Keywords Satellite · Remote sensing · In-situ sensor · Field survey ·
Crowdsourcing · Cloud sourcing
4.1 Introduction
Agro-geoinformatics uses the techniques of geoinformatics to study agricultural
problems (Di and Yang 2014). In common sense, geoinformatics refers to three
techniques: remote sensing (RS) (Richards and Richards 1999), geographical information system (GIS) (Jones 2014), and global positioning system (GPS) (HofmannWellenhof et al. 1994).
The data sources used by agro-geoinformatics are extremely various. Unlike the
old data-thirsty times, today’s researchers or engineers have plenty of choices on
datasets due to the rapid development of both remote and in-situ sensors. However,
as many scientists said, data are never enough. In many scenarios, especially with
high-precision requirements, people have to observe manually at the scene. For
example, the current research about precision agriculture (Mulla 2013; Pierce and
Nowak 1999) relies on the images captured by unmanned aerial vehicles (UAV)
(Valavanis 2008) on demand. When people want to study a crop field, they bring
equipment and stay at the field to observe. They manually operate those machines,
e.g., drones, to obtained data. To avoid the heavy labor duties, in situ sensors (many
are solar powered) are invented and planted in the fields to observe real-timely,
which require no human intervention. The data will be transferred back to the
receiver station at a certain time interval. But they face risks such as flood, animal
damage, rain (shelter camera lens), and communication and power failure.
Also, satellites could be usable but less competitive than UAV in term of spatial
resolution. The highest spatial resolution of satellite images is at the decimeter level
(optical bands, worse for hyper-spectral bands), while the resolution of UAV or
airplanes can reach the centimeter level. Most high-resolution datasets are owned by
commercial corporations and government agencies and are not free for
use (Wikipedia 2014a). Sourcing channels including discounts for big customers
are offered to facilitate buying. A great quantity of freely available RS datasets are
medium to low resolutions, like the famous Landsat (30 meters) (USGS 2014;
Wikipedia 2014b), MODIS (250 m) (NASA 2014), and Sentinel (range from 5 m
to 60 m) (Drusch et al. 2012). But it is not equal to meaning that the free data has no
playground in the market. Medium-to-low- resolution data can monitor crop fields
and predict yields in large scale, e.g., the entire United States, and generate different
kinds of indicators to support annual agricultural policy making.
Besides the remotely sensed data, the in-situ measured vegetation, soil moisture,
and meteorological data are in great demand. Such data are ground truth and very
important. But they are non-continuously captured on specific fields because of the
point-oriented character and high expense of the monitoring devices. A device can
only observe one object and the observations are intermittent in space. Spatial
interpolation, which adds more errors to the results, has to be conducted in order
42
Z. Sun et al.
Crowdsourcing · Cloud sourcing
4.1 Introduction
Agro-geoinformatics uses the techniques of geoinformatics to study agricultural
problems (Di and Yang 2014). In common sense, geoinformatics refers to three
techniques: remote sensing (RS) (Richards and Richards 1999), geographical information system (GIS) (Jones 2014), and global positioning system (GPS) (HofmannWellenhof et al. 1994).
The data sources used by agro-geoinformatics are extremely various. Unlike the
old data-thirsty times, today’s researchers or engineers have plenty of choices on
datasets due to the rapid development of both remote and in-situ sensors. However,
as many scientists said, data are never enough. In many scenarios, especially with
high-precision requirements, people have to observe manually at the scene. For
example, the current research about precision agriculture (Mulla 2013; Pierce and
Nowak 1999) relies on the images captured by unmanned aerial vehicles (UAV)
(Valavanis 2008) on demand. When people want to study a crop field, they bring
equipment and stay at the field to observe. They manually operate those machines,
e.g., drones, to obtained data. To avoid the heavy labor duties, in situ sensors (many
are solar powered) are invented and planted in the fields to observe real-timely,
which require no human intervention. The data will be transferred back to the
receiver station at a certain time interval. But they face risks such as flood, animal
damage, rain (shelter camera lens), and communication and power failure.
Also, satellites could be usable but less competitive than UAV in term of spatial
resolution. The highest spatial resolution of satellite images is at the decimeter level
(optical bands, worse for hyper-spectral bands), while the resolution of UAV or
airplanes can reach the centimeter level. Most high-resolution datasets are owned by
commercial corporations and government agencies and are not free for
use (Wikipedia 2014a). Sourcing channels including discounts for big customers
are offered to facilitate buying. A great quantity of freely available RS datasets are
medium to low resolutions, like the famous Landsat (30 meters) (USGS 2014;
Wikipedia 2014b), MODIS (250 m) (NASA 2014), and Sentinel (range from 5 m
to 60 m) (Drusch et al. 2012). But it is not equal to meaning that the free data has no
playground in the market. Medium-to-low- resolution data can monitor crop fields
and predict yields in large scale, e.g., the entire United States, and generate different
kinds of indicators to support annual agricultural policy making.
Besides the remotely sensed data, the in-situ measured vegetation, soil moisture,
and meteorological data are in great demand. Such data are ground truth and very
important. But they are non-continuously captured on specific fields because of the
point-oriented character and high expense of the monitoring devices. A device can
only observe one object and the observations are intermittent in space. Spatial
interpolation, which adds more errors to the results, has to be conducted in order
42
Z. Sun et al.