Christine A. Hlavka

Multivariate analysis of AVIRIS data for canopy biochemical estimation along the oregon transect

Abstract High spectral resolution data from the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) acquired over several forest stands in west-central Oregon were analyzed with respect to variations in forest canopy biochemical content (kg ha−1), foliar biochemical concentration (mg cm2 leaf area), and leaf area index (LAI). The Lowtran-7 atmospheric radiative transfer code was used to convert AVIRIS at-sensor radiance into reflectance factor. The correlation of reflectance factor and first-difference AVIRIS spectra with total nitrogen (TN), lignin, starch, total chlorophyll content (kg ha−1), and LAI is presented. Correlogram structure in the 1200–2400 nm region is compared with the occurrence of known absorption features associated with organic molecular bonds which comprise foliar biochemical constituents. Regression equations relating the first-difference AVIRIS data to chemical variables and LAI were developed. The chlorophyll red edge spectral region was strongly related to LAI, canopy TN content, and canopy chlorophyll content. Generally good correspondence was found between wavelength selections for TN and lignin (concentration and content) and overtones of fundamental CH and NH absorptance features. The AVIRIS data were not significantly correlated with either starch concentration or starch content, although wavelength selections for starch were consistent with those made in previous laboratory-based studies.

LEAFMOD: A New Within-Leaf Radiative Transfer Model

Abstract We describe the construction and verification of a within-leaf radiative transfer model called LEAFMOD (Leaf Experimental Absorptivity Feasibility MODel). In the model, the one-dimensional radiative transfer equation in a slab of leaf material with homogeneous optical properties is solved. When run in the forward mode, LEAFMOD generates an estimate of leaf reflectance and transmittance given the leaf thickness and optical characteristics of the leaf material (i.e., the absorption and scattering coefficients). In the inverse mode, LEAFMOD computes the total within-leaf absorption and scattering coefficient profiles from measured reflectance, transmittance, and leaf thickness. Inversions with simulated data demonstrate that the model appropriately decouples scattering and absorption within the leaf, producing fresh leaf absorption profiles with peaks at locations corresponding to the major absorption features for water and chlorophyll. Experiments with empirical input data demonstrate that the amplitude of the fresh leaf absorption coefficient profile in the visible wavebands is correlated with pigment concentrations as determined by wet chemical analyses, and that absorption features in the near-infrared wavebands related to various other biochemical constituents can be identified in a dry-leaf absorption profile.

LCM2: A coupled leaf/canopy radiative transfer model

Abstract Two radiative transfer models have been coupled to generate vegetation canopy reflectance as a function of leaf chemistry, leaf morphology (as represented by leaf scattering properties), leaf thickness, soil reflectance, and canopy architecture. A model of radiative transfer within a leaf, called LEAFMOD, treats the radiative transfer equation for a slab of optically uniform leaf material, providing an estimate of leaf hemispherical reflectance and transmittance as well as the radiance exiting the leaf surfaces. The canopy model then simulates radiative transfer within a mixture of leaves, with each having uniform optical properties as determined by LEAFMOD, assuming a bi-Lambertian leaf scattering phase function. The utility of the model, called LCM2 (Leaf/Canopy Model version 2), is demonstrated through predictions of radiometric measurements of canopy reflectance and sensitivity to leaf chlorophyll and moisture content.

Unmixing AVHRR imagery to assess clearcuts and forest regrowth in Oregon

Advanced Very High Resolution Radiometer imagery provides frequent and low-cost coverage of the Earth, but its coarse spatial resolution (1.1 km by 1.1 km) does not lend itself to standard techniques of automated categorization of land cover classes because the pixels are generally mixed; that is, the extent of the pixel includes several land use/cover classes. Unmixing procedures were developed to extract land use/cover class signatures from mixed pixels, using Landsat Thematic Mapper data as a source for the training set, and to estimate fractions of class coverage within pixels. Application of these unmixing procedures to mapping forest clearcuts and regrowth in Oregon indicated that unmixing is a promising approach for mapping major trends in land cover with AVHRR bands 1 and 2. Including thermal bands by unmixing AVHRR bands 1-4 did not lead to significant improvements in accuracy, but experiments with unmixing these four bands did indicate that use of weighted least squares techniques might lead to improvements in other applications of unmixing. >

Thematic Mapper Image Quality: Registration, Noise, And Resolution

This paper provides an assessment of Thematic Mapper data quality in terms of band-to-band registration, periodic noise, and spatial resolution. Based on the Thematic Mapper images analyzed so far, the band-to-band registration accuracy is very good. For bands within the same focal plane, the mean misregistrations are well within the specification, 0.2 pixel, except for the thermal band. The thermal band was misregistered by three pixels in each direction in early data products. The error in the across-scan direction was close to zero in later data products. For bands between the cooled and uncooled focal planes, there was a consistent mean misregistration of 0.5 pixels along-scan and 0.2-0.3 pixels across-scan, larger than the specified 0.3 pixel error for bands between focal planes. An analysis of the standard deviation of the misregistration indicated all band combinations would meet the registration specifications if the mean misregistrations were removed by the data processing software. Analysis of the periodic noise in one image indicated a noise component in bands 1-4 with a spatial frequency of 0.31 cycles/pixel. Other lower amplitude periodic components were also present. The periodic noise components obscured detail in areas of low contrast. Modulation transfer function (MTF) analysis in a comparative mode showed no difference in MTF between the forward and backward scans. The difference in MTF between radiometrically corrected data and geometrically corrrected data appeared to be attributable largely to the cubic convolution resampling used to derive the geometrically corrected data.

Spectral-Temporal Classification Using Vegetation Phenology

In this paper we describe a multitemporal classification procedure for crops in Landsat scenes. The method involves the creation of crop signatures which characterize multispectral observations as functions of phenological growth states. The phenological signature models spectral reflectance explicitly as a function of crop maturity rather than a function of date. This means that instead of stacking spectral vectors of one observation on another, as is usually done for multitemporal data, for each possible crop category a correspondence of time to growth state is established which minimizes the smallest difference between the given multispectral multitemporal vector and the category mean vector indexed by growth state. The results of applying it to winter wheat show that the method is capable of discrimination with about the same degree of accuracy as more traditional multitemporal classifiers. It shows some potential to label degree of maturity of the crop without crop condition information in the training set.

Land-Use Mapping Using Edge Density Texture Measures on Thematic Mapper Simulator Data

Texture analysis was performed as part of an investigation of the information content of Landsat Thematic Mapper (TM) imagery. High-altitude aircraft scanner imagery from the Airborne Thematic Mapper (ATM) instrument was acquired over central California and used to simulate TM data. Edge density texture images were constructed by computation of proportions of edge pixels in a 31×31 moving window on a near-infrared ATM band. A training technique was employed to select computational parameters to maximize the difference between edge density measurements in urban and in rural areas. The results of classification of the texture images showed that urban and rural areas could be distinguished with texture alone, indicating that inclusion of texture in automated classification procedures could significantly improve their accuracy.

Analysis Of Forest Disturbance Using TM and AVHRR Data

A methodology that will be used to determine the proportions of undisturbed, successional vegetation and recently disturbed land cover within coniferous forests using remotely sensed data from the advanced very high resolution radiometer (AVHRR) is presented. The method uses thematic mapper (TM) data to determine the proportions of the three stages of forest disturbance and regrowth for each AVHRR pixel in the sample areas, and is then applied to interpret all AVHRR imagery. Preliminary results indicate that there are predictable relationships between TM spectral response and the disturbance classes. Analysis of ellipse plots from a TM classification of the disturbed forested landscape indicates that the forest classes are separable in the red (0.63-0.69 micron) and near-infrared (0.76-0.90 micron) bands, providing evidence that the proportion of disturbance classes may be determined from AVHRR data.

Analysis of forest foliage spectra using a multivariate mixture model

Wet chemical measurements and near infrared spectra of dry ground leaf samples were analysed to test a multivariate regression technique for estimating component spectra. The technique is based on a linear mixture model for log(1/R) pseudoabsorbance derived from diffuse reflectance measurements. The resulting unmixed spectra for carbohydrates, lignin and protein resemble the spectra of extracted plant carbohydrates, lignin and protein. The unmixed protein spectrum has prominent absorption peaks at wavelengths that have been associated with nitrogen bonds. It therefore appears feasible to incorporate the linear mixture model in whole leaf models of photon absorption and scattering so that effects of varying nitrogen and carbon concentration on leaf reflectance may be simulated.

Application of Geostatistical Simulation to Enhance Satellite Image Products

With the deployment of Earth Observing System (EOS) satellites that provide daily, global imagery, there is increasing interest in defining the limitations of the data and derived products due to its coarse spatial resolution. Much of the detail, i.e. small fragments and notches in boundaries, is lost with coarse resolution imagery such as the EOS MODerate-Resolution Imaging Spectroradiometer (MODIS) data. Higher spatial resolution data such as the EOS Advanced Spaceborn Thermal Emission and Reflection Radiometer (ASTER), Landsat and airborne sensor imagery provide more detailed information but are less frequently available. There are, however, both theoretical and analytical evidence that burn scars and other fragmented types of land covers form self-similar or self-affine patterns, that is, patterns that look similar when viewed at widely differing spatial scales. Therefore small features of the patterns should be predictable, at least in a statistical sense, with knowledge about the large features. Recent developments in fractal modeling for characterizing the spatial distribution of undiscovered petroleum deposits are thus applicable to generating simulations of finer resolution satellite image products. We will present example EOS products, analysis to investigate self-similarity, and simulation results.