ecosystem models has been demonstrated in several land cover types, such as
grasslands (Cayrol et al. 2000; forests (Hasenauer et al. 2012; Liu et al. 1997;
Ranson et al. 2001), and croplands (Bouman 1992; Clevers and van Leeuwen 1996;
Guerif and Duke 2000; Maas 1988; Weiss et al. 2001).
For Canadian mixed grassland ecosystems, a conceptual remote sensing-based
BIOME-BGC model simulating spatially explicit mixed grassland productivity has
already been developed (He 2008). However, there still remains a question to be
addressed within this modeling framework: to what extent and with what limitations can the critical parameter(s) required by the model be derived from available
remote sensing data? A critical parameter in a model is one in which minimal
changes to its value would generate major changes in model output (Makler-Pick
et al. 2011). When high uncertainty in a parameter coincides with high sensitivity of
the model to that parameter, model predictions may not be reliable.
Foliar carbon to nitrogen ratio (C:N) drives terrestrial biogeochemical processes
such as decomposition and mineralization, and is thus one of the most important
parameters that significantly controls NPP in the BIOME-BGC model (White
et al. 2000). Research conducted by Psomas et al. (2008) in semi-natural grassland
types in the Central region of the Swiss Plateau also indicated that NPP estimates
using spatial estimates of foliar C:N derived from remote sensing data are significantly different from those produced when single C:N values representing individual land cover classes were used.
Since foliar C:N ratio within the current Biome-BGC model is assumed to be
constant for a given biome and given that it varies dramatically over space for
different species (Psomas et al. 2008), more spatially accurate information regarding grassland spatial heterogeneity of this key parameter (i.e. foliar C:N ratio)
obtained from remote sensing is needed to improve model predictions. Thus, the
objectives of this research are to: (1) develop methodology for the estimation of
spatially distributed foliar C:N ratio from remote sensing data; (2) test the sensitivity of the model to foliar C:N ratio; and (3) evaluate spatial C:N ratio driven
modelling results.
8.2 Study Area and Field Data
The study area is located in the West Block of the Grasslands National Park (GNP)
and its surroundings in southwest Saskatchewan, Canada (Fig. 8.1). A detailed
description of this area could be found from He (2014). Field data were collected in
mid-June 2005, the approximate date of peak growing season. A total of 24 randomly selected sites were visited, 10 of which are located in upland areas, and the
remaining 14 in sloped areas. Each field site is limited to a homogeneous area of at
least 1 ha in size in order to accommodate positional errors. In each field site, fresh
biomass samples were collected from ten 50 Â 50 cm quadrats. A detailed description of field sampling design and field data collection protocol can be found in
8 Grassland Productivity Simulation: Integrating Remote Sensing. . .
157
grasslands (Cayrol et al. 2000; forests (Hasenauer et al. 2012; Liu et al. 1997;
Ranson et al. 2001), and croplands (Bouman 1992; Clevers and van Leeuwen 1996;
Guerif and Duke 2000; Maas 1988; Weiss et al. 2001).
For Canadian mixed grassland ecosystems, a conceptual remote sensing-based
BIOME-BGC model simulating spatially explicit mixed grassland productivity has
already been developed (He 2008). However, there still remains a question to be
addressed within this modeling framework: to what extent and with what limitations can the critical parameter(s) required by the model be derived from available
remote sensing data? A critical parameter in a model is one in which minimal
changes to its value would generate major changes in model output (Makler-Pick
et al. 2011). When high uncertainty in a parameter coincides with high sensitivity of
the model to that parameter, model predictions may not be reliable.
Foliar carbon to nitrogen ratio (C:N) drives terrestrial biogeochemical processes
such as decomposition and mineralization, and is thus one of the most important
parameters that significantly controls NPP in the BIOME-BGC model (White
et al. 2000). Research conducted by Psomas et al. (2008) in semi-natural grassland
types in the Central region of the Swiss Plateau also indicated that NPP estimates
using spatial estimates of foliar C:N derived from remote sensing data are significantly different from those produced when single C:N values representing individual land cover classes were used.
Since foliar C:N ratio within the current Biome-BGC model is assumed to be
constant for a given biome and given that it varies dramatically over space for
different species (Psomas et al. 2008), more spatially accurate information regarding grassland spatial heterogeneity of this key parameter (i.e. foliar C:N ratio)
obtained from remote sensing is needed to improve model predictions. Thus, the
objectives of this research are to: (1) develop methodology for the estimation of
spatially distributed foliar C:N ratio from remote sensing data; (2) test the sensitivity of the model to foliar C:N ratio; and (3) evaluate spatial C:N ratio driven
modelling results.
8.2 Study Area and Field Data
The study area is located in the West Block of the Grasslands National Park (GNP)
and its surroundings in southwest Saskatchewan, Canada (Fig. 8.1). A detailed
description of this area could be found from He (2014). Field data were collected in
mid-June 2005, the approximate date of peak growing season. A total of 24 randomly selected sites were visited, 10 of which are located in upland areas, and the
remaining 14 in sloped areas. Each field site is limited to a homogeneous area of at
least 1 ha in size in order to accommodate positional errors. In each field site, fresh
biomass samples were collected from ten 50 Â 50 cm quadrats. A detailed description of field sampling design and field data collection protocol can be found in
8 Grassland Productivity Simulation: Integrating Remote Sensing. . .
157