparral Vegetation from AVIRIS Data

Spectral imaging data acquired with Advanced Visible Infrared Imaging Spectrometer over Point Dume (Los Angeles County, CA, USA) were used to assess the ability of hyperspectral reflectance data to estimate canopy Relative Water Content (RWC) at the landscape level. The study was performed on 23 vegetation stands comprised of three characteristic chaparral plant communities, with contrasting phenological stages and canopy cover. Several estimates of water content based on the near-infrared (NIR; reflectance indices and water thickness derived from reflectance and radiance data) and shortwave infrared (SWIR) water absorption bands were compared to measurements of vegetation structure and water content made on the ground. The Water Index (WI) and Normalized Difference Water Index (NDWI), reflectance indices formulated from the NIR water absorption bands, were the best indicators of canopy RWC estimated from combining leaf relative water content with measures of canopy structure. A stepwise multiple regression revealed that canopy structure explained 36% and 41% of the variation in WI and NDWI, respectively. The explained variance in WI and NDWI increased to 44% and 48% when leaf relative water content was included in the model. By contrast, the inclusion of leaf relative water content did not contribute significantly to the explained variance in indices formulated using SWIR water absorption bands and in those based on water thickness. The relationship between WI and the canopy RWC significantly improved when only data from plots with green vegetation cover >70% were considered ( r 2=0.88, p<0.001). All the indices studied had an important structural component (as indicated by the strong correlation with NDVI), yet only the indices WI and NDWI additionally responded to water content. These results indicate that the WI and NDWI are sensitive to variations in canopy relative water content at the landscape scale.

Spectrometry of pasture condition and biogeochemistry in the central Amazon

Regional analyses of Amazon cattle pasture biogeochemistry are difficult due to the complexity of human, edaphic, biotic and climatic factors and persistent cloud cover in satellite observations. We developed a method to estimate key biophysical properties of Amazon pastures using hyperspectral reflectance data and photon transport inverse modeling. Remote estimates of live and senescent biomass were strongly correlated with plant‐available forms of soil phosphorus and calcium. These results provide a basis for monitoring pasture condition and biogeochemistry in the Amazon Basin using spaceborne hyperspectral sensors.