46
Steven W. Running, Peter E. Thornton, Ramakrishna Nemani, and Joseph M. Glassy
tissue. Prince (1991) summarized values of E for
herbaceous vegetation to be from 1.0 to 1.8
g C MJ - 1 for plants with the C 3 photosynthetic
pathway, and higher E for plants with the C 4 photosynthetic pathway. Running and Hunt (1993) and
Hunt (1994) found published E values for woody
vegetation were lower, from about 0.2 to 1.5
g C MJ - 1. and hypothesized that this was the result
of respiration from the living cells in the sapwood
of woody stems.
The second source of variability in E is due to
suboptimal climatic conditions. To extrapolate
Monteith's original theory, designed for wellwatered crops only during the growing season, to
perennial plants living year around, certain severe
climatic constraints must be treated. Evergreen vegetation, such as conifer trees or schlerophyllous
shrubs, absorb PAR all during the nongrowing season, yet during subfreezing temperatures execute
minimal photosynthetic activity, because stomata
are closed by frozen water in the leaves (Waring et
al. 1995; Waring and Running 1998). So, as a
global generalization, we truncate GPP on days
when the minimum temperature is below O°c. Additionally, high vapor pressure deficits, >20 hPa,
have been shown to induce stomatal closure in unirrigated native plants. This level of daily atmospheric water deficit is commonly reached in semiarid regions of the world for much of the growing
season. So, our algorithm mimics this physiological
control by progressively limiting daily GPP when
high vapor pressure deficits are computed from the
surface meteorology.
Other biophysical factors regularly constrain the
conversion of APAR into biomass growth. All
plants have an optimum temperature for photosynthesis, but acclimation is so rapid that a global general temperature limit cannot be easily defined. Progressive seasonal soil water stress also limits plant
growth, but cannot be quantified without a water
balance calculation. For our global logic here, we
assume that continuous soil water stress ultimately
results in reduction in vegetation leaf area, which
is then reflected in the NDVI or FPAR terms of
Equations 3.5 and 3.6 (Nemani and Running 1989).
Finally, we also assume nutrient constraints on vegetation growth to be quantified by limiting leaf area,
rather than attempting to compute a constraint
through E. This is not entirely accurate, as ranges
of leaf nitrogen and photosynthetic capacity occur
in all vegetation types (Schulze et al. 1994).
Parameterization of € with Global
BIOME-BGC Simulations
In order to quantify this biome- and climateinduced range of E, we have simulated global NPP
in advance with a complex ecosystem model,
BIOME-BGC, and computed the E, or conversion
efficiency, from APAR to final NPP (Fig. 3.1). This
complete global range of E is incorporated into a
biome parameter look-up table (BPLUT) (Table
3.1). Operating this simple PAR conversion efficiency model rather than the complex ecosystem
model to compute NPP is roughly 100 times faster
in computer execution, an important consideration when 150 million cells must be computed each
day.
To capture as much of the global range of variation in climate and landcover as possible for determining E, we performed daily simulations over
a global lOx 1 ° grid. Meteorological input was
from the 14-year daily data set developed at Scripps
Oceanographic Institute (Piper and Stewart 1996).
Gridded soil physical parameters were estimated
from the Zobler soil texture database (Zobler,
1986). The standard at-launch EOS global l-km
landcover data set was obtained (Hansen et al.
1999). This consisted of two separate data sets: a
discrete classification of biome types that is also the
at-launch MODIS landcover product, and a continuous fields product that describes the fractional
cover of woody species in each grid cell and
also discriminates woody cover fraction by leaf
longevity and leaf morphology (DeFries et al.
1999). For each grid cell, a separate (independent)
BIOME-BGC simulation was performed for each
functional vegetation type having greater than 1%
of the total vegetated land area within the cell. The
Scripps meteorological data set contains 14 years
of daily gridded surfaces for maximum and minimum temperature and precipitation. We used these
variables to estimate daily humidity (Kimball et al.
1997a) and radiation (Thornton and Running 1999)
over the grid. Day length was calculated for each
grid cell from standard geometrical relationships.
The interactions between the carbon and water
budgets of the BIOME-BGC model have been described in considerable detail elsewhere (Running
Steven W. Running, Peter E. Thornton, Ramakrishna Nemani, and Joseph M. Glassy
tissue. Prince (1991) summarized values of E for
herbaceous vegetation to be from 1.0 to 1.8
g C MJ - 1 for plants with the C 3 photosynthetic
pathway, and higher E for plants with the C 4 photosynthetic pathway. Running and Hunt (1993) and
Hunt (1994) found published E values for woody
vegetation were lower, from about 0.2 to 1.5
g C MJ - 1. and hypothesized that this was the result
of respiration from the living cells in the sapwood
of woody stems.
The second source of variability in E is due to
suboptimal climatic conditions. To extrapolate
Monteith's original theory, designed for wellwatered crops only during the growing season, to
perennial plants living year around, certain severe
climatic constraints must be treated. Evergreen vegetation, such as conifer trees or schlerophyllous
shrubs, absorb PAR all during the nongrowing season, yet during subfreezing temperatures execute
minimal photosynthetic activity, because stomata
are closed by frozen water in the leaves (Waring et
al. 1995; Waring and Running 1998). So, as a
global generalization, we truncate GPP on days
when the minimum temperature is below O°c. Additionally, high vapor pressure deficits, >20 hPa,
have been shown to induce stomatal closure in unirrigated native plants. This level of daily atmospheric water deficit is commonly reached in semiarid regions of the world for much of the growing
season. So, our algorithm mimics this physiological
control by progressively limiting daily GPP when
high vapor pressure deficits are computed from the
surface meteorology.
Other biophysical factors regularly constrain the
conversion of APAR into biomass growth. All
plants have an optimum temperature for photosynthesis, but acclimation is so rapid that a global general temperature limit cannot be easily defined. Progressive seasonal soil water stress also limits plant
growth, but cannot be quantified without a water
balance calculation. For our global logic here, we
assume that continuous soil water stress ultimately
results in reduction in vegetation leaf area, which
is then reflected in the NDVI or FPAR terms of
Equations 3.5 and 3.6 (Nemani and Running 1989).
Finally, we also assume nutrient constraints on vegetation growth to be quantified by limiting leaf area,
rather than attempting to compute a constraint
through E. This is not entirely accurate, as ranges
of leaf nitrogen and photosynthetic capacity occur
in all vegetation types (Schulze et al. 1994).
Parameterization of € with Global
BIOME-BGC Simulations
In order to quantify this biome- and climateinduced range of E, we have simulated global NPP
in advance with a complex ecosystem model,
BIOME-BGC, and computed the E, or conversion
efficiency, from APAR to final NPP (Fig. 3.1). This
complete global range of E is incorporated into a
biome parameter look-up table (BPLUT) (Table
3.1). Operating this simple PAR conversion efficiency model rather than the complex ecosystem
model to compute NPP is roughly 100 times faster
in computer execution, an important consideration when 150 million cells must be computed each
day.
To capture as much of the global range of variation in climate and landcover as possible for determining E, we performed daily simulations over
a global lOx 1 ° grid. Meteorological input was
from the 14-year daily data set developed at Scripps
Oceanographic Institute (Piper and Stewart 1996).
Gridded soil physical parameters were estimated
from the Zobler soil texture database (Zobler,
1986). The standard at-launch EOS global l-km
landcover data set was obtained (Hansen et al.
1999). This consisted of two separate data sets: a
discrete classification of biome types that is also the
at-launch MODIS landcover product, and a continuous fields product that describes the fractional
cover of woody species in each grid cell and
also discriminates woody cover fraction by leaf
longevity and leaf morphology (DeFries et al.
1999). For each grid cell, a separate (independent)
BIOME-BGC simulation was performed for each
functional vegetation type having greater than 1%
of the total vegetated land area within the cell. The
Scripps meteorological data set contains 14 years
of daily gridded surfaces for maximum and minimum temperature and precipitation. We used these
variables to estimate daily humidity (Kimball et al.
1997a) and radiation (Thornton and Running 1999)
over the grid. Day length was calculated for each
grid cell from standard geometrical relationships.
The interactions between the carbon and water
budgets of the BIOME-BGC model have been described in considerable detail elsewhere (Running