Estimation of Tree Canopy Leaf Area Index by Gap Fraction Analysis

Scott N. Martensa,* Susan L. Ustinb, Robert A. Rousseaub
aDepartment of Botany
University of California, Davis
Davis, CA 95616, USA
 
bDepartment of Land, Air and Water Resources
University of California, Davis
Davis, CA 95616, USA
 
Submitted to Forest Ecology and Management
(Accepted 15 April 1993)
Elsevier Science Publishers B.V., Amsterdam
61:91-108

*Corresponding author:
Environmental Science
EES-15, Mailstop J-495
Los Alamos National Laboratory
Los Alamos, NM 87545
USA

ABSTRACT

We estimated leaf area index (LAI) in a needle-leaved forest and a broad-leaved orchard using four instruments which measure fractional light penetration through the canopy. Gap fraction data were analyzed by a one-dimensional inversion model or by using the Beer-Lambert Law. Instrument and analytical technique both had a strong influence on calculated LAI. There was no consistent pattern of LAI results among instruments so no simple cross-calibration can be offered. Three of the four instruments accurately estimated orchard LAI when used with one or the other analytical technique. In addition to performance, practical considerations including cost, sampling, and data analysis were also compared among the instruments. Because the instruments do not provide consistently accurate LAI estimates, LAI should be independently corroborated before use in any particular situation. These instruments may be most useful for relative LAI comparisons in a specific canopy type when a single combination of instrument and analytical technique is employed.

INTRODUCTION

Leaf area index (LAI, leaf area per unit ground area) is an important canopy parameter needed for many physiological and ecosystem studies (Nemani and Running, 1989). For example, it is important in understanding canopy gap dynamics and disturbance (Brokaw,1985; Denslow, 1987; Clark, 1990; Lawton, 1990). Leaf area has been shown to be highly correlated with productivity in a variety of ecosystems, including forests (Gholz, 1982; Waring, 1983; Webb et al., 1983). LAI is commonly used to compare canopy development or structure over time, under different environmental conditions, or among species. Waring (1985) suggested monitoring LAI as an indicator of stress in forests. Long-term monitoring would require robust measurements of LAI, as do comparisons of LAI among stands or communities.

Direct measurement of LAI in large heterogeneous plant canopies, such as natural forests, is difficult (Baldocchi et al., 1984; Neumann et al., 1989; Daughtry, 1990; Martens et al., 1991). Destructive harvest of leaves is usually undesirable, if not impractical, so methods based on allometry have been developed (Whittaker and Woodwell, 1968; Marshall and Waring, 1986). Further refinements have yielded the sapwood area to leaf area estimate (Grier and Waring, 1974; Whitehead et al., 1984), which still requires much laborious fieldwork before its application at any particular site.

Gap fraction analysis is another indirect method of estimating LAI (Campbell and Norman, 1989; Norman and Campbell, 1989). It has been successfully applied to canopies of small plants and also to tree canopies. Recently, advances in instrumentation for measuring gap fraction in plant canopies and the development of gap fraction inversion models, from which canopy characteristics such as LAI and mean leaf inclination angle can be inferred, have increased the attractiveness of this method (Campbell, 1986).

The basic tenet of gap fraction analysis is that canopy leaf area (or more correctly the one-sided surface area of all canopy elements) can be inferred from measurements of canopy gap area. Gap area can be inferred from canopy photographs, measurement of sunfleck area on the ground, or by estimation of the fraction of the direct solar beam that penetrates the canopy. New optical instruments that can readily provide gap fraction data are the Decagon Ceptometer, the Li-Cor Line Quantum Sensor and the Li-Cor LAI-2000 Plant Canopy Analyzer (Welles, 1990). Additionally, the computerized digitization and analysis of hemispheric photographs for canopy gap fraction is readily available (Rich, 1989, 1990). Functionally, these four instruments are either linear (Ceptometer and Line Quantum Sensor) or hemispherical sensors (Plant Canopy Analyzer and hemispheric photographs ).

The robustness of LAI estimates, especially when various instruments, analytical assumptions, and sampling schemes are used, needs careful evaluation. Only a few combinations of instruments and analytical techniques have been applied to tree canopies; for example, the Ceptometer was used with the Beer-Lambert Law (Pierce and Running, 1988), and the Plant Canopy Analyzer with a one-dimensional inversion model (Gower and Norman, 1991). We have assessed the suitability and performance of four instruments in both broad-leaved and needle-leaved tree canopies. We have also explored practical considerations in field application and data analysis that are important to a typical user of each instrument, and have included a comparison of instrument performance using either of the two analytical methods. Finally, we have compared two methods (Beer-Lambert Law and a one-dimensional inversion model) for analyzing gap fraction data from each instrument.

ANALYSIS OF GAP FRACTION DATA

Gap fraction data have been analyzed by several methods but two are most common: application of the Beer-Lambert Law (Jarvis and Leverenz, 1983; Marshall and Waring, 1986) or application of a one-dimensional inversion model (Norman and Campbell, 1989). Both analytical methods assume that canopy elements are randomly dispersed in space. The Beer-Lambert Law is simpler because the exponential extinction of light as it passes through the canopy is assumed to be adequately described by the extinction coefficient, k. The light below the canopy, Qi, is related to the light above the canopy, Q0, and LAI, by the relationship
 
(1)
and solving for LAI gives
 
(2)
Besides the gap fraction (Qi / Q0), only the extinction coefficient, k, for the canopy need be known or estimated for application of this method. For a canopy with a spherical distribution of leaf inclination angles, k is well approximated by 0.50. Extinction coefficients reported for 13 needle- and broad-leaved tree species ranged from 0.28 to 0.65 and averaged 0.47 (Jarvis and Leverenz, 1983).

The one-dimensional inversion model (Campbell and Norman, 1989; Norman and Campbell, 1989) has the same assumptions regarding canopy geometry as the Beer-Lambert Law. However, it requires estimation of canopy gap fraction at two or more zenith angles, from which it also yields the mean leaf inclination angle (or mean tip angle). The estimation of gap fraction as a function of zenith angle is relatively easy with hemispheric photographs or the Plant Canopy Analyzer, as these instruments acquire data for a range of zenith angles simultaneously. With the Ceptometer and the Line Quantum Sensor, data at more than one zenith angle must be obtained by waiting for the sun zenith angle to change and then re-measuring the gap fraction at each sampling point.

METHODS

Study sites

We measured the broad-leaved tree canopy in a 5-year-old walnut orchard (Juglans regia L. cv. ‘Chico’) at the University of California Kearney Agricultural Center, Parlier, CA (36.60° N, 119.50° W). Rows of trees were oriented north-south, with nominal tree spacing of 3.35 m along rows and 6.7 m between rows. The canopies were continuous along the rows but there was a continuous gap in the aisle between the rows. Average maximum tree height was 4.8 m. Forty randomly located sample points were marked with stakes in an area of approximately 330 m2. For comparison with the indirect assessments of LAI, we directly measured leaf area of four walnut trees by a complex, hierarchical subsampling scheme (Martens et al., 1991; Ustin et al., 1991). LAI calculated from these data was 3.29. The calculated stem area index ( (length x diameter) /unit ground area) of 0.11 demonstrates that stems contributed relatively little to light interception by the orchard canopy.

The needle-leaved forest site was a mixed-conifer forest stand located near Lava Butte in Sierra National Forest, CA (T13S R28E S25; 36.77° N, 118.88° W). The forest was dominated by Pinus ponderosa Laws. and Calocedrus decurrens (Torn ) Florin, but also included Abies concolor (Gord. & Glend. ) Lindl. and Pinus lambertiana Dougl. In the forest, 40 sample points were located at 20 m intervals along two randomly placed 400 m line transects.
 

Sampling and analytical methods

We used the gap fraction data from each instrument to calculate LAI by two methods: a one-dimensional inversion using Campbell's elliptical model of leaf angle distribution (Campbell, 1986; Campbell and Norman, 1989), and the Beer-Lambert Law with an assumed extinction coefficient (k) of 0.5. Each instrument required different sampling and analytical techniques to yield estimates of gap fraction. A description of the operation of each instrument and the techniques used for sampling and data analyses follows.
 

Ceptometer

The Ceptometer (Decagon SF-80, Pullman, WA) has 80 adjacent 1 cm2 photosynthetically active radiation (PAR) sensors along a bar. A small, attached data logger records the mean PAR as well as the percentage cover of sunflecks on the 80 cm bar. The percentage cover of sunflecks can be used directly as a measure of canopy gap fraction, but we did not use the sunfleck cover data because in tall canopies with small leaves large penumbral effects make distinguishing between fleck and non-fleck difficult. To estimate the fraction of direct beam transmitted through the canopy (gap fraction), we made four observations at each sample point: below-canopy incident PAR (Eb), below-canopy diffuse only PAR (Ebd), above-canopy incident PAR (Ea), and above-canopy diffuse only PAR (Ead). Gap fraction (F) was calculated by subtracting diffuse PAR as
 
(3)
Gap fraction was also calculated as the simple ratio of Eb to Ea. which includes the diffuse PAR in each measurement.

Measurements of below-canopy incident PAR (Eb) were made while holding the Ceptometer level (using a bubble level) at four compass directions around each sample point. The average of the four values was used for subsequent calculations. Diffuse PAR was measured by holding a 3 cm by 1 m board approximately 20 cm above the instrument to block all incident direct solar beam. In the orchard, below-canopy diffuse PAR (Ebd) was measured at five points along a transect from directly below the center of the canopy (along a row) to the aisle center. Ebd along the transect was measured before and after measuring Eb at the set of 40 sampling points. For each of the 40 sampling points, Ebd was spatially and temporally interpolated from the transect data. We measured Ea and Ead in an adjacent open field. In the forest, Ebd was measured at each of the 40 sampling points, and Ea and Ead were measured in a large gap (about 30 m diameter) within the forest stand. At least two sets of PAR measurements were made for each canopy type to yield gap fraction at more than one solar zenith angle.
 

Line Quantum Sensor

The Line Quantum Sensor (Li-Cor 191, Lincoln, NE) consists of a narrow, 1 m rod containing a single PAR sensor. Data are recorded by a separate Li-Cor data logger. Line Quantum Sensor PAR measurements were taken using two perpendicular orientations of the instrument per sample point. Average PAR values were used for further calculations. All other aspects of sampling were as described above for the Ceptometer.
 

Plant Canopy Analyzer

The Li-Cor Plant Canopy Analyzer sensing head has five concentric silicon ring detectors beneath a nearly hemispherical (150° field of view) lens. A blocking filter restricts radiation of less than 490 nm, which minimizes the effect of light scattered by foliage (Welles, 1990 ). The five rings provide data at five zenith angles for use in the inversion model. A separate dedicated controller-data logger is attached to the sensing head and records light readings for each ring. Readings of above- and below-canopy light conditions are necessary for the calculation of LAI and average leaf inclination angle by the built-in program. Below-canopy data were taken at eight compass directions and the mean of these values was stored in the instrument. In the orchard, above-canopy data were taken by walking to an adjacent open field immediately (less than 2 min) after recording each below-canopy reading. In the forest, above-canopy readings were acquired by another Plant Canopy Analyzer unit atop a nearby butte, which was about 1 km south and at 50 m higher elevation. Data were recorded every 5 min during the period in which below-canopy readings were made. Above-canopy values were then linearly interpolated to provide an appropriate value for each below-canopy reading.
 

Hemispherical photographs

Hemispherical photographs (180° field of view) were mace on black end white film (Kodak Tri-X, ISO 400) with a Canon 9 mm fisheye lens. Normally developed negatives were digitized using hardware and software (CANOPY program) described by Rich (1990). Values for unweighted openness in each photograph were divided by those for a blank negative to obtain gap fraction for each of 160 segments (eight azimuth sectors by 20 zenith classes). When using Campbell's inversion model, LAI was calculated for each of the eight azimuth sectors and then these values were averaged to provide the LAI for the image. This is essentially the same as using a log-average of the gap fractions about the azimuth, as suggested by the work of Lang and Yuequin (1986) .
 

RESULTS

Calculation of gap fraction

Line sensors
The LAI results for the Ceptometer and Line Quantum Sensor are shown in Table 1. The influence of the methods of gap fraction calculation (including or subtracting diffuse PAR) and analysis (Beer-Lambert Law or Campbell's inversion model) can be examined, as well as the influence of each instrument. We used paired t-tests for statistical analyses because each instrument was used at the same 40 points in each canopy type. Paired t-tests for all possible combinations within each row (Table 1) yielded highly significant differences in the forest (all P<0.0001 ) and the orchard (P<0.0003, except for comparison of subtracting vs. including diffuse PAR with the Ceptometer using Campbell's method, when P=0.0074).

Mean LAI values were always lower (paired t-tests, P<0.0001-0.0074) when gap fraction was calculated by including diffuse PAR as compared with subtracting diffuse from total PAR. The decrease in LAI was about the same in forest and orchard canopies (19-29% reduction). The downward bias in LAI that results from including diffuse PAR is expected to increase as diffuse PAR becomes a greater proportion of total below-canopy PAR. Thus, this bias could be significant in many forests.

When diffuse PAR was subtracted to avoid this bias, the numerator of Eq. (3) occasionally became zero when the measured below-canopy diffuse PAR was equal, within the resolution of the instrument, to total below-canopy PAR. This resulted in a gap fraction of zero for which no LAI can be calculated, as the logarithm of zero is undefined. We examined two ways to solve this problem -- by discarding zero gap fraction data points, or by substituting a small number (which reflected the sensitivity of the instrument) for the zero numerator of Eq. (3) when calculating gap fraction.

The small value we chose to substitute was 0.5 m mol m-2 s-1 PAR flux because the Ceptometer PAR readout is in whole units. Hence, when the numerator of Eq. (3) was zero we substituted 0.5. (For comparison, standard PAR in full sunlight is approximately 2000, m mol m-2 s-1.) When gap fraction values of zero were excluded from the calculations (Table 1), the resulting mean LAI values were lower in all eight comparisons than when 0.5 m mol m-2 s-1 was substituted.

The sensitivity of calculated LAI to the substitution of a small value for zero gap fractions is shown for the forest data in Fig. 1. LAI results for two data sets, which differed in number of zero gap fraction values, are plotted. The Line Quantum Sensor data set had 14 zero gap fraction values (17.5% of data set ), and LAI (Campbell's or Beer-Lambert) was more sensitive to this substitution than in the Ceptometer data set, which had only three zero values (3.8% of data set). LAI calculated using the Beer-Lambert Law is more sensitive (steeper slope) than LAI from Campbell's inversion model for both tree canopy data sets. From the regression equations for these lines, we calculated that our substitution of 0.5 m mol m-2 s-1 PAR in the numerator of Eq. (3) is equivalent to substituting a value of about 10-3 for gap fraction.

The Beer-Lambert Law always yielded a higher LAI than Campbell's method (Table 1). The increase in LAI with the Beer-Lambert method ranged from 26 to 88%. Subtracting or including diffuse PAR had little effect on the magnitude of the increase. Line Quantum Sensor results showed a greater response (55-88%increase)than Ceptometer LAI results (26-43%increase).

LAI calculated from the Ceptometer and Line Quantum Sensor differed significantly for the forest regardless of analytical method or whether diffuse PAR was subtracted or included (paired t-tests, 0.0001<P<0.017). Ceptometer LAI was greater than Line Quantum Sensor LAI when Campbell's method was used but Ceptometer LAI was lower when the Beer-Lambert Law was used. For the orchard, the LAI calculated from the Ceptometer and Line Quantum Sensor with Campbell's method did not differ significantly (paired t-tests, P=0.80). However, LAI from these two instruments differed when the Beer-Lambert Law was used with diffuse PAR subtracted (P=0.03) but not when diffuse PAR was included (P=0.09).

Hemispherical photographs
Gap fraction analysis of hemispherical photographs revealed that occasionally some segments (of the 160 segments per photograph) have gap fractions of zero. Rather than discard these zero values (which leads to a downward biased LAI estimate) or substitute a wholly arbitrary value we attempted to use a value close to the resolution of the digitizing procedure. When the unweighted openness of a segment was zero we assigned 0.5 pixel of openness to that segment. Because of the hemispheric projection onto a flat surface, the number of pixels per segment is lowest at the center (near the zenith) and highest at the perimeter (near the zenith horizon). Therefore the substituted gap fraction value (0.5 pixel/number of pixels per segment) was highest for segments near the center and lowest for those at the perimeter.

A sensitivity analysis of LAI to the gap fraction value substituted for zero is presented in Fig. 2. LAI values calculated by the Beer-Lambert Law are more sensitive than those calculated from Campbell's inversion model, following the trend found for the Ceptometer and Line Quantum Sensor data. The forest data set had 23.5% zeroes and was slightly more sensitive than the orchard data set, which had 21.6% zeroes. From the linear regression equations of this LAI sensitivity analysis, we calculated that our 'half-pixel' substitution for zero was equivalent to substituting a gap fraction of 5 x 10-9 for Campbell's inversion method and 1.2 x 10-10 for the Beer-Lambert Law calculation.
 

LAI comparisons among instruments

A summary of mean LAI values for the four instruments, two analytical techniques, and two canopy types used is presented in Table 2. Mean values differed nearly two-fold among instruments. Plant Canopy Analyzer means were consistently among the lowest, and Line Quantum Sensor means were consistently among the highest of the four instruments. The line sensors tended to have higher means than the hemispheric sensors. LAI calculated from hemispheric photographs showed a strong influence of analytical method; for both canopy types, Beer-Lambert LAI was more than twice that calculated with Campbell's model.

Three instruments produced values which compared favorably with directly measured LAI for the orchard (Table 2). The Line Quantum Sensor (LAI=3.29) and Ceptometer (LAI=3.23), when used with Campbell's inversion method, or Plant Canopy Analyzer data analyzed with the Beer-Lambert method (LAI=3.25) were not significantly different from the direct measurement (LAI=3.29). Hemispheric photograph data used with Campbell's method (LAI=2.68) underestimated the direct LAI (3.29), but overestimated LAI when used with the Beer-Lambert method (LAI=6.56) .

Correlations among instruments for a given canopy type and analytical technique are given in Table 3. Correlations between instruments of the same type (line or hemispheric sensors) are generally significant (Bonferroni criterion, P<0.002). The highest correlations are between the Ceptometer and Line Quantum Sensor in the orchard (r=0.489-0.852). The correlations between the two hemispheric instruments (r=0.373-0.492) are less strong.

Correlations across instrument types are significant in only one instance. An exceptionally high correlation (r=0.707) was obtained between the hemispheric photographs and Ceptometer for the Beer-Lambert Law technique in the forest. This Ceptometer data set was also correlated with the Plant Canopy Analyzer LAI values (r=0.419, P< 0.01).
 

LAI comparisons between analytical techniques

LAI results calculated by the Beer-Lambert method were higher than those from Campbell's inversion model for both data sets over all four instrument comparisons (paired t-tests, P<0.0002; Table 2). However, the Campbell and Beer-Lambert techniques were always highly correlated (r>0.91 with one exception, P<0.0001; Table 4). Intercepts were usually less than 0.7 (most were statistically significantly different from zero) except for the Ceptometer forest data. The slopes of the equations relating Campbell LAI to Beer-Lambert LAI ranged from 0.44 to 0.66, which shows the greater responsiveness of the Beer-Lambert technique. The hemispheric instruments demonstrate the most consistent equations between the forest and orchard data sets. The hemispheric photographs yield intercepts near zero and identical intercepts for the comparison of techniques in the orchard and forest. Both line sensor data sets give very different intercepts between the forest and orchard, although slopes are relatively consistent.

The exception to the extremely high correlation between the analytical techniques is with the Ceptometer forest data set (r=0.675, P<0.0001; Table 4). Even when three points that had zero transmission values are eliminated, the correlation between Campbell and Beer-Lambert techniques rises to only 0.736. The large intercept (2.16) decreases to 1.07 when these three values are omitted.
 

DISCUSSION

LAI results

The tree canopies used in this study violated the assumption of randomly dispersed canopy elements (e.g. leaves, stems, fruits) necessary for both analytical methods used here. There was aggregation of canopy elements within trees (e.g. leaves in fascicles) as well as aggregation of individual trees (e.g. row structure in the orchard canopy). Such aggregation is expected to let more light penetrate the canopy than if elements are randomly distributed in space and leads to underestimates of LAI. Nevertheless, three of the four instruments provided LAI measurements which were not significantly different from the directly measured value for the orchard canopy. Campbell's elliptical method with either line sensor provided accurate results, as did the Plant Canopy Analyzer when the Beer-Lambert method was used. We found the combination of the Ceptometer with the Beer-Lambert Law to overestimate LAI in the orchard, in contrast to the results of Pierce and Running (1988) for coniferous forests. The hemispheric sensors used with Campbell's method underestimated LAI, which is consistent with the results of Chen et al. (1991).

The Plant Canopy Analyzer may underestimate LAI in heterogeneous canopies, and the use of view restrictors (which were not available for this study ) is recommended to overcome this problem (Li-Cor,1989). Lang et al. (1985) found that inverting transmission measurements in a sorghum canopy underestimated directly measured LAI. Separately averaging the logarithms of transmission (or averaging LAI ) for distinctly different regions in the canopy (e.g. large gaps between rows) yields more accurate results (Lang and Yuequin, l 986). The proper use of view restrictors on the Plant Canopy Analyzer has the same effect. Gower and Norman (1991) found that multiplying Plant Canopy Analyzer LAI values by the ratio of total projected needle area to shoot silhouette area (about 1.5) improved the accuracy of LAI estimates of four needle-leaved species. Their technique accounts for the underestimation of LAI caused by the clumping of needles along stems. The accurate results we obtained for orchard LAI using the Plant Canopy Analyzer with the Beer-Lambert Law might be due to offsetting errors: LAI is underestimated by the instrument but overestimated by the analytical technique. With respect to precision, the Plant Canopy Analyzer had the highest precision (lowest coefficient of variation) of any instrument (Table 2).

Hemispheric photographs did not yield an accurate orchard LAI. Inaccuracies in gap sizes may result from subjective thresholding of photographs (Becker et al., 1989). However, a post hoc investigation of our subjective thresholding method showed that mean differences between high and low cover estimates were less than 5%. This difference is too small to account for the inaccuracy of the hemispheric photograph orchard LAI estimates. Hemispheric photographs appear to yield data which are very sensitive to the analytical method used. Hemispheric photograph LAI values calculated with Campbell's method were generally lowest, and Beer-Lambert method were usually highest, among the four instruments. Each hemispheric photograph provides data for 20 zenith angles in eight azimuth directions, which is far more detail than provided by any other instrument. Perhaps the analytical methods, especially Campbell's inversion method, are sensitive to the number or range of zenith angles used for calculation of LAI.

In general, calculated LAI was notably influenced by canopy type and analytical technique, so that results among instruments varied in a complex manner. There was no simple, consistent ranking of LAI values among instruments across the combinations of canopy type and analytical technique (Table 2). This precludes a single cross-calibration among instruments for comparison of LAI among different canopies or analytical techniques.

The failure of a one-dimensional inversion model for canopies that grossly violate assumptions of the model is not surprising. A more complex, three-dimensional model (Norman and Welles, 1983) that considers canopies to be an arrangement of ellipsoids within which canopy elements are randomly distributed may yield better results. However, this model requires more information on the geometry of a canopy than is usually obtained, although some suitable data are available for tree canopies (e.g. Martens et al., 1991; Ustin et al., 1991).

Nevertheless, LAI estimates with these techniques may be improved in some cases by using a sampling strategy that assumes a two-phase system -- gap and non-gap (canopy) regions. This is especially suitable for row-structured crops and forests before canopy closure. Our LAI results show greater coefficients of variation in the orchard than in the forest perhaps because our one-phase sampling was more appropriate for the randomly distributed trees in the forest than for the row-structured orchard. When sampling a two-phase system, optical measurements are concentrated in the canopy region and combined with an estimate of the fraction of canopy region relative to the area studied to yield an estimate of LAI for the entire area.
 

Practical considerations

Besides differences in accuracy or precision among the instruments there are practical differences among them with respect to equipment, sampling and data processing. On the basis of our experience with these instruments, we have compiled some of these practical considerations in Table 5. Welles ( 1990 ) has also discussed some of these considerations.

The two linear PAR sensors differ little in most respects, except that the Ceptometer is easier for one person to use under field conditions than the Line Quantum Sensor. Both sensors require sunny, cloudless sky conditions for use, and data must be collected at least twice at each point to obtain the gap fraction at two solar zenith angles. Data should be taken near solar noon, especially in dense canopies, because the fraction of direct beam transmitted at low sun angles approaches zero (numerator of Eq. (3 ) ) and requires assumption of an arbitrary value for direct beam PAR. With both instruments, data processing is relatively difficult because the resulting PAR values require much manipulation to obtain the gap fraction. Gap fractions must then be input to a user-supplied program to calculate LAI. Both line sensors are less expensive than either of the hemispheric types and each can be used as a PAR sensor for other applications.

The Plant Canopy Analyzer is relatively expensive and has limited use other than for LAI measurements. When used in extensive tree canopies without large gaps, it is necessary to have a second instrument to record simultaneously above-canopy (or outside-of-canopy) reference readings. The Plant Canopy Analyzer does not require the user to provide additional data processing to obtain LAI (Campbell's Method) for each point. This is a significant advantage over the other instruments because of savings in time and skilled labor. Plant Canopy Analyzer measurements (and hemispheric photographs) should be made with the sun at or below the horizon, or with a diffuse sky, to avoid mistaking brightly sunlit leaves for gaps. This restricts the time available for data collection to about  h near sunrise or sunset, and could increase the number of field days necessary to acquire sufficient data. In contrast, use of the line sensors may be restricted to a few hours near solar noon on sunny days, for reasons described above.

Hemispheric photographs can be processed by hand, but are most efficiently processed using digitizing hardware and software (e.g. CANOPY program) in a personal computer (Rich, 1989 ). Because of the necessity for film processing, digitizing, and computer analysis, there is a relatively long lag time between data acquisition and LAI, and each step may introduce errors (Rich, 1990). Although the CANOPY program calculates unweighted openness from the digitized images, user-supplied software is still necessary to calculate LAI. However, the gap fraction is more readily obtained from the CANOPY program results than from the linear PAR sensing instruments. Hemispheric photography inherently provides a permanent photographic record of the canopy that may be valuable for other research purposes (e.g. calculating solar tracks and canopy structure).

The equipment necessary for the hemispheric photograph technique represents a substantial investment (about equal to two Plant Canopy Analyzers). However, some equipment may already be on hand (e.g. camera and computer) and a computerized digitizing set-up has many applications other than LAI estimation.
 

CONCLUSION

The ease of use of all of these instruments for indirectly measuring LAI in tree canopies makes them attractive. However, calculated LAI values were not consistent across instruments. A simple cross-calibration among instruments would not suffice because canopy type and analytical method greatly influenced LAI results in a complex manner. The magnitude of variation among instruments appears to preclude quantitative comparisons of results. If relative comparisons of LAI in time or space are to be made, then the use of a single combination of instrument and analytical technique is recommended.
 

ACKNOWLEDGMENTS

We thank Erik Ustin, Jennifer Fitzgerald, Brian Pedersen and Steve Cline for field assistance. Discussions with John Norman, Paul Rich, and Jon Welles provided useful insight. Rahman Azari provided statistical advice. Funding was provided by Environmental Protection Agency (EPA) grant R8 14274010 and NASA NAGW-1101 subcontract 204272. This paper has not been subject to EPA review and should not be construed to represent the policy of the Agency.
 

REFERENCES

Baldocchi, D.D., Matt, D.R., Hutchison, B.A. and McMillen, R.T., 1984. Solar radiation within an oak-hickory forest: an evaluation of the extinction coefficients for several radiation components during the fully-leafed and leafless periods. Agric. For. Meteorol., 32: 317-322.

Becker, P., Erhart, D.W. and Smith, A.P., 1989. Analysis of forest light environments. Part I. Computerized estimation of solar radiation from hemispherical canopy photographs. Agric. For. Meteorol., 44: 217-232.

Brokaw, N., 1985. Treefalls, regrowth, and community structure in tropical forests. In: S.T.A. Pickett and P. White (Editors), The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, Orlando, FL, pp. 53-69.

Campbell, G.S., 1986. Extinction coefficients for radiation in plant canopies calculated using an ellipsoidal inclination angle distribution. Agric. For. Meteorol., 36: 317-321.

Campbell, G.S. and Norman, J.M., 1989. The description and measurement of plant canopy structure. In: G. Russell, B. Marshall and P.G. Jarvis (Editors), Plant Canopies: Their Growth, Form, and Function. Society for Experimental Biology: 31. Cambridge University Press, Cambridge, pp. 1 -19.

Chen, J.M., Black, T.A. and Adams, R.S., 1991. Evaluation of hemispherical photography for determining plant area index and geometry of a forest stand. Agric. For. Meteorol., 56: 129143.

Clark, D.B., 1990. The role of disturbance in the regeneration of neotropical moist forest. In: K.S. Bawa and M. Hadley (Editors), Reproductive Ecology of Tropical Forest Plants. Man and the Biosphere Series, 7. Parthenon, Park Ridge, NJ, pp. 291 -315.

Daughtry, C.S.T., 1990. Direct measurements of canopy structure. Remote Sensing Rev., 5: 4560.

Denslow, J., 1987. Tropical rainforest gaps and tree species diversity. Annul Rev. Ecol. Syst., 18: 431 -452.

Gholz, H.L., 1982. Environmental limits on aboveground net primary production, leaf area, and biomass in vegetation zones of the Pacific Northwest. Ecology, 63: 469-481.

Gower, S.T. and Norman, J.M., 1991. Rapid estimation of leaf area index in conifer and broad-leaved plantations. Ecology, 72: 1896-1900.

Grier, C.C. and Waring, R.H., 1974. Conifer mass related to sapwood area. For. Sci., 20: 205206.

Jarvis, P.G. and Leverenz, J.W., 1983. Productivity of temperate, deciduous and evergreen forests. In: O.L. Lange, P.S. Nobel, C.B. Osmond and H. Ziegler (Editors), Encyclopedia of Plant Physiology, New Series, Vol. 12D, Physiological Plant Ecology, IV. Springer-Verlag, Berlin, pp. 233-280.

Lang, A.R.G. and Yuequin, X., 1986. Estimation of leaf area index from transmission of direct sunlight in discontinuous canopies. Agric. For. Meteorol., 37: 220-243.

Lang, A.R.G., Yuequin, X. and Norman, J.M., 1985. Crop structure and the penetration of direct sunlight. Agric. For. Meteorol., 35: 83-101.

Lawton, R.O., 1990. Canopy gaps and light penetration in wind-exposed tropical lower montane rain forest. Can. J. For. Res., 20: 659-667.

Li-Cor, 1989. LAI-2000 Plant Canopy Analyzer Operating Manual. Li-Cor, Lincoln, NE, 180 pp.

Marshall, J.D. and Waring, R.H., 1986. Comparison of methods of estimating leaf-area index in old-growth Douglas-fir. Ecology, 67: 975-979.

Martens, S.N., Ustin, S.L. and Norman, J.M., 1991. Measurement of tree canopy architecture. Int. J. Remote Sensing, 12: 1525-1545.

Nemani, R.R. and Running, S.W., 1989. Testing a theoretical climate-soil-leaf area hydrologic equilibrium of forests using satellite data and ecosystem simulation. Agric. For. Meteorol., 44: 245-260.

Neumann, H.H., den Hartog, G. and Shaw, R.H., 1989. Leaf area measurements based on hemispheric photographs and leaf litter collection in a deciduous forest during autumn leaf fall., Agric. For. Meteorol., 45: 325-345.

Norman, J.M. and Campbell, G.S., 1989. Canopy structure. In: R.W. Pearcy, J. Ehleringer, H.A. Mooney and P.W. Rundel (Editors), Plant Physiological Ecology: Field Methods and Instrumentation. Chapman and Hall, New York, pp. 301-325.

Norman, J.M. and Welles, J.M., 1983. Radiative transfer in an array of canopies. Agron. J., 75: 481-488.

Pierce, L.L. and Running, S.W., 1988. Rapid estimation of coniferous forest leaf area index using a portable integrating radiometer. Ecology, 69: 1762-1767.

Rich, P.M., 1989. A manual for analysis of hemispherical canopy photography. Los Alamos National Laboratory, Los Alamos, NM, Rep. LA- 11733-M, 80 pp.

Rich, P.M., 1990. Characterizing plant canopies with hemispherical photographs. Remote Sensing Environ., 5: 13-29.

Ustin, S.L., Martens, S.N. and Vanderbilt, V.C., 1991. Canopy architecture of a walnut orchard. IEEE Trans. Geosci. Remote Sensing, 29: 843-851.

Waring, R.H., 1983. Estimates of forest growth and efficiency in relation to canopy leaf area. Adv. Ecol. Res., 13: 327-354.

Waring, R.H., 1985. Imbalanced forest ecosystems: assessments and consequences. For. Ecol. Manage., 12: 93-112.

Webb, W.L., Lauenroth, W.K., Szarek, S.R. and Kinerson, R.S., 1983. Primary production and abiotic controls in forests, grasslands, and desert ecosystems in the United States. Ecology, 64: 134-151.

Welles, J.M., 1990. Some indirect methods of estimating canopy structure. Remote Sensing Rev., 5: 31-43.

Whitehead, D.W., Edwards, R.N. and Jarvis, P.B., 1984. Conducting sapwood area, foliage area, and permeability in mature trees of Picea sitchensis and Pinus contorta. Can. J. For. Res., 14: 940-947.

Whittaker, R.H. and Woodwell, G.M., 1968. Dimension and production relations of trees and shrubs in the Brookhaven forest, New York. J. Ecol., 56: 1-25.

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