Monitoring data processing method of water conservancy rain measuring radar for rainwater condition monitoring and forecasting

Abstract

The invention relates to the field of data processing, in particular to a monitoring data processing method of a water conservancy rain measuring radar for rainwater condition monitoring and forecasting, and the method comprises the steps: collecting the monitoring data of the water conservancy rain measuring radar; performing terrain complexity assessment according to the feature data of all the radar body scanning units, acquiring a first optimization factor of distance measurement according to the terrain complexity of the monitoring range of the water conservancy rain measurement radar, and performing first Euclidean distance assessment between the radar body scanning units according to the first optimization factor; evaluating a second optimization factor according to the mahalanobis distance between the radar body scanning units, and measuring and evaluating the optimization distance between the radar body scanning units according to the second optimization factor, the first Euclidean distance and the mahalanobis distance; clustering is carried out according to the optimized distance measurement; and performing local self-adaptive Z-R relation fitting evaluation according to a clustering division result. Therefore, the problem that the rainfall estimation precision is low due to the fact that existing distance measurement cannot accurately describe the spatial difference of mountainous radar data is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a method for processing monitoring data of a hydraulic rain gauge radar for rain and water regime monitoring and forecasting. Background Art

[0002] Rainfall inversion based on a hydraulic rain gauge radar usually relies on the empirical relationship between the radar reflectivity factor (Z) and the rainfall intensity (R). Due to the complexity and spatio-temporal variability of the rainfall microphysical process, the Z-R relationship is not fixed, but varies significantly in different regions and different rainfall types. Especially under the complex terrain conditions in mountainous areas, the spatial distribution of rainfall is strongly affected by the terrain, showing stronger non-uniformity. The traditional rainfall inversion method based on a fixed Z-R relationship is difficult to accurately capture this non-uniformity, resulting in a decrease in rainfall estimation accuracy. To solve this problem, the local adaptive technology of the Z-R relationship has been developed in recent years. This technology divides the radar observation area into partitions, and determines the Z-R relationship within each partition respectively, thereby improving the accuracy of rainfall inversion. Among them, the clustering algorithm is the key technology to achieve the local adaptability of the Z-R relationship. Through the clustering algorithm, regions with similar rainfall characteristics can be divided into one category, and then the Z-R relationship can be established within each category respectively. However, due to the complex terrain in mountainous areas, different terrain features have different degrees of influence on rainfall, and there is also a correlation between different statistical characteristics of the radar reflectivity data. In the existing clustering algorithms, it is difficult to accurately reflect the true spatial differences of rainfall data under the complex terrain in mountainous areas through the traditional Euclidean distance metric, resulting in poor clustering results. Therefore, how to obtain an accurate division of the radar monitoring area in the local adaptability of the Z-R relationship based on clustering has become an urgent problem to be solved. Summary of the Invention

[0003] In view of this, an embodiment of the present invention provides a method for processing monitoring data of a hydraulic rain gauge radar for rain and water regime monitoring and forecasting to solve the problem of obtaining an accurate division of the radar monitoring area in the local adaptability of the Z-R relationship based on clustering.

[0004] An embodiment of the present invention provides a method for processing monitoring data of a hydraulic rain gauge radar for rain and water regime monitoring and forecasting. The method for processing monitoring data of a hydraulic rain gauge radar for rain and water regime monitoring and forecasting includes the following steps:

[0005] Collect the monitoring data of the hydrological rainfall radar, including radar reflectivity factor, rainfall intensity data observed by rain gauges, digital elevation model data, and radar volume scan unit feature data; evaluate the terrain complexity according to all the radar volume scan unit feature data to obtain the terrain complexity of the monitoring range of the hydrological rainfall radar, obtain the first optimization factor for the distance measurement between the radar volume scan units according to the terrain complexity of the monitoring range of the hydrological rainfall radar, evaluate the first Euclidean distance between the radar volume scan units according to the first optimization factor for the distance measurement between the radar volume scan units, and obtain the first Euclidean distance between the radar volume scan units; obtain the Mahalanobis distance measurement between the radar volume scan units, evaluate the second optimization factor for the distance measurement between the radar volume scan units according to the Mahalanobis distance between the radar volume scan units, and obtain the second optimization factor for the distance measurement between the radar volume scan units; evaluate the optimized distance measurement between the radar volume scan units according to the second optimization factor for the distance measurement between the radar volume scan units, the first Euclidean distance between the radar volume scan units, and the Mahalanobis distance between the radar volume scan units, and obtain the optimized distance measurement between the radar volume scan units; perform clustering according to the optimized distance measurement between the radar volume scan units to obtain the clustering division result of all the volume scan units of the hydrological rainfall radar; perform a locally adaptive Z-R relationship fitting evaluation according to the clustering division result of all the volume scan units of the hydrological rainfall radar.

[0006] Preferably, the collection of the monitoring data of the hydrological rainfall radar, including radar reflectivity factor, rainfall intensity data observed by rain gauges, digital elevation model data, and radar volume scan unit feature data, includes:

[0007] The digital elevation model data is represented as a two-dimensional matrix, where the two dimensions respectively represent the row index and column index of the grid cell, the matrix value corresponding to the row index and column index is the elevation value of the grid cell, and the elevation value of the grid cell can be used to extract the feature data of each radar volume scan unit. The feature data of the radar volume scan unit is represented as a three-dimensional feature vector, where the three dimensions are the mean value of radar reflectivity data, the variance of radar reflectivity data, and the texture of radar reflectivity data.

[0008] Preferably, the evaluation of the terrain complexity according to all the radar volume scan unit feature data to obtain the terrain complexity of the monitoring range of the hydrological rainfall radar includes:

[0009] Obtain the standard deviation and mean of the slope values of all the radar volume scan units, obtain the aspect values of all the radar volume scan units, convert the aspect values into complex number forms, and take the modulus calculation result of the complex standard deviation of the complex number forms of the aspect values as the standard deviation of the aspect values of all the radar volume scan units. Similarly, obtain the aspect values of all the radar volume scan units, convert the aspect values into complex number forms, and take the modulus calculation result of the mean of the complex number forms of the aspect values as the mean of the aspect values of all the radar volume scan units. Obtain the standard deviation and mean of the elevation values of all the radar volume scan units;

[0010] Take the calculation result of dividing the standard deviation of the slopes of all the radar volume scan units by the mean of the slopes of all the radar volume scan units as the slope variation coefficient; take the calculation result of dividing the standard deviation of the aspect values of all the radar volume scan units by the mean of the aspect values of all the radar volume scan units as the aspect variation coefficient; take the calculation result of dividing the standard deviation of the elevation values of all the radar volume scan units by the mean of the elevation values of all the radar volume scan units as the elevation variation coefficient; take the calculation result of adding the slope variation coefficient, the aspect variation coefficient and the elevation variation coefficient and dividing by three as the terrain complexity of the monitoring range of the hydrological rainfall radar.

[0011] Preferably, obtaining the first optimization factor for the distance metric between the radar volume scan units according to the terrain complexity of the monitoring range of the hydrological rainfall radar includes:

[0012] Obtain the terrain complexity of the monitoring range of the hydrological rainfall radar, and obtain the slope values, aspect values and elevation values of any two volume scan units among all the volume scan units of the hydrological rainfall radar; take the calculation result of dividing the absolute value of the difference in slope values between the any two volume scan units by the maximum slope value as the slope optimization evaluation value, take the calculation result of dividing the absolute value of the difference in aspect values between the any two volume scan units by the maximum aspect value as the aspect optimization evaluation value, and take the calculation result of dividing the absolute value of the difference in elevation values between the any two volume scan units by the maximum elevation value as the elevation optimization evaluation value; take the calculation result of summing the slope optimization evaluation value, the aspect optimization evaluation value and the elevation optimization evaluation value and dividing by the sum of a constant one and the terrain complexity of the monitoring range of the hydrological rainfall radar and multiplying by negative one as the first evaluation factor between any two volume scan units among all the volume scan units of the hydrological rainfall radar, and take the result of mapping the first evaluation factor between any two volume scan units among all the volume scan units of the hydrological rainfall radar through an exponential function with the natural constant as the base as the first optimization factor for the distance metric between the radar volume scan units.

[0013] Preferably, evaluating the first Euclidean distance between the radar volume scan units according to the first optimization factor for the distance metric between the radar volume scan units to obtain the first Euclidean distance between the radar volume scan units includes:

[0014] Obtain a first optimization factor for the distance metric between the radar volume scan units, and use the calculation result of multiplying the first optimization factor for the distance metric between the radar volume scan units by the Euclidean distance between the feature vectors between the radar volume scan units as the first Euclidean distance between the radar volume scan units.

[0015] Preferably, obtain the Mahalanobis distance metric between the radar volume scan units, evaluate a second optimization factor for the distance metric between the radar volume scan units according to the Mahalanobis distance between the radar volume scan units, and obtain the second optimization factor for the distance metric between the radar volume scan units, including:

[0016] Obtain the Mahalanobis distance between the radar volume scan unit and the cluster center volume scan unit in the clustering process, and use the calculation result of dividing the reciprocal of the Mahalanobis distance between the radar volume scan unit and the cluster center volume scan unit of the cluster during the clustering iteration process by the sum of the reciprocals of the Mahalanobis distances between the radar volume scan unit and all cluster center volume scan units during the clustering iteration process as the second optimization factor for the distance metric between the radar volume scan units.

[0017] Preferably, evaluate the optimized distance metric between the radar volume scan units according to the second optimization factor for the distance metric between the radar volume scan units, the first Euclidean distance between the radar volume scan units, and the Mahalanobis distance between the radar volume scan units, and obtain the optimized distance metric between the radar volume scan units, including:

[0018] Obtain the second optimization factor for the distance metric between the radar volume scan units, obtain the first Euclidean distance between the radar volume scan units, and obtain the Mahalanobis distance between the radar volume scan unit and the cluster center volume scan unit in the clustering process; use the calculation result of multiplying the second optimization factor by the Mahalanobis distance between the radar volume scan unit and the cluster center volume scan unit in the clustering process as the first optimized distance metric; use the calculation result of multiplying the subtraction of one from the second optimization factor by the first Euclidean distance as the second optimized distance metric; use the calculation result of adding the first optimized distance metric and the second optimized distance metric as the optimized distance metric between the radar volume scan units.

[0019] Preferably, perform clustering according to the optimized distance metric between the radar volume scan units, and obtain the clustering division result of all volume scan units of the hydrological rainfall radar, including:

[0020] Obtain the number of target cluster classes in the clustering process. During the iterative process in the clustering process, first perform the initial cluster class division in the iterative process according to the traditional Euclidean distance metric in the original clustering process. After obtaining the initial cluster class division result in the iterative process, obtain the covariance matrix of each cluster class in the initial cluster class division result according to the initial cluster class division result in the iterative process. Obtain the optimized distance metric between each radar volume scan unit and all cluster center radar volume scan units in the iterative process of the clustering process according to the covariance matrix of each cluster class in the initial cluster class division result for secondary division, obtain the secondary division result, and update the cluster center in the clustering process and the clustering label of the radar volume scan unit according to the secondary division result to complete the iterative process in the clustering process; repeat the iterative process in the clustering process until the clustering result of the clustering process no longer changes, and obtain the clustering division result of all volume scan units of the hydrological rainfall radar corresponding to the clustering process.

[0021] Preferably, the local adaptive Z-R relationship fitting evaluation according to the clustering division result of all volume scan units of the hydrological rainfall radar includes:

[0022] Obtain the clustering division result of all volume scan units of the hydrological rainfall radar, determine the Z-R relationship coefficient of the cluster class by the least square method according to the radar reflectivity factor data of the hydrological rainfall radar volume scan units in each cluster class in the clustering division result, and obtain the local adaptive Z-R relationship fitting evaluation result.

[0023] The beneficial effects of the embodiments of the present invention compared with the prior art are:

[0024] The present invention collects monitoring data of a hydraulic rainfall radar, including radar reflectivity factor, rainfall intensity data observed by a rain gauge, digital elevation model data, and radar volume scan unit feature data; evaluates the terrain complexity according to all the radar volume scan unit feature data to obtain the terrain complexity of the monitoring range of the hydraulic rainfall radar, obtains a first optimization factor for distance measurement between the radar volume scan units according to the terrain complexity of the monitoring range of the hydraulic rainfall radar, evaluates the first Euclidean distance between the radar volume scan units according to the first optimization factor for distance measurement between the radar volume scan units, and obtains the first Euclidean distance between the radar volume scan units; obtains the Mahalanobis distance measurement between the radar volume scan units, evaluates a second optimization factor for distance measurement between the radar volume scan units according to the Mahalanobis distance between the radar volume scan units, and obtains the second optimization factor for distance measurement between the radar volume scan units; evaluates the optimized distance measurement between the radar volume scan units according to the second optimization factor for distance measurement between the radar volume scan units, the first Euclidean distance between the radar volume scan units, and the Mahalanobis distance between the radar volume scan units, and obtains the optimized distance measurement between the radar volume scan units; performs clustering according to the optimized distance measurement between the radar volume scan units to obtain the clustering division result of all the volume scan units of the hydraulic rainfall radar; and performs a local adaptive Z-R relationship fitting evaluation according to the clustering division result of all the volume scan units of the hydraulic rainfall radar. Among them, the Euclidean distance between the radar volume scan units is optimized based on the terrain complexity, so that the clustering process can fully consider the influence of terrain factors on rainfall distribution, and the accuracy and adaptability of the clustering result are improved; on this basis, further through the second optimization factor, the optimized Euclidean distance and the Mahalanobis distance are comprehensively evaluated for distance, which not only more comprehensively measures the difference between the radar volume scan units, but also effectively reduces the clustering deviation caused by terrain differences, thereby providing more reliable data support for the subsequent local adaptive Z-R relationship fitting, and making the rainfall and water regime monitoring and forecasting more accurate and practical. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0026] Figure 1 It is a flowchart of a method for processing monitoring data of a hydraulic rainfall radar for rainfall and water regime monitoring and forecasting according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] Embodiments of the present disclosure will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where like or similar reference numerals denote like or similar elements or elements having like or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present disclosure, and should not be construed as a limitation of the present disclosure.

[0028] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned accompanying drawings are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this toothpaste can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0029] In order to illustrate the technical solution of the present invention, it will be described below through specific embodiments.

[0030] The specific scenario targeted by the present invention is: local adaptive evaluation of the Z-R relationship through the monitoring data of the volume scanning unit of a hydrological rainfall radar.

[0031] See Figure 1 , which is a flowchart of a method for processing monitoring data of a hydrological rainfall radar for rainfall and water regime monitoring and forecasting provided in Embodiment 1 of the present invention. As Figure 1 shown, a method for processing monitoring data of a hydrological rainfall radar for rainfall and water regime monitoring and forecasting may include:

[0032] Step S101, collect the monitoring data of the hydrological rainfall radar, evaluate the terrain complexity according to all the radar volume scanning unit characteristic data, and obtain the terrain complexity of the monitoring range of the hydrological rainfall radar.

[0033] Collect the monitoring data of the hydrological rainfall radar, including radar reflectivity factor, rainfall intensity data observed by rain gauges, digital elevation model data, and radar volume scanning unit characteristic data. The digital elevation model data is represented as a two-dimensional matrix, where the two dimensions respectively represent the row index and column index of the grid unit, and the matrix value corresponding to the row index and column index is the elevation value of the grid unit. The elevation value of the grid unit can be used to extract each radar volume scanning unit characteristic data, and the radar volume scanning unit characteristic data is represented as a three-dimensional feature vector, where the three dimensions are the mean value of radar reflectivity data, the variance of radar reflectivity data, and the texture of radar reflectivity data.

[0034] After obtaining the monitoring data of the hydrological rainfall radar, the terrain complexity can be evaluated according to all the radar volume scan unit characteristic data, and the terrain complexity of the monitoring range of the hydrological rainfall radar can be obtained. This process includes: obtaining the standard deviation and mean of the slope values of all the radar volume scan units, obtaining the aspect values of all the radar volume scan units, converting the aspect values into complex number forms, and taking the modulus calculation result of the complex standard deviation of the complex number forms of the aspect values as the standard deviation of the aspect values of all the radar volume scan units. Similarly, obtaining the aspect values of all the radar volume scan units, converting the aspect values into complex number forms, and taking the modulus calculation result of the mean of the complex number forms of the aspect values as the mean of the aspect values of all the radar volume scan units, and obtaining the standard deviation and mean of the elevation values of all the radar volume scan units; taking the calculation result of dividing the standard deviation of the slopes of all the radar volume scan units by the mean of the slopes of all the radar volume scan units as the slope variation coefficient; taking the calculation result of dividing the standard deviation of the aspect values of all the radar volume scan units by the mean of the aspect values of all the radar volume scan units as the aspect variation coefficient; taking the calculation result of dividing the standard deviation of the elevation values of all the radar volume scan units by the mean of the elevation values of all the radar volume scan units as the elevation variation coefficient; taking the calculation result of adding the slope variation coefficient, the aspect variation coefficient and the elevation variation coefficient and dividing by three as the terrain complexity of the monitoring range of the hydrological rainfall radar.

[0035] In one embodiment, assume that the mean of the slope values of all the volume scan units of the hydrological rainfall radar is Mean slope , and the standard deviation of the slope values of all the volume scan units of the hydrological rainfall radar is Std slope ; the mean of the aspect values of all the volume scan units of the hydrological rainfall radar is Mean aspect , and the standard deviation of the aspect values of all the volume scan units of the hydrological rainfall radar is Std aspect ; the mean of the elevation values of all the volume scan units of the hydrological rainfall radar is Mean elevation , and the standard deviation of the elevation values of all the volume scan units of the hydrological rainfall radar is Std elevation , then the calculation expression of the terrain complexity of the monitoring range of the hydrological rainfall radar is:

[0036]

[0037] where CV terrain represents the terrain complexity of the monitoring range of the hydrological rainfall radar; represents the slope variation coefficient of the monitoring range of the hydrological rainfall radar; represents the aspect variation coefficient of the monitoring range of the hydrological rainfall radar; represents the elevation variation coefficient of the monitoring range of the hydrological rainfall radar.

[0038] Step S102: Obtain a first optimization factor for distance measurement between the radar volume scan units according to the terrain complexity of the monitoring range of the hydrological rainfall radar, evaluate the first Euclidean distance between the radar volume scan units according to the first optimization factor for distance measurement between the radar volume scan units, and obtain the first Euclidean distance between the radar volume scan units.

[0039] After obtaining the terrain complexity of the monitoring range of the hydrological rainfall radar, the first optimization factor for distance measurement between the radar volume scan units can be evaluated according to the terrain complexity of the monitoring range of the hydrological rainfall radar. Obtain the terrain complexity of the monitoring range of the hydrological rainfall radar, and obtain the slope value, aspect value and elevation value of any two volume scan units among all the volume scan units of the hydrological rainfall radar; Divide the absolute value of the difference in slope values between any two volume scan units by the maximum value of the slope value, and use the calculation result as the slope optimization evaluation value. Divide the absolute value of the difference in aspect values between any two volume scan units by the maximum value of the aspect value, and use the calculation result as the aspect optimization evaluation value. Divide the absolute value of the difference in elevation values between any two volume scan units by the maximum value of the elevation value, and use the calculation result as the elevation optimization evaluation value; Divide the sum of the slope optimization evaluation value, aspect optimization evaluation value and elevation optimization evaluation value by the sum of a constant one and the terrain complexity of the monitoring range of the hydrological rainfall radar, and multiply the result by negative one as the first evaluation factor between any two volume scan units among all the volume scan units of the hydrological rainfall radar. Use the result of mapping the first evaluation factor between any two volume scan units among all the volume scan units of the hydrological rainfall radar with an exponential function with the natural constant as the base as the first optimization factor for distance measurement between the radar volume scan units.

[0040] In one embodiment, the calculation expression of the first optimization factor is:

[0041]

[0042] where α ij represents the first optimization factor between the i-th radar volume scan unit and the j-th radar volume scan unit; slope i , slope j respectively represent the slope values of the i-th radar volume scan unit and the j-th radar volume scan unit; aspect i , aspect j respectively represent the aspect values of the i-th radar volume scan unit and the j-th radar volume scan unit; elevation i , elevation j respectively represent the elevation values of the i-th radar volume scan unit and the j-th radar volume scan unit; slope max , aspect max , elevationmax respectively represent the maximum slope value, the maximum aspect value, and the maximum elevation value among all radar volume scan units; CV (errai+ represents the terrain complexity of the monitoring range of the hydrological rainfall radar; 1 represents the constant 1; || represents absolute value calculation; e represents the natural constant e.

[0043] It should be noted that in the specific calculation process, considering the periodicity of the aspect, so define |aspect i -aspect j | = min(|aspect i -aspect j |, 360 - |aspect i -aspect j |); The first optimization factor α ij comprehensively considers three terrain features: slope, aspect, and elevation. The form of the exponential function is used to map the terrain feature differences to the interval [0, 1], and it is monotonically decreasing. The greater the terrain difference, the smaller α ij is, which conforms to the definition of terrain similarity. The difference term: partially calculates the differences between two volume scan units in three terrain features. Dividing by the maximum values slope max , aspect max , elevation max is to normalize the terrain feature differences with different dimensions to the range [0, 1], so that the differences of different features are comparable. Dividing by (1 + CV terrain ) ensures that the terrain difference term is adaptively adjusted according to the terrain complexity of the study area. The role of CV terrain is: the larger CV terrain , the more complex the terrain. The spatial distribution of rainfall in complex terrain areas may be affected by more other factors, and the influence of terrain is relatively weakened. Therefore, dividing by (1 + CV terrain ) can adaptively adjust the terrain difference term. When CV terrain is larger (terrain is complex), the terrain difference term is reduced, and the weight of the terrain factor is decreased; when CV terrain is smaller (terrain is flat), the influence on the terrain difference term is smaller, and the weight of the terrain factor is relatively higher.

[0044] After obtaining the first optimization factor of the distance metric between the radar volume scan units, the first Euclidean distance can be evaluated according to the first optimization factor of the distance metric between the radar volume scan units. Multiply the first optimization factor of the distance metric between the radar volume scan units by the calculation result of the Euclidean distance between the feature vectors of the radar volume scan units as the first Euclidean distance between the radar volume scan units.

[0045] In one embodiment, the eigenvalue of the p-th dimension of the i-th radar volume scanning unit is x i,p , and the eigenvalue of the p-th dimension of the j-th radar volume scanning unit is x j,p . Then, the calculation expression of the first Euclidean distance between the two radar volume scanning units is:

[0046]

[0047] where D α (i,j) represents the first Euclidean distance between the i-th radar volume scanning unit and the j-th radar volume scanning unit; a ij represents the first optimization factor between the i-th radar volume scanning unit and the j-th radar volume scanning unit; P represents the number of dimensions; x i,p represents the eigenvalue of the p-th dimension of the i-th radar volume scanning unit; x j,p represents the eigenvalue of the p-th dimension of the j-th radar volume scanning unit.

[0048] Step S103: Obtain the Mahalanobis distance metric between the radar volume scanning units, evaluate the second optimization factor for the distance metric between the radar volume scanning units according to the Mahalanobis distance between the radar volume scanning units, and obtain the second optimization factor for the distance metric between the radar volume scanning units.

[0049] It can be understood that when obtaining the first Euclidean distance between radar volume scan units, using the first Euclidean distance for distance measurement in the clustering process can better consider the influence of terrain factors. However, traditional clustering algorithms based on Euclidean distance still assume that different statistical features of radar reflectivity data are independent of each other and have the same variance. In practical applications, especially under complex terrain conditions in mountainous areas, there may be correlations between different statistical features of radar reflectivity data. For example, areas with heavy precipitation usually correspond to larger mean values, variances, and texture features of radar reflectivity. In addition, the variances of different statistical features may also be different. For example, the range of variation of the mean value of radar reflectivity may be larger than that of the variance. Therefore, after obtaining the first Euclidean distance between radar volume scan units, further, the Mahalanobis distance between radar volume scan units is used for distance measurement, so as to consider the covariance relationship between each dimension in the multi-dimensional feature data of radar volume scan units, and thus more accurately reflect the true differences between data points. However, the standard Mahalanobis distance evaluates the distance through the global covariance matrix. In mountain rainfall monitoring, due to the strong spatial heterogeneity of rainfall, the global covariance matrix may not accurately reflect the local covariance relationship of each data point. For example, in areas with high rainfall intensity, the correlation between various statistical features of radar reflectivity data may be stronger, while in areas with low rainfall intensity, the correlation may be weaker. If the global covariance matrix is used to calculate the Mahalanobis distance between the multi-dimensional feature data of all radar volume scan units, errors may be introduced, resulting in inaccurate clustering results.

[0050] Therefore, in the clustering process, first, the first Euclidean distance is used for preliminary clustering. In the clustering process, the local covariance of each cluster is obtained through the cluster division results in each iteration, and the Mahalanobis distance is evaluated based on the local covariance. Then, the second optimization factor is evaluated according to the membership degree of each multi-dimensional feature data point corresponding to each radar volume scan unit to each cluster and the reliability of the covariance matrix of each cluster, so as to combine the first Euclidean distance and the Mahalanobis distance to accurately measure the distance between the volume scan units of the hydrological rainfall radar.

[0051] Specifically, obtain the Mahalanobis distance metric between the radar volume scan units, and evaluate the second optimization factor for the distance metric between the radar volume scan units according to the Mahalanobis distance between the radar volume scan units, so as to obtain the second optimization factor for the distance metric between the radar volume scan units, including: obtaining the Mahalanobis distance between the radar volume scan unit and the cluster center volume scan unit in the clustering process, and taking the calculation result of dividing the reciprocal of the Mahalanobis distance between the radar volume scan unit and the cluster center volume scan unit of the cluster during the clustering iteration process by the sum of the reciprocals of the Mahalanobis distances between the radar volume scan unit and all cluster center volume scan units in the clustering iteration process as the second optimization factor for the distance metric between the radar volume scan units.

[0052] In one embodiment, assume that the cluster center point of the k-th cluster of the multi-dimensional monitoring data points of the i-th radar volume scan unit during the clustering iteration process is C k , and the local covariance matrix of the k-th cluster is S k , then the calculation expression of the second optimization factor is:

[0053]

[0054] where, β i represents the second optimization factor for the distance metric between the radar volume scan units; X i represents the feature vector of the i-th radar volume scan unit (including statistical features such as the mean, variance, and texture of the radar reflectivity data); C k represents the feature vector corresponding to the cluster center volume scan unit of the k-th cluster; S k represents the feature vector corresponding to the cluster center volume scan unit of the k-th cluster; K represents the number of target clusters in the clustering process.

[0055] Step S104, evaluate the optimized distance metric between the radar volume scan units according to the second optimization factor for the distance metric between the radar volume scan units, the first Euclidean distance between the radar volume scan units and the Mahalanobis distance between the radar volume scan units, so as to obtain the optimized distance metric between the radar volume scan units.

[0056] After obtaining the second optimization factor, according to the second optimization factor for distance measurement between the radar volume scanning units, the first Euclidean distance between the radar volume scanning units and the Mahalanobis distance between the radar volume scanning units are used to evaluate the optimized distance measurement between the radar volume scanning units, and the optimized distance measurement between the radar volume scanning units is obtained, including: obtaining the second optimization factor for distance measurement between the radar volume scanning units, obtaining the first Euclidean distance between the radar volume scanning units, and obtaining the Mahalanobis distance between the radar volume scanning unit and the cluster center volume scanning unit during the clustering process; multiplying the second optimization factor by the calculation result of the Mahalanobis distance between the radar volume scanning unit and the cluster center volume scanning unit during the clustering process as the first optimized distance measurement; subtracting one from the second optimization factor and multiplying the calculation result by the first Euclidean distance as the second optimized distance measurement; adding the calculation results of the first optimized distance measurement and the second optimized distance measurement as the optimized distance measurement between the radar volume scanning units.

[0057] In one embodiment, it is assumed that the Mahalanobis distance between the i-th radar volume scanning unit and the k-th cluster center volume scanning unit during the iteration of the clustering process is D m (i,C k ), then the calculation expression for the optimized distance measurement between the radar volume scanning units is:

[0058] D(i,k) = β i ·D m (i,C k )+(1 - β i )·D α (i,C k )

[0059] Wherein, D(i,k) represents the optimized distance measurement between the radar volume scanning units; β i represents the second optimization factor of the i-th radar volume scanning unit; D m (i,C k ) represents the Mahalanobis distance between the i-th radar volume scanning unit and the k-th cluster center volume scanning unit during the iteration of the clustering process; D α (i,C k ) represents the first Euclidean distance between the i-th radar volume scanning unit and the k-th cluster center volume scanning unit.

[0060] It should be noted that the calculation formula of the second optimization factor β i can be understood as the ratio of the reciprocal of the Mahalanobis distance between the multi-dimensional monitoring data point i of the radar volume scanning unit and its affiliated cluster center to the sum of the reciprocals of the Mahalanobis distances between this point and all cluster centers. The reciprocal of the Mahalanobis distance indicates that the greater the Mahalanobis distance, the lower the similarity between the data point and the cluster center. The larger the value after taking the reciprocal, the higher the similarity. β iThe larger it is, the smaller the Mahalanobis distance between the data point i and its affiliated cluster center (the higher the similarity), and the more reliable the local covariance matrix of this cluster is for this point. Specifically, when β i is relatively large, the weight of the Mahalanobis distance is large, and the final distance metric is closer to the Mahalanobis distance. When β i is relatively small (the data point i is far from its affiliated cluster center or near the cluster boundary, and the local covariance matrix is unreliable), the weight of the first Euclidean distance (D α ) is large, and the final distance metric is closer to the first Euclidean distance.

[0061] Step S105: Cluster according to the optimized distance metric between the radar volume scan units, and obtain the clustering division result of all the volume scan units of the hydrological rainfall radar; perform a locally adaptive Z-R relationship fitting evaluation according to the clustering division result of all the volume scan units of the hydrological rainfall radar.

[0062] After obtaining the optimized distance metric between the radar volume scan units, clustering can be performed according to the optimized distance metric between the radar volume scan units to obtain the clustering division result of all the volume scan units of the hydrological rainfall radar, including: obtaining the target number of cluster classes in the clustering process. During the iterative process in the clustering process, first perform an initial cluster class division in the iterative process according to the traditional Euclidean distance metric in the original clustering process. After obtaining the initial cluster class division result in the iterative process, obtain the covariance matrix of each cluster class in the initial cluster class division result according to the initial cluster class division result. Obtain the optimized distance metric between each radar volume scan unit and all the cluster center radar volume scan units in the iterative process of the clustering process according to the covariance matrix of each cluster class in the initial cluster class division result for secondary division, obtain the secondary division result, and update the cluster class center and the clustering label of the radar volume scan unit in the iterative process of the clustering process according to the secondary division result to complete the iterative process in the clustering process; repeat the iterative process in the clustering process until the clustering result of the clustering process no longer changes, and obtain the clustering division result of all the volume scan units of the hydrological rainfall radar corresponding to the clustering process.

[0063] After obtaining the clustering division result of all the volume scan units of the hydrological rainfall radar, a locally adaptive Z-R relationship fitting evaluation can be performed according to the clustering division result of all the volume scan units of the hydrological rainfall radar. The process includes: obtaining the clustering division result of all the volume scan units of the hydrological rainfall radar, determining the Z-R relationship coefficient of each cluster class according to the radar reflectivity factor data of the hydrological rainfall radar volume scan units in each cluster class in the clustering division result by the least squares method, and obtaining the locally adaptive Z-R relationship fitting evaluation result.

[0064] It should be noted that determining the Z-R relationship coefficient by the least squares method is a prior art, so it will not be elaborated here.

[0065] In summary, the embodiments of the present invention collect the monitoring data of a hydrological rainfall radar, including radar reflectivity factor, rainfall intensity data observed by a rain gauge, digital elevation model data, and radar volume scan unit feature data; evaluate the terrain complexity according to all the radar volume scan unit feature data to obtain the terrain complexity of the monitoring range of the hydrological rainfall radar, obtain a first optimization factor for the distance metric between the radar volume scan units according to the terrain complexity of the monitoring range of the hydrological rainfall radar, evaluate the first Euclidean distance between the radar volume scan units according to the first optimization factor for the distance metric between the radar volume scan units to obtain the first Euclidean distance between the radar volume scan units; obtain the Mahalanobis distance metric between the radar volume scan units, evaluate a second optimization factor for the distance metric between the radar volume scan units according to the Mahalanobis distance between the radar volume scan units to obtain the second optimization factor for the distance metric between the radar volume scan units; evaluate the optimized distance metric between the radar volume scan units according to the second optimization factor for the distance metric between the radar volume scan units, the first Euclidean distance between the radar volume scan units, and the Mahalanobis distance between the radar volume scan units to obtain the optimized distance metric between the radar volume scan units; perform clustering according to the optimized distance metric between the radar volume scan units to obtain the clustering division result of all the volume scan units of the hydrological rainfall radar; perform a locally adaptive Z-R relationship fitting evaluation according to the clustering division result of all the volume scan units of the hydrological rainfall radar. Among them, the Euclidean distance between the radar volume scan units is optimized based on the terrain complexity, so that the clustering process can fully consider the influence of terrain factors on rainfall distribution, improving the accuracy and adaptability of the clustering result; on this basis, further through the second optimization factor, the optimized Euclidean distance and the Mahalanobis distance are comprehensively evaluated, which not only more comprehensively measures the differences between the radar volume scan units, but also effectively reduces the clustering deviation caused by terrain differences, thus providing more reliable data support for the subsequent locally adaptive Z-R relationship fitting, making the rainfall and water regime monitoring and forecasting more accurate and practically valuable.

[0066] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A monitoring data processing method for rain monitoring and forecasting using a hydraulic rain radar, characterized in that: include: Collect monitoring data of water conservancy rainfall radar, including radar reflectivity factor, rainfall intensity data observed by rain gauge, digital elevation model data and radar volume scanning unit characteristic data; Evaluate the terrain complexity according to the characteristic data of all the radar volume scanning units, obtain the terrain complexity of the monitoring range of the water conservancy rain measuring radar, obtain the first optimization factor of the distance measurement between the radar volume scanning units according to the terrain complexity of the monitoring range of the water conservancy rain measuring radar, evaluate the first Euclidean distance between the radar volume scanning units according to the first optimization factor of the distance measurement between the radar volume scanning units, and obtain the first Euclidean distance between the radar volume scanning units; obtain the Mahalanobis distance measurement between the radar volume scanning units, evaluate the second optimization factor of the distance measurement between the radar volume scanning units according to the Mahalanobis distance between the radar volume scanning units, and obtain the second optimization factor of the distance measurement between the radar volume scanning units; According to the second optimization factor of the distance metric between the radar volume scanning units, the first Euclidean distance between the radar volume scanning units and the Mahalanobis distance between the radar volume scanning units, the optimized distance metric between the radar volume scanning units is evaluated to obtain the optimized distance metric between the radar volume scanning units; clustering is performed according to the optimized distance metric between the radar volume scanning units to obtain the clustering division results of all the volume scanning units of the water conservancy rainfall measuring radar; and local adaptive ZR relationship fitting evaluation is performed according to the clustering division results of all the volume scanning units of the water conservancy rainfall measuring radar.

2. The monitoring data processing method of a rainwater radar for rainwater monitoring and forecasting according to claim 1 is characterized in that: The monitoring data collected by the water conservancy rainfall radar includes radar reflectivity factor, rainfall intensity data observed by rain gauge, digital elevation model data and radar volume scanning unit characteristic data, including: The digital elevation model data is represented as a two-dimensional matrix, in which the two dimensions respectively represent the row index and column index of the grid cell, and the matrix values ​​corresponding to the row index and the column index are the elevation values ​​of the grid cell. The elevation values ​​of the grid cell can be used to extract the feature data of each radar volume scanning unit. The radar volume scanning unit feature data is expressed as a three-dimensional feature vector, in which the three dimensions are respectively the mean of radar reflectivity data, the variance of radar reflectivity data and the texture of radar reflectivity data.

3. The monitoring data processing method of a rainwater radar for rainwater monitoring and forecasting according to claim 1 is characterized in that: The evaluating the terrain complexity according to the characteristic data of all the radar volume scanning units to obtain the terrain complexity of the monitoring range of the water conservancy rainfall measuring radar includes: Obtain the standard deviation and mean of the slope values ​​of all the radar volume scan units, obtain the aspect values ​​of all the radar volume scan units, convert the aspect values ​​into complex form, and use the modulus calculation result of the complex standard deviation of the complex form of the aspect values ​​as the standard deviation of the aspect values ​​of all the radar volume scan units. Similarly, obtain the aspect values ​​of all the radar volume scan units, convert the aspect values ​​into complex form, and use the modulus calculation result of the complex form mean of the aspect values ​​as the mean of the aspect values ​​of all the radar volume scan units, and obtain the standard deviation and mean of the elevation values ​​of all the radar volume scan units; The calculation result of dividing the standard deviation of the slopes of all the radar volume scan units by the mean value of the slopes of all the radar volume scan units is used as the slope variation coefficient; the calculation result of dividing the standard deviation of the slope aspect values ​​of all the radar volume scan units by the mean value of the slope aspect values ​​of all the radar volume scan units is used as the slope variation coefficient; the calculation result of dividing the standard deviation of the elevation values ​​of all the radar volume scan units by the mean value of the elevation values ​​of all the radar volume scan units is used as the elevation variation coefficient; the calculation result of adding the slope variation coefficient, the slope aspect variation coefficient and the elevation variation coefficient and dividing the sum by three is used as the terrain complexity of the monitoring range of the water conservancy rain measuring radar.

4. The monitoring data processing method for using a hydraulic rain measuring radar for rain monitoring and forecasting according to claim 1 is characterized in that: The method of obtaining the first optimization factor of the distance measurement between the radar volume scanning units according to the terrain complexity of the monitoring range of the water conservancy rainfall measuring radar includes: Obtain the terrain complexity of the monitoring range of the water conservancy rain measuring radar, obtain the slope value, aspect value and elevation value of any two body scanning units in all the body scanning units of the water conservancy rain measuring radar; divide the absolute value of the slope value difference between any two body scanning units by the maximum slope value as the slope optimization evaluation value, divide the absolute value of the aspect value difference between any two body scanning units by the maximum aspect value as the aspect optimization evaluation value, and divide the absolute value of the elevation value difference between any two body scanning units by the maximum elevation value as the aspect optimization evaluation value. is the elevation optimization evaluation value; the slope optimization evaluation value, the aspect optimization evaluation value and the elevation optimization evaluation value are summed and divided by the constant one and the sum of the terrain complexity of the monitoring range of the water conservancy rain measuring radar and multiplied by negative one, and the result of the calculation is used as the first evaluation factor between any two body scanning units of all the body scanning units of the water conservancy rain measuring radar; the result of exponential function mapping the first evaluation factor between any two body scanning units of all the body scanning units of the water conservancy rain measuring radar with the natural constant as the base is used as the first optimization factor for the distance measurement between the radar body scanning units.

5. The monitoring data processing method of a hydraulic rainfall radar for rain monitoring and forecasting according to claim 1 is characterized in that: The step of evaluating the first Euclidean distance between the radar volume scanning units according to the first optimization factor of the distance measurement between the radar volume scanning units to obtain the first Euclidean distance between the radar volume scanning units includes: A first optimization factor of the distance measurement between the radar volume scanning units is obtained, and a calculation result of multiplying the first optimization factor of the distance measurement between the radar volume scanning units by the Euclidean distance between the feature vectors between the radar volume scanning units is used as the first Euclidean distance between the radar volume scanning units.

6. The monitoring data processing method of a hydraulic rain measuring radar for rain monitoring and forecasting according to claim 1 is characterized in that: The obtaining of the Mahalanobis distance metric between the radar volume scanning units, evaluating the second optimization factor of the distance metric between the radar volume scanning units according to the Mahalanobis distance between the radar volume scanning units, and obtaining the second optimization factor of the distance metric between the radar volume scanning units includes: The Mahalanobis distance between the radar volume scanning unit and the cluster center volume scanning unit in the clustering process is obtained, and the reciprocal of the Mahalanobis distance between the radar volume scanning unit and the cluster center volume scanning unit of the cluster in the clustering iteration process is divided by the sum of the reciprocals of the Mahalanobis distances between the radar volume scanning unit and all the cluster center volume scanning units in the clustering iteration process as a second optimization factor for the distance measure between the radar volume scanning units.

7. The monitoring data processing method for using a hydraulic rain measuring radar for rain monitoring and forecasting according to claim 1 is characterized in that: The step of evaluating the optimized distance metric between the radar body scanning units according to the second optimization factor of the distance metric between the radar body scanning units, the first Euclidean distance between the radar body scanning units, and the Mahalanobis distance between the radar body scanning units, and obtaining the optimized distance metric between the radar body scanning units includes: Obtain a second optimization factor of the distance metric between the radar volume scanning units, obtain a first Euclidean distance between the radar volume scanning units, and obtain a Mahalanobis distance between the radar volume scanning unit and the cluster center volume scanning unit in the clustering process; multiply the second optimization factor by the Mahalanobis distance between the radar volume scanning unit and the cluster center volume scanning unit in the clustering process as the first optimized distance metric; subtract a constant one from the second optimization factor and multiply the result by the first Euclidean distance as the second optimized distance metric; and use the result of adding the first optimized distance metric to the second optimized distance metric as the optimized distance metric between the radar volume scanning units.

8. The monitoring data processing method for using a hydraulic rainfall radar for rain monitoring and forecasting according to claim 1 is characterized in that: The clustering is performed according to the optimized distance metric between the radar volume scanning units to obtain the clustering division results of all volume scanning units of the water conservancy rainfall measurement radar, including: The target number of clusters in the clustering process is obtained. In the iterative process of the clustering process, the initial clustering division in the iterative process is first performed according to the traditional Euclidean distance metric in the original clustering process. After the initial clustering division result in the iterative process is obtained, the covariance matrix of each cluster in the initial clustering division result is obtained according to the initial clustering division result in the iterative process. According to the covariance matrix of each cluster in the initial clustering division result, the optimized distance metric between each radar volume scanning unit and all cluster center radar volume scanning units in the iterative process of the clustering process is obtained for secondary division, and the secondary division result is obtained. According to the secondary division result, the cluster center in the clustering process and the clustering label of the radar volume scanning unit are updated to complete the iterative process in the clustering process. The iterative process in the clustering process is repeated until the clustering result of the clustering process no longer changes, and the clustering division result of all volume scanning units of the water conservancy rainfall measurement radar corresponding to the clustering process is obtained.

9. The monitoring data processing method for using a hydraulic rainfall radar for rain monitoring and forecasting according to claim 1 is characterized in that: The locally adaptive ZR relationship fitting evaluation is performed according to the clustering division results of all volume scanning units of the water conservancy rainfall measuring radar, including: The clustering division results of all volume scanning units of the water conservancy rainfall measuring radar are obtained, and the ZR relationship coefficient of the cluster is determined by the least squares method according to the radar reflectivity factor data of the volume scanning units of the water conservancy rainfall measuring radar in each cluster in the clustering division results, so as to obtain a local adaptive ZR relationship fitting evaluation result.

Images