Can a satellite-derived estimate of the fraction of PAR absorbed by chlorophyll (FAPARchl) improve predictions of light-use efficiency and ecosystem photosynthesis for a boreal aspen forest?
Introduction
Realistic models of plant canopy photosynthesis are necessary for obtaining accurate estimates of the carbon cycle for use in land surface models (LSMs) and atmospheric general circulation models (GCMs) (Sellers et al., 1996a, Sellers et al., 1996b). In vegetative canopies, photosynthetically active radiation (PAR) is absorbed from sunlight by photosynthetic pigments, primarily chlorophyll a and its accessory pigments (chlorophyll b, carotenoids). When ecosystem photosynthesis is calculated with a process model, it is referred to as Gross Primary Production (GPP). When it is calculated from flux tower data, it is referred to as Gross Ecosystem Production, designated here as GEPtower.
Plant production efficiency models (PEMs) have been developed to estimate GPP at canopy, landscape, regional and global scales, utilizing optical remote sensing to provide the fraction of absorbed PAR (FAPAR). Examples include GLO-PEM (Prince et al., 1995, Prince et al., 2000, Prince and Goward, 1995, Prince and Goward, 1996), TURC (Ruimy et al., 1994, Ruimy et al., 1996a, Ruimy et al., 1996b), 3-PG (Landsberg and Waring, 1997, Law et al., 2000) and PSN (Running et al., 1994, Running et al., 1999a, Running et al., 1999b, Running et al., 2000, Running et al., 2004). This latter model is a satellite-based global photosynthesis product derived from the MODerate resolution Imaging Spectroradiometer (MODIS) on the Terra and Aqua platforms.
All of these models estimate GPP as the product of three terms: (1) the light use efficiency of the canopy (LUEcanopy), which is a measure of the PAR conversion efficiency into photosynthetically fixed CO2; (2) the FAPAR of the canopy (FAPARcanopy), which is estimated using radiative transfer models and remote sensing data or using empirical relationship between FAPARcanopy and the normalized difference vegetation index (NDVI, Tucker, 1979); and (3) the incident PAR where:
Consequently, accurate estimates of FAPAR and LUE for ecosystems are essential for obtaining accurate GPP.
The LUE concept was initially developed for agricultural crops at harvest to determine the conversion efficiency of available light into biomass (g C dry mass) over a full growing season and is typically expressed in units such as g C MJ− 1 PAR (Monteith, 1972, Monteith, 1977). This seasonal crop-level LUE represents a direct measure of the average conversion efficiency of all above ground plant material (Gower et al., 1999), which is dominated by foliage for agricultural crops. Eddy covariance flux towers have the capability to provide near-continuous measurements of GEP — denoted as GEPtower, and absorbed PAR — denoted as APARtower (see Section 3.2.3 for more details, also see Krishnan et al., 2006), for an entire ecosystem for time periods as short as 30 min. Consequently, these instrumented flux towers also provide near-continuous measurements of LUE, denoted as LUEtower, over these same time periods as:
The LUEtower is typically expressed as µmol CO2 µmol− 1 PAR or µmol C µmol− 1 PPFD, where PPFD is the photosynthetic photon flux density. For these tower-based calculations, incident PAR is measured directly by radiometers attached to the flux tower and the FAPAR estimate is based both on detailed canopy structural measurements and on the radiometer measurements (Barr et al., 2007). An underlying assumption supporting the LUE retrieval through the MODIS modeling approach is that the LUEcanopy used in the models is a good approximation of LUEtower, at least when the measurement footprint of the tower is roughly equivalent to the area of the remote sensing pixel. Apparent ecosystem LUE obtained from flux tower measurements (i.e., LUEtower) directly describes the integrated response of the whole ecosystem to the prevailing environmental conditions, as do remotely acquired spectral snapshots although these latter are limited to specific acquisition times and viewing configurations.
On a canopy or ecosystem scale, GEP and APAR are typically linearly related (e.g., Waring et al., 1995), so that LUE can be determined from the slope of this relationship. This apparent linearity results from multiple scattering within the canopy, which involves 3-D contributions from foliage of multiple species and illumination conditions, as well as non-photosynthetic material (e.g., limbs, trunks, cones, litter). On the other hand, comparable light response curves for individual leaves of selected species yield non-linear responses for which the initial slope of the linear portion of the curve describes the quantum efficiency (Mohr et al., 1995). The quantum efficiency of individual leaves can also serve as an input to carbon cycle models but a means of scaling it to the canopy level is still required.
A common modeling approach is to set a maximum LUE for optimal environmental conditions (i.e., unstressed vegetation) and to simulate ecosystem responses when unfavorable environmental conditions occur (e.g., limitations of temperature, humidity, soil moisture, etc.) through down-regulation of the maximum LUE to achieve an apparent LUE (Medlyn, 1998). This is the approach used for the MODIS GPP product, an output of the PSN model. This GPP product has been compared with measurements made at flux towers by several research groups. For instance, Turner et al., 2003, Turner et al., 2004, Turner et al., 2006 found that the annual MODIS GPP totals calculated using MODIS standard photosynthesis products for a deciduous forest in Massachusetts, USA, matched well with the annual GEP totals from the flux tower. However, the seasonal time course of MODIS GPP dynamics differed significantly from the GEP measured by the flux tower (GEPtower) suggesting that a more physiologically realistic method of estimating GPP could be useful.
Even though maximum leaf LUE can be strongly influenced by leaf chlorophyll concentration (e.g., Waring et al., 1995), it is less clear how canopy chlorophyll concentration might influence apparent LUE at the ecosystem scale. Laboratory studies (Yoder and Waring, 1994) have shown that variation in canopy total chlorophyll content of miniature Douglas fir canopies was significantly correlated with their photosynthesis, although the correlation was higher for canopies exposed to full sun. Several other studies have shown a relationship between leaf or canopy nitrogen concentration and light use efficiency at the ecosystem scale (Kergoat et al., 2008, Ollinger et al., 2008), we believe that remote sensing techniques that evaluate chlorophyll rather than nitrogen could have even greater potential for estimating ecosystem light use efficiency and GPP.
From a biochemical perspective, only the PAR absorbed by photosynthetic pigments (designated as APARchl) enables photosynthetic processes, whereas the PAR absorbed by non-photosynthetic components such as boles, branches, stems, and litter is not used for CO2 fixation. We designate chlorophyll-based FAPAR here as FAPARchl. By definition, APARcanopy (the product of FAPARcanopy and PAR) is greater than APARchl (the product of FAPARchl and PAR). For linking to remote sensing applications, estimates of APARchl should provide more realistic GEPtower and LUEtower values than similar estimates using APARcanopy. We define LUE based on APARchl versus APARcanopy as follows:
In earlier studies (Zhang et al., 2005, Zhang et al., 2006), an approach to estimate FAPARchl was proposed using daily MODIS data. Since then, we have refined our algorithm to retrieve FAPARchl from MODIS imagery using the modified PROSPECT-SAIL2 model, PROSAIL-2 (Zhang et al., 2005, Zhang et al., 2006). The new version of this algorithm provides a statistical distribution of likely FAPARchl values for each cloud-free MODIS observation.
In this article, we combine five years of flux, meteorological, and remote sensing data from a boreal aspen flux site to attain the following four objectives: (1) to present a method for estimating FAPARchl and FAPARcanopy using single-date, scaled-up MODIS observations; (2) to apply the FAPARchl and FAPARcanopy algorithms to MODIS data acquired for 2001–2005 over this aspen flux site in Saskatchewan; (3) to link our estimates of MODIS FAPARchl and FAPARcanopy to the tower-based observations of PAR and GEP so as to derive LUE on both a unit chlorophyll area basis (LUEchl, Eq. (3) above) and for the whole canopy (LUEcanopy, Eq. (4) above); and (4) to compare our LUEchl, LUEcanopy and tower-based LUE estimates (i.e., LUEtower) to see if the LUEchl could provide a more physiologically realistic input to land surface process models. For this latter objective, we test the hypotheses that: (i) LUEcanopy = LUEchl; (ii) LUEchl = LUEtower; and (iii) LUEcanopy = LUEtower.
Section snippets
Southern Old Aspen
The Southern Old Aspen forest (SOA) was established in 1919 after a forest fire in Prince Albert National Park at the southern edge of the Canadian boreal forest (Barr et al., 2007). The eddy flux tower site (53.7°N, 106.2°W, 600 m elevation) is located ~ 50 km northwest of Prince Albert, Saskatchewan. SOA originated as part of the BOReal Ecosystem Atmosphere Study (BOREAS), and has continued operations under the Boreal Ecosystem Research and Monitoring Sites (BERMS) project and the Canadian
Spatial integration of MODIS data
A possible source of discrepancy between tower-based photosynthesis and the MODIS standard GPP is the way the MODIS reflectance products are used to calculate FAPARcanopy and the MODIS standard GPP (Justice et al., 1998, Wolfe et al., 1998). This occurs because the areal coverage of the MODIS products used in the GPP calculations are not constant over the growing season and may not match the footprint of the flux tower site. The MODIS observations made at multiple times over a target actually
PROSAIL-2 derived canopy variables including APARchl and APARcanopy
The posterior distributions of SOA canopy variables from the PROSAIL-2 model display seasonal variation, as shown for nine of the seventeen MODIS daily 5 × 5 observations in 2005 (Table 3). The other observations had similar distributions as those shown in Table 3 so they are not presented. Several canopy variables are shown: LAI, CF, total photosynthetic pigment content, water content, dry matter content, FAPARcanopy, and FAPARchl. Differences in total chlorophyll concentration were related to
Discussion
LUEchl captured more seasonal (Table 3) and inter-annual variation than LUEcanopy and provided an improved overall relationship to GEP (Fig. 3, Fig. 4). Krishnan et al. (2006) reported that the annual average LUEtower at the SOA tower for 2001–2005 was 0.0229–0.0302 µmol C µmol− 1 PPFD. Their in situ average LUEtower estimate matched well with our average MODIS-derived LUEchl (0.0229–0.0302 versus 0.0225–0.0310 µmol C µmol− 1 PPFD from Fig. 4) but not with the MODIS-derived LUEcanopy over five
Acknowledgments
This project was supported by the NASA Carbon Cycle Science Program (Dr. Diane Wickland, Program Manager) under the auspices of the North American Carbon Program. Support was provided to the FCRN and BERMS programs through several Canadian funding sources (Environment Canada, the Canadian Forest Service, NSERC, CFCAS, BIOCAP, Action Plan 2000 and PERD).
References (45)
- et al.
Leaf area index measurements at Fluxnet-Canada forest sites
Agricultural and Forest Meteorology
(2006) - et al.
Direct and indirect estimation of leaf area index, fAPAR and net primary production of terrestrial ecosystems
Remote Sensing of Environment
(1999) - et al.
Ecophysiological controls on the carbon balances of three southern boreal forests
Agricultural and Forest Meteorology
(2003) - et al.
Development of vegetation and soil indexes for Modis-Eos
Remote Sensing of Environment
(1994) - et al.
A generalized model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning
Forest Ecology and Management
(1997) - et al.
A global terrestrial monitoring network integrating tower fluxes, flask sampling, ecosystem modeling and EOS satellite data
Remote Sensing of Environment
(1999) - et al.
Age-related decline in forest productivity: Pattern and process
Advances in Ecological Research
(1997) Red and photographic infrared linear combinations for monitoring vegetation
Remote Sensing of Environment
(1979)- et al.
Scaling Gross Primary Production (GPP) over boreal and deciduous forest landscapes in support of MODIS GPP product validation
Remote Sensing of Environment
(2003) - et al.
The normalized difference vegetation index of small Douglas-fir canopies with varying chlorophyll concentrations
Remote Sensing of Environment
(1994)