temperature sensor fails, past temperature measurement patterns in the neighboring
stations and temperature correlated other measurements in the same station can be
used for temporary data reconstruction (Altan and Üstündağ 2012).
Acquired raw datasets are also used to calculate agricultural indices. For example,
growing degree days (GDD) and vapor pressure deficit (VPD) are two important
parameters for the plant growth rate. Temperature measurement is used in both of
them. Automated diagnostics for the quality of service (QoS) management in DaaS
business models also consider data fusion methods. On the other hand, probably the
most important use of data fusion in agricultural applications is the prediction.
Prediction can either be performed as nowcasting or forecasting. Data fusion is
used for nowcasting when target data is not feasible for direct monitoring due to
physical restrictions, operational reasons, or cost-related issues. Forecasting is based
on the fusion of data patterns depending on machine learning methods, adaptation of
analytic models, or hybrid methods for the estimation of future value in time, space,
or both. Risk management in agriculture may require both nowcasting and forecasting models. For example, if a disease is known to occur at some known locations,
nowcasting can help the probable distribution of damage, and forecasting may help
for the development of risk. In this case, the known interpolatable parameters like
temperature and air humidity and location-specific remote sensing data can be used
in a fusion scheme for nowcasting and forecasting purposes.
Automated cropland cover identification system requires training patterns and
terrestrial reference observations for supervised classification besides the
terrain data.
Geo-statistical yield prediction models are based on probabilistic distribution of
yield concerning spatial surveys and terrain models. Their accuracy depends on
sampling size and terrain complexity. On the other hand, their spatial resolution is
limited as a region or province. Crop yield prediction and mapping is a good
example to demonstrate how data fusion improves spatial accuracy with respect to
pure statistical models. Data fusion is also used for spatial interpolations of
agrometeorological parameters. Some of them are interpolatable based on kriging
methods and some models, while many surface parameters such as soil moisture and
phenological stage distribution are not interpolatable. Data fusion methods can be
used for mapping of uninterpolatable parameters together with correlated
interpolatable parameters as an adaptive model solution. Some of the spatial information also intersects with temporal observation locations (Fig. 7.1). Spatial and
temporal characteristics at those intersection points are used in the calibration or
adaptation of data fusion models. Hence spatial and temporal data can be used in the
generation of multitemporal spatial data by using adapted spatiotemporal data fusion
models.
Data fusion models increase the service quality and capabilities of the vertical
integration platforms. Sectoral integration in agriculture consists of five major
components (Fig. 7.2):
(a) Asset management
(b) Efficiency management
7 Data Fusion in Agricultural Information Systems
107
stations and temperature correlated other measurements in the same station can be
used for temporary data reconstruction (Altan and Üstündağ 2012).
Acquired raw datasets are also used to calculate agricultural indices. For example,
growing degree days (GDD) and vapor pressure deficit (VPD) are two important
parameters for the plant growth rate. Temperature measurement is used in both of
them. Automated diagnostics for the quality of service (QoS) management in DaaS
business models also consider data fusion methods. On the other hand, probably the
most important use of data fusion in agricultural applications is the prediction.
Prediction can either be performed as nowcasting or forecasting. Data fusion is
used for nowcasting when target data is not feasible for direct monitoring due to
physical restrictions, operational reasons, or cost-related issues. Forecasting is based
on the fusion of data patterns depending on machine learning methods, adaptation of
analytic models, or hybrid methods for the estimation of future value in time, space,
or both. Risk management in agriculture may require both nowcasting and forecasting models. For example, if a disease is known to occur at some known locations,
nowcasting can help the probable distribution of damage, and forecasting may help
for the development of risk. In this case, the known interpolatable parameters like
temperature and air humidity and location-specific remote sensing data can be used
in a fusion scheme for nowcasting and forecasting purposes.
Automated cropland cover identification system requires training patterns and
terrestrial reference observations for supervised classification besides the
terrain data.
Geo-statistical yield prediction models are based on probabilistic distribution of
yield concerning spatial surveys and terrain models. Their accuracy depends on
sampling size and terrain complexity. On the other hand, their spatial resolution is
limited as a region or province. Crop yield prediction and mapping is a good
example to demonstrate how data fusion improves spatial accuracy with respect to
pure statistical models. Data fusion is also used for spatial interpolations of
agrometeorological parameters. Some of them are interpolatable based on kriging
methods and some models, while many surface parameters such as soil moisture and
phenological stage distribution are not interpolatable. Data fusion methods can be
used for mapping of uninterpolatable parameters together with correlated
interpolatable parameters as an adaptive model solution. Some of the spatial information also intersects with temporal observation locations (Fig. 7.1). Spatial and
temporal characteristics at those intersection points are used in the calibration or
adaptation of data fusion models. Hence spatial and temporal data can be used in the
generation of multitemporal spatial data by using adapted spatiotemporal data fusion
models.
Data fusion models increase the service quality and capabilities of the vertical
integration platforms. Sectoral integration in agriculture consists of five major
components (Fig. 7.2):
(a) Asset management
(b) Efficiency management
7 Data Fusion in Agricultural Information Systems
107