3.1.2 Satellite images

In order to study satellite images, data will be provided by a specialized company, by means of its intelligent environmental knowledgebase (i-ekbase), and made available via web tool, with limited and free use, for research related to the topic. The web tool will provide data from the area chosen initially for the study. The intelligent environmental knowledgebase (i-ekbase)<sup>2</sup> is an autonomous Big Data Analytics engine with a CLOUD system, and a fully automated geographic information system (GIS) [89]. Figure 8 illustrates an example image provided by the i-ekbase tool, while Table 1 shows the data generated by the web tool in the CSV (Comma Separated Values) format.

#### Figure 8.

Land surface temperature image by the i-ekbase web tool. Source: Adapted from the i-ekbase system.


#### Table 1.

Data exported by the i-ekbase web tool. Source: Adapted from the i-ekbase system.

<sup>2</sup> http://iekbase.com/

### Integrating Remote Sensing Data into Fuzzy Control System for Variable Rate Irrigation Estimates DOI: http://dx.doi.org/10.5772/intechopen.87023

The i-ekbase system services provide larges area-wise resource management maps, with supporting remote digital scouting for decision support systems and rapid intervention of issues. For developing the experimental system were processed 12 months of Data, these remote sensing imageries were acquired by Landsat (with a spatial resolution of 30 m, but for this experiment the Data was upscale to 10 m) and Sentinel (with a spatial resolution of 10 m) satellites. Data that constitute this image have more than 14,000 georeferenced points, containing at each point or pixel the attributes of the agricultural analysis. Due to the extension of the data, only a few lines are shown in Table 1.

In order to apply this approach to the commercial field scale, the remote sensing data required to describe the soil-plant-atmosphere relationship can be acquired from satellite [90] and aircraft images [91, 92]. However, high costs, spatial resolution, data frequency and data availability [93, 94], in addition to cloudless satellite imagery, are a challenge for the correct execution of models based on remote sensing [95]. These issues can limit the efficiency of real-time variable rate irrigation management.

From the remote sensing data, those that best describe the soil-plant-atmosphere relationship for the intelligent irrigation system of the plantation site will be selected. In this phase, the correct selection of these data is fundamental to correctly calculate the results. A simple but promising approach uses crop coefficients derived from the normalized difference vegetation index (NDVI) along with local climate data to infer quantities of evapotranspiration (ETc) from variable crops almost in real time [57, 83, 96].

Based on the choice of planting site and type of crop to be irrigated, in relation to plant type data, the crop coefficient will be used along with information from the satellite images. In this case, the reading values of NDVI, near soil moisture and vegetative canopy temperature will be used. The latter is an important parameter for irrigation management and should be adjusted according to local growing conditions.

#### 3.2 Location of the study

with quantifiable crop yields [88], such as productivity, water use efficiency, seasonal evapotranspiration, leaf water potential at noon time, irrigation rates and

Irrigation - Water Productivity and Operation, Sustainability and Climate Change

In order to study satellite images, data will be provided by a specialized company, by means of its intelligent environmental knowledgebase (i-ekbase), and made available via web tool, with limited and free use, for research related to the topic. The web tool will provide data from the area chosen initially for the study. The intelligent environmental knowledgebase (i-ekbase)<sup>2</sup> is an autonomous Big Data Analytics engine with a CLOUD system, and a fully automated geographic information system (GIS) [89]. Figure 8 illustrates an example image provided by the i-ekbase tool, while Table 1 shows the data generated by the web tool in the

Land surface temperature image by the i-ekbase web tool. Source: Adapted from the i-ekbase system.

NDVI (%)

15.2464 54.0157 0.0 0.0 13.49 0.0 3.35 13.52 36.48 15.2464 54.0156 0.09 0.0 15.24 0.14 3.32 13.19 36.81 15.2464 54.0155 0.41 0.0 15.36 0.15 3.39 13.93 36.07 15.2464 54.0159 3.36 0.0 22.76 0.76 3.16 11.61 38.39 15.2464 54.0158 4.96 0.0 26.68 1.09 3.10 11.00 39.00 15.2463 54.0162 7.37 0.0 31.78 1.52 2.87 8.65 41.35 15.2463 54.0162 9.30 1.0 36.34 1.89 2.80 8.03 38.97 15.2463 54.0161 11.59 1.0 41.42 2.32 2.68 6.84 40.16

Biomass (tn/ha

Soil salinity (dS/m)

Soil moisture (%)

Canopy temp. (°C)

Leaf area index (m<sup>2</sup> /m<sup>2</sup>

Data exported by the i-ekbase web tool. Source: Adapted from the i-ekbase system.

damage caused by herbicides.

CSV (Comma Separated Values) format.

3.1.2 Satellite images

Figure 8.

Table 1.

38

<sup>2</sup> http://iekbase.com/

Lat Long Canopy

nitrogen (%)

The study site is a farm located in the municipality of Primavera do Leste, MT, latitude 15° 14<sup>0</sup> 24.73 "S and longitude 54 ° 0'53.29" W. This site has areas of cultivation irrigated by central pivot, and the crops planted are soybean, cotton and second-crop corn. The delimited area presents a total of 140 ha, in a radius of 667 m, see Figure 9. The area delimited by the red circle has central pivot irrigation, and the information used in the case study is from a 2015/2016 second-crop corn cycle. Irrigation in maize crop means to meet the minimum water requirements for the development of the crop.

Maize expresses high sensitivity to droughts. Therefore, the incidence of periods with reduction of the water supply to the plants at critical moments of the development of the crop, from flowering to physiological ripeness, can cause a direct reduction in the final harvest. In order to obtain maximum output, maize planting requires approximately 650 mm of water during its cycle [97], which can vary from 110 to 140 days in medium-cycle hybrids. For this preliminary analysis, data on daily average precipitation were used, provided by INMET (National Institute of Meteorology), from April to September 2016, to the city of Primavera do Leste, in the State of Mato Grosso, Brazil. Figure 10 shows the data obtained.

These readings recorded during the development of the plantation under study corroborate the supposition of water stress due to lack of rainfall (from June to September), which would indicate the possibility of complementing water demand by irrigation.

#### Figure 9.

RGB images of the location of study area in in the municipality of primavera do Leste, state of Mato Grosso, Brazil.

movement of the pivot in relation to the amount of water sprayed by the sprinklers. The decision unit or inference machine to perform rule-based inference operations will be implemented using the Mamdani method, with crisp inputs and crisp output

Canopy temperature (°C) <14 14 < ϕ < 27 >24 Upper layer soil moisture (%) <14 12 < ϕ < 24 >21 NDVI (%) <16 12 < ϕ < 27 >27

Corresponding membership functions for each system entry, (a) canopy temperature, (b) upper layer soil

Integrating Remote Sensing Data into Fuzzy Control System for Variable Rate Irrigation Estimates

Low Average High

Input variables Linguistic variables

In this first stage of development, the water depth that the irrigation system provides will be considered constant, and the database, which defines the functions of association of the sets used in the fuzzy rules, will be implemented as shown in

With the remote sensing data, it is possible to construct the universes of discourse of each input variable and thus transform the database into linguistic vari-

ables, such as those presented in the table above. Each of these inputs was previously limited in the universe of discourse in question and associated with a degree of pertinence in each fuzzy set by means of specialist knowledge. In this manner, in order to obtain the degree of pertinence of a given crisp input, it is necessary to search for this value in the knowledge base of the fuzzy system. The fuzzification of the decision-making system is shown in Figure 11, and it is possible

<sup>3</sup> https://www.agritempo.gov.br/agritempo/jsp/Grafico/graficoMicrorregiao.jsp?siglaUF=MT

value<sup>3</sup> .

41

Table 2.

Figure 11.

moisture, (c) NDVI.

Table 2 and Figure 11.

Fuzzy input set for the fuzzy inference system.

DOI: http://dx.doi.org/10.5772/intechopen.87023

Figure 10. Daily average precipitations obtained in the period of 2016. Fonte: INMET.

### 3.3 Fuzzy systems

In this step, a fuzzy system will be used, which in this case will be capable to infer the variations of linear speeds of the pivot according to the images provided by the satellite. For the creation of the control map, a system with artificial intelligence will be developed, capable of manipulating data and knowledge.

Three input variables (NDVI, near-soil moisture and canopy temperature) were used to infer the speed that the pivot should have to improve the level of irrigation within the management area, so that an adequate speed could be found for the

Integrating Remote Sensing Data into Fuzzy Control System for Variable Rate Irrigation Estimates DOI: http://dx.doi.org/10.5772/intechopen.87023

#### Figure 11.

Corresponding membership functions for each system entry, (a) canopy temperature, (b) upper layer soil moisture, (c) NDVI.


#### Table 2.

Fuzzy input set for the fuzzy inference system.

movement of the pivot in relation to the amount of water sprayed by the sprinklers. The decision unit or inference machine to perform rule-based inference operations will be implemented using the Mamdani method, with crisp inputs and crisp output value<sup>3</sup> .

In this first stage of development, the water depth that the irrigation system provides will be considered constant, and the database, which defines the functions of association of the sets used in the fuzzy rules, will be implemented as shown in Table 2 and Figure 11.

With the remote sensing data, it is possible to construct the universes of discourse of each input variable and thus transform the database into linguistic variables, such as those presented in the table above. Each of these inputs was previously limited in the universe of discourse in question and associated with a degree of pertinence in each fuzzy set by means of specialist knowledge. In this manner, in order to obtain the degree of pertinence of a given crisp input, it is necessary to search for this value in the knowledge base of the fuzzy system. The fuzzification of the decision-making system is shown in Figure 11, and it is possible

3.3 Fuzzy systems

Figure 10.

40

Figure 9.

Brazil.

In this step, a fuzzy system will be used, which in this case will be capable to infer the variations of linear speeds of the pivot according to the images provided by the satellite. For the creation of the control map, a system with artificial intelligence

RGB images of the location of study area in in the municipality of primavera do Leste, state of Mato Grosso,

Irrigation - Water Productivity and Operation, Sustainability and Climate Change

Three input variables (NDVI, near-soil moisture and canopy temperature) were used to infer the speed that the pivot should have to improve the level of irrigation within the management area, so that an adequate speed could be found for the

will be developed, capable of manipulating data and knowledge.

Daily average precipitations obtained in the period of 2016. Fonte: INMET.

<sup>3</sup> https://www.agritempo.gov.br/agritempo/jsp/Grafico/graficoMicrorregiao.jsp?siglaUF=MT

to visualize the corresponding membership functions, considering these intervals as the universe of discourse of these variables.

Triangular membership functions were chosen because they simplify the calculation of the fuzzy inference mechanism. Well distributed triangular membership functions transform the input values into fuzzy values (low, medium and high), as shown in Figure 5, as well as the values of soil moisture and NDVI (Figure 11b and c, respectively). The fuzzy output set, which represents the rotational speed of the central pivot, was built on five linguistic variables: very low (VL), low (L), normal (N), high (H) and very high (VH). These sets were interpreted by means of their degrees of pertinence, illustrated in Figure 12.

If the center of gravity method is used for defuzzification, the fuzzy set produced after aggregation will be a numerical output composed of the union of all rule contributions. This calculation is made according to Eq. (5):

$$\mu^{\*} = \frac{\sum\_{i=1}^{n} \mu\_{i} \bullet \mu\_{out}(\mu\_{i})}{\sum\_{i=1}^{n} \mu\_{out}(\mu\_{i})} \tag{5}$$

The values μout μ<sup>i</sup> ð Þ represent the area of a pertinence function modified by the result of fuzzy inference, and μ<sup>i</sup> ð Þ is the position of the centroid of the individual pertinence function.

Finally, the basis of fuzzy rules IF-THEN was elaborated and presented in Table 3, the fuzzy rule relating to rotation speed contains 27 rules, thus, the Mamdani inference method for a set of conjunctive rules is given by Eq. (3), for example: IF NDVI is Low AND Canopy temperature is Low AND Near-soil moisture is Low THEN Rotation Speed is Low.

This set of rules is based on the basic knowledge about irrigation, according to a methodology adopted by [37, 39].

The rules were constructed with the connective "AND", and are based on the supposition that where there is little leaf growth, there is soil water deficit. Together with the characteristic of the high canopy temperature, indicating a lower evapotranspiration, that is, water stress of the plants, the values of near-soil moisture provided by the web tool are readings of the locations where there are few leaves, and it is possible to estimate their value.

4. Control maps of central pivot motion speed

Fuzzy rules for central pivot speed control.

The development of the crop is evidenced captured by the images throughout the crop, and the information contained in this sensing is the NDVI values. By analyzing the information contained in Figure 13, it is possible to verify the similarity between the values attributed to Kc. It is noticed that as the crop evolves, the greater the exposure of the leaf area, and thus it is possible to establish a relation of NDVI. This process is described in [56], with ratios to calculate the base crop coefficient (Kcb) for cotton as a function of NDVI. When we look closely at each stage of the development of the plantation, two distinct areas are noticed: one with

Inputs Output

DOI: http://dx.doi.org/10.5772/intechopen.87023

Low

Medium

High

Table 3.

43

NDVI Temperature Near-soil moisture Rotation speed

Integrating Remote Sensing Data into Fuzzy Control System for Variable Rate Irrigation Estimates

Low Low Low

Medium Low Low

High Low Low

Low Low Normal

Medium Low Normal

High Low Low

Low Low High

Medium Low Very high

High Low Very high

Medium Low High Very low

Medium Low High Very low

Medium Low High Very low

Medium Normal High High

Medium Normal High High

Medium Normal High Normal

Medium Very high High Very high

Medium Very high High Very high

Medium Very high High Very high

Figure 12. Function of pertinence of the speed corresponding to the defuzzification of the system.


Integrating Remote Sensing Data into Fuzzy Control System for Variable Rate Irrigation Estimates DOI: http://dx.doi.org/10.5772/intechopen.87023

Table 3.

to visualize the corresponding membership functions, considering these intervals as

Irrigation - Water Productivity and Operation, Sustainability and Climate Change

If the center of gravity method is used for defuzzification, the fuzzy set produced after aggregation will be a numerical output composed of the union of all

∑<sup>n</sup>

Finally, the basis of fuzzy rules IF-THEN was elaborated and presented in Table 3, the fuzzy rule relating to rotation speed contains 27 rules, thus, the Mamdani inference method for a set of conjunctive rules is given by Eq. (3), for example: IF NDVI is Low AND Canopy temperature is Low AND Near-soil

The values μout μ<sup>i</sup> ð Þ represent the area of a pertinence function modified by the result of fuzzy inference, and μ<sup>i</sup> ð Þ is the position of the centroid of the individual

This set of rules is based on the basic knowledge about irrigation, according to a

The rules were constructed with the connective "AND", and are based on the supposition that where there is little leaf growth, there is soil water deficit. Together with the characteristic of the high canopy temperature, indicating a lower evapotranspiration, that is, water stress of the plants, the values of near-soil moisture provided by the web tool are readings of the locations where there are few leaves,

<sup>i</sup>¼<sup>1</sup>μ<sup>i</sup> <sup>∙</sup> <sup>μ</sup>out <sup>μ</sup><sup>i</sup> ð Þ

<sup>i</sup>¼<sup>1</sup>μout <sup>μ</sup><sup>i</sup> ð Þ (5)

rule contributions. This calculation is made according to Eq. (5):

<sup>μ</sup><sup>∗</sup> <sup>¼</sup> <sup>∑</sup><sup>n</sup>

Triangular membership functions were chosen because they simplify the calculation of the fuzzy inference mechanism. Well distributed triangular membership functions transform the input values into fuzzy values (low, medium and high), as shown in Figure 5, as well as the values of soil moisture and NDVI (Figure 11b and c, respectively). The fuzzy output set, which represents the rotational speed of the central pivot, was built on five linguistic variables: very low (VL), low (L), normal (N), high (H) and very high (VH). These sets were interpreted by means of their

the universe of discourse of these variables.

degrees of pertinence, illustrated in Figure 12.

moisture is Low THEN Rotation Speed is Low.

methodology adopted by [37, 39].

and it is possible to estimate their value.

Function of pertinence of the speed corresponding to the defuzzification of the system.

pertinence function.

Figure 12.

42

Fuzzy rules for central pivot speed control.
