The photosynthetic activity of vegetation is a crucial parameter in global carbon balance models, as well as for predictions of the effects of climate change. Unfortunately, direct measurement of canopy and landscape level photosynthetic gas fluxes remains expensive and limited to a few sites. Remote sensing of spectral reflectance is more easily applied over large land areas and may be useful for estimation of photosynthetic activity. However, most work to date with remote sensing has focused on measurement of total vegetation cover, rather than the physiological activity of that vegetation. We measured photosynthetic rates of individual leaves from a wide range of species and compared these to spectral indices, calculated from hyperspectral reflectance - reflectance with many adjacent narrow bands, of single leaves, single species canopies and 100 m transects. The spectral indexes used were; the normalized difference vegetation index (NDVI, estimates total vegetation), the water band index (WBI, estimates liquid water content of vegetation), and the photochemical reflectance index (PRI, estimates photosynthetic light use efficiency). Using spectral indices measured at the leaf level, leaf photosynthetic rates were significantly correlated only with PRI. However, using spectral indexes measured at the canopy and transect levels, significant correlations with leaf photosynthetic rate were found for both WBI and NDVI. In fact the best correlation at the canopy level was with NDVI. Correlations with PRI and WBI at the canopy level were significant but weaker. The best predictor of leaf photosynthetic rate was an index that combined both WBI and PRI, suggesting that these indexes describe different aspects of the physiological status of the plants. Similar relationships were found between leaf photosynthetic rate and the indices measured at the transect level, although the relationships varied with stand age. The significance of these results for estimation of canopy and landscape level photosynthetic fluxes from remote sensing data will be discussed.
 
 

For prediction of canopy and ecosystem level CO₂ fluxes from remote sensing data we need estimates not only of absorbed light but also the efficiency of light utilization for photosynthesis. NDVI has proved to be a robust predictor of absorbed light but relatively less is known about prediction of photosynthetic efficiency.

1- In this study we test two ways of estimating photosynthetic efficiency:
PRI is related to the efficiency of PSII biochemistry and is thus a direct measure of photosynthetic efficiency. However, it remains unclear how well this index can be applied at the landscape scale.

2 - It is also possible that photosynthetic efficiency is related to the total biomass of plants in an ecosystem. Water or nutrient limitations tend to have parallel effects on leaf photosynthetic rates and total vegetation amounts. Thus, remote sensing measures of total biomass, such as NDVI and WBI may also provide an indirect estimate of photosynthetic efficiency, particularly in vegetation types such as California chaparral where water stress is common
 

1. To compare the utility of PRI, vs indices related to biomass (NDVI and WBI), for prediction of leaf photosynthetic rates.

2. To determine how the spatial scale of measurement affects the optimal choice of indexes for prediction of leaf photosynthetic rates.

3. To assess the generality of these relationships across a wide range of species and plant functional types.
 

Photosynthesis:

All photosynthesis measurements were made with a LICOR 6400 and conifer chamber. For the canopies, measurements were made on single leaves within or close to the section that was harvested. Because of the very small leaf size for the dominant shrubs on the transects, photosynthetic rates of short branch segments of Ceanothus and Adenostoma were measured and the value for the more dominant species on a given transect is used in the correlations.
 

Leaf reflectance:

Hyperspectral reflectance (reflectance with many narrow bands) of individual leaves was measured with a Unispec spectrometer, a bifurcated fiberoptic and leaf clip. The measurement area of the leaf was illuminated through one side of the bifurcated fiber with light from a halogen lamp in the Unispec while reflectance was detected through the other side of the fiber. The useable spectral range of this instrument is 400-1000 nm.

Canopy and Transect reflectance:

Canopy and transect reflectance was measured with the same spectrometer but utilized a single fiberoptic cable attached to a pole so that it could be held above the vegetation. Canopies were defined as a single plant of one species or a dense stand of individuals of the same species in the case of the smaller plants. Transects were 100 meters long and were measured at 1 m intervals.

Harvest:

Following the measurement of photosynthesis and reflectance, sections of the canopies were harvested for determination of leaf area and water content. This was done only for the individual canopies, not for the transects.
 
 

Table 1: Species

We measured 19 species representing a wide range of plant functional types found in California (annual, herbaceous perennial, winter deciduous, drought deciduous and evergreens). For the canopy measurements, individuals or single species groups of plants that had developed a full, closed canopy were selected. The transects were dominated by Ceanothus and/or Adenostoma but also included a wide range of other species
 

Annual	Herbaceous perennial	Winter deciduous	Drought deciduous	Evergreen
Avena	Marah macrocarpus	Populus fremontii	Artemisia californica	Adenostoma fasciculatum
Helianthus annuus		Ribes aureum	Eriogonum fasciculatum	Ceanothus megacarpus
Lolium		Sambucus mexicana	Malacothamnus fasciculatus	Heteromeles arbutifolia
Vicia villosa		Vitis girdiana	Salvia mellifera	Malosma laurina
			Tetradymia spp	Quercus agrifolia

 

Field sites in the Santa Monica Mts:

Spectral reflectance indices:

We calculated the following spectral indices:

1. Water Band Index (WBI), (R₉₀₀+(R₉₀₀-R₈₃₀))/R₉₇₀
2. Normalized Difference Vegetation Index (NDVI), (R₇₅₀-R₇₀₅)/(R₇₅₀+R₇₀₅)
3. Photochemical Reflectance Index (PRI), (R₅₃₁-R₅₇₀)/(R₅₃₁+R₅₇₀)
4. Combined index, (WBI-1)+PRI/2

Estimation of leaf PRI from transect measurements:

When measurements are made at the leaf level, PRI is related to the efficiency of light utilization by the photosynthetic apparatus. However, application of these relationships at the canopy and transect levels is complicated by the background reflectance. PRI values for soil and dead vegetation are in the middle of the normal range for leaf PRI. This is in contrast to NDVI and WBI where the values for soil and dead vegetation are almost always lower than for live vegetation.

Figure 1:
Representative plots of PRI vs NDVI for a 100 m transect measured at 1 m intervals in two seasons. Note that the value of PRI at NDVI=0.1 does not change between seasons. This value is an estimate of the soil and dead vegetation PRI. PRI at higher NDVIs changes much more dramatically with season and this variation is related to changes in PRI measured for individual leaves along the transect (green points at right).
 
 

Figure 2:

The same data as in Fig 1 but averaged over 16 m intervals to represent a typical spatial resolution of aircraft or satellite sensors. Note that regression lines are quite similar to those for the 1 m data, suggesting that this analysis is also applicable to lower resolution data.
 
 

Figure 3: The relationship between transect and leaf PRI can be significantly improved by estimation of PRI at full plant cover, thus eliminating the effect of the background. This was done by extrapolating the regression line in plots such as Fig 1 to estimate PRI at NDVI = 0.9 . Different colors represent different vegetation ages; green is 3 years following fire, blue is 20 years, red is 40+ years. Changing amounts of soil background at different ages affects the relationship between transect mean PRI and leaf PRI. Estimation of PRI at full cover provides a much more robust estimate of leaf PRI. Remaining differences between canopy and leaf PRI may result from variation in PRI between canopy leaves that was not adequately represented in the single leaf measurements (these measurements were made only on fully sunlit leaves).
 
 

Relationships between leaf photosynthetic rate and reflectance indices measured at 3 spatial scales:

Each figure shows relationships between leaf photosynthetic rate and one of the reflectance indices measured at the leaf (~1mm), canopy (~1m), and transect (~100 m) scales. Leaf and canopy data were collected on the same set of plants (see Table 1 for a list). Each point represents means for a different species.

Transects were dominated by Adenostoma and Ceanothus and are grouped by time following fire, new burns are first year following fire, developing canopies are 3 years following fire and closed canopies are 20+ years following fire. For the transects, leaf photosynthesis was measured on Adenostoma for new burn and developing canopies and on Ceanothus for closed canopies. Different points represent replicate transects and measurements at different times of year.

Figure 4:  PRI is the most robust predictor of leaf photosynthetic rate across all scales. It is the only index which shows a significant relationship to leaf photosynthesis at the leaf level. It also shows less variation between stand ages at the transect level.
 
 

Figure 5:  WBI was correlated with leaf photosynthetic rate at both the canopy and transect levels but not at the leaf level. This relationship changed with canopy development for the transects. The highest canopy point represents Helianthus. This species had very high water contents but because of the density of the canopy this was not reflected in the WBI.
 
 

Figure 6:  NDVI was well correlated with leaf photosynthetic rate only at the canopy level. There was no significant relationship at the leaf level and the relationships at the transect level were generally poor.
 
 

Figure 7:  Using the canopy level data, we empirically developed a new index combining PRI and WBI (see models) that significantly improved the correlation with leaf photosynthetic rate. This index also performed well for closed canopy transects although the correlation was not better than for PRI alone.
 

1.  Generality: All of the species and plant functional types appeared to fall on the same general relationships. Thus there seems to be the potential to generalize these relationships between habitats without the requirement of knowing the exact species composition.
 

2.  Spatial scaling: Somewhat surprisingly, the relationships between spectral indexes and leaf photosynthetic rate tended to become better as the scale of measurement increased. It should be emphasized that photosynthetic rate was still measured at the leaf level in all cases so there is an increasing mismatch in scales between the measurements as the scale of the reflectance measurements is increased. This suggests that leaf photosynthetic rate is somehow related to processes occurring at larger scales.
 

3.  Comparison of indices: PRI was the most robust predictor of leaf photosynthetic rate. It was the only index that was significantly correlated with leaf photosynthesis when reflectance was measured at the leaf level. At the canopy level, the relationship was good below a photosynthetic rate of around 15 but tended to saturate above this level. Results of this study as well as other published data suggest that this saturation level is fairly consistent across a wide range of species and measurement conditions. We also demonstrated that the effect of the background can be eliminated, significantly increasing the utility of this index at larger scales. However, it remains unclear whether the relatively small PRI signal can be reliably resolved in aircraft and satellite images where there is significant atmospheric interference. NDVI and WBI have an advantage here since these signals are larger and less subject to atmospheric interference. Both NDVI and WBI were correlated with leaf photosynthesis for measurements made at the canopy and transect scales. However, these relationships differed between stand ages suggesting that the biomass of the younger canopies had not had sufficient time to fully equilibrate to the environmental conditions. For the canopy harvest data, leaf photosynthesis was in fact correlated with measures of total biomass (water content and leaf area) of the canopies (Fig 8 and 9). However, the relationship between leaf photosynthesis and canopy leaf area (Fig 8) was considerably weaker than the relationship to water content (Fig 9). The reasons for this difference remain unclear but suggest interesting avenues for further research.

Figure 8:  Relationships between leaf photosynthetic rate and green area index (total projected area of leaves, green fruits and green stems per unit ground area) for the canopy harvests (see Table 1 for list of species). Each point represents the mean of 3 measurements on one species.
 
 

Figure 9:  Relationships between leaf photosynthetic rate and green water (total water content of leaves, green fruits and green stems per unit ground area) for the canopy harvests (see Table 1 for list of species). Each point represents the mean of 3 measurements on one species.
 
 

Models of canopy flux rates:

Leaf level photosynthetic rates are only one half the total canopy flux equation. Changes in light absorption due to differences in vegetation cover can also contribute a large fraction of the variation in flux rates. Light absorption can be estimated from NDVI and combined with an estimate of photosynthetic light use efficiency to predict canopy flux rates. The simplest form of such a model is a Big leaf model, where photosynthetic efficiency of the whole canopy is assumed to be proportional to photosynthetic rates of top canopy leaves.

In this study we measured photosynthetic rates of top canopy leaves and explored the potential to predict these rates from spectral reflectance indices. All of the indices had some degree of correlation with photosynthetic rates but the relationships varied widely. Introducing canopy light absorptance would be expected to further change these relationships. At very low cover, and thus low canopy light absorption, total canopy flux rates will be low and variation in leaf photosynthetic rates will have little effect on total fluxes. Conversely, at high vegetation covers, variation in leaf photosynthesis will have a much greater effect on total fluxes. Consequently, indices that predict photosynthetic rates well at high vegetation cover will have advantages over those that only predict photosynthesis well at low covers.

To compare how the different indices might function for prediction of canopy flux rates we developed some simple models based on the product of light absorption and an estimate of photosynthetic efficiency. The data used were from the transects. In all cases, light absorption was estimated from NDVI. In the Big Leaf Model we used the measured leaf photosynthesis as an estimate of photosynthetic efficiency. The PRI and WBI models assumed a linear relationship (roughly correct for the transect data), between these indices and photosynthetic efficiency. Units are relative and not intended to accurately represent actual canopy flux rates.

Note that the PRI model was better correlated with the big leaf model than was the WBI model when data from all transect ages was included. However, the WBI model performed better when only the closed canopies were considered. This is in contrast to the relationships with leaf photosynthesis by itself (Fig 4 and 5) where PRI performed better. PRI predicted leaf photosynthesis best at low photosynthetic rates when canopy cover was also low. In contrast, WBI was somewhat better at predicting leaf photosynthesis at higher photosynthetic rates and canopy covers.