ABSTRACT

Anisotropy of thermal infrared exitance above and within a relatively closed fully irrigated sunflower canopy is detailed. Azimuthal variation in thermal infrared exitance above canopies was weakly (statistically) related to solar position and was comparable to or larger than errors in satellite-based canopy estimates. Anisotropy within canopies was significantly lower and decreased with canopy closure and depth into the canopy. Measured azimuthal isotropy within canopies supports the use of this assumption in radiative transfer models. Significant differences in canopy temperature measurements were found depending upon whether the instruments were within or above the canopy. These differences could produce errors of 20-35% in latent energy estimates during periods of high evapotranspiration (ET) and greater errors in periods of restricted ET.

INTRODUCTION

The angle dependence of emitted thermal longwave radiation as viewed from above a plant canopy was noted by Fuchs et al. (1967), and has been studied in more detail since then (Kimes et al., 1980, 1981; Heilman et al., 1981; Parson, 1985; Balick and Hutchinson, 1986; Balick et al., 1987; Matthias et al., 1987). The non-Lambertian (anisotropic) nature of longwave emissions from a canopy was noted to occur in two possible vector rotations, one in the azimuthal plane and the other in the vertical plane. Anisotropy results from the role of canopy geometry on the composite thermal exitance. It is azimuthally dependent on interplant spacing and plant size, and the vertical distribution of canopy elements. Generally, the anisotropy has been attributed to the differential illumination of different canopy parts (soil, leaves, branches, inflorescences) inducing a variation in surface temperature of these parts (Fuchs et al., 1967; Kimes et al., 1980, 1981; Parsons, 1985; Balick and Hutchinson, 1986; Matthias et al., 1987). Although the vertical anisotropy and spatial variation of upwelling thermal infrared radiation from canopies has been studied extensively, few studies have addressed the issue of anisotropy of radiation within and below canopies. Generally considered to be of lower magnitude than vertical anisotropy, errors in this assumption have significant consequences in energy budget analyses.

Our objectives in this research were to characterize the anisotropy of thermal infrared exitance within a sunflower canopy in relation to leaf temperature and to longwave radiation measurements above the canopy. We also examined the anisotropy of the thermal infrared exitance above the canopy but in a less detailed fashion.

MATERIALS AND METHODS

Field/crop/weather description

Vegetation measurements

Thermal infrared exitance and leaf temperature measurements

The instruments were taken into representative areas of the field, and the pipe assembly was oriented vertically. The first vertical angle was fixed and the pipe was rotated through all azimuthal angles, which were multiples of 45°, thus splitting the azimuthal circle into eight points for each vertical angle. Readings were taken for each point; when the circle was complete the vertical angle was adjusted to the next position and the pipe was rotated azimuthally again. For the zenith or nadir angles, the pipe was still rotated in an azimuthal circle of eight points. This was to obtain a better average of canopy and sky radiation or canopy and soil radiation. After all five angle multiples of the vertical plane were sampled, the pipe was lifted above the canopy such that the lowest instrument was at 2.0 m and the highest instrument at 2.5 m above the soil surface. Measurements were then made at all azimuthal angles for +45° (upwards to the sky) and -45° (downward at the canopy).

At the start and end of each data set (which took approximately 30 min), both infrared thermometers were calibrated with a black body target (Everest model 1000). Effective radiative temperatures (ERT) measured by the sensors were corrected to ETT (effective target temperature) by linear offset of the average of the two corrections obtained during calibrations. The corrections were small; they averaged 0.06 and 0.10° C for each instrument, respectively, with a standard deviation of 0.02° C, and a maximum value of 0.35° C, We define the effective radiative temperature (ERT) as the temperature measured by the infrared thermometers, the effective target temperature (ETT) as the temperature of the target calibration standard, and the effective black body temperature (EBT)) as the temperature of the target if it was a true black body with an emissivity of 1.00. It was assumed that the ETT was close to the effective black body temperature (EBT), i.e. that the target was almost a black body. This was because of two reasons: the manufacturer of the calibration target estimated its emissivity as 0.98, and because the ERTs of the leaves were usually close to that of the calibration target. This meant that the reflected canopy radiation from the target (a reflectivity of 0.02 by Kirchoff's law) was at approximately the same value as the emitted target radiation. For example, if the calibration target was at 300 K, and the surrounding leaves had EBTs of 10 K lower (290 K), the error due to the reflection effect would be only 0.3% of the thermal infrared exitance, associated with an EBT error of 0.2 K. In all cases the calibration target was within 5 K of the surrounding leaf ERTs. Thermal infrared exitance estimations were initially made employing finite integration over the wavelengths to which the instruments were sensitive, consistent with correction factors reported by Amiro et al. (1983) and Singh (1985). Comparison of infrared exitances calculated using these factors with those using the calibration target revealed that the calibration procedure described above could be used effectively; therefore the thermal infrared exitance was directly computed with the Stefan-Boltzmann equation from EBT.

Measurements were made between 0800 and 1800 h on 13-20 August 1987, with most measurements taken near peak solar irradiance, between mid-morning and mid-afternoon. Similar measurements were made in 1986, but in that field experiment two different types of infrared thermometers (AG-42 and Linear Laboratories C-500F) were used at the two heights on the pipe assembly. Because the results of that experiment were consistent with the present results and because dissimilar infrared thermometers were used, they are omitted here.

Simultaneous measurements of leaf temperatures were made with most of the thermal infrared exitance data sets. Leaf temperature measurements were made with a Linear Laboratories model C-500F infrared thermometer of 1° F repeatability. Calibration of the thermometer was identical to the procedure previously described. The ETT of every leaf on four adjacent plants was estimated based on measured ERTs. These plants were approximately 2.5 m from the location of the simultaneous thermal infrared exitance measurements, a distance chosen to minimize interference with the thermal exitance measurements. The height of each leaf above the soil surface was measured with a meter stick to an accuracy of approximately 0.02 m.

RESULTS AND DISCUSSION

Vegetation profiles

Leaf senescence was vertically distributed in the canopy with dead leaves below 0.3 m, both green and senescent leaves between 0.3 and 0.9 m, and green leaves above 0.9 m. Leaf area and dry weight are linearly related (r=0.83, P<0.001). Both leaf dry weight and leaf area vary significantly with height (Fig. 1) best fit by a second order polynomial. The bulk of the canopy biomass occurs between 0.4 and 1.3 m height were it is relatively uniformly distributed. Similarly, most of the canopy leaf area occurs between 0.3 and 1.2 m, with few leaves below this height and smaller leaves above 1.2 m.
 

Radiation and leaf temperature

The azimuthal rotation of the instruments looking down at the canopy (-45°) from above yielded the expected anisotropy of thermal infrared exitance with respect to solar position. In general, in the azimuthal direction opposite the sun, higher leaf temperatures and higher thermal infrared exitances were found (see Fig. 3). This supports the hypotheses that preferential viewing of sunlit canopy parts relative to shaded parts produces higher readings. The general location of the maximum thermal infrared exitance and EBT shifted from the west in the morning to a northerly direction near noon, and eventually to the east near sunset. However, a one-way analysis of variance with a t-test showed that the thermal infrared exitance in each azimuthal direction was not significantly different from the overall mean, at the 0.01 level, even with the visually discernable solar position affect. This implies the solar effect is statistically weak. With a less homogeneous, incomplete canopy with variable patch size, it is likely that greater variance would occur.

The maximum EBT difference from one azimuthal angle to another was 9.3° C, which was less than the 13° C reported by Kimes et al. (1980), for measurements above a wheat canopy. This is consistent with the fact that the sunflower canopy was dense with little exposed row structure during the experimental period and because our rotation above the canopy was in the azimuthal plane and not in the vertical plane. Kimes et al. (1980) found their greatest temperature variations when the sensor was rotated in the vertical such that the relative views of the wheat crop row structure and the soil varied greatly. In our case, much of the observed temperature difference was probably caused by differential illumination and physiological response. Also, it is possible that transient leaf temperatures responding to intermittent, coherent gust structures could have been partially responsible. This possibility was caused by the time (3-5 min) it took to rotate the instruments through the azimuthal angles; during this time numerous coherent structures could occur (unpublished data shows a period of 15-60 s for these structures over plant canopies). These results confirm that if only one or two canopy temperature measurements are taken, very large errors may result, but if the readings are averaged over a range of azimuthal directions, errors will be minimized.

Although the maximum EBT difference was high, the azimuthal standard deviation in thermal infrared exitance was only 0.51% of the mean, or 0.37° C in EBT. The standard deviation can be expressed as an equivalent error in the energy budget flux density of sensible heat (which is frequently used in the estimation of evapotranspiration, and represents the error in evapotranspiration estimates due to sensible heat estimation errors); the result is only 22 W m^-2, assuming an aerodynamic resistance of 20 s m^-1. The error of 22 W m^-2 could represent less than 10% error in ET estimation near maximum midday and mid-afternoon conditions, but much more relative error under low ET conditions. The maximum EBT difference of 9.3° C could result in sensible heat estimate differences of over 500 W m^-2, which could represent a 100% error in midday ET estimation.

Although the variance in the thermal infrared exitance due to azimuthal rotation at a vertical angle of -45° above the canopy showed a weak relationship to solar position (although not statistically significant), the relative standard deviation was less than the relative standard deviation in the longwave radiation flux density from the sky, showing that for a closed canopy under well-irrigated conditions (i.e. no soil moisture limits on transpiration), the emitted radiation can be considered as statistically isotropic as the sky radiation. It was surprising that relative statistical isotropy existed for the above canopy -45° measurements, considering the variation of the leaf temperatures were found to be greatest near the top of the canopy (see below) and the visually discernable solar position effect. However, because the leaf variation was at the individual plant level, this would result in above-canopy measurements averaging the individual leaf variation at each azimuthal angle, reducing the overall azimuthal variance.

Azimuthal rotation within the canopy yielded even less anisotropy, especially in the lower half of the canopy (see Fig. 4, Examples 1, 3, 4 and 6). In Fig. 4, Examples 2 and 5 are for the sensors pointed towards the sky, and reflect a greater anisotropy caused by incomplete canopy cover ("skyflecks"); however the anisotropy is still greater at 1.00 m (Fig. 4, Example 5) than at 0.50 m (Fig. 4, Example 2). This was expected because few sunflecks penetrate to the lower canopy, and senescent and dead leaves presented little variation in physiological response. Conversely, greater anisotropy was seen in the upper half of the canopy because of the larger number of sunflecks and greater variability in leaf physiological response. There was no apparent direct relationship to solar position, in contrast to the azimuthal measurements of canopy thermal infrared exitance taken above the canopy. Not surprisingly, within the canopy plant architecture strongly controls the spatial distribution of the radiant exchange process.

The standard deviation in thermal infrared exitance was 0.35% of the mean at 1.0 m above the soil, and fell to 0.12% of the mean at 0.5 m, when the sensors were rotated azimuthally at a horizontal angle of 0°. When the instruments were rotated at the +90° vertical angle, a large variation was seen in the measurements as a result of viewing either sky or leaf, which have greatly different EBTs (thermal infrared exitance standard deviation 2.9% of the mean). In marked contrast, the instrument rotation at a -90° angle showed the soil temperature to have little variation (thermal infrared exitance standard deviation 0.54% of the mean, at the 1 m height). The soil results should not be considered universal; no measurements were taken when the soil surface was very dry or under different canopy closure conditions.

Vertical rotation of the sensors for a given azimuthal angle yielded a large anisotropy, with isotropy increasing near the soil surface (see Fig. 5). This was no surprise, because the plant canopy and soil were at similar EBT levels, in contrast to the sky which was at a much lower EBT. Above the canopy, the downwelling longwave energy flux density was only 46% of the upwelling longwave energy flux density emitted by the plant canopy (note this number could be 10-20% too low if the instruments were not accurately measuring the lower, shorter wavelength sky thermal infrared radiation). At 50 cm above the soil surface, however, the downwelling longwave energy flux density was 96% of the upwelling longwave energy flux density. An analysis of variance with a t-test showed that all vertical rotations had at least one mean longwave radiation which was significantly different at the 0.01 level from the other mean longwave radiation values at other angles.

It has been postulated that tree canopy temperatures might be measured from underneath the canopy to ease the problems of above-canopy measurements. Our results show that above-canopy (-45°) temperatures averaged 24.07° C EBT in contrast to the within-canopy looking-up (+45°) EBT average of 22.57° C at 0.5 m and 15.89° C at 1.0 m. An EBT error of 1.5° C corresponds to a 90 W m^-2 error in latent energy estimates, at an aerodynamic resistance of 20 s m^-1. This is a significant error (approximately 20-35%), even during periods of maximum ET. Our results for crop canopies should not be considered greatly different from tree canopies, because the radiative transfer physics are similar despite the vertical canopy structure being different. If anything, the errors were minimized in our experiment because the crop was more homogeneous and closed compared to natural forest ecosystems. A more sparse canopy would create greater errors due to the sensor viewing more of the sky and an increased incidence of sunflecks.

The measurements summarized above could be averaged over the upper hemisphere and lower hemisphere to form a vertical profile of upwelling, downwelling, and net longwave radiation flux density within and above the canopy (see Fig. 6). The profile shows the upwelling longwave radiation flux to be relatively constant in comparison to the downwelling longwave radiation flux which decreases rapidly with increasing height above the soil surface within the canopy, reaching a minimum at the upper canopy surface. The net longwave radiation profile reflects this differential by dramatically increasing in magnitude (becoming increasingly negative) with increasing height.

A summary of leaf ETTs, presented in Fig. 7, shows a profile as a function of height. It is interesting that the profile is almost constant with height in the lower part of the canopy, but changes in nature at the canopy top. There, the variance increases greatly and the mean temperature changes. The increased variance is probably the result of two factors. The first is the large number of sunflecks and the orientation of the leaves with respect to the direct solar radiation vector. This results in leaf energy budget variability coupled to the plant's physiological response to light. The second is the effect of turbulence (especially gust structures) at the canopy surface, which causes time variation in leaf temperatures. Near the canopy bottom, lower wind speeds and decreased turbulence are expected to reduce the variation. The measurement techniques used here (and those generally used by researchers employing infrared thermometers) cannot distinguish time variation from variation of the individual measurements.

Figure 7 confirms that there should be a difference between above-canopy and below-canopy infrared temperature measurements. Mean leaf effective temperatures within the canopy were notably lower than those at the top of the canopy. Our discussion assumes that the leaf emissivity did not vary greatly with height.

CONCLUSIONS

The concept of a two-stream radiation field which is azimuthally isotropic within the upper and lower hemispheres, employed by such radiative models as those of Norman (1979) and Kimes et al. (1981), is supported by our results. Such models have recently been employed in sophisticated soil-plant-atmosphere models (Paw U et al., 1985; Meyers and Paw U, 1987), as well as in similar previous pioneering schemes (Norman, 1979). Our results show a large difference in the exitance of the upper and lower hemisphere, showing anisotropy in the vertical rotation sense.

ACKNOWLEDGMENTS

REFERENCES