KR20120075948A - Method of presumption for quality of water using multiple regression - Google Patents

Abstract

PURPOSE: A water quality environment estimation method through multiple regression analysis is provided to estimate the water quality by using reliable independent variable observed from an observatory without requiring a high cost system. CONSTITUTION: Data of water quality environment is collected(S1). A multiple regression analysis is performed by using the independent variable. A multicollinearity diagnosis is performed(S3). The independent variable is at least more than two out of water temperature, temperature, pressure conductivity, salinity, dissolved oxygen, chlorophyll content. The multicollinearity diagnosis uses the dispersion expansion index and is evaluated. If the dispersion expansion index is over ten, the explanation variable is removed(S5).

Description

Method of Presumption for Quality of Water using Multiple Regression}

The present invention relates to a method for estimating water quality through multiple regression analysis, and to a method for estimating water quality of aquaculture farms using independent variable data.

In the process of fisheries using aquaculture farms, water quality information such as water temperature and salinity is very important. Generally, a fishery producer installs a sensor in a tank at a production site and monitors water quality information such as water temperature and salinity and takes action when there is an abnormality. However, the monitoring system has problems such as the purchase cost of equipment including a sensor for water environment information, and there is a burden of continuously repairing and managing the system.

The present invention has been made to solve the problems of the prior art and the necessity of technology development, and provides a method for estimating the water environment through multiple regression analysis that can estimate the water environment without expensive costs Has its purpose.

In addition, an object of the present invention is to estimate the water quality environmental information based on reliable independent variables.

In order to achieve the above object, the method for estimating water quality through multiple regression analysis according to the present invention comprises: collecting data of independent variables corresponding to the water quality environment; Performing a multiple regression analysis using the independent variable; And performing a multicollinearity diagnosis.

The independent variable may be at least two of water temperature, air temperature, barometric conductivity, salinity, dissolved oxygen content and chlorophyll content.

Multicollinearity diagnosis is evaluated using the dispersion expansion index, and if the dispersion expansion index is 10 or more, it may be determined that the explanatory variable causes a multicollinearity problem, and thus the explanatory variable may be removed.

And multiple regression analysis for water temperature prediction can use the following equation (1).

[Equation 1]

Aquaculture water temperature = -4.402 + 0.906 * foaming water temperature + 0.198 * foaming salinity

In addition, multiple regression analysis for dissolved oxygen prediction can use the following equation (2).

&Quot; (2) "

Dissolved oxygen = 15.833-0.209 * water temperature-0.142 * salinity

In addition, multiple regression analysis for dissolved oxygen prediction may use the following [Equation 3].

&Quot; (3) "

Dissolved oxygen = 11.488-0.2097 * water temperature

As described above, according to the present invention, the water quality environment can be estimated using reliable independent variables observed at the station without requiring a costly system.

1 is a flow chart showing a water quality estimation method according to the present invention.
2 is a graph showing a temperature change trend of an independent variable acquisition sample area.
3 is a graph showing the change in salinity of the independent variable acquisition sample area.
4 shows aquaculture farm temperature and dissolved oxygen content;
5 shows the estimated and actual temperatures.
6 shows the estimated dissolved oxygen and actual observations.

Although the terms used in the present invention have been selected as general terms that are widely used at present, there are some terms selected arbitrarily by the applicant in a specific case. In this case, the meaning described or used in the detailed description part of the invention The meaning must be grasped.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flow chart showing a water quality estimation method through multiple regression analysis according to the present invention.

Referring to FIG. 1, the method for estimating water quality through multiple regression analysis according to the present invention first collects data of independent variables corresponding to the water quality environment (S1). Independent variables include at least two of water temperature, air temperature, barometric conductivity, salinity, dissolved oxygen content and chlorophyll content. Then, multiple regression analysis is performed using the independent variable (S2). Then, multicollinearity diagnosis is performed (S3). Multicollinearity diagnosis can be evaluated using the dispersion expansion index (S4). In this case, when the variation inflation factor (VIF) is 10 or more, it is determined that the explanatory variable causes a multicollecting problem, and the explanatory variable is removed (S5). After the multicollinearity diagnosis results, the dispersion expansion index does not exceed 10 or after the explanatory variable is removed, the estimated value and the observed value are compared (S7). Subsequently, when the difference between the estimated value and the observed value is not large, a water environment estimation model may be established (S8). The result difference between the estimated value and the observed value for establishing the water quality estimation model may be set by the user.

Looking at the water environment estimation method through the multiple regression analysis in more detail through the following examples.

The survey areas selected for performing the actual independent variable data and the multi-session analysis were applied to three areas including Geumho Fisheries, a land tank farm in Goheung-gun, Goheung-gun, Jeollanam-do, Yeosu Odongdo, and Joheung Observatory. Since November 2008, Kumho Fishery Farm has been selected as a pilot farm for environmental information monitoring system and has installed sensors and monitoring system in the farm. The water temperature, salinity, dissolved oxygen, chlorophyll, pH, etc., which are the environmental information of the tank, have been accumulated in the system every minute from the intelligent sensor installed in the tank. The foaming and odongdo tidal observatory data were requested from the National Maritime Research Institute, and the tide, wind speed, gust, wind direction, temperature, barometric pressure, conductivity, water temperature, and salinity data measured every hour were used for this study.

In addition, data of water quality items were used from January 2009 to July 2010. We want to perform statistical analysis by classifying data from three regions into five sets as shown in [Table 1]. In [Table 1], A represents Kumho farm, B represents firing area, and C represents Yeosu area. The first data set is to examine the characteristics of the water quality information of the Kumho Fishery Farm, and the second data set is to examine the characteristics of the water quality, which is a measure of the tide station. To do salinity difference analysis. Fourth, the data set is for estimating the environmental information of Kumho Fishery Farm using the bubbling tide station data. Fifth, the data set is composed of data sets for plotting the predicted and measured values by the regression estimation equation.

division Item term area Set1 Water temperature, salinity, DO, chlorophyll, pH, oxygen saturation January 2009 ~ December 2009 A Set2 Tide, wind speed, gust, wind direction, temperature, barometric pressure, conductivity, water temperature, salinity January 2009 ~ December 2009 A, B Set3 Water temperature, salinity January 2009 ~ December 2009 A, B, C Set4 A: water temperature, salinity, pH, dissolved oxygen
B: water temperature, air pressure, conductivity, wind speed, salinity January 2009 ~ December 2009 A, B Set5 A: water temperature, dissolved oxygen
B: water temperature, salinity 2010.2.4 ~
2010.7.30 A, B

Statistical processing for statistical analysis was performed using SAS 9.2, a representative statistical package. The daily average value of water quality factor data measured every minute at Kumho Fisheries farm was used for analysis. Most of the statistical analysis except the correlation analysis for the data measured every hour at the tidal station, for example, analysis of variance, factor analysis, multiplexing The daily mean value of the tide station data was calculated and used for the regression analysis and the PLOT of the predicted value.

Then, factor analysis is performed on nine water quality items at the tide station to find several potential variables by grouping the highly correlated variables of the data.The number of factors is selected by selecting an eigenvalue corresponding to the mean explanatory power of each factor of 1 or more. Let's count.

[Table 2] shows the water quality statistics for the 2009 Kumho Fisheries Farm.

variable
month Temp. Sal. DO Chl. pH N X ± SD X ± SD X ± SD X ± SD X ± SD (day) One 7.37 ± 0.52 30.75 ± 0.56 10.05 ± 0.24 1.31 0.52 8.14 0.06 30 2 8.82 ± 0.42 30.08 ± 0.93 9.78 ± 0.16 1.22 0.33 8.18 0.03 26 3 10.26 ± 0.94 29.14 ± 0.85 9.40 ± 0.34 1.64 0.50 8.19 0.04 30 4 12.92 ± 0.94 30.02 ± 1.48 8.33 ± 0.40 1.51 0.55 7.99 0.20 26 5 14.88 ± 0.86 27.50 ± 0.18 8.21 ± 0.15 1.28 0.32 8.18 0.07 5 6 22.07 ± 1.53 29.54 ± 0.36 6.26 ± 1.26 1.48 0.66 8.18 0.08 13 7 21.55 ± 0.50 27.99 ± 0.78 6.86 ± 0.27 2.91 1.37 7.98 0.13 27 8 24.69 ± 1.87 28.44 ± 0.81 6.31 ± 1.15 1.06 0.22 8.27 0.17 28 9 24.10 ± 0.70 30.57 ± 0.62 6.41 ± 0.20 1.68 0.79 8.29 0.05 30 10 20.32 ± 1.65 30.17 ± 0.48 7.24 ± 0.33 2.98 0.47 8.17 0.06 29 11 14.80 ± 1.97 28.60 ± 0.87 8.09 ± 0.38 3.30 0.68 7.97 0.13 26 12 10.62 ± 1.29 28.50 ± 0.77 9.26 ± 0.66 2.85 0.39 8.19 0.08 15

As shown in [Table 2], the water temperature was the lowest in January and the highest in August due to seasonal factors. The concentration of dissolved oxygen required for the survival of fish is more than 4ppm, and dissolved oxygen, which is inversely related to the water temperature, is observed to be 6.2 or more. The hydrogen ion concentration (pH) is alkaline at most pH 7.0 or more. There is no big fluctuation in chlorophyll concentration when the nutrients are introduced in large quantities, but July, October, November, and December were slightly higher than other months. N is an observation day, and problems such as aquaculture tank cleaning and system blackouts are insufficient in May and June due to the lack of observation days.

Table 3 is a table showing seasonal water quality statistics. In Table 3, the seasons were classified as December, January, February in spring, March, April and May in summer, June, July and August in summer, and September, October and November in autumn. This can then be compared with the seasonal characteristics of the station area.

N Tp Sa Do Cp Ph Fall 85 19.97 ± 4.08 29.83 ± 1.07 7.21 ± 0.75 2.62 ± 0.97 8.15 ± 0.16 Spring 61 11.78 ± 1.83 29.38 ± 1.33 8.85 ± 0.65 1.56 ± 0.51 8.10 ± 0.17 Summer 68 22.94 ± 2.03 28.47 ± 0.92 6.52 ± 0.96 1.87 ± 1.25 8.14 ± 0.19 winter 71 8.59 ± 1.43 30.03 ± 1.13 9.78 ± 0.46 1.60 ± 0.78 8.17 ± 0.06

[Table 4] shows the results of correlation analysis on five water quality items to determine the linear correlation between variables.

Temp. Sal. DO Chl. pH Temp. 1,000 Sal.  -0.16849 1,000 DO  -0.95471 0.13080 1,000 Chl. 0.12519 -0.27459  -0.12736 1,000 pH  0.14362 0.14475 0.03082 -0.31983 1,000

Referring to Table 4, it can be seen that the water temperature shows a high negative correlation with the dissolved oxygen (r = -0.95).

Table 5 shows the results of factor analysis based on the 2009 yearly data for the firing area.

Variables Factor1 Factor2 Factor3 wtp 0.91315 0.20188 -0.18942 tp 0.91443 0.19596 -0.24771 ap -0.72324 -0.27681 0.30463 ws -0.34217 0.90026 0.00177 CD 0.88747 0.20831 0.31764 jw 0.25248 0.03425 0.53549 sa 0.24825 0.07999 0.86757 gu -0.39153 0.89023 0.00736 wd -0.34946 0.3449 0.02625 Eigen value 3.4985850 1.92867033 1.33111390 Variance 3.4985851 1.9286703 1.3311139

Referring to [Table 5], the eigenvalue is selected to be one or more, and the number of factors is three. Factor 1 includes water temperature, temperature, barometric pressure, and conductivity, and short-term pressure is inversely related to other variables. Factor 2 includes wind speed and gust, and factor 3 is salinity.

[Table 6] is a table showing the correlation of environmental factors in Goheung firing area.

wtp tp ap ws CD jw sa gu wd wtp One tp 0.91764 One ap -0.63697 -0.73697 One ws -0.12329 -0.13937 0.04741 One CD 0.83519 0.76135 -0.52205 -0.11839 One jw 0.14573 0.12769 -0.07769 -0.04394 0.23723 One sa 0.0259 0.015 0.00291 -0.02751 0.56689 0.21588 One gu -0.17531 -0.19025 0.0687 0.951 -0.16322 -0.05308 -0.03207 One wd -0.24122 -0.22238 0.06796 0.24123 -0.20515 -0.0684 -0.01525 0.29196 One

Referring to [Table 6], in the correlation analysis, the water temperature indicates a positive correlation with the air temperature (r = 0.92), a negative correlation with the air pressure (r = -0.64), and a positive correlation with the conductivity (r = 0.84). Is positively correlated with silver conductivity (r = 0.76), air pressure is negatively correlated with conductivity (r = -0.52), wind direction is positively correlated with gust (r = 0.95), and conductivity and salinity (r = 0.57) are positive It can be seen that the correlation is shown.

Table 7 shows the mean and standard deviation of the seasonal water quality data for the foaming port and Yeosu Odongdo.

wtp tp ap ws CD sa gu B C B C B C B C B C B C B C spring 19.41 ± 4.71 20.40 ± 3.82 15.65 ± 5.59 16.91 ± 5.29 1016.3 ± 5.75 1016.3 ± 5.64 3.04 ± 1.01 3.54 ±
1.65 43.17 ± 3.81 37.75 ± 6.49 31.59 ± 0.51 29.34 ± 2.99 4.07 ± 1.55 4.55 ± 1.99 summer 13.06 ± 3.17 13.17 ± 3.23 11.78 ± 4.23 13.44 ± 4.42 1013.8 ± 5.63 1014.0 ± 5.68 2.94 ± 1.39 2.81 ±
1.65 37.80 ± 3.82 31.73 ± 4.64 31.93 ± 2.02 26.54 ± 3.97 4.10 ± 2.13 3.89 ± 2.17 autumn 22.38 ± 2.23 22.05 ± 2.04 21.82 ± 2.34 23.32 ± 2.20 1005.9 ± 3.80 1006.5 ± 3.86 2.57 ± 1.06 2.79 ±
1.37 45.57 ± 1.94 39.65 ± 4.40 31.30 ± 0.92 26.11 ± 4.07 3.40 ± 1.42 3.89 ± 1.78 winter 7.02 ± 1.97 8.12 ± 2.07 3.59 ± 3.81 4.87 ± 3.82 1020.0 ± 4.89 1019.8 ± 4.90 3.31 ± 1.29 3.72 ±
1.71 31.58 ± 3.36 30.24 ± 6.40 31.01 ± 2.74 28.97 ± 6.25 4.72 ± 2.01 4.90 ± 2.35

At this time, in [Table 7] B represents the foaming area, C represents the Yeosu area.

The data of tide station measured every hour have information of tide, wind speed, wind gust, wind direction, temperature, barometric pressure, conductivity, water temperature, salinity. , pH, chlorophyll.

The same observation items in the three areas, Kumho Fishery Farm, Foaming Area, and Yeosu Area, were used for variance analysis.

Figure 2 is a graph showing the water temperature between the three regions in 2009, Figure 3 is a graph showing the salinity. In FIGS. 2 and 3, the dividing lines of the graphs indicate February, March, April, May, June, July, August, September, October, November, and December start dates, respectively.

2 and 3, it can be seen that the difference in water temperature between the three regions is very small and the salinity is different.

And Table 8 is a table showing the variance analysis of these FIG. 2 and FIG.

Dependent variable Type III sum of squares Degrees of freedom Mean square F Significance Water temperature 57.7084 2 28.8542 0.69 0.503 Salinity 3226.1824 2 1613.0912 185.39 <.0001

In the results of the analysis of variance in Table 8, it can be seen that the water temperature does not differ between the three regions at the significance level of 5%, and the salinity is different. Yeosu Odongdo, in particular, has a lower salinity than other regions because of its close proximity to urban areas.

Table 9 shows the results of the post test on the months with significant differences.

Mon. Para. Pr> F Duncan grouping Ave. N Sites Jan. Temp.
0.0001
A
B
C 7.37
6.29
4.86 30
31
31 A
C
B Sal.
0.0369
A
A
B 30.83
30.75
28.09 31
30
31 B
A
C Feb. Temp.
0.0001
A
B
B 8.82
7.83
7.63 26
28
28 A
B
C Sal.
0.0079
A
A
B 30.08
29.53
26.49 26
28
28 A
B
C Mar. Temp.
0.002
A
B
B 10.26
9.52
9.46 30
31
31 A
C
B Sal.
0.0001
A
B
C 31.30
29.14
28.01 31
30
31 B
A
C Sep. Temp.
0.0600
A
AB
B 24.46
24.35
24.10 30
30
30 B
C
A Sal.
0.0001
A
A
B 31.14
30.57
27.79 30
30
30 B
A
C Oct. Temp.
0.015
A
AB
B 21.10
20.32
19.97 31
29
31 C
A
B Sal.
0.0001
A
B
C 31.51
30.17
28.64 31
29
31 B
A
C Nov. Temp.
0.0048
A
AB
B 15.72
14.80
13.80 30
26
30 C
A
B Sal.
0.0001
A
A
B 32.13
31.60
28.60 30
30
26 B
C
A Dec. Temp.
0.0001
A
A
B 10.62
10.39
8.45 15
31
31 A
C
B Sal.
0.0001
A
B
C 32.54
32.09
28.50 31
31
15 B
C
A

Referring to [Table 9], there was no difference in water temperature by region of 4, 5, 6, 7, 8, and September at the significance level of 5% when comparing the three regions monthly. There was a significant difference in December. However, the observation dates of Kumho Fisheries in May and June were 5 days and 13 days, respectively, so there was no difference in water temperature between Yeosu and firing. Salinity was also different among the three regions in the monthly analysis, and foaming had the highest salinity and the lowest Yeosu.

In order to predict the water temperature in the near-farm, using the data from Goheung foam tide station, multiple regression variables were performed using independent variables as water temperature, wind speed, and salinity based on the results of factor analysis and correlation analysis. As a result, it was found that the p-value of the F test statistic was less than the significance level of 0.05, so that at least one of the water temperature, salinity, and wind speed affects the farm water temperature. However, in the t-test for significance test of the regression coefficient, the significant explanatory variables at the significance level of 0.05 were the station temperature and salinity. Accordingly, the results obtained by excluding the wind speed are shown in [Table 10].

Item Non-standardized coefficient Normalization factor t Significance VIF β Error a constant -4.402 1.667 -2.64 <.0088 0 Effervescent water temperature 0.906 0.007 1.014 121.75 <.0001 1.5667 Effervescent salinity 0.198 0.050 0.033 3.97 <.0001 1.5667 R ² = .988 Adjusted R ² = .988 F (p) = 11167.8 (<.0001)

The regression model was found to be significant because F = 11167.8 (p <.001) was smaller than the significance level of 0.05 as a result of the F-test and the coefficient of determination (R ² ). In addition, the value of the coefficient of determination is close to 1, and it can be estimated that the regression model accounts for most of the entire variance. The t-test on the regression coefficients is also significant and the VIF value is close to 1, indicating no multicollinear effects. The standardized regression coefficient values indicate that the water temperature is 1.014, which explains most of the overall regression model.

Therefore, the regression equation for farm water temperature prediction through multiple regression analysis can be expressed as Equation 1 below.

[Equation 1]

Aquaculture Water Temperature = -4.402 + 0.906 * Foaming Water Temperature + 0.198 * Foaming Salt

And Table 11 is a regression model for pH, salinity, dissolved oxygen.

Multiple linear regression Model Sig pH = 5.368-0.024 * salin + 0.003 * atm + 0.003 * water temperature 0.06 ** Sal = 30.151-0.044 * Temperature + 0.004 * Salinity-0.056 * Blast 0.07 ** DO = 15.833-0.209 * water temperature-0.142 * salinity 0.89 **

In this case, pH and salinity are very low, so the regression model is not appropriate, but [2], which is another regression equation for dissolved oxygen, has a linear regression relationship at the 89% crystallinity.

[Equation 2]

Dissolved oxygen = 15.833- 0.209 * water temperature -0.142 * salinity

Therefore, it is known that the water temperature and dissolved oxygen can be used to estimate the water quality information of farms using Goheung tide station data. That is, in the estimation of the dissolved oxygen using the 2009 Kumho Fisheries measurement data, it can be expressed as the following regression equation [Equation 3].

Aquaculture Dissolved Oxygen = 11.488-0.2097 * Aquaculture Water Temperature

Based on Equation 3, the correlation between the farm water temperature and dissolved oxygen is shown in FIG. 4.

Referring to FIG. 4, in the regression linear relationship at the level of 96% of the coefficient of determination, the dissolved oxygen regression equation of [Equation 3] can be seen that the observed values are concentrated in a straight line than the dissolved oxygen regression equation of [Equation 2]. .

FIG. 5 is a diagram illustrating estimation of water temperature and dissolved oxygen value based on Equation 1, which is a regression equation estimated based on an independent variable as described above. In FIG. 5, red is an estimated value by a regression equation, and black is an actual observed water temperature.

6 is a view showing the dissolved oxygen amount estimated based on [Equation 2] and [Equation 3]. In FIG. 6, the red color indicates dissolved oxygen amount estimated based on [Equation 2], and the green color indicates dissolved oxygen amount estimated based on [Equation 3].

5 and 6, it can be seen that the green data, which is the dissolved oxygen estimate of Equation 3 using the water temperature estimate, and the red data estimated by Equation 2 are almost similar. That is, it can be seen that the water environment prediction method using Equation 3 is suitable for predicting dissolved oxygen in aquaculture farms.

As described above, the present invention has been described with reference to preferred embodiments, but is not limited to the above-described embodiments, and various modifications may be made by those skilled in the art without departing from the gist of the present invention. Changes and corrections will be possible.

Claims (6)

Collecting data of independent variables corresponding to the water environment;
Performing a multiple regression analysis using the independent variable; And
A method of estimating water quality through multiple regression analysis, comprising: performing a multicollinearity diagnosis.
The method of claim 1,
The independent variable is at least two or more of water temperature, air temperature, barometric conductivity, salinity, dissolved oxygen content, chlorophyll content, water environment estimation method through multiple regression analysis.
The method of claim 1,
The multicollinearity diagnosis is evaluated by using a variance expansion index, and when the variance expansion index is 10 or more, it is determined that the explanatory variable causes a multicollinearity problem, and the multivariate regression analysis is performed. Method of estimating water quality through water.
The method of claim 1,
The multiple regression analysis for water temperature prediction is characterized by using the following equation.
(Mathematical formula)
Aquaculture water temperature = -4.402 + 0.906 * foaming water temperature + 0.198 * foaming salinity
The method of claim 1.
The multiple regression analysis for the prediction of dissolved oxygen uses the following equation.
(Mathematical formula)
Dissolved oxygen = 15.833-0.209 * water temperature-0.142 * salinity
The method of claim 1,
The multiple regression analysis for the prediction of dissolved oxygen uses the following equation.
(Mathematical formula)
Dissolved oxygen = 11.488-0.2097 * water temperature

Images