8.1 Introduction
Globally, grasslands are important in the study of terrestrial ecosystems as they
cover nearly 20 % of the Earth’s surface (Lieth 1978); contain approximately 30 %
of global carbon stocks (Ojima et al. 1996; Parton et al. 1996); and store at least
10 % of the global soil organic matter (Eswaran et al. 1993). There are approximately 24 M ha of mixed grasslands in Canada, serving a variety of economic,
environmental, and ecological purposes. In recent years, grassland degradation has
become a worldwide problem due to intense human activities and environmental
changes, and the mixed grasslands of Canada are no exception. As a result, mixed
grasslands have frequently been associated with fluctuating, and unreliable productivity (Curll et al. 1985a, b; Fothergill et al. 2000; He 2014; Laws and Newton 1992;
Orr et al. 1990; Schwinning and Parsons 1996a, b). To ensure the sustainable
development of Canadian mixed grasslands and to predict the cascading effects
of human activities and climate change on these grasslands, ecosystem process
modeling is required because it can simulate and predict vegetation productivity
and also project ecosystem response to a wide range of environmental conditions.
Over the past 30 years, a considerable number of ecosystem process models such
as BIOME-BGC (Running and Hunt 1993) and CENTURY (Parton et al. 1993)
have been developed to investigate many different aspects of ecosystems, including
vegetation productivity, changing vegetation distributions, and land carbon sinks
(Adams et al. 2004). These models have significantly improved our understanding
of the possible consequences and responses of terrestrial ecosystems to different
environmental conditions (e.g. Cramer et al. 1999; Song and Woodcock 2003). At
the core of most of these models is a net primary productivity (NPP) sub-model,
which can be used to simulate or predict global vegetation productivity for a
specific ecosystem. However, these NPP sub-models are typically site-specific,
meaning they assume vegetation is homogenous within the ecosystem under study.
When applied in a spatially distributed mode, ecosystem process models can
effectively integrate a diverse assemblage of data and simulate ecosystem conditions with spatial details (Turner et al. 2004). Over landscape or regional scales,
remote sensing provides the only practical source of spatial information that is
required to parameterize, drive and validate process-based models (Psomas
et al. 2008; Turner et al. 2004). Many of the relevant data on vegetation ecosystems
are now available from remotely sensed platforms, and the integration of remote
sensing derived variables and process modeling is a rapidly evolving field (Cohen
and Goward 2004). Examples of ecological variables that can be obtained from
remote sensing data are: (1) biophysical parameters (the leaf area index and the
minimum canopy resistance to evaporation), which can be assessed by spectral
indices to aid biological processes that control fluxes of mass; (2) surface temperature, which can be achieved from various satellite sensors to improve simulation of
energy balance components, and (3) surface soil moisture content, which can be
derived from microwave data to improve the process modeling of bare soil and
sparsely vegetated surfaces. The feasibility of using remote sensing data in
156
Y. He et al.
Globally, grasslands are important in the study of terrestrial ecosystems as they
cover nearly 20 % of the Earth’s surface (Lieth 1978); contain approximately 30 %
of global carbon stocks (Ojima et al. 1996; Parton et al. 1996); and store at least
10 % of the global soil organic matter (Eswaran et al. 1993). There are approximately 24 M ha of mixed grasslands in Canada, serving a variety of economic,
environmental, and ecological purposes. In recent years, grassland degradation has
become a worldwide problem due to intense human activities and environmental
changes, and the mixed grasslands of Canada are no exception. As a result, mixed
grasslands have frequently been associated with fluctuating, and unreliable productivity (Curll et al. 1985a, b; Fothergill et al. 2000; He 2014; Laws and Newton 1992;
Orr et al. 1990; Schwinning and Parsons 1996a, b). To ensure the sustainable
development of Canadian mixed grasslands and to predict the cascading effects
of human activities and climate change on these grasslands, ecosystem process
modeling is required because it can simulate and predict vegetation productivity
and also project ecosystem response to a wide range of environmental conditions.
Over the past 30 years, a considerable number of ecosystem process models such
as BIOME-BGC (Running and Hunt 1993) and CENTURY (Parton et al. 1993)
have been developed to investigate many different aspects of ecosystems, including
vegetation productivity, changing vegetation distributions, and land carbon sinks
(Adams et al. 2004). These models have significantly improved our understanding
of the possible consequences and responses of terrestrial ecosystems to different
environmental conditions (e.g. Cramer et al. 1999; Song and Woodcock 2003). At
the core of most of these models is a net primary productivity (NPP) sub-model,
which can be used to simulate or predict global vegetation productivity for a
specific ecosystem. However, these NPP sub-models are typically site-specific,
meaning they assume vegetation is homogenous within the ecosystem under study.
When applied in a spatially distributed mode, ecosystem process models can
effectively integrate a diverse assemblage of data and simulate ecosystem conditions with spatial details (Turner et al. 2004). Over landscape or regional scales,
remote sensing provides the only practical source of spatial information that is
required to parameterize, drive and validate process-based models (Psomas
et al. 2008; Turner et al. 2004). Many of the relevant data on vegetation ecosystems
are now available from remotely sensed platforms, and the integration of remote
sensing derived variables and process modeling is a rapidly evolving field (Cohen
and Goward 2004). Examples of ecological variables that can be obtained from
remote sensing data are: (1) biophysical parameters (the leaf area index and the
minimum canopy resistance to evaporation), which can be assessed by spectral
indices to aid biological processes that control fluxes of mass; (2) surface temperature, which can be achieved from various satellite sensors to improve simulation of
energy balance components, and (3) surface soil moisture content, which can be
derived from microwave data to improve the process modeling of bare soil and
sparsely vegetated surfaces. The feasibility of using remote sensing data in
156
Y. He et al.