Effects of Ozone Exposures on the Spectra of Foliage of Loblolly Pine Measured at the University of Georgia Experimental Forest

Brian Curtiss
CIRES/CSES University of Colorado
Boulder, CO 80309
Susan L. Ustin
Dept. of Land, Air, and Water Resources
University of California, Davis, CA 95616

Address for correspondence:
Susan L. Ustin
Department of Land, Air, and Water Resources
University of California
Davis, CA 95616
(530) 752-0621

Abstract

Spectral characteristics were measured for loblolly pine (Pinus taeda) foliage from mature trees exposed to ozone fumigation for one growing season. Branches of mature trees were exposed to ozone concentrations ranging from charcoal filtered to 2.5 times the ambient concentration. In addition, some of the trees were irrigated to avoid soil moisture stress. Reflectance was monitored using a nondestructive field procedure. Foliage was divided into age classes including two flushes for the current year and the second flush of the previous year. No spectral differences were observed between trees receiving supplemental irrigation and trees grown under field conditions. Abundant rainfall during the 1989 measurement season may have contributed to this observation. Principal component analysis was used to analyze the effect of ozone treatment. Spectral changes caused by needle aging were about three times the effect measured for ozone treatments. At the time of measurement, branches had been exposed to one season of ozone exposure (about three months duration). It is possible that chronic exposure may alter this pattern. Although the effects measured were small, they were consistent with the growth and physiological observations made at the time of measurement and they indicated the potential for increased damage in future years.

I. Introduction

The early detection and monitoring of pollutant stress on forested ecosystems is of considerable global significance. Efficient monitoring of forests at the regional and global scale will require the use of satellite and aircraft remote sensing because of the complexity of the processes and the spatial scales involved. The use of satellite or aircraft sensor technology for mapping the aerial extent of forest disturbance has been well documented. However, development of methods for the early detection of pollutant injury are less well developed. In part, this limitation has been due to the inability of relatively coarse spectral and spatial resolution scanners to detect the subtle changes in ecosystem processes and functioning that are evident before mortality and consequent changes in community structure occur. The recent development of high spectral resolution field instruments make such measurements feasible under environmental conditions. Spectroscopic assays are the primary analytical method currently used in laboratory research for identification and quantification of plant pigments, nitrogen, lignin, cellulose, and other biochemical components of leaves. Direct spectral assessment of canopy biochemical properties may lead to process-directed monitoring of forest ecosystems. It is now possible to acquire high resolution spectra non-destructively outside a laboratory and several instruments are now available for field-based or spaceborne research. Reflectance characteristics of forest canopies in the visible (500-700 nm) spectral region are dominated by photosynthetic pigment absorption, the near-infrared (800-1100 nm) by leaf structure, and in the shortwave-infrared (1550-1850 and 2000-2400 nm) by water content. Chlorophyll, nitrogen, lignin and cellulose have been detected using high spectral resolution sensors. Spectral changes produced by changes in leaf chemistry resulting from altered metabolic processes or carbon allocation may be used to infer both qualitative and quantitative information with which to assess forest condition.

This study was specifically formulated to examine the changes in the spectral reflectance of needles from mature Loblolly pine (Pinus taeda L.) trees in response to exposure to controlled levels of ozone. The exposure study was carried out at the University of Georgia by the research group headed by Dr. R. Teskey. Ozone treatment of mature tissue was achieved using branch exposure chambers or BECs (Figure 1). "Branches of mature trees growing in the field are contained inside chambers in which the ambient air can be controlled" (1). Figure 2 shows the layout of the study site. Six branches from each of six trees were exposed to varying levels of ozone within the BECs. Several reflectance spectra (usually 3) were measured for each needle age class, from up to two branch tips from each branch chamber. Because of the need to measure the spectra nondestructively, the branches were not clipped from the trees but measured in situ and were measured at night so that illumination could be controlled. Needles from each age class, while still attached to the branch, were attached to a frame so that the viewing and illumination geometry could be held constant between all spectra. This apparatus made it difficult to have access and measure the reflectance of branches in some of the chambers.

Ozone is a strong oxidant and produces several symptomatic physiological effects in coniferous forests. Ambient ozone concentrations of 40 to 70 ppb are common over much of the United states, concentrations well above the 10 to 40 ppb concentrations seen in clean air. There is considerable variation in the response of different genotypes of trees to ozone. Some genotypes of loblolly pine have shown response at concentrations as low as 50 ppb (2). Thus, ozone is suspected to be a contributing cause, and in many areas may be one of the leading causes, of forest damage in most of the world's industrialized countries.

Studies have indicated that cell membranes, including chloroplast membranes, suffer the most injury. The granulation of the chloroplast stroma are the first observed cellular anatomical changes following fumigation in beans. General disruption of chloroplast function, including loss of grana and thylakoid membranes, may proceed other anatomical cellular changes. Because chlorophyll is one of the major light absorbing compounds, these changes are directly observable in the reflectance spectra of the plant canopy.

The cellular damage that results from exposure to ozone or other pollutants result in an increase in the allocation of metabolic energy for repair. At some exposure threshold, the cellular damage becomes greater than the plants ability to respond through increased repair and maintenance respiration. At this stage it often becomes inefficient to maintain older tissue and senescence ensues. In the case of conifers, this results in the retention of fewer needle flushes. Once this threshold is exceeded, visible damage, in the form of chlorosis and foliar loss, becomes evident.

Numerous changes occur in photosynthetic pigment concentrations during stress. Losses of chlorophyll a and b often are not proportional and the chlorophyll a/b ratio tends to increase with stress. The ratio of chlorophyll a/b is generally greater in photosystem I than photosystem II. The damage produced as a result of ozone exposure disrupts the PS II - PS I pathway and causes increased fluorescence and photo-oxidation of PS II which may produce a shift in chlorophyll a/b ratios. Such effects have been noted in response to environmental stresses and to pollution effects (3,4). Another possible explanation for the increase in the chlorophyll a/b ratio is the large developmentally controlled difference between the a/b ratio for young and old foliar tissue. Repair associated with increased pollutant induced damage will result in a higher proportion of young foliage growth, thus the higher chlorophyll a/b ratio.

The concentration of accessory pigments also change under ozone exposure. The concentrations, and possibly the species, of carotenoids and xanthophylls, having absorption features in the 400-500 nm wavelength region, increase with stress. Increased accessory pigments provide a mechanism for protecting photosynthetic reaction centers from photo-oxidation. Although specific roles for accessory pigment shifts remains uncertain, Gamon et al. reported spectral changes due to a xanthophyll conversion under high irradiance conditions that are related to photosynthetic oxidation state(5). Such spectral changes may be important factors for monitoring physiological states in foliage and canopies. Coincident with accessory pigment changes is a corresponding linear increase in fluorescence from photosystem II as energy for carbon fixation is lost. Such changes indicate the extent of altered light harvesting capability in chloroplasts and the affect on carbon assimilation. Gamon et al. found fluorescence quenching occurred when the xanthophyll cycle pigment was converted from violaxanthin to zeaxanthin. Quantification of such accessory pigment responses to pollutant stress could provide a means to link mechanistic carbon balance models to spectral measurements.

II Methods

Branch chambers were installed for ozone exposures at the beginning of the growing season 1988 in the mid-canopy of mature loblolly pine trees as part of the University of Georgia's research effort under the National Acid Precipitation Research Program (NAPAP). Branch chambers were located above scaffolding placed around the tree crowns about 20 ft above ground level (Figure 1). Access to the fumigation experiment located at the University of Georgia's experimental forest were provided by Dr. Robert Teskey and Dr. Phil Dorrety, School of Forest Resources, University of Georgia, Athens, Georgia.

Measurements required non-destructive spectral sampling of the foliage within the ozone exposure chambers. We needed constant lighting and sample geometry for measurements so we constructed a foliage sample holder attached to the terminus of the fiberoptic cable that also held a 1000 watt power-regulated lamp in constant orientation (Figure 3). The sample holder was U-shaped, approximately three inches square with Velcro strips along the sides. This permitted us to rapidly place the foliage across the field-of-view of the sensor, and to hold it in place, without damage during measurement. Measurements were made at night to reduce extraneous light that might have been reflected from adjacent surfaces within the chamber during the day, and to keep solar lighting conditions constant, while at the same time keeping the entire measurement apparatus small enough to fit easily into the ozone chambers without damage to the branches. The lamp was attached to the sample holder at a 25° angle between the sample and sensor. The sample was oriented nadir to the sensor at 10 cm from the terminus of the fiber optic probe producing a field of view about 1 cm2. Spectra were measured using a Personal Spectrometer II model Analytical Spectral Devices Spectrometer and recorded to disk using a Zenith laptop computer. Data were acquired during the nights between Aug. 2-5,1989.

Spectral measurements were made on foliage from accessible branches within the chambers. Foliage was divided into age classes, with the most recent growth (second flush, 1989) = age class 1; first flush, 1989 = age class 2; and second flush 1988 = age class 3. Not all chambers had 1988 foliage represented on the branches. Three replicate spectra of each age class from two branches per chamber were measured in each chamber unless otherwise noted. Foliage was oriented horizontally and held in place to fully cover the field of view. Different amounts of foliage were present, depending on treatment and the specific orientation of the branches within the chambers.

III Data Analysis

A. Summary of Reflectance Spectra

The plots of the mean normalized reflectance spectra of foliage by treatment and age class are listed in Appendix A. Initially, each spectral plot was examined to identify those spectra which should be excluded from further analysis due to measurement errors. Less than a dozen spectra were excluded and represent those spectra for which either the instrument malfunctioned or the sample was not properly aligned in the sample holder.

The foliage spectra from the trees were divided into two groups: irrigated and non-irrigated. Because of the level of rainfall during the summer, no significant physiological differences were observed in the two groups (R. Teskey, University of Georgia, pers. comm.). Based on an initial examination of the reflectance spectra, no significant spectral differences attributable to irrigation treatment were observed either. For this reason, the spectra from all trees were combined for data analyses presented here.

Because of the difficulty of nondestructively measuring the reflectance of individual whorls, the amount of foliage within the field-of-view varies between samples independent of the treatment. While much of the total foliage biomass variability is associated with reduced needle retention in the older whorls, there was also random variance associated with the measurement procedure. Because this random variance would obscure the variance associated with ozone exposure, the data was normalized to constant reflectance in the region of 840 nanometers. The 840 nm region was selected because both chlorophyll and water have small absorption coefficients at this wavelength. These normalized mean spectral plots are shown in Appendix A.

The spectra in Appendix A were grouped both by ozone treatment and by whorl. The "by treatment" plot shown in Figure 4 for the high ozone exposures (2.5 times ambient) clearly shows increased reflectance in the chlorophyll absorption region (650-700 nm) with increasing needle age. This increased reflectance, or decreased absorptance, can be attributed to foliar chlorophyll loss as the needles age.

The relationships seen in the "by Whorl" (i.e., by age-class) plots are less clear. The "by Whorl" youngest needle-age averages (whorl 1) for all trees is plotted in Figures 5 and 6. While increases in reflectance with ozone treatment level are expected, this relationship is not consistently observed in all cases (see Table 1). In particular, the non-filtered air treatment (NFA) does not fit the trend defined by the other exposure treatments: it has higher reflectance across the spectrum than is expected for nonozone exposed foliage. This albedo offset possibly relates to differences between chamber conditions and ambient air conditions. For this reason, the non-filtered air treatment has not been included in further analysis and the carbon-filtered air treatments were used for comparisons with the ozone treatments.
 

B. Band Analysis

Six spectral bands plus a normalized difference vegetation index (NDVI) I) were selected to determine if, through multiple linear regression techniques, the ozone exposure level could be predicted by reflectance from those bands. The bands were selected to best describe the observed variability for the spectrum and are listed in Table 2. The NDVI is a frequently used spectral measure that enhances the vegetation characteristics in spectra and is useful because the magnitude of the ratio is thought to correlate with the absorption of solar energy in the region of photosynthetically active radiation. The correlation of these bands with the level of ozone exposure is listed in Table 3. From inspection of this table, it can be seen that the reflectance spectra of the first and third whorls are most highly correlated to ozone exposure level. The branches exposed to the higher levels of ozone tend to retain fewer needles in the older age classes (i.e., previous years needles). Because the higher exposure classes have fewer old needles to maintain, the low correlation of the second whorl to ozone exposure may be due to increased pigment production in needles grown in the higher ozone exposure chambers. This suggests some compensation for increased photosynthetic capacity or efficiency in exposed needles, as might occur if nitrogen and other mobile essential nutrients were translocated from senescent growth to new growth. This type of foliar adaptive mechanism has been observed in ozone chamber experiments in Ponderosa pine (John Hom, USDA Forest Service, pers. comm.). The values for these bandpasses was used to predict the ozone exposure levels (CFA=0, 1, 1.5, 2 and 2.5) using multiple linear regression. The correlations for these regressions range form 0.44 (second whorl) to 0.72 (third whorl). The predicted verses measured reflectance in ozone exposures are shown in Figures 7, 8, and 9.
 

C. Principal Component Analysis

The reflectance data was further analyzed using principal component analysis to determine if spectral changes or trends associated with ozone exposure could be identified. Prior to PCA analysis, the mean of the normalized dataset was subtracted from each spectrum. This is done so as to reduce the probability that the direction of the first eigenvector would be related to albedo, generally the largest source of variation in spectral data. Because of the measurement requirements, variability in albedo due to varying amounts of biomass within the field-of-view of the spectrometer could not be assumed to be due to the treatments. The use of nonmean corrected data results in a first eigenvector whose direction is determined by the mean of the data. Because subsequent eigenvectors are constrained to be orthogonal to the first eigenvector, they may not accurately represent the spectral trends that exist in the data. In some cases, trends within the data volume are not entirely independent and may have biologically significant non-orthagonal correlations.

While the first eigenvector from the analysis of mean corrected data is comparable to the second eigenvector from the analysis of non-mean corrected data, it is not necessarily identical. Eigenvalues in the analysis of mean corrected data represent the variance about the mean, rather than about the origin, their relative values give a clear indication of the importance of the spectral trends defined by the eigenvectors. As before, the data was analyzed using two different groupings: by treatment and by needle age. The analysis of the data grouped by treatment reveals spectral changes associated with increasing needle age. While the analysis of the data grouped by needle age was used to identify spectral changes associated with the various ozone treatments.

For all but the "non-filtered air" (NFA) treatment, the first principal component of the "by treatment" grouped data accounts for more than 90% of the variance (see Table 4). As was discussed earlier, the data for the NFA treatment does not follow the trends defined from the other treatments. The results from the NFA treatment were not used in further analysis and were not included in the "by needle age class" grouped analysis.

The dominant spectral trends defined in the PCA "by ozone treatment" is plotted in Figure 10 and can be primarily attributed to changes in needle chlorophyll concentration. As chlorophyll absorption increases (reflectance decreases) the features get broader. This type of band broadening (over the region between 550-750 nary), occurs with increased chlorophyll absorption, that is with increased band depth, and is consistent with a pattern of chlorophyll concentration increasing with needle age (this study included only three age classes: the two whorls from the current seasons growth and the second whorl from last years growth). Chlorosis generally produces a decrease in chlorophyll absorption well depth and a narrowing of the chlorophyll well, but in this case, it produces a more rapid loss of absorption in the 560 to 640nm portion of the chlorophyll absorption well than at 680nm and longer wavelengths. This asymmetry appears to result from chloroplast damage or alterations that accompanies the chlorophyll loss. To visualize the changes in reflectance properties as foliage ages we show the mean spectrum modified by adding or subtracting the first principal component for age classes one and three. In general, the reflectance spectra from older needles have a spectrum which corresponds to the mean reflectance spectrum plus the first eigenvector spectrum multiplied by a constant (the thin line in Figure 10). Those spectra from younger needles correspond to the mean spectrum minus the first eigenvector spectrum multiplied by a constant (the thick line in Figure 10). This presentation allows comparison of how foliar aging effects differ from the mean spectrum. Although the PCA modified spectra are similar, the effects of ozone treatment on foliage of differing ages is not identical. In general for a given ozone exposure, younger foliage exhibits less asymmetry in the 575 nm to 680 nm region than do older needles. The region about 550 nm shows the least discrimination in response to ozone exposure between foliage of differing ages.

The dominant spectral trends in the reflectance spectra grouped "by needle age" are plotted in Figure 11 and can be also attributed chlorosis and chloroplast membrane damage. In this case, the feature also gets broader, but this change occurs as chlorophyll absorption decreases (reflectance increases). This type of band broadening, coupled with a decrease in chlorophyll absorption, is consistent with a combination of decreasing chlorophyll concentration and chloroplast membrane damage. These spectral changes have been observed in previous studies of the effects of ozone on the spectra of Ponderosa pine (6) where the band broadening was attributed to membrane changes and/or chloroplast damage. For both the first and third whorl, the level of ozone exposure is highly correlated to the weighting of this first component. In general, The reflectance spectra of needles exposed to low levels of ozone have a spectrum that corresponds to the mean spectrum plus the first eigenvector spectrum multiplied by a constant (the thin line in Figure 11). Those spectra from needles exposed to high levels of ozone exhibit spectra that correspond to the mean spectrum minus the first eigenvector spectrum multiplied by a constant (the thick line in Figure 11).

The ozone exposure effects of greatest magnitude, as identified in the PCA, are in the wavelength regions centered around 680 nm and 550 nm. Foliage exposed to low ozone concentrations compared to foliage exposed to higher ozone concentrations, have a more steeply negative slope in the 550-680 nm chlorophyll absorption region compared to the flatter slope for foliage exposed to higher ozone concentrations, consistent with greater absorption of photosynthetically active radiation in foliage having less ozone exposure. Such patterns would result if ozone exposures in differing proportions of chlorophyll a and b, and changes in the pool sizes of carotein and xanthophyll accessory pigments. An additional difference is the apparent shift in the wavelength position of the "red edge," the inflection point for chlorophyll absorption on the 700 nm side of the feature. The shift in this inflection point to longer wavelengths has been attributed to higher chlorophyll concentrations (7), however, the effect of ozone in this study was to broaden the chlorophyll well at lower chlorophyll concentrations. Close inspection of Figures 10 and 11 reveal that the spectral trends defined by the first principal components are not the same between foliar aging and ozone exposure, despite the fact that both result in chlorosis due to loss of foliar chlorophyll, changes in relative proportions and pool sizes of other accessory pigments, and to membrane damage or other changes. This is an important observation as it provides a basis to separate both biologically important changes in plant spectra due to aging and ozone exposure.

IV. Conclusions

Changes in both needle phenology and ozone exposure produce significant dynamic changes in the reflectance characteristics of foliage measured at the branch level. Needle aging produces a change that is about three time the intensity of that produced by the range of ozone exposures used in this study. These exposures ranged from charcoal filtered air to 2.5 times the nominal ambient ozone concentration near Athens, Georgia for a duration of one growing season. the reflectance changes produced by needle aging resulted primarily from changes in needle chlorophyll and accessory pigment concentrations while ozone exposure produces spectral changes that are consistent with greater chloroplast damage. These changes, though small, are consistent with growth and physiological observations made later by the University of Georgia (R. Teskey, personal communication), who reports that their measurements did not show significant ozone treatment differences until the following growing season. However, the trends reported here are consistent with the patterns they subsequently observed. These types of spectral observations provide a basis for rapid early screening of foliar damage due to ozone exposure and such field techniques could possibly be extended to assess other air pollutants.

It is not clear whether the relatively small spectral effects of chloroplast alteration observed in this study on foliage could be seen as clearly at the canopy level. Nonetheless, trees exposed to high levels of ozone retain fewer whorls of needles than those growing in areas of lower ozone concentration. Because of the large effect of needle age on reflectance, this change in needle retention and mean canopy needle age, will assist in the detection of ozone exposure at the canopy scale.

Because of the limited range of spectral changes observed, and the fact that airborne sensors would be subject to greater levels of extraneous measurement error, it is unclear whether such canopy scale measurements could be successful. It is important to keep in mind that the responses reported here are from a single growing season of ozone exposure and the maximum dose, 2.5 times ambient, represents a relatively moderate exposure. Under natural conditions cumulative effects may be greater, and at least in some regions of the United States, ozone concentrations may currently exceed the levels used in this study. In an examination of image data, images can be compared for changes over time, allowing control over some external sources of variation. Both the mean and the variance of spectral properties are expected to change under chronic canopy ozone exposures, and where spatially coherent patterns occur, detecting small absolute changes in reflectance may not be impossible. At the field level, spectral measures such as those reported here, offer a non-destructive sampling technique to monitor canopy physiological properties and cumulative responses to environmental stress agents such as ozone exposure.

V. References

1998, Center for Spatial Technologies and Remote Sensing (CSTARS)
University of California, Davis