50
Steven W. Running, Peter E. Thornton, Ramakrishna Nemani, and Joseph M. Glassy
To reiterate, these intensive global simulations
with BIOME-BGC have been done in advance of
EOS launch to define the BPLUT with as much
ecological accuracy as possible (see Table 3.1).
Next, we will cover the simple implementation algorithm that will be executed within the EOS computing system to compute global NPP.
Algorithm Implementation Logic
in EOS
Each day the EOS data system will ingest calibrated
and atmospherically corrected reflectances from
each spectral channel of the MODIS sensor for each
cloud-free pixel. In order to make the daily GPP
computation for each global terrestrial pixel, three
other variables must first be retrieved: the biome
type, which changes no more than annually; the
FPAR, which can change weekly during rapid vegetation growth and senescence; and the daily surface climate conditions, which change diurnally.
Satellite-Derived Input Variables
Biome Classes
From earlier analyses, we define the major differences in vegetation biogeochemistry by defining
key biome classes with markedly different nutritional constraints and carbon cycling dynamics, often called functional types (Running et al. 1995).
The difference in our analysis, though, was to define these biome attributes from characteristics that
would be amenable to remote sensing (Nemani and
Running 1997). For example, deciduous foliage
and evergreen foliage have different lignin/nitrogen
ratios that control the rate of litter decomposition
and N mineralization. Grasslands and annual crops
have very aerodynamically smooth surfaces,
whereas forests have very aerodynamically rough
surfaces, affecting the energy, momentum, water
and CO 2 fluxes. Forests have large respiring masses
of living wood (sapwood), whereas grasslands and
annual crops do not. Broadleaf forests have different radiative transfer properties compared with
needleleaf forests. These differences in vegetation
biophysics and biogeochemistry are defined implicitly by the biome classification.
From sensltlVlty analyses with BIOME-BGC
and remote sensing data, we determined that seven
fundamental biome classes were needed to represent essential differences in biogeochemical cycling of different global vegetation types (Running
et al. 1995; Nemani and Running 1996). The standard global l-km biome classification for EOS defines 11 classes, incorporating some additional fractional cover combinations (see Table 3.1) (Fig. 3.2,
top) (Hansen et al. 1999). Grasslands may have either the C 3 or C 4 photosynthetic pathways, controlling Egross. We assume that C4 photosynthesis
occurs in tropical and arid climates and C 3 photosynthesis occurs in moist, temperate climates, and
use long-term climatic conditions to make this distinction globally (Piper and Stewart 1996).
The biome class for each pixel is planned to
be updated once each year to reflect changes in
globallandcover (John Townshend, personal communication). Consequently, each pixel will be reevaluated for biome changes resulting from deforestation and reforestation, new agriculture or
urbanization, large-scale disturbances, and successional development.
FPAR and LAI
For each pixel, an algorithm using the MODIS sensor measures the FPAR (the radiometric equivalent
of LA I) each day, and this is averaged over an 8-day
period (see Fig. 3.2, middle) (Myneni et al. 1997b).
As the spring growing season begins around the
world, emerging leaves will be quantified by increasing FPAR at each l-km pixel. Conversely, at
fall senescence or crop harvest, the FPAR will decline. Consequently, this FPAR will also provide a
measure of continental phenology in addition to
quantifying the vegetation canopy every 8 days for
the NPP algorithm (White et al. 1997).
Climatology Inputs from the DAD
The NASA Data Assimilation Office (DAO) , ingests all available surface weather observations
globally every 3 hours. The DAO then interpolates and grids these point data, runs a short-time
sequence global climate model, and produces surface climatic conditions at lOx 1
0
for the world.
From this data output, the NPP algorithm
retrieves
Steven W. Running, Peter E. Thornton, Ramakrishna Nemani, and Joseph M. Glassy
To reiterate, these intensive global simulations
with BIOME-BGC have been done in advance of
EOS launch to define the BPLUT with as much
ecological accuracy as possible (see Table 3.1).
Next, we will cover the simple implementation algorithm that will be executed within the EOS computing system to compute global NPP.
Algorithm Implementation Logic
in EOS
Each day the EOS data system will ingest calibrated
and atmospherically corrected reflectances from
each spectral channel of the MODIS sensor for each
cloud-free pixel. In order to make the daily GPP
computation for each global terrestrial pixel, three
other variables must first be retrieved: the biome
type, which changes no more than annually; the
FPAR, which can change weekly during rapid vegetation growth and senescence; and the daily surface climate conditions, which change diurnally.
Satellite-Derived Input Variables
Biome Classes
From earlier analyses, we define the major differences in vegetation biogeochemistry by defining
key biome classes with markedly different nutritional constraints and carbon cycling dynamics, often called functional types (Running et al. 1995).
The difference in our analysis, though, was to define these biome attributes from characteristics that
would be amenable to remote sensing (Nemani and
Running 1997). For example, deciduous foliage
and evergreen foliage have different lignin/nitrogen
ratios that control the rate of litter decomposition
and N mineralization. Grasslands and annual crops
have very aerodynamically smooth surfaces,
whereas forests have very aerodynamically rough
surfaces, affecting the energy, momentum, water
and CO 2 fluxes. Forests have large respiring masses
of living wood (sapwood), whereas grasslands and
annual crops do not. Broadleaf forests have different radiative transfer properties compared with
needleleaf forests. These differences in vegetation
biophysics and biogeochemistry are defined implicitly by the biome classification.
From sensltlVlty analyses with BIOME-BGC
and remote sensing data, we determined that seven
fundamental biome classes were needed to represent essential differences in biogeochemical cycling of different global vegetation types (Running
et al. 1995; Nemani and Running 1996). The standard global l-km biome classification for EOS defines 11 classes, incorporating some additional fractional cover combinations (see Table 3.1) (Fig. 3.2,
top) (Hansen et al. 1999). Grasslands may have either the C 3 or C 4 photosynthetic pathways, controlling Egross. We assume that C4 photosynthesis
occurs in tropical and arid climates and C 3 photosynthesis occurs in moist, temperate climates, and
use long-term climatic conditions to make this distinction globally (Piper and Stewart 1996).
The biome class for each pixel is planned to
be updated once each year to reflect changes in
globallandcover (John Townshend, personal communication). Consequently, each pixel will be reevaluated for biome changes resulting from deforestation and reforestation, new agriculture or
urbanization, large-scale disturbances, and successional development.
FPAR and LAI
For each pixel, an algorithm using the MODIS sensor measures the FPAR (the radiometric equivalent
of LA I) each day, and this is averaged over an 8-day
period (see Fig. 3.2, middle) (Myneni et al. 1997b).
As the spring growing season begins around the
world, emerging leaves will be quantified by increasing FPAR at each l-km pixel. Conversely, at
fall senescence or crop harvest, the FPAR will decline. Consequently, this FPAR will also provide a
measure of continental phenology in addition to
quantifying the vegetation canopy every 8 days for
the NPP algorithm (White et al. 1997).
Climatology Inputs from the DAD
The NASA Data Assimilation Office (DAO) , ingests all available surface weather observations
globally every 3 hours. The DAO then interpolates and grids these point data, runs a short-time
sequence global climate model, and produces surface climatic conditions at lOx 1
0
for the world.
From this data output, the NPP algorithm
retrieves