(remote sensing and in situ) and the reference soil moisture data available only at
some reference points. The neural network is trained by the reference soil moisture
data. This kind of mapping is proposed for adaptive irrigation planning. The same
method can be used for missing data reconstruction in reference measurement
points too.
When the sampling period T is chosen as 1 day, reference evapotranspiration
(ET 0 ), SNDVI, irrigation, and precipitation patterns are used as 5-point data
sequences, each representing today’s and last 4 days’ values. The soil moisture
was measured in terms of centibars (cb) in training datasets, and soil moisture
estimation in the test set example (Fig. 7.16) is in centibars. Here, 200 cb indicates
the minimum moisture (dry) and 0 cb indicates the maximum moisture (wet soil). It
should be noted that different crop types may require different training set. On the
other hand, if field-specific crop type is known either by using remote sensing–based
cropland cover classification methods or gathered from farm registry systems, then
this kind of fusion scheme can be used for crop-specific estimations in the region. If
the integration level goes down to irrigation automation systems or the farmer
informs the system with daily irrigation amount, then it can be used for irrigation
schedule optimization.
0
50
100
150
200
250
a
b
cb
soil moisture 45cm
predicted soil moisture 45cm
0
50
100
150
200
250
cb
Fig. 7.16 Monitored and predicted soil moisture at 45-cm depth at two different locations within
the same time interval, crop type, and the region (a) training site and (b) test site patterns
7 Data Fusion in Agricultural Information Systems
127
some reference points. The neural network is trained by the reference soil moisture
data. This kind of mapping is proposed for adaptive irrigation planning. The same
method can be used for missing data reconstruction in reference measurement
points too.
When the sampling period T is chosen as 1 day, reference evapotranspiration
(ET 0 ), SNDVI, irrigation, and precipitation patterns are used as 5-point data
sequences, each representing today’s and last 4 days’ values. The soil moisture
was measured in terms of centibars (cb) in training datasets, and soil moisture
estimation in the test set example (Fig. 7.16) is in centibars. Here, 200 cb indicates
the minimum moisture (dry) and 0 cb indicates the maximum moisture (wet soil). It
should be noted that different crop types may require different training set. On the
other hand, if field-specific crop type is known either by using remote sensing–based
cropland cover classification methods or gathered from farm registry systems, then
this kind of fusion scheme can be used for crop-specific estimations in the region. If
the integration level goes down to irrigation automation systems or the farmer
informs the system with daily irrigation amount, then it can be used for irrigation
schedule optimization.
0
50
100
150
200
250
a
b
cb
soil moisture 45cm
predicted soil moisture 45cm
0
50
100
150
200
250
cb
Fig. 7.16 Monitored and predicted soil moisture at 45-cm depth at two different locations within
the same time interval, crop type, and the region (a) training site and (b) test site patterns
7 Data Fusion in Agricultural Information Systems
127