Even when the daily observations at medium spatial resolution are available, near
real-time mapping of crop growth stages is still challenging (Liu et al. 2018). For
example, crop emergence is defined as the first appearance of crop leaves. Remote
sensing data from the early growing stage could be affected by the change of soil
moisture (e.g., snow/ice melts). The subtle changes in crop emergence may not be
sensitive to sensors. Near real-time (or within the season) crop phenology mapping
approaches have been developed (Zhang et al. 2012; Liu et al. 2018; Gao et al.
2020a, b). These approaches can run in the near real-time mode using any period of
an imagery time series. Results show that crop emergence dates and cover crop
termination dates may be reliably detected within 1–3 weeks using high-temporal
and spatial-resolution remote sensing data (Gao et al. 2020a, b). Figure 2.3 shows
green-up dates detected using the Vegetation and Environment monitoring New
MicroSatellite (VENμS, 5-m, 2-day revisit) time series from three dates in 2019.
Later green-up events were detected by including more recent observations.
2.3.3 Crop Yield Estimation
Accurate estimation of crop yield before harvest is critical for sustaining agricultural
markets and ensuring food security. Remote sensing data have been demonstrated
useful for estimating crop yield for several decades. More than three decades ago,
Tucker et al. (1980) used field observation and reported that the normalized difference vegetation index (NDVI) for a 5-week period from stem elongation to anthesis
explained about 64% of grain yield variation of wheat.
In the recent era of rich satellite data availability, numerous studies have been
published using satellite imagery to estimate crop yields. Many of these used
empirical relationships between yields and various vegetation indices (VIs). The
empirical approach builds the relationship between ground yield survey samples and
the remote sensing-derived parameters and then applies the relationship to remote
sensing imagery to map yield over the entire area. VI-based metrics (e.g., maximum
VI, integral VI from the entire growing season or for a specific growth period) have
been used for estimating crop yield. Although an empirical model built for a specific
region has a limited applicability to different areas or years, the empirical model is
simple and effective for the local region if ground survey samples are representative
and accurate.
Another front of the effort has been to incorporate remote sensing data into
physiology-based crop growth modeling. Conventional models simulate crop
growth and yield (or biomass) through crop biophysical processes. Remote sensing
variables like leaf area index (LAI) can be integrated into these models via direct
replacement or through data assimilation techniques. Physiological crop growth
models typically require a large number of inputs and computing resources. To
reduce the input data requirements and computing costs, Lobell et al. (2015)
developed a scalable satellite-based crop yield mapper (SCYM) to relate weather
18
F. Gao
real-time mapping of crop growth stages is still challenging (Liu et al. 2018). For
example, crop emergence is defined as the first appearance of crop leaves. Remote
sensing data from the early growing stage could be affected by the change of soil
moisture (e.g., snow/ice melts). The subtle changes in crop emergence may not be
sensitive to sensors. Near real-time (or within the season) crop phenology mapping
approaches have been developed (Zhang et al. 2012; Liu et al. 2018; Gao et al.
2020a, b). These approaches can run in the near real-time mode using any period of
an imagery time series. Results show that crop emergence dates and cover crop
termination dates may be reliably detected within 1–3 weeks using high-temporal
and spatial-resolution remote sensing data (Gao et al. 2020a, b). Figure 2.3 shows
green-up dates detected using the Vegetation and Environment monitoring New
MicroSatellite (VENμS, 5-m, 2-day revisit) time series from three dates in 2019.
Later green-up events were detected by including more recent observations.
2.3.3 Crop Yield Estimation
Accurate estimation of crop yield before harvest is critical for sustaining agricultural
markets and ensuring food security. Remote sensing data have been demonstrated
useful for estimating crop yield for several decades. More than three decades ago,
Tucker et al. (1980) used field observation and reported that the normalized difference vegetation index (NDVI) for a 5-week period from stem elongation to anthesis
explained about 64% of grain yield variation of wheat.
In the recent era of rich satellite data availability, numerous studies have been
published using satellite imagery to estimate crop yields. Many of these used
empirical relationships between yields and various vegetation indices (VIs). The
empirical approach builds the relationship between ground yield survey samples and
the remote sensing-derived parameters and then applies the relationship to remote
sensing imagery to map yield over the entire area. VI-based metrics (e.g., maximum
VI, integral VI from the entire growing season or for a specific growth period) have
been used for estimating crop yield. Although an empirical model built for a specific
region has a limited applicability to different areas or years, the empirical model is
simple and effective for the local region if ground survey samples are representative
and accurate.
Another front of the effort has been to incorporate remote sensing data into
physiology-based crop growth modeling. Conventional models simulate crop
growth and yield (or biomass) through crop biophysical processes. Remote sensing
variables like leaf area index (LAI) can be integrated into these models via direct
replacement or through data assimilation techniques. Physiological crop growth
models typically require a large number of inputs and computing resources. To
reduce the input data requirements and computing costs, Lobell et al. (2015)
developed a scalable satellite-based crop yield mapper (SCYM) to relate weather
18
F. Gao