Spatial distribution of soil nutrient content for sustainable rice agriculture using geographic information system and Naïve Bayes classifier

This device is accurate, easy, and relatively fast in analyzing the nutrient content in the soil. Nitrogen, phosphorus, potassium, and soil pH are the primary measurement targets for the PUTS design. The Paddy Soil Test Kit (PUTS) kit consists of several chemicals used for extracting soil nutrients, a leaf color chart (BWD), and instructions for use, along with fertilizer recommendations. The image of a PUTS device is shown in Figure 1.

Figure 1

Probability and statistics are classification methods found in the Naïve Bayes method. This method was proposed by Thomas Bayes and is used as a reference to predict future opportunities based on previous experience [46], [47]. Naïve Bayes Classifier requires a small amount of training data to achieve classification of the parameters. The variables in this method are assumed to determine each class [42]. Naïve Bayes classification is obtained by using the following expression: (1) P(x)=P(c)P(c)P(x), P(x) = {{P(c)P(c)} \over {P(x)}}, where P(c) = class prior probability; P(x|c) = likelihood; P(x) = predictor prior probability; and P(c|x) = posterior probability.

Figure 2 shows the block diagram of this system.

Figure 2

The block diagram in Figure 2 shows the working of the paddy soil phosphorus meter. The initial stage is to get the color of soil phosphorus in the paddy soil samples extracted using PUTS. The Soil phosphorus determination are obtained using the TCS3200 sensor through the serial peripheral interface (SPI) communication line connected to the Wemos D1 Mini board. The resulting data of the TCS3200 sensor are processed on the Wemos D1 Mini board, which is connected to a 5 V voltage source. The data processed on the Wemos board and classified using the Naïve Bayes method are displayed on a 16 × 2 liquid-crystal display (LCD) and then sent to the ceerduad.com website. Wemos is used because it is a microcontroller that has an integrated Internet of Things (IoT) system and a wireless fidelity (WiFi) module with 4 MB of storage memory [48]. The data are displayed on a 16 ཌ 2 LCD, and the web map is then created using ArcGIS software.

The Wemos D1 Mini system control is a WIFI-based development board module designed as a sensor controller. Sensor reading data are processed on the sensor controller. The outline of the series of systems designed for measuring the phosphorus levels in paddy soil is shown in Figure 3.

Figure 3

The design of this measuring instrument program uses Arduino Uno. The program results are then uploaded to the Wemos D1 Mini board. The software design is shown in the flowchart in Figure 4. In the workflow of this paddy soil phosphorus meter, when the measuring instrument is running, the sensor will immediately read the room value, which will be used as the sensor' s default value for calibration. If the calibration is successful, it will display a command to insert the extracted soil sample to read the red, green, and blue (RGB) values. The classification process is carried out with the Naïve Bayes algorithm. The results of the classification process are displayed on the LCD, and the RGB values of the soil samples are sent to ceerduad.ac.id, a web server, in the form of low, medium, or high phosphorus status. The design of the phosphorus measurement tool is shown in Figure 5.

Figure 4

Figure 5

From the readings of the phosphorus level obtained using the TCS3200 sensor for the 20 soil samples extracted, this study obtained the RGB values shown in Table 1.

The RGB values in Table 1 constitute the data snippet from the 200 training data used for reading phosphorus levels on the PUTS color chart. For determining the phosphorus level status, 20 soil samples were used, and each soil sample contained 10 training data for sensor readings. The graph of RGB values obtained from these 20 samples is shown in Figure 6. The soil sample measurements produced 200 experimental data values, out of which 194 measurement data obtained using a phosphorus measuring instrument based on PUTS measurements were valid and six experimental results were not suitable. From the results obtained, the error value can be calculated using Eq. (2) as follows: (2) Error=Numb. of PUTS Test−Numb. of Testing of MeasuringNumb. of PUTS Tests×100%Error=3% \matrix{{Error = {{Numb.\,of\,PUTS\,Test - Numb.\,of\,Testing\,of\,Measuring} \over {Numb.\,of\,PUTS\,Tests}} \times 100\%} \hfill \cr {Error = 3\%} \hfill \cr}

Figure 6

Using the Naïve Bayes equation, we found that the error rate of the measuring instrument was 3%, and the instrument accuracy was 97%. This result was compared with previous research that was applied for nitrogen monitoring [44], with an accuracy of 87.5%. To measure the accuracy, other studies used the coefficient of determination (R²), as shown previously [49] for monitoring of the leaf nitrogen concentration of wheat, with accuracy of 0.91%. Another study [50] calculated the accuracy of the sensor by using the coefficient of determination (R²) for measuring the total nitrogen content in agricultural runoff. [53] also proposed the monitoring of wheat grain nitrogen content with the coefficient of determination (R²), with a result of 0.42%.

The Naïve Bayes method is used to calculate the probability of an event occurring in the future based on previous experience [51], [52]. Naïve Bayes analysis was used to classify the test results of the 20 soil samples to obtain the paddy soil phosphorus status. Based on the graph in Figure 6, the result obtained was as follows:

P(c(MEDIUM)) = 68 / 200 = 0.34 ;

Tables 2–4 show the RGB value probabilities from the research results.

Table 2 shows the probability of phosphorus status in each class of red values divided into four ranges (1–63, 64–127, 128–190, and 192–255). Classification results based on red odds indicate the possibility of a medium or high level of phosphorus content. The probability of the green value class is shown in Table 3.

Table 3 shows the probability of the phosphorus status in each green value class divided into four classes with moderate- and high-phosphorus-status probabilities. The RGB reading value ranges from 1 to 255. The probability of the blue value status is shown in Table 4.

Table 4 shows the probability of phosphorus status in each blue value class divided into four classes with moderate and high phosphorus-status probabilities. The RGB reading value ranges from 1 to 255.

Example: The RGB value read by the sensor in the phosphorus meter: (3) x={R=126; G=111; B=81}. x = \{R = 126;\,G = 111;\,B = 81\}.

Then, the phosphorus status of the rice fields is determined using Eq. (1).

The first step is to determine the probability of medium- and high-value P(c) of all training data used for reading the paddy soil sample: (4) P(S)=68/200;P(T)=132/200. P(S) = 68/200;P(T) = 132/200.

Next, we determine the P(x|c) value of the medium- and high-phosphorus-status probabilities of the known data classes:

Medium-phosphorus-status probability (5) P(rse)=30/68;P(gse)=68/68; P(bse)=70/68. P(rse) = 30/68;P(gse) = 68/68;\,P(bse) = 70/68.

High-phosphorus-status probability (6) P(rti)=122/132;P(gti)=70/132; P(bse)=109/132. P(rti) = 122/132;P(gti) = 70/132;\,P(bse) = 109/132.

For determining the P(x|c) P(c) value:

MEDIUM (7) P(c)P(se)=3068×6868×7068×68200=0.15. P(c)P(se) = {{30} \over {68}} \times {{68} \over {68}} \times {{70} \over {68}} \times {{68} \over {200}} = 0.15.

HIGH (8) P(c)P(ti)=122132×70132×109132×132200=0.26. P(c)P(ti) = {{122} \over {132}} \times {{70} \over {132}} \times {{109} \over {132}} \times {{132} \over {200}} = 0.26.

The occurrence of the set probability value x in the entire data set is as follows: (9) P(x)=152200×138200×179200=0.58. P(x) = {{152} \over {200}} \times {{138} \over {200}} \times {{179} \over {200}} = 0.58.

For determining the probabilities of moderate and high phosphorus status in the set value x, we proceed as follows: (10) Moderate:P(se|x)=0.150.58=0.25; \matrix{{{\rm{Moderate}}:} \hfill & {P(se|x) = {{0.15} \over {0.58}} = 0.25;} \hfill \cr} (11) High:P(ti|x)=0.260.58=0.44; \matrix{{{\rm{High}}:} \hfill & {P(ti|x) = {{0.26} \over {0.58}} = 0.44;} \hfill \cr}

So, the value P(se|x) > P(ti|x).

It can be concluded that for RGB value (12) x={R=126; G=111; B=81}=Fosfor Tinggi. x = \{R = 126;\,G = 111;\,B = 81\} = Fosfor\,Tinggi.

The data obtained from the average RGB values of the 20 samples of paddy fields in Figure 6 that have been tested and classified using the Naïve Bayes algorithm are shown in Table 5.

Table 5 shows the average RGB values of all soils that were tested and classified using the Naïve Bayes algorithm. The classification results yielded seven moderate phosphorus statuses and 13 high phosphorus statuses.

A graph of the average RGB values of the test samples can be obtained from the classification results, as shown in Figure 7 for the 20 soil samples. Soils with medium phosphorus values are present in the 3rd, 4th, 10th, 11th, 12th, 17th, and 20th samples. High phosphorus values are indicated by the RGB values of the 1st, 2nd, 5th, 6th, 7th, 8th, 9th, 13th, 14th, 15th, 16th, 18th, and 19th samples.

Figure 7

The data are sent to the ceerduad.com web server using the esp8266 WiFi module for sending the sensor reading data. The view of the ceerduad.com server is shown in Figure 8.

Figure 8

After successfully logging into the ceerduad.com website, a graph of the measurements of the paddy soil sample, which were read by the TCS3200 sensor, will be displayed according to the data that have been sent by the phosphorus-measuring instrument. The graphic image of the RGB value data that were successfully sent to the website is shown in Figure 9.

Figure 9

The graph in Figure 9 represents the sensor data sent from the phosphorus meter to the website based on the phosphorus meter reading. The graphic data in Figure 9 can be downloaded in Excel form. The graph shows the RGB values from the sampling at 20 rice fields in Lindahl District. Soil sample data are then extracted with PUTS and tested using a phosphorus meter that has been developed.

Mapping is carried out to map the results of reading of the phosphorus levels in the soil samples at a specific location using several steps, as shown in Figure 10.

Figure 10

ArcGIS mapping was carried out by taking satellite imagery of the mapped location using Google Earth. The location pictures captured by the satellite are then used to create an SHP data (shapefile) map. The shapefile data that have been created are given an input of several attribute data scores, which then enter the stage of merging of several spatial elements into new spatial elements (overlay). The process of obtaining the cartographic map layout is completed in various steps, starting from the overlay process to the merging stage of several spatial elements into one spatial element without changing the combined spatial elements (union).

The sampling of paddy soil was carried out in the Lendah district, divided into 20 test locations, where each point represented one sample of paddy soil. The map of the paddy soil sampling distribution is shown in Figure 11.

Figure 11

ArcGIS software created a map of the sampling locations in Figure 11. ArcGIS is a processing software based on geographic data, which can present, manipulate, and save geographic information data. Some main features of the ArcGIS software in mapping include the ArcMap and ArcCatalog. ArcMap is used in data management, including visualization, editing, and map-making, while ArcCatalog is a feature used for creating vector data, raster data, and grouping according to its function.

The mapping of the paddy fields' phosphorus status was accomplished based on the soil sampling location and phosphorus status obtained during the testing of the paddy soil samples, as shown in Figure 12.

Figure 12

From the mapping of the paddy soil phosphorus status in Figure 12, it can be seen that of the 20 locations of paddy soil sampling, 13 locations were of dark blue paddy fields and seven light blue locations. A dark blue indicator on the paddy soil phosphorus status map above indicates a high status of phosphorus, while the light blue color indicator on the map above indicates the moderate status of phosphorus.

This geographical map contains information that researchers prepared to facilitate use by readers, such as rivers, roads, village boundaries, inland waters, plantation villages, and rice fields. This map aims to focus on the nutrient content in terms of soil phosphorus in the rice fields of Lendah district. The final result of mapping of phosphorus levels in the Lendah District is shown in Figure 13.

Figure 13

In making the phosphorus content map, Figure 13 used the raster and vector data. Raster data is in the form of squares or cells. Raster data like a image data presented in the form of jpg and unitary coordinates in square. Vector data are in the form of spatial data such as points, lines, and areas. On a map, raster data form an image in jpg format, while vector data are in the form of roads, rivers, village boundaries (in the form of lines), villages, and rice fields (data that have a large area).