decades (Foley et al. 2011; White et al. 2011). In the past decades, crop growth
modeling and yield forecasting have attracted increasing attention in both scientific
researches and agricultural practices. Many scientific studies have been carried out to
improve the capabilities of crop growth modeling and yield forecasting by using
various data sources and method like statistical models, crop growth simulation
models, and remote sensing (Bennett et al. 2017; Gowda et al. 2014; Prasad et al.
2006). In addition, many studies also have demonstrated the advantages of integrating remotely sensed data and crop growth models by data assimilation in crop
growth modeling and yield forecasting (de Wit and van Diepen 2007; Fang et al.
2008; Huang et al. 2015).
The majority contents of crop growth monitoring focus on the dynamics parameters during crop growth, such as like leaf area index, leaf nitrogen accumulation,
dry matter content, and soil moisture. The timely and accurate crop growth variables
can help to learn the state of the crop growth period. A few months before harvest
crop yield forecasting can be important to national food trade and security. The
success of crop yield forecasting strongly depends on the crop growth monitoring’s
ability (Horie et al. 1992). Agricultural research community has developed many
crop growth modeling approaches that can be broadly divided into two categories to
forecasting crop yield. The first category comprises statistical models, which establish the contact between climate, remote sensing, or other variables and yield using
statistical methods such as regression analysis (Basso et al. 2013; Kogan et al. 2013;
Michel and Makowski 2013). Statistical modeling plays a key role in current
research studies on yield forecasting, while most of these models are locally calibrated, not easily used in the other cropland region. The second category contains
physiological/physical-based crop growth models. They can simulate the main crop
growth and production processes such as photosynthesis, solar radiation absorption,
phenology, carbon, and nitrogen balances (Möller and Müller 2012). Numerous crop
growth models have been developed, and some models are widely used in crop
growth monitoring and yield forecasting (Eitzinger et al. 2004; Palosuo et al. 2011a).
Although crop models can reproduce the main physiological/physical processes
occurring during the crop growth, the use of models often limited by the uncertainties of input variables such as filed management, soil, and initial conditions, and
the large model parameters calibration are also challenging. Using these models to
predict regional crop yield should be complex, because many of these crop modeling
approaches were developed at the farm (point) scale. Cropland spatial and temporal
heterogeneity, the available grid soil, and climate driving databases are other challenge using the crop growth models at a regional scale. However, with the development of remote sensing and GIS techniques, many studies have been successfully
carried out using a crop growth model to simulate regional crop growth dynamics
and yield (Bastiaanssen and Ali 2003; Ren et al. 2011; Therond et al. 2011).
Remote sensing has already been proven as an effective measure for crop area
statistics and crop mapping, quantitative inversion of key crop growth variables,
crop phenology, and yield forecasting. After decades’ development, the application
of remote sensing in agriculture has undergone tremendous changes, from simple
images to quantitative agricultural parameters. Currently, crop growth–related
206
H. Pan and Z. Chen
modeling and yield forecasting have attracted increasing attention in both scientific
researches and agricultural practices. Many scientific studies have been carried out to
improve the capabilities of crop growth modeling and yield forecasting by using
various data sources and method like statistical models, crop growth simulation
models, and remote sensing (Bennett et al. 2017; Gowda et al. 2014; Prasad et al.
2006). In addition, many studies also have demonstrated the advantages of integrating remotely sensed data and crop growth models by data assimilation in crop
growth modeling and yield forecasting (de Wit and van Diepen 2007; Fang et al.
2008; Huang et al. 2015).
The majority contents of crop growth monitoring focus on the dynamics parameters during crop growth, such as like leaf area index, leaf nitrogen accumulation,
dry matter content, and soil moisture. The timely and accurate crop growth variables
can help to learn the state of the crop growth period. A few months before harvest
crop yield forecasting can be important to national food trade and security. The
success of crop yield forecasting strongly depends on the crop growth monitoring’s
ability (Horie et al. 1992). Agricultural research community has developed many
crop growth modeling approaches that can be broadly divided into two categories to
forecasting crop yield. The first category comprises statistical models, which establish the contact between climate, remote sensing, or other variables and yield using
statistical methods such as regression analysis (Basso et al. 2013; Kogan et al. 2013;
Michel and Makowski 2013). Statistical modeling plays a key role in current
research studies on yield forecasting, while most of these models are locally calibrated, not easily used in the other cropland region. The second category contains
physiological/physical-based crop growth models. They can simulate the main crop
growth and production processes such as photosynthesis, solar radiation absorption,
phenology, carbon, and nitrogen balances (Möller and Müller 2012). Numerous crop
growth models have been developed, and some models are widely used in crop
growth monitoring and yield forecasting (Eitzinger et al. 2004; Palosuo et al. 2011a).
Although crop models can reproduce the main physiological/physical processes
occurring during the crop growth, the use of models often limited by the uncertainties of input variables such as filed management, soil, and initial conditions, and
the large model parameters calibration are also challenging. Using these models to
predict regional crop yield should be complex, because many of these crop modeling
approaches were developed at the farm (point) scale. Cropland spatial and temporal
heterogeneity, the available grid soil, and climate driving databases are other challenge using the crop growth models at a regional scale. However, with the development of remote sensing and GIS techniques, many studies have been successfully
carried out using a crop growth model to simulate regional crop growth dynamics
and yield (Bastiaanssen and Ali 2003; Ren et al. 2011; Therond et al. 2011).
Remote sensing has already been proven as an effective measure for crop area
statistics and crop mapping, quantitative inversion of key crop growth variables,
crop phenology, and yield forecasting. After decades’ development, the application
of remote sensing in agriculture has undergone tremendous changes, from simple
images to quantitative agricultural parameters. Currently, crop growth–related
206
H. Pan and Z. Chen