Prediction of water quality parameters using an artificial neural network for irrigation


The result of this study is presented in three categories, namely; descriptive statistics, water quality test result and ANN model and model evaluation performance, respectively.

The result of descriptive statistics is presented in Tables 1, 2, 3, 4. It describes the basic characteristics of the data from this study. They provide simple summaries on the sample and measures such as mean, median, maximum, minimum, and standard deviation, respectively.

Table 1 Descriptive statistics of the water quality analyzed in point 1.
Table 2 Descriptive statistics of the water quality analyzed in point 2.
Table 3 Descriptive statistics of the water quality analyzed in point 3.
Table 4 Descriptive statistics of the quality of the water analyzed in point 4.

The descriptive statistics in Tables 1, 2, 3, 4 show that the mean values ​​of the dataset range from 6.29 to 6.34, 1956.21 to 2458.19, 3.35 to 7.39 and 39 , 13 to 51.06 for Ph, TDS (mg / l), EC (dS / m) and Na (mg / l), respectively. Median values ​​for the dataset range from 6.31 to 6.39, 2010.00 to 2439.50, 3.14 to 4.24, and 39.13 to 51.06 for pH, TDS (mg / l), EC (dS / m) and Na (mg / l), respectively. The maximum data set is 6.48-6.64, 2286.00-2742.00, 2.21-5.82, and 64.50-88.45 for Ph, TDS (mg / l), EC (dS / m) and Na (mg / l), respectively. The minimum data set ranges from 6.00 to 6.09, 1367.00 to 2199.00, 2.01 to 3.18 and 21.21 to 40.24 for Ph, TDS (mg / l) , EC (dS / m) and Na (mg / l), respectively. Standard deviation values ​​range from 0.08 to 0.16, 114.47 to 213.04, 0.23 to 31.49 and 14.06 to 8.16 for Ph, TDS (mg / l), EC (dS / m) and Na (mg / l), respectively. The low standard deviation values ​​recorded in this study show that the dataset was very close to the average of the dataset.

The result of the water quality analysis test indicates the level of concentrations of TDS (mg / l), EC (dS / m) and Na (mg / l) in the Ele river in Nnewi, in the Anambra State in Nigeria. The FAO standard for irrigation water quality for TDS, EC and Na is 0–2000, 0–3 and 0–40, respectively. The water quality results show that the pH values ​​ranging from 6.01 to 6.87 were within the FAO standard at all points for the rainy and dry seasons, while the TDS (mg / l), EC (dS / m) and Na (mg / l) parameter values ​​range from 2001 to 2506, 3.01 to 5.76 and 40.42 to 73.45 respectively, were above the norm of FAO from point 1 to point 3 and fall within the FAO standard in point 4 with values ​​ranging from 1003 to 1994, 2.01 to 2.78 and 31.24 to 39.44, respectively. However, during the dry season, the TDS, EC and Na values ​​ranging from 2002 to 2742, 3.04 to 5.82 and 40.14 to 88.45 respectively, were all above the FAO standard. Anthropogenic pollution emitted in water bodies has recently been identified as a significant source of pollutants which require immediate action in order to avoid serious environmental effects.11.

The results also showed that concentrations decrease along the downstream sampling points. It should be noted that irrigation water with a pH outside the normal range may cause nutritional imbalance or may contain toxic ion which is harmful to crops.19. The high concentrations of TDS observed in this study are likely to increase the salinity of the river water, alter the taste of the water and thus decrease the level of dissolved oxygen in the surface water, which which makes it difficult for plants and aquatic organisms to survive. organizationsseven.

In addition, these anions and cations which increase the electrical conductivity in the water negatively affect the irrigation because the salts are deposited in the root zones of the crops, making it difficult for infiltration, absorption of moisture and soil. nutrients needed for agricultural production.

The ANN model and forecasts for water quality parameters are shown from Figs. 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19. Given the permissible range of water quality, modeling and River quality forecasts show different seasonal variations such that the level of pollution during the dry season was higher than during the rainy season.

Figure 4

(A and B): pH model and forecast graph at point 1.

Figure 5
digit5

(A and B): TDS model and forecast graph at point 1.

Figure 6
digit6

(A and B): EC model and forecast graph at point 1.

Figure 7
number7

(A and B): Na model and forecast graph at point 1.

Figure 8
figure 8

(A and B): Ph model and forecast graph at point 2.

Figure 9
number9

(A and B): TDS model and forecast graph at point 2.

Figure 10
number10

(A and B): EC model and forecast graph in point 2.

Figure 11
figure11

(A and B): Na model and forecast graph at point 2.

Figure 12
figure12

(A and B): Ph model and forecast graph at point 3.

Figure 13
figure13

(A and B): TDS model and forecast graph in point 3.

Figure 14
figure14

(A and B): EC model and forecast graph in point 3.

Figure 15
number15

(A and B): Na model and forecast graph at point 3.

Figure 16
number16

(A and B): pH model and forecast graph at point 4.

Figure 17
number17

(A and B): TDS model and forecast graph in point.

Figure 18
number18

(A and B): EC model and forecast graph in point 4.

Illustration 19
number19

(A and B): Na model and forecast graph at point 4.

Typically, the artificial neural network models the actual data set very well. At various sample points, the developed ANN models descriptively show insignificant deviation values ​​for the actual data set. There were continuous variations in developed models and forecasts over time.

Multilayer Neural Network Model (FFMNN) Performance Assessment Results are shown in Table 5. The performance evaluation of the model was carried out based on the training of the developed ANN model, testing and forecasting, respectively. The performance evaluation of the model was carried out using the multiple coefficient of determination R2 and root mean square error (RMSE).

Table 5 Statistical measurement of the model formed, test and forecast.

The R2 The values ​​were generally observed to vary to the second decimal place for the training, testing and forecasting model, respectively.

The evaluation of training performance shows that R2 the values ​​range from 0.981 to 0.990, 0.981 to 0.988, 0.981 to 0.989 and 0981 to 0.989, for pH, TDS, EC and Na, respectively. The results of the training show that the pH model has the best performance, followed by EC and Na.

In addition, test performance shows that the R2 the value ranges from 0.952 to 0.967, 0.953 to 0.970, 0.951 to 0.967 and 0.953 to 0.968, for pH, TDS, EC and Na, respectively. However, the evaluation of test performance shows that TDS had the best performance. The evaluation of the forecast performance shows that the R2 values ​​range from 0.945 to 0.968, 0.946 to 0.968, 0.944 to 0.967 and 0.949 to 0.965 for pH, TDS, EC and Na respectively. However, it was found that TDS made the best prediction followed by pH. The performance of the water quality forecast was assessed using the root mean square error (RMSE) which ranges from 0.022 to 0.088, 0.012 to 0.087, 0.015 to 0.085 and 0.014 to 0.084 for pH, TDS, EC and Na, respectively. The ANN model worked very well because their coefficient of multiple determinations R2 were very close to 1, which is in agreement with the study by Awu et al. (2017) and Abrahart et al. (2005). By comparing the performance of the training model to the test model and the predictions, it shows that the training set performed better than the test set followed by the prediction as a coefficient of multiple determinations, R2, was much closer to 1.

Previous Chocolate company made in Ghana exporting to Europe
Next Intel apologizes for asking suppliers to avoid products from China's Xinjiang region