Retrieval of the Pri to Assess Xanthophyll Cycle Activity Printer-friendly Version Interactive Discussion Retrieval of the Photochemical Reflectance Index for Assessing Xanthophyll Cycle Activity: a Comparison of Near-surface Optical Sensors Bgd Retrieval of the Pri to Assess Xanthophyll Cycle Activ

Unattended optical sensors are increasingly being deployed on eddy covariance flux towers and are often used to complement existing vegetation and micrometeorolog-ical measurements to enable assessment of biophysical states and biogeochemical processes over a range of spatial scales. Of particular interest are sensors that can 5 measure the photochemical reflectance index (PRI), which can provide information pertaining to leaf pigments and photosynthetic activity. This interest has facilitated the production of a new range of lower-cost sensors specifically designed to measure temporal changes in the PRI signal. However, little is known about the characteristics (spectral, radiometric and temporal) of many of these PRI sensors, making it difficult 10 to compare data obtained from these sensors across time, geographical locations and instruments. Furthermore, direct testing of the capability of these sensors to actually detect the conversion of the xanthophyll cycle, which is the original biological basis of the PRI diurnal signal, is largely absent, which often results in an unclear interpretation of the signal, particularly given the wide range of factors now known to influence 15 PRI. Through a series of experiments, we assess the sensitivity of one of the leading brands of PRI sensor (Skye SKR 1800) to changes in vegetation photosynthetic activity in response to changing irradiance. We compare the results with those obtained using a more expensive industry-standard spectrometer (PP-systems UniSpec) and determine the radiometric compatibility of measurements made by the different instruments. 20 Results suggest that the lower cost SKR 1800 instrument is able to track rapid (seconds to minutes) and more gradual diurnal changes in photosynthetic activity associated with xanthophyll cycle pigment conversion. Measurements obtained from both the high and lower cost instrument were significantly linearly correlated but were subject to a large systematic bias, illustrating that small differences in instrument configuration 25 can have a large impact on the PRI measurement values obtained. Despite differences in absolute PRI values, significant correlations were observed between the PRI derived from the SKR 1800 and the epoxidation state of the xanthophyll cycle (r 2 = 0.46, p < 0.05), although the dynamic range of the SKR 1800 PRI signal was often lower than more expensive instruments and thus the lower cost instrument may be less sensitive to pigment dynamics related to photosynthetic activity. Based on our findings, we make a series of recommendations for the effective use of such sensors under field conditions.

2.2.4 Correcting for differences in instrument spectral response function 1 The UniSpec instruments and each of the SKR 1800 sensors have different spectral response 2 functions (SRF) (e.g. band centres and full width at half maximum; FWHM), which result in 3 differences between instruments in the PRI wavelengths (Fig. 1). The spectral resolution 4 (FWHM) of the UniSpec instruments is approximately 10 nm with a ~3 nm sampling interval. 5 The data are subsequently interpolated to 1 nm intervals during processing and the HCRF at 531 6 nm and 570 nm is used for the calculation of PRI in both the SC and DC configurations. For the 7 SKR 1800 sensors the PRI wavelengths are centred at 530 nm and 569 nm for the reflected 8 radiation (i.e. downward facing sensor) and 531 nm and 567 nm for the upward looking sensor 9 recording incoming irradiance. As noted above, these values are a function of the particular filters 10 chosen by the manufacturer and can vary from one batch of sensors to the next. To understand 11 how instrument SRFs may influence the PRI, we convolved the UniSpec spectra with the 12 manufacturer supplied SRFs for the SKR 1800 downward-facing sensors and compared SKR 13 1800 PRI values to those obtained from the UniSpec before and after spectral convolution. As the 14 SKR 1800 sensor only contains two wavebands we used the UniSpec (1 nm spectral data) to 15 simulate each of the bands separately using the method outlined by Robinson and MacArthur 16 (2011) so that the output value for each band is the integral of the product of the SKR 1800 SRF 17 and the UniSpec input spectrum. All calculations were implemented in the R statistical 18 environment (R Development Core Team, 2012). 19

Experimental set-up 20
All experiments were performed at the University of Alberta campus, Edmonton, Canada during a series of midday shade removal studies, which provided an abrupt transition from low to high 23 light intensities over a range of plant canopies; and (ii) a single diurnal study that followed PRI 24 change under ambient sunlight. Both types of experiment were designed to observe the canopies 25 as they underwent physiological transitions from their dark state to full illumination, and 26 facilitated an investigation of the sensitivity of the different instruments to changes in vegetation 27 photosynthetic activity and the radiometric compatibility of the PRI measurements. A series of 28 custom sensor mounts were designed to ensure that on each occasion all instruments used for 29 inter-comparisons were viewing a similar portion of the plant canopy as was physically possible. The area viewed by each instrument was approx. 20 cm in diameter. Similar experimental 2 protocols were applied to all sets of measurements collected. 3 4 Dark-to-light transition experiments were performed over four different plant canopies ( Table 2)  5 following a similar approach to that described in Gamon et al. (1990) On each sampling occasion plants were covered with a black shade cloth the evening prior to the 19 experimental measurements. Near solar noon the following day the shade cloth was abruptly 20 removed, exposing the plants to full sunlight. The rapid response of plants to excess 21 photosynthetic photon flux density (PPFD) is such that changes in sun angle, canopy structure 22 and leaf movement, which can confound PRI measurements, will have limited influence on the 23 spectral measurements (Gamon et al., 1990(Gamon et al., , 1992. Over the next ~30 minutes, canopy HCRF 24 was collected from multiple instruments ( Table 2) at ~10 second intervals over the previously 25 shaded plant canopy. The PRI from the UniSpec DC was obtained using equations 1 and 2 26 whereas the PRI from the SKR 1800 sensors was derived from equation 4 (see section 2.2 for 27 equations and details of spectral processing). 28

Experiment 1: Dark-to-light transitions
Time series plots were used to visually examine dynamic changes in the canopy PRI. The mean 1 difference (MD) between the values of PRI derived from the SKR 1800 and UniSpec DC 2 instruments, along with the standard deviation of the differences (SDs) and of the mean 3 difference (or standard error, SE) were computed as a quantitative measure of discrepancies: 4

11
The diurnal experiment was undertaken to explore the relationships between PRI measurements 12 and the epoxidation state of the xanthophyll cycle pigments under naturally changing sunlight. 13 The influence of changing sun angle and low light levels on measured PRI and diurnal 14 dependencies of sensor differences were also assessed. Measurements were collected from 06: 30 15 to 19:50 on 25th July 2013 over a potted lodgepole pine (Pinus contorta) closed-canopy synthetic 16 stand (1 x 1 m plot). The pine saplings were approximately 4 years old, well-watered and located 17 on the roof of the Biological Sciences building at the University of Alberta. 18 Canopy PRI was measured at 1 minute intervals from the automated SKR 1800 sensors and at 19 ~15 minute intervals for the UniSpec SC instrument. In addition, hourly leaf-level HRCFs were 20 measured using a separate UniSpec SC instrument fitted with a needle leaf clip, bifurcated fiber 21 optic and an internal light source. On each occasion, leaf spectral measurements were recorded 22 from the same four plants, one located in each of the four corners of the study plot (n = 40). 23 Sample leaves were randomly chosen from the top of each plant canopy. Needles with a similar 24 orientation and sun exposure to those used for leaf reflectance factor measurements were also 25 sampled (2 needles per plant, each 3 cm long) and immediately frozen in liquid nitrogen for 1 analysis of xanthophyll cycle pigments using high-performance liquid chromatography (HPLC; 2 1260 Infinity, Agilent Technologies, Santa Clara, CA, USA) and the procedure of Thayer and 3 Björkman (1990). The epoxidation state (EPS) was calculated from the area-based molar 4 concentrations of the three xanthophyll cycle pigments, violaxanthin (V), antheraxanthin (A), and 5 zeaxanthin (Z) using Eq. (9): 6 (9) 7 Incident PPFD was recorded throughout the experiment with a quantum sensor (LI190SB, LI-8 COR, Lincoln NE, USA). Time series were used to visually examine relationships between PRI 9 and the epoxidation state of the pine canopy as a function of illumination conditions, and 10 regression relationships were formulated between PRI and EPS.  Fig. 2 shows an example of the observed changes in the PRI as an alfalfa canopy was suddenly 16 exposed to high light levels. The dynamic pattern of the PRI was similar for both instruments, 17 although the actual values of the index measured by the SKR 1800 sensors were much higher. 18 When data from all plant canopies used in the dark-to-light experiments were pooled, there was a 19 near-linear relationship between the PRI recorded by both sensors (r 2 = 0.98; Fig. 3a), although 20 the values obtained from the SKR 1800 sensor-pair exhibited a lower dynamic range and were 21 consistently and significantly higher (p < 0.0001, Student's t-test) than the UniSpec DC, with a 22 mean difference (MD) of 0.1. After normalising for instrument configuration differences, using 23 the SRFs for the SKR 1800, the SKR 1800 PRI remained consistently and significantly higher (p 24 < 0.0001, Student's t-test) than those derived from the UniSpec DC, but the values were closer to 25 the 1:1 line, and the MD was reduced by a factor of 10 (MD=0.01, Fig. 3b). Fig. 4 summarizes 26 the differences between the two instruments by plant species, after SRF corrections had been applied. Mean PRI instrument differences were similar for alfalfa, aspen and strawberry canopies 1 (~10-15%), but SKR 1800 PRI values were often more than twice as high than those measured by 2 the UniSpec DC over the ponderosa pine canopy. 3

4
The full results of the dark-to-light experiment, after normalising for different instrument SRFs, 5 can be seen in Fig. 5. For most species, the PRI rapidly decreased upon initial removal of the 6 shade cloth. Largest decreases occurred within the first 5 minutes after exposure to sunlight. 7 After the initial reduction, the PRI for aspen and ponderosa pine began to gradually increase as 8 the leaves became acclimatised to the light. The fluctuating nature of the PRI response for aspen 9 can be explained by intermittent cloud cover that was present during the latter part of the 10 experiment (data not shown). Additionally, aspen leaves are prone to fluttering in the wind 11 (Roden and Pearcy 1992), which may have caused additional fluctuation in the PRI response. 12 Slight differences in the FOV of the two sensors over the aspen canopy may have led to the more 13 evident fluctuations in PRI measured by the UniSpec than measured by the SKR 1800 sensors. 14

22
Diurnal PRI profiles for the pine canopy (UniSpec SC canopy and SKR 1800) and individual 23 pine needles (UniSpec SC leaf) are shown in Fig. 7a. The PRI was highest in the morning and 24 early evening and lowest during the early to mid-afternoon when both temperature and 25 illumination were greatest ( Fig. 6a and 6c). Leaf-level PRI followed a similar trend to that of the 26 canopy but did not replicate the high values measured at the canopy-level during the early part of the day when solar zenith angles (SZA) were high (> ~ 60°). These anomalously high canopy 1 PRI values are unlikely to accurately indicate physiological state. A closer inspection of the full 2 VIS-NIR reflectance spectrum for data collected with the canopy UniSpec instrument, illustrated 3 that the observed artefacts at high SZAs were not confined to reflectance factors at 531 nm and 4 570 nm (data not shown), but were probably general responses to high SZAs and low light. When 5 differences in the SRF of each instrument were not taken into account, the SKR 1800 PRI values 6 were significantly higher than those obtained from either of the UniSpec instruments, and the PRI 7 values measured at the leaf-level were generally higher than those recorded with the UniSpec 8 instrument over the canopy. 9 To further investigate the reasons surrounding the comparatively high PRI values derived from 10 the SKR 1800 sensor-pair, we used white panel measurements collected throughout the course of 11 the day to perform an in situ cross-calibration of the sensors. The cross-calibration enabled the 12 HCRF to be derived from each of the two SKR 1800 wavelength channels (see section 2.2.3). 13 The diurnal pattern of reflectance factors for the 531 nm and 570 nm channels, in comparison to 14 those measured by the UniSpec canopy instrument, are shown in Fig. 8. The figure clearly 15 illustrates differences in the HCRFs measured by each instrument. The 531 nm reflectance factor 16 recorded by the SKR 1800 sensors is consistently higher than that recorded at 570 nm. However, 17 the opposite is true for both the UniSpec canopy measurements (Fig. 8), and the UniSpec leaf 18 measurements (data not shown). Using Eq. (1) to obtain PRI for these data resulted in a lower 19 PRI for data collected with the UniSpec instruments than those obtained by the SKR 1800 20 sensors, as shown in Fig. 7a. 21 in the diurnal pattern of the PRI were not purely a consequence of differences in the spectral 24 response. Small differences can be seen between the PRI obtained by the SKR 1800 sensor-pair 3.2.2 Tracking physiological change 1 Fig. 9 illustrates the diurnal PRI and EPS patterns of individual needle leaves sampled from each 2 of the four corners of the pine canopy. Temporal changes were most pronounced in leaves 3 located in the southern corners of the sampling plot. Over the course of the experiment these 4 leaves were exposed to higher light levels for a longer duration and showed a clear decrease in 5 both PRI and EPS as illumination increased during the early to mid-afternoon, before gradually 6 increasing towards the early evening. A similar decrease in the PRI during the early to mid-7 afternoon was observed at the canopy-scale by the SKR 1800 sensors, although a less prominent 8 pattern was observed by the UniSpec instrument (Fig. 9). 9 The PRI was significantly correlated with EPS both at the leaf and at the canopy-level (Fig. 10). 10 The strongest correlations were observed at the canopy-scale when PRI was measured with the 11 UniSpec instrument (r 2 = 0.76), and weakest when using the SKR 1800 sensors (r 2 = 0.46). 12 Differences in instrument SRFs did not influence the strength of the correlations between EPS 13 and the UniSpec canopy PRI, and UniSpec leaf PRI (after SRF corrections were applied, r 2 = 14 0.77, p < 0.001 and r 2 = 0.56, p < 0.01; respectively (data not shown)). 15 We calculated the NDVI from the UniSpec canopy data to explore whether the PRI was 16 influenced by diurnal changes in plant canopy architecture. The results showed that there was no 17 correlation between NDVI and EPS (r 2 = 0.007; data not shown), indicating that the diurnal 18 variation observed in the canopy PRI was not simply a consequence of changing canopy 19 architecture but instead reflected actual changes in the xanthophyll cycle related to altered 20 photosynthetic activity. 21 22

Discussion 23
Our results suggest that under environmental conditions (e.g. temperature and relative humidity) 24 similar to those observed in the current study, both the UniSpec and SKR 1800 instruments are 25 able to track changes in the PRI signal in response to short-term (or facultative) plant responses 26 to changing illumination conditions. Differences between the values of the PRI obtained from architectures. Although the centre wavelengths of each of the SKR 1800 channels were located 1 very close to the standard 531 nm and 570 nm wavelengths commonly used to calculate the PRI, 2 the SRFs of the two instruments were different (Fig. 1). For the SKR 1800 sensor-pair, these 3 differences resulted in a higher HCRF at 531 nm, a region where changing absorption takes place 4 due to the activity of xanthophyll pigments, than at the reference wavelength of 570 nm; which is 5 opposite to that observed by the UniSpec instruments. Simulating the SKR 1800 measurements 6 from the UniSpec via convolution resulted in PRI values from both instruments that were more 7 similar, although statistically significant differences remained. Differences in the SRFs between 8 instruments are common and not confined to the two instruments used in this study. Castro-Esau in the current study, may be that the corresponding spectral channels on the upward and 13 downward facing SKR 1800 sensors are not identical i.e. their SRFs differ (Fig. 1). This was not 14 accounted for in the spectral convolution, which only used the SRFs generated for the downward 15 facing sensor. Even though great care was taken to match the ground resolution element observed 16 by each instrument, small differences in the area of the canopy that was observed may remain. 2014). All species showed a decline in the PRI as plants were exposed to rapid increases in 28 illumination, suggesting changes in the epoxidation state of the xanthophyll cycle in response to observed for the ponderosa pine canopy (Fig. 5d) was most likely a consequence of the saplings 1 becoming excessively hot under the black cloth prior to the experimental measurements, which 2 lead to the visible death of many top canopy leaves by the end of the experiment. 3 After normalisation for different instrument SRFs, instrument differences between PRI values 4 were similar for all canopies, apart from the ponderosa pine where the PRI measured by the SKR 5 1800 sensors was often double of that recorded by the UniSpec (Fig. 4). Due to the death of many 6 of the top canopy leaves during the canopy shading, the PRI remained extremely low throughout 7 the experiment (Fig. 5d) and thus the large between-instrument differences in index values may 8 be a consequence of a low signal-to-noise ratio for the SKR 1800 sensors under conditions where 9 the reflectance signal is weak. 10 Even though the PRI is often used as an indicator of photosynthetic efficiency in many remote 11 sensing studies, few studies actually relate the index to the changes in the xanthophyll pigment 12 pool, which it aims to detect (e.g. Gamon et al., 1990Gamon et al., , 1992Gamon et al., , 2001; Filella et al., 1996; Gamon 13 and Berry, 2012). Significant correlations were observed between diurnal changes in EPS and 14 PRI at both the canopy-and leaf-level (Fig. 10), and indicate leaf responses are also detectable at 15 the canopy scale with both instruments. These results are similar to previous diurnal studies by 16 Gamon et al. (1992) and Filella et al. (1996). 17 The strength of the relationship between PRI and EPS measured at the leaf-level was weaker than 18 that measured at the canopy-scale using a similar UniSpec instrument. Diurnal PRI patterns at the 19 leaf-level were largely dominated by leaves sampled from plants facing south, but also included 20 measurements from leaves exposed to lower levels of illumination where diurnal changes in EPS 21 and PRI were minimal (Fig. 9). Consequently relationships between leaf-level PRI and EPS were 22 stronger for south facing leaves (r 2 = 0.73, p < 0.05 and r 2 = 0.43, p < 0.05; for the SW and SE 23 facing leaves respectively) than those facing north (r 2 = 0.07, p = 0.5 and r 2 = 0.12, p = 0.4; for 24 the NW and NE facing leaves respectively). Both temporal patterns were incorporated into the 25 mean values of EPS and PRI, which introduced scatter into the leaf-level EPS-PRI regression and 26 thus weakening the overall relationship (Fig. 10). Differences in the linear regression coefficients 27 of the EPS-PRI relationship for the leaf and canopy, when using similar UniSpec instruments, 28 were also apparent. Such differences may have resulted from the use of two different UniSpec instruments, which were not cross-calibrated, but may also be a consequence of comparing leaf-1 level measurements made under controlled illumination conditions with those obtained from an 2 entire canopy under natural sunlight. Similar offsets in PRI values between leaf and canopy-3 levels have been reported previously (e.g. Gamon and Qiu 1999). 4 Canopy PRI obtained by both the SKR 1800 and UniSpec instruments predominantly reflected 5 changes in the sun-exposed canopy. However, the SKR 1800 sensors recorded a prominent 6 decrease in the PRI during the early afternoon (Fig. 7). This pattern was also reflected in some of 7 the more southerly facing leaf-level measurements and coincided with increased variability in the 8 measures of the EPS. Consequently the observed between-sensor differences are likely due to 9 each sensor having a slightly different FOV. The complexity of the conifer canopy is such that 10 even relatively small differences in the FOV between instruments may result in each instrument 11 measuring parts of the canopy that may have been exposed to different levels of illumination (i.e. 12 levels of sun and shade; Gamon and Bond 2013). 13 Although differences in the spectral configuration of the two instruments resulted in significantly 14 different PRI values, the strength of the PRI-EPS relationship for the UniSpec instruments prior 15 to and after applying the SRF corrections, was not significantly different. Early work on the 16 initial formulation of the PRI (e.g. Gamon et al, 1992) showed that there may not be a single 17 optimum reference wavelength for the PRI equation. In these early studies, using a Spectron 18 instrument (FWHM ~10 nm), Gamon et al. (1992) showed that significant correlations between 19 EPS and PRI could be obtained using a reference wavelength within the ~ 550 nm to ~570 nm 20 range. Consequently, whilst the SRF centred at ~570 nm differed between the SKR 1800 and 21 UniSpec instruments, resulting in differences in the wavelengths and relative contribution of light 22 to the 570 nm radiance measurements, the light contributing to the reference wavelength for each 23 instrument was within the optimum range reported by Gamon et al. (1992). However, even when 24 the SKR 1800 sensors were simulated by the UniSpec canopy instrument, the diurnal pattern did 25 not match that of the actual SKR 1800 sensors and the PRI-EPS relationship was consistently 26 stronger than the relationship observed between EPS and the SKR 1800-measured PRI 27 suggesting that instrument differences other than the spectral response (e.g. signal-to-noise ratio, 28 FOV, the use of a cosine diffuser compared to a Spectralon panel, quality of the optics) may also 29 have contributed to the observed divergence in diurnal PRI values between instruments. Further inter-comparison studies, which utilise more uniform vegetation canopies (e.g. Anderson et al. Near-surface optical sampling can be used to complement existing vegetation and 5 micrometeorological measurements to enable assessment of biogeochemical processes over a 6 range of spatial scales. Sensors that are capable of measuring reflectance across narrow spectral 7 bands are of particular interest for monitoring changes in plant physiological processes (e.g. 8 photochemical reflectance index; PRI) linked to carbon exchange and photosynthetic 9 downregulation via xanthophyll cycle pigments. The cost of unattended optical instruments is 10 now such that these instruments are increasingly being deployed for long term temporal 11 monitoring. However, a full characterisation of these sensors is necessary if the data are to be 12 compared across geographical locations, over time and between instruments. Specifically, it is 13 critical that the SKR 1800 sensors being used have matching wavelengths and the same spectral 14 response. Ideally, this could be confirmed by the manufacturer or by independent laboratory tests. 15 If independent, automated spectrometers were also on site, then it would be possible to simulate 16 the SKR 1800 response to understand the sources of any differences that might occur. All sensors 17 deployed should be mounted at similar distances from the canopy and at similar angles. They 18 should be checked and cleaned annually and according to the manufacturer's recommendations, 19 returned for laboratory calibration every two years. Additional corrections, for dark-current drift 20 and temperature drift in response to large variations in temperature, may also be required 21 (Eklundh et al. 2011). Regular cross-calibration can be used to assess and possibly correct for 22 such instrument drift. 23 In this paper, we compared the physical capabilities of two brands of field-portable narrow-band 24 instruments commonly used to measure PRI; namely UniSpec spectroradiometer (PP Systems, 25 USA) and SKR 1800 (Skye Instruments, UK). The shade-removal experiments revealed that both 26 instruments were able to track rapid apparent changes in the epoxidation state of xanthophyll 27 cycle pigments, although the dynamic range of the PRI was lower for the SKR 1800 sensors 28 suggesting a lower sensitivity to changes in xanthophyll cycle pigments related to photosynthetic 29 activity. The PRI values measured from each instrument were subject to a systematic difference 1 (bias), the magnitude of which appeared to be generally consistent across the range of species 2 studied and could primarily be explained by differences in the spectral configuration of each 3 instrument. 4 Despite the recent proliferation in the use of SKR 1800 unattended PRI sensors, to the best of our 5 knowledge there are no published data reporting relationships between the PRI measurements 6 obtained from this instrument and actual changes in xanthophyll pigments. In this study, the 7 diurnal course of the PRI obtained from both the UniSpec and SKR 1800 instruments compared 8 well with leaf-level HCRF measurements and physical measures of the EPS. However, both 9 instruments were susceptible to the well-documented issues associated with the collection of 10 spectral data at high solar zenith angles (> 60°) and under fluctuating illumination conditions 11 independent of whether the mode of operation was dual or single beam (SKR 1800 and UniSpec 12 SC, respectively). The findings suggest that the SKR 1800 sensors can be used for tracking short-  values for a fixed PVC cable. 5 Table 2. Type of experiment undertaken, the species over which measurements were performed, 1 the sensors that were used and whether spectral data were collected at the canopy-or leaf-level 2