KR20230144249A - How to construct an optimal water quality observation network - Google Patents

Abstract

The present invention relates to the construction of an optimal water quality observation network. Based on water quality data measuring the water quality of water bodies, water quality trends at each observation point are analyzed and similarity between observation points is calculated to construct an optimal water quality observation network for water quality management. This is to support you. The present invention includes a collection step of collecting water quality data of the target water body, and a similarity analysis step of performing water quality similarity analysis for each observation point based on the application of principal component analysis, correlation coefficient matrix, and multidimensional scaling method for each item of the water quality data. , discloses a method of configuring an optimal water quality observation network, which includes a decision step of determining the number of observation points and representative observation points by hierarchically classifying them based on the similarity analysis results.

Description

{How to construct an optimal water quality observation network}

The present invention relates to a method of configuring an optimal water quality observation network. More specifically, the water quality trend at each observation point is analyzed based on water quality data, and the similarity between observation points is calculated using this to create an optimal water quality observation network for water quality management. It is about a method to configure .

Since the Rio Conference on the Environment in 1992, the international community has been strengthening environmental conservation obligations through various international norms to achieve environmentally sustainable development, and countries around the world have established policies for integrated management and sustainable development for environmental conservation. Aiming.

In response to the international trend for environmental conservation, Korea is also establishing and promoting an integrated management plan to prevent environmental pollution, and preventing an increase in pollution load due to indiscriminate development of the region by placing the burden of environmental improvement on the polluter. To this end, a total pollution load system is being implemented for the four major rivers across the country. However, due to rapid changes in land use, the load of pollutants emitted at once over a short period of time may greatly exceed the self-cleaning capacity that nature can provide. This phenomenon can occur anywhere in the country, and countermeasures against it are urgently needed.

Unlike rivers that naturally flow by gravity, reservoirs and lakes have the characteristic of accumulating and preserving pollutants within the lake when they are introduced in large quantities or for a long period of time. Therefore, it takes a huge amount of time and resources to restore a reservoir that has once lost its function. It takes. To make up for these problems, efforts have recently been made to improve the management system, such as the development of a comprehensive water quality information system for reservoirs and reclaimed lakes and the expansion of measurement networks for water pollution monitoring. However, measures related to reservoir water quality management have not yet gone beyond the level of establishing countermeasures and improvement plans, and in fact, efforts to approach scientific problem solving, such as the reaction within the lake caused by inflow contaminants, which are more important in reservoirs, are very passive.

Meanwhile, the water quality measurement network comprehensively understands the water quality status of public waters subject to water quality conservation, such as rivers and lakes across the country, identifies water quality change trends, analyzes the effects of major policy projects that have already been implemented, and provides basic data for establishing future water quality conservation policies. It is operated to secure. However, the construction of the water quality measurement network that has been operated so far in Korea has been made based on the experience and subjective judgment of managers as the need for management increases, rather than based on a systematic design method according to scientifically established measurement purposes. Therefore, it can be seen that the currently operating measurement network has a problem in that it does not meet the need for utilization or the production of various water quality information required for water system management.

Moreover, water quality has a variety of investigated items, diverse sources of pollution, and is affected by seasons and changes in flow rate. Therefore, it is necessary to form an optimal water quality observation network by evaluating the similarity of observation points considering water quality characteristics and seasonal trends, determining the number of observation points using this, and selecting representative observation points.

Therefore, the purpose of the present invention to solve the above-mentioned problem is to calculate the similarity between observation points based on time series data measured at each survey point of the water body and hierarchically classify observation points with high similarity to achieve the optimal water quality required by the user. The goal is to provide a method to easily configure an observation network.

According to one aspect of the present invention, a collection step of collecting water quality data of a target water body; A similarity analysis step of performing water quality similarity analysis for each observation point based on application of principal component analysis, correlation coefficient matrix, and multidimensional scaling method to each item of the water quality data; A decision step of hierarchically classifying based on the similarity analysis results to determine the number of observation points and representative observation points; Comprising: performing water quality fluctuation trend analysis including correlation analysis of selected variables, time series graph, and scatter plot for data for each point of the target water body included in the collected water quality data; It further includes.

The similarity analysis step uses the Euclidean distance of the principal component score that abbreviates the water quality of the target water body or the distance of the coordinates obtained by applying the correlation coefficient matrix and multidimensional scaling method to consider time series changes in the water quality pattern of the target water body. A step of calculating similarity, wherein the determining step includes classifying the similarity analysis result using a hierarchical cluster analysis method and determining an appropriate number of clusters according to a dendrogram method; determining a representative observation point by calculating the center point of the determined cluster; It further includes.

When applying the hierarchical cluster analysis method, in order to maximize the distance between clusters after cluster analysis, minimize the intra-cluster distance, single connection using the closest of the object distances of the two clusters, full connection using the farthest distance, and Using at least one of average linking using the distance average, central linking using the distance between the centers of objects in the cluster, and Wald variance minimization linking, which has been proposed as a method of minimizing information sacrifice that occurs during the clustering process; It further includes.

According to the present invention as described above, it is possible to evaluate collected water quality data, analyze water quality trends at each observation point, and calculate similarity between observation points, thereby providing an optimal water quality measurement network configuration for water quality management.

1 is a flowchart illustrating a method of configuring an optimal water quality observation network according to the present invention.

The present invention provides a method for configuring an optimal water quality observation network.

Figure 1 is a diagram for explaining a method of configuring an optimal water quality observation network according to the present invention.

In the method of configuring the optimal water quality observation network of the present invention, first, in the data collection stage, the observation network selection device collects water quality data from each observation point of the water body. Here, the water quality data may be observation point time series water quality data. Since water quality has a variety of investigated items, diverse sources of pollution, and is affected by seasons and flow changes, it is necessary to collect water quality and hydrological data that can apply both water quality characteristics and seasonal trends to calculate accurate observation points. For this purpose, it is desirable that a structure be prepared or investigated at the observation point of the relevant water body to collect water quality data including water quality characteristics and seasonal trends.

Here, the water quality data may include data on various measurement items related to water quality. For example, basic items may include pH, water temperature, dissolved oxygen, turbidity, electrical conductivity, etc. In addition, various types of items such as COD, TN, NO3-N, NH3-N, TP, PO4-P, SS, transparency, chl-a, etc. may be added to the water quality data depending on the purpose of use of the data.

Next, in the similarity calculation step, the observation network selection device calculates the similarity between observation points. To this end, the observation network selection device performs water quality change trend analysis using a time series graph of selected variables for data at each point of the target water body. In addition, the observation network selection device calculates a principal component score by performing principal component analysis based on the analyzed water quality change trend analysis results, and can be performed by calculating a correlation coefficient matrix for each water quality item and applying multidimensional scale analysis.

In particular, the observation network selection device calculates the similarity between observation points using a correlation coefficient matrix and multidimensional scale analysis in a way that can consider time series changes in water quality patterns using observations of water quality survey points, that is, water quality data.

To explain this in more detail, first, the observation network selection device calculates the overall average of the observed water quality data to calculate the main component score and uses this to calculate the main component score for each observation point. The Euclidean distance of the principal component scores of observation points can be the similarity between observation points. Principal component analysis is a multivariate analysis method that reduces the order of the original data by converting a number of linearly correlated raw variables into principal component variables that are independent of each other and are a linear combination of the original variables.

Eigenvalues are calculated from the covariance matrix of the original variables, and the main component variables are obtained using the corresponding eigenvectors as linear coefficients. The number of dimensions to be reduced is determined by selecting the main component whose cumulative contribution rate to explaining the change in the original variable is 80% or more or whose eigenvalue is 1 or more.

On the other hand, if the measurement units of the measured variables are different, the size of the variance also changes. As a result, the variance and covariance for variables with large measurement units increase, so the observation network selection device of the present invention obtains eigenvalues and eigenvectors from a correlation coefficient matrix dividing the variance of each variable in the covariance matrix and uses them to obtain the main component. You can.

On the other hand, if the measurement units of the measured variables are different, the size of the variance also changes. As a result, the variance and covariance for variables with large measurement units increase, so the observation network selection device of the present invention obtains eigenvalues and eigenvectors from a correlation coefficient matrix dividing the variance of each variable in the covariance matrix and uses them to obtain the main component. You can.

A similar observation point grouping process is performed in the grouping step using the calculated similarity. At this time, the observation network selection device measures the similarity between objects based on variables of interest in order to classify objects based on similarity, and uses this to apply clustering analysis to classify similar objects.

The cluster analysis is a hierarchical method that groups objects in order of high similarity according to the structure of the given data, and groups objects based on a certain level of similarity or by the number of clusters determined in advance. There is a non-hierarchical way to classify entities. The hierarchical cluster analysis method is easy to use, but once an object belongs to a cluster, it cannot be moved to another cluster, and abnormal objects are not removed but must belong to a certain cluster. Non-hierarchical cluster analysis guarantees a certain level of clustering, but can only be used if the number of clusters is determined in advance. Therefore, the general order of cluster analysis is to determine the appropriate number of clusters by using the hierarchical cluster analysis method.

In the case of a hierarchical cluster analysis method that groups groups in order of highest similarity, the distance between objects and clusters must be calculated, which is called the linkage method. The above connection methods include single connection using the closest of the distances of the objects in the two clusters, complete connection using the farthest distance, and average (median) using the average distance (median) of the objects between the clusters. ) connection, centroid connection using the distance between the centroids of objects in the cluster, and the Wald variance minimization connection method proposed as a method to minimize the loss of information that occurs during the clustering process (Johnson et al. . al., 2007). The appropriate connection method is determined by the dendrogram obtained as a result of hierarchical cluster analysis. In this observation network selection device, the most frequently used average connection method was applied as default, but this can be upgraded to allow device users to set options to select the connection method in the future.

Claims (3)

A collection step of collecting water quality data of the target water body;
A similarity analysis step of performing water quality similarity analysis for each observation point based on application of principal component analysis, correlation coefficient matrix, and multidimensional scaling method to each item of the water quality data;
A decision step of hierarchically classifying based on the similarity analysis results to determine the number of observation points and representative observation points; includes
Performing water quality change trend analysis including correlation analysis of selected variables, time series graph, and scatter plot for data for each point of the target water body included in the collected water quality data; A method of configuring an optimal water quality observation network, further comprising: According to clause 1,
The similarity analysis step uses the Euclidean distance of the principal component score that abbreviates the water quality of the target water body or the distance of the coordinates obtained by applying the correlation coefficient matrix and multidimensional scaling method to consider time series changes in the water quality pattern of the target water body. This is the step of calculating the similarity of
The determining step includes classifying the similarity analysis result using a hierarchical cluster analysis method and determining an appropriate number of clusters according to a dendrogram method; determining a representative observation point by calculating the center point of the determined cluster; A method of configuring an optimal water quality observation network further comprising: According to clause 2,
When applying the hierarchical cluster analysis method, in order to maximize the distance between clusters after cluster analysis, minimize the intra-cluster distance, single connection using the closest of the object distances of the two clusters, full connection using the farthest distance, and Using at least one of average linking using the distance average, central linking using the distance between the centers of objects in the cluster, and Wald variance minimization linking, which has been proposed as a method of minimizing information sacrifice occurring in the clustering process; A method of configuring an optimal water quality observation network further comprising: