Paul J. Curran

Remote sensing of foliar chemistry

Abstract Remotely sensed data are being used to estimate foliar chemical content as a result of our need for the information and our increasing ability to understand and measure foliar spectra. This paper reviews how stepwise multiple regression and deconvolution have been used to extract chemical information from foliar spectra, and concludes that both methods are useful, but neither is ideal. It is recommended that the focus of research be modeling in the long term and experimentation in the short term. Long-term research should increase our understanding of the interaction between radiation and foliar chemistry so that the focus of research can move from leaf model to canopy model to field experiment. Short-term research should aim to design experiments in which remotely sensed data are used to generate unambiguous and accurate estimates of foliar chemical content.

The MERIS terrestrial chlorophyll index

The long wavelength edge of the major chlorophyll absorption feature in the spectrum of a vegetation canopy moves to longer wavelengths with an increase in chlorophyll content. The position of this red-edge has been used successfully to estimate, by remote sensing, the chlorophyll content of vegetation canopies. Techniques used to estimate this red-edge position (REP) have been designed for use on small volumes of continuous spectral data rather than the large volumes of discontinuous spectral data recorded by contemporary satellite spectrometers. Also, each technique produces a different value of REP from the same spectral data and REP values are relatively insensitive to chlorophyll content at high values of chlorophyll content. This paper reports on the design and indirect evaluation of a surrogate REP index for use with spectral data recorded at the standard band settings of the Medium Resolution Imaging Spectrometer (MERIS). This index, termed the MERIS terrestrial chlorophyll index (MTCI), was evaluated using model spectra, field spectra and MERIS data. It was easy to calculate (and so can be automated), was correlated strongly with REP but unlike REP was sensitive to high values of chlorophyll content. As a result this index became an official MERIS level-2 product of the European Space Agency in March 2004. Further direct evaluation of the MTCI is proposed, using both greenhouse and field data.

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.

A new technique for interpolating the reflectance red edge position

The point of maximum slope on the reflectance spectrum of vegetation between red and near-infrared wavelengths, termed the red edge position (REP), is correlated strongly with foliar chlorophyll content and provides a very sensitive indicator of, among other things, vegetation stress. The high spectral resolution of airborne imaging spectrometers now offers the potential for determining the REP of vegetation canopies at regional scales. However, the accurate estimation of the REP is dependent upon sensor band positions and widths. Various techniques have been developed to minimize the error in estimating the REP, such as linear interpolation or inverted Gaussian curve fitting in the region of the red edge which requires an a priori knowledge of the spectrum under investigation. This technical note presents a simple technique known as Lagrangian interpolation which is applied to the first-derivative transformation of the reflectance spectrum. The technique fits a second-order polynomial curve to three band...

LIBERTY—Modeling the Effects of Leaf Biochemical Concentration on Reflectance Spectra

Abstract The conifer leaf model LIBERTY (Leaf Incorporating Biochemistry Exhibiting Reflectance and Transmittance Yields) is an adaptation of radiative transfer theory for determining the optical properties of powders. LIBERTY provides a simulation, at a fine spectral resolution, of quasiinfinite leaf reflectance (as represented by stacked leaves) and single leaf reflectance. Single leaf reflectance and transmittance are important input variables to vegetation canopy reflectance models. A prototype parameterization of LIBERTY was based upon measurements of pine needles and known absorption coefficients of pure component leaf biochemicals. The estimated infinite-reflectance output was compared with the spectra of both dried and fresh pine needles with root mean square errors (RMSE) of 2.87% and 1.73%, respectively. The comparisons between measured and estimated reflectance and transmittance values for single needles were also very accurate with RSME of 1.84% and 1.12%, respectively. Initial inversion studies have demonstrated that significant improvements can be made to LIBERTY by utilizing in vivo absorption coefficients which have been determined by the inversion process. These results demonstrate the capability of LIBERTY to model accurately the spectral response of pine needles.

Geostatistics and remote sensing

In geostatistics, spatial autocorrelation is utilized to estimate optimally local values from data sampled elsewhere. The powerful synergy between geostatistics and remote sensing went unrealized until the 1980s. Today geostatistics are used to explore and describe spatial variation in remotely sensed and ground data; to design optimum sampling schemes for image data and ground data; and to increase the accuracy with which remotely sensed data can be used to classify land cover or estimate continuous variables. This article introduces these applications and uses two examples to highlight characteristics that are common to them all. The article concludes with a discussion of conditional simulation as a novel geostatistical technique for use in remote sensing.

DYNAMICS OF CANOPY STRUCTURE AND LIGHT INTERCEPTION IN PINUS ELLIOTTII STANDS, NORTH FLORIDA'

In order to develop a model of the carbon cycle for mature slash pine (Pinus elliottii) stands in north Florida, we studied seasonal variation in leaf area index (LAI, all- sided), aboveground biomass increment and litterfall, and light penetration through the forest canopy, over a 3-yr period. The primary approach to establishing monthly LAI included annual destructive analyses and monthly measurements of needle fall and elon- gation. Imagery from the Landsat Thematic Mapper (TM) and patterns of light penetration were also used in attempts to derive less arduous estimates; the TM imagery was most promising. LAIs ranged from 3.0 to 6.5 on control plots over the 3 yr, with repeated fertilization increasing maximum LAI by >40%. Seasonal variation was high (40%), as was variation from year to year. An average of 3 1% of the incident photosynthetically active radiation (PAR) penetrated the canopies annually, ranging from 18 to 42% seasonally. Seasonal light penetration could not be described using a simple application of the Beer-Lambert law, perhaps due to the highly aggregated nature of the canopies. Models incorporating more information on canopy structure are necessary to predict light penetration through slash pine stands accurately. A model of needle litterfall was derived that could account for much of the seasonal and annual variation using stand basal area and climate conditions from the spring of the previous year; this model may be useful for developing climate-driven predictions of LAI. Efficiencies of use of incoming and intercepted PAR were low compared to other forest types. Low light interception and high nutrient-use efficiencies (demonstrated in earlier studies) are important adaptive characteristics of slash pine stands to these relatively warm and nutrient-poor sites.

Identifying terrestrial carbon sinks: Classification of successional stages in regenerating tropical forest from Landsat TM data

Abstract Remote sensing has generally been used to study the role of tropical forests as a source of atmospheric carbon, primarily through land-use change, such as deforestation, and biomass burning. Regeneration of forest on previously cleared areas, however, is a significant carbon sink . The strength of this carbon sink is dependent on the age and composition of the regenerating forest. The ability to identify regenerating forest classes that may differ in terms of carbon sink strength was investigated with Landsat TM data of a test site near Manaus, Brazil. A number of forest age classes were defined from a time series of Landsat sensor data, and their separability in Landsat TM data was assessed by maximum likelihood classifications. A high level of class separability was observed with a weighted kappa coefficient of 0.8569 obtained for a classification of six forest regeneration classes. Of the classification errors observed most were found to be associated with the youngest forest age class. At the test site, however, two main successional pathways were followed and the differences between areas of forest of the same age but on different pathways was most apparent with the youngest forests. Splitting the regenerating forests by the successional pathway was found to increase classification accuracy, with a weighted kappa coefficient of 0.9315 observed for an 11 class classification. A range of tropical forest classes that vary in strength as a carbon sink could therefore be identified accurately from Landsat TM data. Although the broader generality of the results requires further investigation, this indicates the potential to use image classifications to scale-up point measurements of the carbon flux between regenerating forest classes and the atmosphere over large areas.

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.

The integration of spectral and textural information using neural networks for land cover mapping in the Mediterranean

The aim of this study was to develop an efficient and accurate procedure for classifying Mediterranean land cover with remotely sensed data. Combinations of artificial neural networks (ANN) and texture analysis on a per-field basis were used to classify a Landsat Thematic Mapper image of the Cukurova Deltas, Turkey, into eight land cover classes. This study integrated spectral information with measures of texture, in the form of the variance and the variogram. The accuracy of the ANN was greater than that of maximum likelihood (ML) when using spectral data alone and when using spectral and textural data. The use of texture measures through the per-pixel and per-field majority rule approaches were found to reduce classification accuracy because the field boundaries were enlarged and so overwhelmed the measures of texture. In contrast, the per-field approach (where the field was specified prior to analysis) combined with texture information increased significantly classification accuracy. However, the accuracy decreased as the variogram lag increased. The accuracy with which land cover could be classified in this region was maximised at 89% by using a per-field, ANN approach in which semivariance at a lag of 1 pixel was incorporated as textural information. This is 15% greater than the accuracy achieved using a standard per-pixel ML classification. The primary limitation of the use of the per-field approach was noted to be the need for prior knowledge of field boundaries which may be resolved using existing data or through some form of edge-detection routine.