4.5 Prospect of updated crop status forecast for field-level
in-season management
Since most remote sensing algorithms have been adapted to large areas,
making them applicable to the field-level is a priority now, especially
in countries like China (Weiss et al., 2020) where the agricultural
system is dominated by millions of smallholders and N is always
overdosed (Cui et al., 2018). Spatial and temporal predictions ofN leaf (Fig. 5) are crucial for determining the
management of fertilization timing to be performed at specific growth
stages (Weiss et al., 2020). However, the use of remote sensing data
solely is not sufficient to quantify top-dressing requirements (Weiss et
al., 2020), due to the manifold interactions in the soil-crop-atmosphere
continuum. Assimilating supplementary information from a crop model with
remote sensing data has been identified as one of the most promising
methods to make field management decisions (Jin et al., 2018; Weiss et
al., 2020). For instance, Baret et al. (2007) demonstrated that an
optimized in-season N application map can be generated by assimilating
remotely sensed LAI and N above into the crop
model STICS, in which the N application rate in each map pixel (20×20
m2) was optimized by maximizing the productivity,
using the historical meteorological data between the time of decision
and harvest as the unknown future weather conditions. With optimized
real-time fertilizer management, severe environmental issues caused by
overfertilization can be reduced, thereby enabling smart farming and
sustainable agricultural production (Berger et al., 2020b).