Overview..
The net flux of carbon between the atmosphere and terrestrial vegetation can be expressed on an annual basis in terms of net biomass accumulation, or net primary production (NPP). The NPP of boreal forests is particularly important because their carbon balance is suspected of playing a significant role in the global carbon cycle, including their capacity to act as a sink for atmospheric CO2. Boreal forests are also more sensitive than most ecosystems to predicted changes in global climate. A large inter-disciplinary field experiment was recently conducted to address these questions (BOREAS).
Several methods of estimating NPP over large areas have been established. Some are based on complex ecophysiology models that link carbon and nutrient cycles, while others utilize remotely sensed data to provide information on vegetation condition and to monitor changes in leaf area index or canopy light absorption through time.
The linkage of remote sensing and ecophysiology provide a spatially explicit means to monitor short-term variations in photosynthetic capacity through limitations imposed by the current light, moisture and nutrient regimes.
It has been shown that remote sensing can also be used in relatively simple modeling frameworks to estimate global NPP of terrestrial vegetation using the relationship between reflectance properties and absorption of photosynthetically active radiation (PAR), if net conversion efficiencies can be approximated or assumed nearly constant. As a result, the efficiency with which canopy PAR absorption is converted to dry matter (biomass) has become the subject of much recent research. Theoretical arguments suggest that, at least in relatively stable vegetation assemblages, values of whole ecosystem conversion efficiency should converge irrespective of life-form as a result of resource-use optimization driven by evolutionary adaptation (another paper evaluates this hypothesis). The hypothesis is supported by field measurements of net ecosystem CO2 exchange in relation to PAR absorption, which has been shown to be remarkably similar among tropical, temperate and boreal forests.
Contrary to these findings, PAR utilization efficiency (or radiation use efficiency) has been shown to vary due to the combined effects of growth stage, nutrients, pathogens and drought stress. A literature review of published values of above-ground PAR utilization efficiency in a variety of crops demonstrated a relatively wide range of values (0.95 - 3.72 g MJ-1) (review paper). Ecophysiological growth models driven by climate and remotely sensed estimates of leaf area index and leaf duration also simulate a relatively large range in epsilon for total production (0.4 to 3.9 g MJ-1) over a range of ecosystems. These results suggest that ecosystems are not utilizing PAR in a similar manner on annual time scales. However, because values of PAR utilization efficiency reported in the literature are rarely representative of whole ecosystems, it would be premature to reject the functional convergence hypothesis or results of analyses that assume it. For example, understory and ground cover vegetation, herbivory, decomposition, and below-ground production are rarely included in field measurements. Furthermore, estimates of canopy PAR absorption are frequently obtained from a highly simplified exponential absorption model (Bougher-Beers law) which may be significantly in error in some situations.
Objectives and Approach..
The objective of the study described here (see an abstract of the original paper) was to estimate canopy PAR absorption with high spatial-resolution remotely-sensed data, and to determine the degree to which the PAR utilization varies among stands of a broadleaf deciduous species (Populus tremuloides) and a needleleaf evergreen species (Picea mariana) spanning a range of age classes and site conditions in the Superior National Forest of NE Minnesota.
In order to model NPP and PAR utilization with remote sensing observations it is necessary to characterize different components of the systems being studies. These include canopy phenological dyanmics, fractional instantaneous canopy light interception and integrated annual canopy light interception.
Phenological Dynamics
Seasonal greenness spectral vegetation index (Gi) models (profiles) were generated for all stands. The phenological variation in spruce stands as depicted by variations in Gi was negligible in closed-canopy stands, whereas discontinuous stands exhibit a slight seasonality due to variations in the amount of sphagnum moss background illuminated as solar zenith angle varies through the year. In contrast, aspen stands exhibited large phenological variation in Gi, particularly during rapid leaf expansion and gradual leaf senescence.
Some poor goodness-of-fit statistics are associated with spruce phenology models as a result of the small variation in Gi through the growing season, hence a low dependence of Gi on growing degree day. This is particularly true at a few spruce sites which were essentially invariant through the growing season.
Fractional Canopy Light Interception (Fpar)
Canopy model simulations of Gi and related values of the fraction of incident photosynthetically active radiation (PAR) intercepted by the canopy (Fpar) were derived for each tree functional type (spruce vs aspen), as well as for all aspen stands to characterize understory variability in reflectance. In sites with large understory contributions to stand reflectance, canopy Fpar increased less rapidly with increasing Gi than in those sites where both stand Gi and Fpar were dominated by the canopy. In contrast to aspen, spruce stands exhibited an inverse relationship between Gi and Fpar. In spruce, as the density grades from a closed to a discontinuous canopy, the amount of illuminated sphagnum moss increased, which acted to increase Gi. As a result, an increase in Gi reflected a decrease in canopy Fpar. This result is analogous to an opaque media obscuring a highly reflective background.
Variation in both canopy reflectance and Fpar in the spruce stands was much greater between stands than in aspen. Whereas variation in Gi in the aspen stands was driven by phenological dynamics through the growing season, variation in the spruce stands was driven primarily by between-stand differences in stem density. This result is consistent with the observation that stem density and LAI in black spruce stands co-vary, whereas aspen stands maintain nearly constant LAI despite decreasing stem-density with age (self-thinning).
Peak canopy Fpar varied between 0.50 - 0.80 in young stands (20 year or less) and between 0.56 - 0.86 in mature stands (47 years or greater). Those stands with a small understory contribution to stand reflectance were generally characterized by low Gi values relative to stands with a high understory contribution to stand reflectance. The relationships between Fpar and vegetation indices can be concealed by such variations in stand reflectance originating from understory vegetation. Thus canopy Fpar - Gi relationships must incorporate characterization of understory components.
Annual Canopy PAR Interception (IPAR)
Much of the variation in phenological dynamics and between-stand differences in Fpar are reflected in IPAR, and in annual PAR utilization values for aspen and spruce. IPAR varied over a comparable range in aspen (374 - 1089 MJ m-2 yr-1) and spruce (125 - 1057 MJ m-2 yr-1). In initial runs of the canopy Fpar model that did not incorporate the understory component of aspen stands, there was relatively little variation in IPAR. Only after the understory contribution to stand reflectance was characterized did a range of IPAR emerge, therefore only simulations incorporating an understory component were used. Similar observations have been noted in the estimation of aspen LAI with remotely sensed data.
In spruce, IPAR also covaried with stand stem density, with a gradual increase in PAR interception as stem density increased. In contrast we found that the self-thinning nature of aspen stands resulted in variations in canopy PAR interception that were largely unrelated to variations in stem density.
PAR Utilization Efficiency
In examining the relationship between IPAR and NPP for all stands it is evident that the variation in NPP relative to a given value of IPAR was greater in aspen than in spruce, particularly in the mature aspen stands. Simple linear regression of IPAR on NPP for all stands provided some predictive capability based on IPAR alone (explaining 59% of variation in NPP), but the residuals were large (averaging 195 g m-2 yr-1). In spruce stands, IPAR was a moderately good predictor of AANPP (explaining 69% of the variation), and the mean error was reduced to 93 g m-2 yr-1. Elimination of a single outlier spruce stand increased the explained variance to 76% and reduced the error to 81 g m-2 yr-1. In aspen stands, IPAR was a relatively poor predictor of NPP (49% of variance explained) and residual errors were large (154 g m-2 yr-1). Stratification of the aspen into young versus mature stands improved the relationship in young stands (77%) but reduced it in the mature stands (18%).
Relatively large variability in IPAR in relation to NPP strongly suggests that factors other than light are limiting production. This conclusion was most apparent in the mature aspen stands, and may be due in part to the the fact that NPP in mature aspen was confined to a smaller range of values (538 - 985 g m-2 yr-1) than the young stands (190 - 1199 g m-2 yr-1). Elimination of less productive stands through competitive replacement may explain this observation. The poor relationship between IPAR and NPP in mature aspen was probably a result of changes in the respiratory requirements of mature trees (discussed more later, and more generally in another paper - see a synopsis).
Total above-ground PAR utilization varied considerably among both spruce and aspen stands. The range in PAR utilization for aspen (0.44 - 1.29 g MJ-1) was fully expressed over the young stands alone. This was comparable to, although higher than, the range in spruce (0.17 - 0.89 g MJ-1). The average values of PAR utilization were 0.92 (+/-0.22 ) in aspen and 0.49 (+/-0.17 ) g MJ-1 in spruce. Differences in PAR utilization between the two species may reflect differences in life history, particularly differences in the energy requirements associated with different resource allocation strategies (an idea explored further here).
Simulation studies suggest that interactions between carbon and nitrogen dynamics are related through climate variables such that temperature and moisture variations result in increasing nitrogen limitation of productivity from tropical to boreal to tundra ecosystems. To determine if this interaction between climate, nutrient availability and production is reflected in PAR utilization we compared our values with those for other forest stands for which such data were available.
All data are reported in g dry matter per MJ PAR intercepted or absorbed, as indicated.
Summary and Implications of Results
Annual estimates of PAR interception derived from high spatial resolution satellite data, Fpar and temperature-driven phenology models of boreal forest stands were used to examine the capability of remote sensing to estimate NPP and dry matter yields of energy for aspen and spruce in a range of age classes and site conditions. Relationships between Gi and Fpar were species-specific and highly dependent on background and understory vegetation. Thus assumptions of canopy PAR interception proportional to SVI may result in large errors in the estimation of NPP in situations where NPP data are not available for all constituants that contribute to stand reflectance.
Errors in the NPP estimates may also result when foliage display is decoupled from annual production, which can occur through intermittent resource limitations (predominantly moisture stress) and respiration losses associated with stand age. The weakest relationship between IPAR and NPP occured in mature aspen stands, a result which is consistent with decreased photosynthetic gains relative to respiratory costs. In contrast, strong positive relationships between IPAR and NPP were noted in lowland spruce stands and in young aspen stands, and there was no age dependence in the relationship for spruce. Remotely sensed estimates of light interception for these stands can be used with confidence to extend NPP estimates over larger areas if canopy interception is properly estimated from SVI.
These results suggest life-form differences (e.g., broadleaf deciduous versus needleleaf evergreen) should be considered in the remote sensing of NPP, not only in the development of Fpar models, but also in quantifying radiation use efficiency. Physiological differences between life-forms has also been noted as a primary determinant of net ecosystem production in ecophysiological modeling of NPP.
Methods to incorporate respiration and stress terms in remote sensing of NPP have been proposed (Prince 1991) and demonstrated to be technically feasible (Global Production Efficiency Model).
The ranges in PAR utilization we observed were within the range of previously reported values for forest ecosystems, particularly temperate species. Reasons for observed variations are discussed above and are being explored in follow-on work. Areal estimates of net carbon uptake over the 2280 km2 study area, derived from the IPAR data and median values of PAR utilization, compare favorably with estimates of net ecosystem production derived for another boreal forest ecosystem.
Within-species variations in PAR utilization of the magnitude reported here suggest that optimization of resource use through "functional convergence" may not result in a convergence on a very narrow range of values. Literature survey and ecophysiology model simulations of PAR utilization values in a variety of ecosystems support this conclusion (see a synopsis of related work), but a definitive test will not be possible until more extensive coincident NPP, IPAR and Fpar measurements are available for a variety of ecosystems. Some of these data, and insight into the unresolved questions raised here, are being addressed by related work in Central Alaska and the Boreal Ecosystem Atmosphere Study (BOREAS) among others.
The research discussed here has been published in the following publications:
S.J.Goetz and S.D.Prince, 1996, Remote sensing of net primary production in boreal forest stands Agricultural and Forest Meteorology, 78:149-179.
Goetz, S. J. and S. D. Prince. 1998. Variability in carbon exchange and light utilization among boreal forest stands: implications for remote sensing of net primary production,, Canadian Journal of Forest Research 28(3) : 375-389.
Goetz, S. J. and S. D. Prince. 1998. Modeling terrestrial carbon exchange and storage: evidence and implications of functional convergence in light use efficiency,, Advances in Ecological Research 28 : 57-92.