Jennifer L. Dungan

Estimating the foliar biochemical concentration of leaves with reflectance spectrometry: Testing the Kokaly and Clark methodologies

In an effort to further develop the methods needed to remotely sense the biochemical concentration of plant canopies, we report the results of an experiment to estimate the concentration of 12 foliar biochemicals (chlorophyll a, chlorophyll b, total chlorophyll, lignin, nitrogen, cellulose, water, phosphorous, protein, amino acids, sugar, starch) from reflectance spectra of dried and ground slash pine needles. The three methodologies employed used stepwise regression and either of the following: (i) standard first derivative reflectance spectra (FDS), (ii) absorption band depths, following continuum removal and normalisation against band depth at the centre of the absorption feature (BNC) or (iii) absorption band depths, following continuum removal and normalisation against the area of the absorption feature (BNA). These latter two methodologies have been proposed in this journal [Remote Sens. Environ., 67 (1999) 267.] on the basis of an experiment using reflectance spectra of dried and ground tree leaves and the concentration of three foliar biochemicals: nitrogen, lignin and cellulose. All three methodologies were implemented on a spectra/biochemical data set from early in the growing season and tested on a similar data set from late in the growing season. The accuracy with which foliar biochemical concentration could be estimated, while high for all methodologies, was highest when using the two proposed by Kokaly and Clark. At an illustrative R2 threshold of .85 (between estimated and observed biochemical concentration), all three methodologies could be used to estimate total chlorophyll, nitrogen, cellulose and sugar; in addition, the BNC methodology could be used to estimate chlorophyll a and b, and in addition to this, the BNA methodology could be used to estimate lignin and water. Given the advantages offered by the Kokaly and Clark methodologies (over and above the standard methodology) for a wide range of foliar biochemicals, it is recommended that their utility is investigated for the estimation of foliar biochemical concentration from field, airborne and spaceborne spectra.

Reflectance spectroscopy of fresh whole leaves for the estimation of chemical concentration

Abstract In an effort to further develop the methods needed to remotely sense the biochemical content of plant canopies, we report the results of an experiment to relate the concentrations of chlorophyll, protein, starch, sugar, amaranthin, and water in fresh whole leaves to their reflectance at wavelengths throughout the visible and near infrared. This is an analysis of laboratory data from a previously reported experiment (Curran et al., 1991) in which 163 freshly excised leaves of the plant Amaranthus tricolor were measured for reflectance and biochemical content. Stepwise regression was used to generate an equation for the estimation of chemical concentration from derivative reflectance in selected wavelengths. The reduction of instrument noise through Fourier filtering and a sample control procedure to minimize spectral overlap had little effect on the correlation between derivative reflectance in selected wavelengths and chemical concentration but did enable absorption features attributable to sugar and protein to be detected. However, the minimization of spectral overlap did increase the number of wavelengths attributable to known absorption features that were selected by stepwise procedures. Using only derivative reflectance in wavelengths that were attributable to absorption by the chemical of interest, the coefficients of determination (R 2 ) between estimated and measured concentrations of chlorophyll, amaranthin, starch, and water were 0.82 or above, with root-mean square errors that were 12.5% of the median or less.

Forest ecosystem processes at the watershed scale: basis for distributed simulation

Abstract A framework is described to compute and map forest evapotranspiration and net primary productivity over complex mountainous terrain. The methodology is based on the interface of geographic information processing and remote sensing with FOREST-BGC, a nonlinear deterministic model designed to simulate carbon, water and nitrogen cycles in a forest ecosystem. The model as input the geographic patterns of leaf area index ( lai ), available soil water capacity ( swc ) and microclimatic parameters over the landscape. These patterns are represented with the use of a template consisting of the set of hillslopes, stream channels and subwatersheds that completely define the landscape. A geo-referenced database containing digital elevation data, remotely sensed information and other environmental data are stratified by this template. We have found that the stratification of the surface data sets by a hillslope or watershed template produces landscape units with low internal variance of the important model parameters but high between unit variance. By producing templates at different levels of resolution, we have the ability to reorganize the model parameter set to different levels of surface generalization. The model is directly parameterized for each of these surface units which can then be simulated in parallel, providing the ability to expand the simulation to large regions.

The effect of a red leaf pigment on the relationship between red edge and chlorophyll concentration

Abstract The point of maximum slope in a reflectance spectrum of vegetation occurs at the boundary between red and near infrared wavelengths and is known as the “red edge”. There is a strong relationship between the red edge and the chlorophyll concentration of leaves and canopies. The aim of this research was to determine the effect of a second leaf pigment, red amaranthin, on the relationship between red edge and chlorophyll concentration. The red edge, chlorophyll concentration, and amaranthin concentration were recorded for 163 amaranth leaves in the laboratory. Experimental treatments with nitrate and salts caused a very large range in red edge (686–724 nm), chlorophyll concentration (0–20 mg/g), and amaranthin concentration (0–0.47 mg/g). There was a near-linear relationship between red edge and chlorophyll concentration for leaves with low amaranthin concentration (

Seasonal LAI in slash pine estimated with LANDSAT TM

The leaf area index (LAI, total area of leaves per unit area of ground) of most forest canopies varies throughout the year, yet for logistical reasons it is difficult to estimate anything more detailed than an annual average LAI. To determine if remotely sensed data can be used to estimate LAI at times throughout the year (herein termed seasonal LAI), field measurements of LAI were compared to normalized difference vegetation index (NDVI) values, derived using Landsat Thematic Mapper (TM) data, for 16 fertilized and control slash pine plots on three dates. Linear relationships existed between NDVI and LAI with R2 values of 0.35, 0.75, and 0.86 for February 1988, September 1988and March 1989, respectively. Predictive relationships based on data from eight of the plots were used to estimate the LAI of the other eight plots with a root-mean-square error of 0.74 LAI, which is 15.6% of the mean LAI. This demonstrates the potential use of Landsat TM data for studying seasonal dynamics in forest canopies.

Estimation of signal-to-noise: a new procedure applied to AVIRIS data

To make the best use of narrowband Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data, an investigator needs to know the signal-to-noise ratio (SNR). The signal is land cover dependent and varies with both wavelength and atmospheric absorption, and random noise comprises sensor noise and intrapixel variability (i.e. variability within a pixel). The three existing methods for estimating the SNR are inadequate, since typical laboratory methods inflate, while typical dark-current and image methods deflate the SNR value. The authors propose a procedure called the geostatistical method that is based on the removal of periodic noise by notch filtering in the frequency domain and the isolation of sensor noise and intrapixel variability using the semivariogram. This procedure was applied easily and successfully to five sets of AVIRIS data from the 1987 flying season and could be applied to remotely sensed data from broadband sensors. >

Kriging in the shadows: geostatistical interpolation for remote sensing

It is often useful to estimate obscured or missing remotely sensed data. Traditional interpolation methods, such as nearest-neighbor or bilinear resampling, do not take full advantage of the spatial information in the image. An alternative method, a geostatistical technique known as indicator kriging, is described and demonstrated using a Landsat Thematic Mapper image in southern Chiapas, Mexico. The image was first classified into pasture and nonpasture land cover. For each pixel that was obscured by cloud or cloud shadow, the probability that it was pasture was assigned by the algorithm. An exponential omnidirectional variogram model was used to characterize the spatial continuity of the image for use in the kriging algorithm. Assuming a cutoff probability level of 50%, the error was shown to be 17% with no obvious spatial bias but with some tendency to categorize nonpasture as pasture (overestimation). While this is a promising result, the methods practical application in other missing data problems for remotely sensed images will depend on the amount and spatial pattern of the unobscured pixels and missing pixels and the success of the spatial continuity model used.

Forest variable estimation from fusion of SAR and multispectral optical data

Radar and optical remote sensing data are used in a unified algorithm to estimate forest variables. The study site is the H. J. Andrews experimental forest in Oregon, which has significant topography and several mature and old-growth conifer stands with biomass values sometimes exceeding 1000 tons/ha. Polarimetric multifrequency Airborne Synthetic Aperture Radar (AIRSAR) backscatter, interferometric C-band Topographic Synthetic Aperture Radar (TOPSAR) coherence, and multispectral Landsat Thematic Mapper (TM) digital numbers are used in a regression analysis that relates them to forest variable measurements on the ground. Parametric expressions are derived and used to estimate the same variables(s) at other locations from the combination of AIRSAR and TM data. It is shown that the estimation accuracy is significantly improved when the radar and optical data are used in combination compared to estimating the same variable from a single data type alone.

Modeling seasonal and interannual variability in ecosystem carbon cycling for the Brazilian Amazon region

Previous field measurements have implied that undisturbed Amazon forests may represent a substantial terrestrial sink for atmospheric carbon dioxide. We investigated this hypothesis using a regional ecosystem model for net primary production (NPP) and soil biogeochemical cycling. Seasonal and interannual controls on net ecosystem production (NEP) were studied with integration of high-resolution (8-km) multiyear satellite data to characterize Amazon land surface properties over time. Background analysis of temporal and spatial relationships between regional rainfall patterns and satellite observations (for vegetation land cover, fire counts, and smoke aerosol effects) reveals several notable patterns in the model driver data. Autocorrelation analysis for monthly vegetation “greenness” index (normalized difference vegetation index, NDVI) from the advanced very high resolution radiometer (AVHRR) and monthly rainfall indicates a significant lag time correlation of up to 12 months. At lag times approaching 36 months, autocorrelation function (ACF) values did not exceed the 95% confidence interval at locations west of about 47°W, which is near the transition zone of seasonal tropical forest and other (nonforest) vegetation types. Even at lag times of 12 months or less, the location near Manaus (approximately 60°W) represents the farthest western point in the Amazon region where seasonality of rainfall accounts significantly for monthly variations in forest phenology, as observed using NDVI. Comparisons of NDVI seasonal profiles in areas of the eastern Amazon widely affected by fires (as observed from satellite) suggest that our adjusted AVHRR-NDVI captures year-to-year variation in land cover greenness with minimal interference from small fires and smoke aerosols. Ecosystem model results using this newly generated combination of regional forcing data from satellite suggest that undisturbed Amazon forests can be strong net sinks for atmospheric carbon dioxide, particularly during wet (non El Nino) years. However, drought effects during El Nino years can reduce NPP in primary forests of the eastern Amazon by 10–20%, compared to long-term average estimates of regional productivity. Annual NEP for the region is predicted to range from −0.4 Pg C yr−1 (net CO2 source) to 0.5 Pg C yr−1 (net CO2 sink), with large interannual variability over the states of Para, Maranhao, and Amazonas. As in the case of predicted NPP, it appears that periods of relatively high solar surface irradiance combined with several months of adequate rainfall are required to sustain the forest carbon sink for positive yearly NEP estimates.

Visualizing spatially varying distribution data

Box plot is a compact representation that encodes the minimum, maximum, mean, median, and quartile information of a distribution. In practice, a single box plot is drawn for each variable of interest. With the advent of more accessible computing power, we are now facing the problem of visualizing data where there is a distribution at each 2D spatial location. Simply extending the box plot technique to distributions over 2D domain is not straightforward. One challenge is reducing the visual clutter if a box plot is drawn over each grid location in the 2D domain. This paper presents and discusses two general approaches, using parametric statistics and shape descriptors, to present 2D distribution data sets. Both approaches provide additional insights compared to the traditional box plot technique.