As shown in Table 5.2, the RMSD values of soybeans were lower than 2 g C
m
-2 d
-1 across multiple years (1.9 and 1.5 g C m
-2 d
-1 for 2002 and 2004,
respectively). The relative error (%RE) values between the seasonal sums of
GPP VPM and GPP EC were 0.4 and -1.8 % in 2002 and 2004, respectively.
The above simulation results show that the VPM, based on the concept of light
absorption by chlorophyll, had great potential to estimate the seasonal dynamics
and interannual variation of GPP for soybean (C 3 ) and maize (C 4 ) cropland
rotation in different years. Another earlier study (Yan et al. 2009) used the VPM
model to estimate GPP of winter wheat (C 3 ) and maize (C 4 ) rotation within a year
in the North China Plain, China, and also reported good agreement between
GPP VPM and GPP EC in the cropland site (Yan et al. 2009).
5.6 Future Research Directions
Remotely sensed data and PEMs have been widely used to estimate GPP and NPP
over the last few decades, as several coarse and moderate resolution sensors (e.g.,
AVHRR, SPOT-VEGETATION, MODIS, and MERIS) provide images for the globe
every day. However, further development and implementation of satellite-based
PEMs still face three major challenges for more accurate estimation of GPP and NPP.
First, the uncertainty concerning the scaling-up of light absorption from chlorophyll and leaf levels to canopy, ecosystem and landscape levels is still significant, and how to accurately estimate the amount of light absorbed by chlorophyll is
a challenging task from the perspectives of both field measurements and radiative
transfer models (Zhang et al. 2005, 2006, 2009). The concept of FPAR at the
canopy level (FPAR canopy ) has been widely adopted by the remote sensing and
ecosystem modeling communities, and FPAR canopy is often calculated as a semiempirical linear function of NDVI and leaf area index (LAI). It is noted that only
light absorbed by chlorophyll pigments is used for photosynthesis, and the concept
of FPAR chl is more meaningful for GPP estimation. Simulation results from a
coupled leaf-canopy radiative transfer model have suggested that FPAR canopy is
much higher than FPAR leaf or FPAR chl (Zhang et al. 2006). Recently, FPAR chl has
been used for the development of several new PEMs and updated versions of
existing PEMs (Potter 2012; Sims et al. 2006b, 2008 Wu et al. 2012; Xiao et al.
2005b; Yan et al. 2009). In these PEMs, FPAR chl is often estimated as the linear
function of EVI or other chlorophyll-related vegetation indices, which have proved
more consistent with the light absorption for photosynthesis at the chlorophyll
level. Nevertheless, difficulties still exist in accurately quantifying leaf chlorophyll
content and FPAR chl . The empirical and semiempirical relationships between
vegetation indices and FPAR chl , (for example, FPARchl = a 9 EVI), need to be
evaluated across various biome types with accurate in situ measurement, better
implementation of radiative transfer models (both model variables and parameters), and selection of satellite images from various sensors (e.g., MODIS, MERIS
and RapidEye). It is necessary to regularly measure both biochemical (chlorophyll
142
X. Xiao et al.
m
-2 d
-1 across multiple years (1.9 and 1.5 g C m
-2 d
-1 for 2002 and 2004,
respectively). The relative error (%RE) values between the seasonal sums of
GPP VPM and GPP EC were 0.4 and -1.8 % in 2002 and 2004, respectively.
The above simulation results show that the VPM, based on the concept of light
absorption by chlorophyll, had great potential to estimate the seasonal dynamics
and interannual variation of GPP for soybean (C 3 ) and maize (C 4 ) cropland
rotation in different years. Another earlier study (Yan et al. 2009) used the VPM
model to estimate GPP of winter wheat (C 3 ) and maize (C 4 ) rotation within a year
in the North China Plain, China, and also reported good agreement between
GPP VPM and GPP EC in the cropland site (Yan et al. 2009).
5.6 Future Research Directions
Remotely sensed data and PEMs have been widely used to estimate GPP and NPP
over the last few decades, as several coarse and moderate resolution sensors (e.g.,
AVHRR, SPOT-VEGETATION, MODIS, and MERIS) provide images for the globe
every day. However, further development and implementation of satellite-based
PEMs still face three major challenges for more accurate estimation of GPP and NPP.
First, the uncertainty concerning the scaling-up of light absorption from chlorophyll and leaf levels to canopy, ecosystem and landscape levels is still significant, and how to accurately estimate the amount of light absorbed by chlorophyll is
a challenging task from the perspectives of both field measurements and radiative
transfer models (Zhang et al. 2005, 2006, 2009). The concept of FPAR at the
canopy level (FPAR canopy ) has been widely adopted by the remote sensing and
ecosystem modeling communities, and FPAR canopy is often calculated as a semiempirical linear function of NDVI and leaf area index (LAI). It is noted that only
light absorbed by chlorophyll pigments is used for photosynthesis, and the concept
of FPAR chl is more meaningful for GPP estimation. Simulation results from a
coupled leaf-canopy radiative transfer model have suggested that FPAR canopy is
much higher than FPAR leaf or FPAR chl (Zhang et al. 2006). Recently, FPAR chl has
been used for the development of several new PEMs and updated versions of
existing PEMs (Potter 2012; Sims et al. 2006b, 2008 Wu et al. 2012; Xiao et al.
2005b; Yan et al. 2009). In these PEMs, FPAR chl is often estimated as the linear
function of EVI or other chlorophyll-related vegetation indices, which have proved
more consistent with the light absorption for photosynthesis at the chlorophyll
level. Nevertheless, difficulties still exist in accurately quantifying leaf chlorophyll
content and FPAR chl . The empirical and semiempirical relationships between
vegetation indices and FPAR chl , (for example, FPARchl = a 9 EVI), need to be
evaluated across various biome types with accurate in situ measurement, better
implementation of radiative transfer models (both model variables and parameters), and selection of satellite images from various sensors (e.g., MODIS, MERIS
and RapidEye). It is necessary to regularly measure both biochemical (chlorophyll
142
X. Xiao et al.