Strong convective weather forecasting method and device based on space-time cross convergence mapping

Abstract

The invention relates to a severe convection weather forecasting method and device based on space-time cross convergence mapping. The method comprises the following steps: firstly, preprocessing radar reflectivity data and meteorological element data during convection nascent; correlation calculation is carried out on the meteorological element data and a strong echo region in radar reflectivity, candidate factors are screened out according to the significance level, and a candidate factor set is constructed; and finally, calculating the causal relationship between the candidate factors and the radar reflectivity through space-time cross convergence mapping. And if the causal relationship exists, further calculating the causal relationship strength, and taking the candidate factor as a severe convection forecasting factor. According to the method, by analyzing the severe convection forecasting factors, input is provided for severe convection forecasting, and the forecasting accuracy and timeliness of severe convection weather can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of non - linear time - series analysis, and particularly to a severe convective weather forecasting method and device based on spatio - temporal cross - convergent mapping. Background Art

[0002] Severe convective phenomena (such as thunderstorms, heavy rains, tornadoes, etc.) are important research objects in atmospheric science. Identifying their causal relationships is of great significance for weather forecasting and disaster prevention. Therefore, in the study of severe convection, causal inference has always been a core issue. In statistics, a correlation relationship indicates that two variables have a certain association or co - variation trend. Affected by confounding variables, a strong correlation cannot determine the causal relationship between two variables. More causal models are needed to identify the characteristics between data. In recent years, Convergent Cross - mapping (CCM) has provided new ideas for causal relationship identification. Its core element is to abstract time series into chaotic systems, and the shadow manifold based on phase - space reconstruction attractors can effectively represent the asymmetric causal relationship between system state variables. However, in the business field of severe convection, high timeliness is pursued, and this method has limitations in processing severe convective data, such as low computational efficiency and difficulty in extending to spatio - temporal data. Summary of the Invention

[0003] Based on this, it is necessary to provide a severe convective weather forecasting method and device based on spatio - temporal cross - convergent mapping for the above - mentioned technical problems.

[0004] A severe convective weather forecasting method based on spatio - temporal cross - convergent mapping, the method includes: Obtain radar reflectivity data and meteorological element data at the beginning of convection, and perform pre - processing.

[0005] Perform a correlation analysis on the pre - processed radar reflectivity data and meteorological element data, screen out candidate factors according to the significance level, and construct a candidate factor set.

[0006] Randomly select prediction pixels in the radar reflectivity, and calculate whether there is a causal relationship between each candidate factor pixel and the radar reflectivity respectively by using the spatio - temporal cross - convergent mapping algorithm according to the delay embedding of each candidate factor pixel and the prediction pixel; if there is a causal relationship, further calculate the strength of the causal relationship, and use this candidate factor as a severe convective forecasting factor.

[0007] Use all severe convective forecasting factors for severe convective weather forecasting to obtain a severe convective weather forecasting result.

[0008] In one embodiment, obtaining radar reflectivity data and meteorological element data at the beginning of convection, and performing pre - processing includes: Obtain the radar reflectivity data and meteorological element data at the initial stage of convection.

[0009] Perform spatio-temporal resampling on the radar reflectivity data and meteorological element data to make their spatio-temporal resolutions consistent.

[0010] Label the pixels with radar reflectivity greater than 35 dBz as the occurrence of the initial stage of convection; a pixel is the smallest spatio-temporal unit of the radar reflectivity data and meteorological element data.

[0011] In one embodiment, perform correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screen out candidate factors according to the significance level, and construct a candidate factor set, including: Extract the radar reflectivity values of the meteorological element data and the convection initial area, and calculate the Pearson correlation coefficient using the numerical relationship between pixels.

[0012] Construct a statistic of the student distribution based on the Pearson correlation coefficient for significance testing.

[0013] Screen out candidate factors according to the significance level to obtain a candidate factor set.

[0014] In one embodiment, randomly select prediction pixels in the radar reflectivity, and use the delay embedding of each candidate factor pixel and the prediction pixel to calculate whether there is a causal relationship between each candidate factor and the radar reflectivity respectively by using the spatio-temporal cross-convergent mapping algorithm; if there is a causal relationship, further calculate the strength of the causal relationship, and use this candidate factor as a severe convection prediction factor, including: Randomly select several prediction pixels in the radar reflectivity to form a prediction pixel set as: ; where, represents the set of true values of the prediction pixels, represents the radar reflectivity value of the th prediction pixel, represents the number of randomly selected prediction pixels, represents the th spatio-temporal index of the prediction pixel, , Lon 、 Lat and Time represent longitude, latitude and time respectively.

[0015] Traverse each candidate factor, use the spatio-temporal index of the prediction pixel to find the corresponding pixel in the current candidate factor; calculate the Euclidean distance between each pixel in the candidate factor and the delay embedding of the corresponding pixel to obtain the delay embedding distance.

[0016] According to the current candidate factors, predicted pixels and the delayed embedding distance of the corresponding pixels, the spatiotemporal cross-convergence mapping is used to judge the causal relationship.

[0017] If there is a causal relationship between the current candidate factor and the radar reflectivity, the strength of the causal relationship is calculated, and the candidate factor is used as a severe convection forecast factor.

[0018] Continue to make causal judgment on the next candidate factor until all candidate factors are traversed.

[0019] In one embodiment, according to the current candidate factor, the predicted pixel and the delayed embedding distance of the corresponding pixel, the causal relationship is determined by using the spatiotemporal cross-convergence mapping, including: Traverse each predicted pixel, and find the closest distance to each predicted pixel from the current candidate factors according to the delayed embedding distance of the corresponding pixel. k pixels, which are called similar pixels; obtain the spatiotemporal index of similar pixels and the delayed embedding distance with the current predicted pixel.

[0020] According to the delayed embedding distance between the spatiotemporal index of similar pixels and the current predicted pixel, the weight of similar pixels is calculated as: ; in, Yes The normalized weights, Indicates selected The delayed embedding distance between the pixel and the current predicted pixel, represents the unnormalized weights.

[0021] The estimated value of each prediction pixel is calculated based on the weight: ; in, represents the estimated value of radar reflectivity predicted only based on the numerical relationship between the current candidate factor and radar reflectivity; express The radar reflectivity value corresponding to the similar pixels under the spatiotemporal index.

[0022] After completing the traversal of the predicted pixels, calculate the difference between the estimated value of the predicted pixel and the true value. Pearson Correlation coefficient and calculate the significance level corresponding to the correlation coefficient.

[0023] A significance test is performed according to the significance level corresponding to the correlation coefficient. If the significance test is passed, there is a causal relationship between the current candidate factor and the radar reflectivity; otherwise, there is no causal relationship between the current candidate factor and the radar reflectivity.

[0024] In one embodiment, the process of calculating the strength of the causal relationship includes: traversing the current candidate factors with a window of a preset size, randomly selecting m windows of a corresponding size from the current candidate factors, and evaluating each window using spatio-temporal cross-convergent mapping to obtain m the mean of the correlation coefficients of the correlation coefficients; the window of the preset size is three-dimensional, including all time segments and part of the longitude and latitude grid, and the shape of the window is represented in the form of L×L×T where L×L represents the geographical spatial range, and T represents all time segments.

[0025] Construct a statistic of the student distribution for the mean of the correlation coefficients for significance testing; the statistic of the student distribution is: ; where represents the statistic of the student distribution, represents the mean of the correlation coefficients, is the number of randomly selected predicted pixels.

[0026] Iteratively select the window and perform significance testing in a binary search manner until the smallest window when the mean of the correlation coefficients passes the significance test is found, and use the L value of the smallest window as the strength of the causal relationship.

[0027] A severe convective weather forecasting device based on spatio-temporal cross-convergent mapping, the device includes: A data preprocessing module, configured to obtain radar reflectivity data and meteorological element data at the time of the initial occurrence of convection, and perform preprocessing.

[0028] A candidate factor set construction module, configured to perform correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screen out candidate factors according to the significance level, and construct a candidate factor set.

[0029] A causal relationship judgment and severe convective forecasting factor determination module, configured to randomly select predicted pixels in the radar reflectivity, and calculate whether there is a causal relationship between each candidate factor and the radar reflectivity respectively according to the delay embedding and predicted pixels of each candidate factor pixel using the spatio-temporal cross-convergent mapping algorithm; if there is a causal relationship, further calculate the strength of the causal relationship, and use the candidate factor as a severe convective forecasting factor.

[0030] A severe convective weather forecasting module, configured to use all severe convective forecasting factors for severe convective weather forecasting to obtain a severe convective weather forecasting result.

[0031] The above-mentioned severe convective weather forecasting method and device based on spatio-temporal cross-convergence mapping, the method includes: first, preprocess the radar reflectivity data and meteorological element data at the initial stage of convection; then calculate the correlation between the meteorological element data and the area of strong echoes in the radar reflectivity, and screen out candidate factors according to the significance level to construct a candidate factor set; finally, calculate the causal relationship between the candidate factors and the radar reflectivity through spatio-temporal cross-convergence mapping; if there is a causal relationship, further calculate the strength of the causal relationship, and use this candidate factor as a severe convective forecasting factor. This method provides input for severe convective forecasting by analyzing severe convective forecasting factors, which helps to improve the forecasting accuracy and timeliness of severe convective weather. Description of the Drawings

[0032] Figure 1 It is a schematic flow chart of a severe convective weather forecasting method based on spatio-temporal cross-convergence mapping in an embodiment; Figure 2 It is a schematic diagram of delay embedding in an embodiment; Figure 3 It is a flow chart of spatio-temporal cross-convergence mapping in an embodiment; Figure 4 It is a schematic overall flow chart of a severe convective weather forecasting method based on spatio-temporal cross-convergence mapping in another embodiment. Detailed Embodiment

[0033] In order to make the purpose, technical solution and advantages of this application clearer, the following further details this application in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain this application and are not used to limit this application.

[0034] In one embodiment, as Figure 1 shown, a severe convective weather forecasting method based on spatio-temporal cross-convergence mapping is provided, and the method includes the following steps: Step 100: Obtain the radar reflectivity data and meteorological element data at the initial stage of convection, and perform preprocessing.

[0035] Specifically, in an observation of the initial stage of convection, operations such as reading, resolution scaling, and annotation are performed on the radar reflectivity data and meteorological element data respectively.

[0036] Step 102: Perform correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screen out candidate factors according to the significance level, and construct a candidate factor set.

[0037] Step 104: Randomly select prediction pixels in the radar reflectivity. According to the delay embedding of each candidate factor pixel and the prediction pixel, use the spatio-temporal cross-convergent mapping algorithm to calculate whether there is a causal relationship between each candidate factor and the radar reflectivity respectively; if there is a causal relationship, further calculate the strength of the causal relationship, and use this candidate factor as a severe convective weather forecasting factor.

[0038] Specifically, randomly select prediction pixels in the radar reflectivity; the spatio-temporal index of a pixel includes: longitude, latitude, and time, that is , where i is the spatio-temporal index of the pixel, Lon 、 Lat and Time represent longitude, latitude, and time respectively.

[0039] Delay embedding distance calculation: Calculate the delay embedding of each pixel of the current candidate factor, and use the spatio-temporal index of the prediction pixel to find the corresponding pixel. The delay embedding distance can be calculated by the Euclidean distance of the delay embedding, which measures the similarity between pixels.

[0040] Causal relationship judgment: Use spatio-temporal cross-convergent mapping to judge the causal relationship. Specifically, with the global area of the current candidate factor as the calculation area, for each prediction pixel, search for similar pixels in the candidate factor. Using the delay embedding distance as the weight, estimate the value of each prediction pixel. Calculate the Pearson correlation coefficient and significance level between the estimated value and the true value. If the significance test is passed, it indicates that there is a causal relationship between the current candidate factor and the radar reflectivity, otherwise there is no.

[0041] Causal relationship strength calculation: After judging that there is a causal relationship, with the window of the current candidate factor as the calculation area, use spatio-temporal cross-convergent mapping to calculate the results for windows of different sizes. Use the smallest window size with significance to characterize the strength of the causal relationship. And use this candidate factor as a severe convective weather forecasting factor.

[0042] Step 106: Use all severe convective weather forecasting factors for severe convective weather forecasting to obtain the severe convective weather forecasting result.

[0043] In the above strong convective weather forecasting method based on spatio-temporal cross-convergence mapping, the method includes: First, preprocess the radar reflectivity data and meteorological element data at the initial stage of convection; Then, calculate the correlation between the meteorological element data and the area of strong echoes in the radar reflectivity, screen out candidate factors according to the significance level, and construct a candidate factor set; Finally, calculate the causal relationship between the candidate factors and the radar reflectivity through spatio-temporal cross-convergence mapping; If there is a causal relationship, further calculate the strength of the causal relationship, and use this candidate factor as a strong convective forecasting factor. This method provides input for strong convective forecasting by analyzing strong convective forecasting factors, which helps to improve the forecasting accuracy and timeliness of strong convective weather.

[0044] In one embodiment, step 100 includes: Obtain the radar reflectivity data and meteorological element data at the initial stage of convection; Perform spatio-temporal resampling on the radar reflectivity data and meteorological element data to make their spatio-temporal resolutions consistent; Label the pixels with a radar reflectivity greater than 35 dBz as the occurrence of the initial stage of convection; A pixel is the smallest spatio-temporal unit of the radar reflectivity data and meteorological element data.

[0045] Specifically, obtaining the radar reflectivity data and meteorological element data at the initial stage of convection and performing preprocessing specifically includes: Step 1-1, data reading: According to the observation of the initial stage of a convection, read the radar reflectivity data and meteorological element data according to time and space.

[0046] Step 1-2, initial convection labeling: Call the smallest unit of the radar reflectivity and meteorological element data a pixel, and label the pixels with a radar reflectivity greater than 35 dBz as the occurrence of the initial stage of convection, otherwise the pixel has not experienced the initial stage of convection.

[0047] Step 1-3, data scaling: Perform spatio-temporal resampling on the radar reflectivity and meteorological element data to make their spatio-temporal resolutions consistent.

[0048] In one embodiment, step 102 includes: Extract the radar reflectivity values of the meteorological element data and the initial convection area, and calculate the Pearson correlation coefficient using the numerical relationship between pixels; Construct a statistic of the student distribution for significance testing according to the Pearson correlation coefficient; Screen out candidate factors according to the significance level to obtain a candidate factor set.

[0049] Specifically, Pearson Correlation coefficient calculation: Extract the radar reflectivity values of the meteorological element data and the initial convection area, and calculate Pearson the correlation coefficient Pearson The correlation coefficient calculation formula is: ; Where,X and Y respectively represent the meteorological elements and radar reflectivity values in the incipient convective region. cov is the covariance, and respectively represent the standard deviations of the meteorological elements and radar reflectivity values in the incipient convective region. r represents Pearson the correlation coefficient.

[0050] Significance test of the correlation coefficient: After calculating the Pearson correlation coefficient, a significance test can be performed by constructing a statistic of the Student's distribution.

[0051] ; wherein, represents the statistic of the Student's distribution, num represents the number of pixels in the incipient convective region.

[0052] In one embodiment, step 104 includes: randomly selecting a number of predicted pixels in the radar reflectivity to form a set of predicted pixels as: ; wherein, represents the set of true values of the predicted pixels, represents the th radar reflectivity value of the predicted pixel, n represents the number of randomly selected predicted pixels, represents the th spatio-temporal index of the predicted pixel, , Lon , Lat and Time respectively represent longitude, latitude and time; Traverse each candidate factor, use the spatio-temporal index of the predicted pixel to find the corresponding pixel in the current candidate factor; calculate the Euclidean distance between each pixel in the candidate factor and the delayed embedding of the corresponding pixel to obtain the delayed embedding distance; according to the current candidate factor, the predicted pixel and the delayed embedding distance of the corresponding pixel, use spatio-temporal cross-convergent mapping to judge the causal relationship; if there is a causal relationship between the current candidate factor and the radar reflectivity, calculate the strength of the causal relationship, and use this candidate factor as a severe convection prediction factor; continue to perform causal judgment on the next candidate factor until all candidate factors are traversed.

[0053] Specifically, delayed embedding means: The delayed embedding of each pixel includes all pixels adjacent to it in space and time. After the delayed embedding operation, each pixel corresponds to a three-dimensional matrix, where the three dimensions respectively correspond to all adjacent pixels on the three axes of longitude, latitude and time. Delayed embedding is asFigure 2 as shown

[0054] Spatio-temporal index matching: Among the current candidate factors, find the pixels whose spatio-temporal indices are consistent with the predicted pixel, and call them "corresponding pixels".

[0055] Euclidean distance quantization: Store the delay embeddings of each corresponding pixel, obtain the delay embeddings of each pixel in the candidate factor, and calculate the Euclidean distance between every two of them. Formally, the value of the Euclidean distance can be expressed as the second-order norm of a matrix: ; where represents the distance between two pixels, X 1 and X 2 respectively represent the delay embeddings of two pixels.

[0056] In one embodiment, according to the delay embedding distances of the current candidate factor, the predicted pixel, and the corresponding pixels, spatio-temporal cross-convergence mapping is used to judge the causal relationship, including: Traverse each predicted pixel. According to the delay embedding distances of the corresponding pixels, find the pixel closest to each predicted pixel from the current candidate factor, and call it the similar pixel; obtain the spatio-temporal index of the similar pixel and the delay embedding distance from the current predicted pixel; according to the spatio-temporal index of the similar pixel and the delay embedding distance from the current predicted pixel, calculate the weight of the similar pixel as: ; where is the weight after normalizing , represents the delay embedding distance between the selected pixels and the current predicted pixel, represents the unnormalized weight.

[0057] Calculate the estimated value of each predicted pixel according to the weight as: ; where represents the estimated value of the radar reflectivity predicted only based on the numerical relationship between the current candidate factor and the radar reflectivity; represents the radar reflectivity values corresponding to the

[0058] After completing the traversal of the predicted pixels, calculate the PearsonCalculate the correlation coefficient and the corresponding significance level of the correlation coefficient; conduct a significance test based on the significance level corresponding to the correlation coefficient. If the significance test is passed, there is a causal relationship between the current candidate factor and the radar reflectivity; otherwise, there is no causal relationship between the current candidate factor and the radar reflectivity.

[0059] The process of spatio-temporal cross-convergence mapping is as Figure 3 shown.

[0060] Specifically, the specific steps for causal relationship judgment include: (1) Similar pixel search: Traverse each predicted pixel, and find the pixel closest to each predicted pixel from the candidate factors according to the calculated delay embedding distance, and obtain their spatio-temporal indices and the distance from the current predicted pixel.

[0061] (2) Distance weight solution: According to the delay embedding distances of the pixels, calculate the weights using the weight expression of similar pixels

[0062] (3) Estimation of predicted pixels: Calculate the estimated value , is obtained by multiplying the radar reflectivity values of these pixels at the corresponding spatio-temporal indices by the corresponding weights and then summing them up with weights. The calculation formula is as shown in the estimated value expression of the predicted pixel above.

[0063] (4) Significance test: After completing the traversal of the predicted pixels, form each into , calculate the between and Pearson correlation coefficient , and calculate the significance level; where represents the estimated value.

[0064] In one embodiment, the process of calculating the strength of the causal relationship in step 104 includes: traversing the current candidate factor with a window of a preset size, randomly selecting m windows of the corresponding size from the current candidate factor, and evaluating each window using spatio-temporal cross-convergence mapping to obtain the m mean value of the correlation coefficients of the L×L×T correlation coefficients; the window of the preset size is three-dimensional, including all time segments and part of the latitude and longitude grid cells, and the shape of the window is represented in the form of L×LRepresents the geographical spatial range, and T represents all time segments; a statistical quantity of the student distribution is constructed for the mean of the correlation coefficient for significance testing; the statistical quantity of the student distribution is: ; Among them, represents the statistical quantity of the student distribution, represents the mean of the correlation coefficient, is the number of randomly selected prediction pixels.

[0065] Iteratively select the window and perform significance testing in a binary search manner until the smallest window when the mean of the correlation coefficient passes the significance test is found, and use the L value of the smallest window as the strength of the causal relationship.

[0066] Specifically, window setting: The window is three-dimensional, including all time segments and part of the longitude and latitude grid, and its shape can be represented in the form of L×L×T, where L×L represents the geographical spatial range and T represents all time segments. After determining the window size, randomly select m windows of the corresponding size from the current candidate factors and traverse these m windows.

[0067] Window prediction result calculation: When a window is selected, repeat the spatio-temporal cross-convergence mapping. When all these m regions of the same size have been traversed, take the mean m of these . Construct a statistical quantity of the student distribution for significance testing of . The statistical quantity of the student distribution is as shown in the above statistical quantity expression of the student distribution. Output of the causal relationship strength: Iteratively set the window size and calculate the window prediction results in a binary search manner until the smallest window when

[0068] passes the significance test is found. Specifically, if the calculated by the current window is significant, then reduce , otherwise increase L . After completing the traversal, output the L value as the representation of the strength of the causal relationship. L The overall process of the severe convective weather forecasting method based on spatio-temporal cross-convergence mapping is as

[0069] shown. Figure 4 It should be understood that although

[0070] Figure 1 ​The steps in the flowchart are sequentially shown according to the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least a part of the steps in Figure 1 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0071] In one embodiment, a severe convective weather forecasting device based on spatio-temporal cross-convergence mapping is provided, including: a data preprocessing module, a candidate factor set construction module, a causal relationship judgment and severe convective forecasting factor determination module, and a severe convective weather forecasting module, where: The data preprocessing module is configured to obtain radar reflectivity data and meteorological element data at the initial stage of convection and perform preprocessing.

[0072] The candidate factor set construction module is configured to perform correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screen out candidate factors according to the significance level, and construct a candidate factor set.

[0073] The causal relationship judgment and severe convective forecasting factor determination module is configured to randomly select prediction pixels in the radar reflectivity, and respectively calculate whether there is a causal relationship between each candidate factor pixel and the radar reflectivity by using the spatio-temporal cross-convergence mapping algorithm according to the delayed embedding and prediction pixels of each candidate factor pixel; if there is a causal relationship, further calculate the strength of the causal relationship, and use the candidate factor as a severe convective forecasting factor.

[0074] The severe convective weather forecasting module is configured to use all severe convective forecasting factors for severe convective weather forecasting to obtain a severe convective weather forecasting result.

[0075] In one of the embodiments, the data preprocessing module is further configured to obtain radar reflectivity data and meteorological element data at the initial stage of convection; perform spatio-temporal resampling on the radar reflectivity data and meteorological element data to make their spatio-temporal resolutions consistent; label the pixels with a radar reflectivity greater than 35 dBz as the occurrence of the initial stage of convection; a pixel is the smallest spatio-temporal unit of the radar reflectivity data and meteorological element data.

[0076] In one embodiment, the candidate factor set construction module is further configured to extract meteorological element data and the radar reflectivity value of the convective initiation region, calculate the Pearson correlation coefficient by using the numerical relationship between pixels; construct a statistic of the student distribution according to the Pearson correlation coefficient for significance test; and screen out candidate factors according to the significance level to obtain a candidate factor set.

[0077] In one embodiment, the causal relationship judgment and severe convection forecasting factor determination module is further configured to randomly select prediction pixels in the radar reflectivity; the prediction pixels are as shown in the above prediction pixel expression; traverse each candidate factor, and use the spatio-temporal index of the prediction pixels to find the corresponding pixels in the current candidate factor; calculate the Euclidean distance between each pixel in the candidate factor and the delayed embedding of the corresponding pixel to obtain the delayed embedding distance; according to the current candidate factor, the prediction pixels, and the delayed embedding distance of the corresponding pixels, use spatio-temporal cross-convergent mapping to judge the causal relationship; if there is a causal relationship between the current candidate factor and the radar reflectivity, calculate the strength of the causal relationship, and use the candidate factor as a severe convection forecasting factor; continue to perform causal judgment on the next candidate factor until all candidate factors are traversed.

[0078] In one embodiment, the causal relationship judgment and severe convection forecasting factor determination module is further configured to traverse each prediction pixel, and according to the delayed embedding distance of the corresponding pixel, find the k pixel with the closest distance to each prediction pixel from the current candidate factor, and call it a similar pixel; obtain the spatio-temporal index of the similar pixel and the delayed embedding distance from the current prediction pixel; according to the spatio-temporal index of the similar pixel and the delayed embedding distance from the current prediction pixel, calculate the weight of the similar pixel by using the weight expression of the similar pixel as described above.

[0079] According to the weight, calculate the estimated value of each prediction pixel by using the estimated value expression of the prediction pixel as described above; after completing the traversal of the prediction pixels, calculate the Pearson correlation coefficient between the estimated value and the true value of the prediction pixel, and calculate the significance level corresponding to the correlation coefficient; perform a significance test according to the significance level corresponding to the correlation coefficient. If the significance test is passed, there is a causal relationship between the current candidate factor and the radar reflectivity; otherwise, there is no causal relationship between the current candidate factor and the radar reflectivity.

[0080] In one embodiment, the process of calculating the strength of the causal relationship in the causal relationship judgment and severe convection forecasting factor determination module includes: traversing the current candidate factor with a window of a preset size, randomly selecting m windows of a corresponding size from the current candidate factor, and performing evaluation within each window by using spatio-temporal cross-convergent mapping to obtain mThe mean correlation coefficient of the correlation coefficients; the window of the preset size is three-dimensional, including all time segments and part of the longitude and latitude grid. The shape of the window is represented in the form of L×L×T , where L×L represents the geographical spatial range, and T represents all time segments; a significance test is performed on the statistic of the student distribution constructed for the mean correlation coefficient; the statistic of the student distribution is as shown in the above expression of the statistic of the student distribution; the window selection and significance test are iterated in a binary search manner until the smallest window when the mean correlation coefficient passes the significance test is found, and the L value of the smallest window is used as the strength of the causal relationship.

[0081] For the specific limitations of the severe convective weather forecasting device based on spatio-temporal cross-convergence mapping, reference can be made to the limitations of the severe convective weather forecasting method based on spatio-temporal cross-convergence mapping in the above text, which will not be elaborated here. Each module in the above severe convective weather forecasting device based on spatio-temporal cross-convergence mapping can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0082] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0083] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation to the scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A severe convective weather forecasting method based on spatio-temporal cross-convergence mapping, characterized in that, The method includes: Obtaining radar reflectivity data and meteorological element data at the initial stage of convection, and performing preprocessing; Performing correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screening out candidate factors according to the significance level, and constructing a candidate factor set; Randomly selecting prediction pixels in the radar reflectivity, and calculating whether there is a causal relationship between each candidate factor pixel and the radar reflectivity respectively by using the spatio-temporal cross-convergent mapping algorithm according to the delay embedding of each candidate factor pixel and the prediction pixel; if there is a causal relationship, further calculating the intensity of the causal relationship, and taking this candidate factor as a severe convection forecasting factor; Using all the severe convection forecasting factors for severe convection weather forecasting to obtain a severe convection weather forecasting result.

2. The severe convective weather forecasting method based on spatio-temporal cross-convergence mapping according to claim 1, wherein Obtaining radar reflectivity data and meteorological element data at the initial stage of convection, and performing preprocessing, including: Obtaining radar reflectivity data and meteorological element data at the initial stage of convection; Performing spatio-temporal resampling on the radar reflectivity data and meteorological element data to make their spatio-temporal resolutions consistent; Labeling the pixels with radar reflectivity greater than 35 dBz as the occurrence of initial convection; the pixel is the smallest spatio-temporal unit of the radar reflectivity data and meteorological element data.

3. The severe convective weather forecasting method based on spatio-temporal cross-convergence mapping according to claim 1, wherein Performing correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screening out candidate factors according to the significance level, and constructing a candidate factor set, including: Extracting the radar reflectivity values of the meteorological element data and the convection initial area, and calculating the Pearson correlation coefficient by using the numerical relationship between pixels; Constructing a statistic of the student distribution according to the Pearson correlation coefficient for significance testing; Screening out candidate factors according to the significance level to obtain a candidate factor set.

4. The severe convective weather forecasting method based on spatio-temporal cross-convergence mapping according to claim 1, characterized in that Randomly selecting prediction pixels in the radar reflectivity, and calculating whether there is a causal relationship between each candidate factor pixel and the radar reflectivity respectively by using the spatio-temporal cross-convergent mapping algorithm according to the delay embedding of each candidate factor pixel and the prediction pixel; If there is a causal relationship, further calculating the intensity of the causal relationship, and taking this candidate factor as a severe convection forecasting factor, including: Randomly selecting several prediction pixels in the radar reflectivity to form a prediction pixel set as: ； Among them, represents the set of true values of the predicted pixels, represents the radar reflectivity value of the th predicted pixel, represents the number of randomly selected predicted pixels, represents the spatio-temporal index of the th predicted pixel, , Lon , Lat and Time represent longitude, latitude and time respectively; Traversing each candidate factor, finding the corresponding pixel in the current candidate factor by using the spatio-temporal index of the prediction pixel; calculating the Euclidean distance between each pixel in the candidate factor and the delay embedding of the corresponding pixel to obtain the delay embedding distance; Judging the causal relationship by using the spatio-temporal cross-convergent mapping according to the current candidate factor, prediction pixel and the delay embedding distance of the corresponding pixel; If there is a causal relationship between the current candidate factor and the radar reflectivity, calculating the intensity of the causal relationship, and taking this candidate factor as a severe convection forecasting factor; Continuing to perform causal judgment on the next candidate factor until all candidate factors are traversed.

5. The severe convective weather forecasting method based on spatio-temporal cross-convergence mapping according to claim 1, wherein, Judging the causal relationship by using the spatio-temporal cross-convergent mapping according to the current candidate factor, prediction pixel and the delay embedding distance of the corresponding pixel, including: Traverse each predicted pixel, and find the pixel with the closest distance to each predicted pixel from the current candidate factors according to the delay embedding distance of the corresponding pixel, and call it the similar pixel; obtain the spatio-temporal index of the similar pixel and the delay embedding distance from the current predicted pixel; k For each predicted pixel, find the pixel with the closest distance to it from the current candidate factors according to the delay embedding distance of the corresponding pixel, and call it the similar pixel; obtain the spatio-temporal index of the similar pixel and the delay embedding distance from the current predicted pixel; Calculating the weight of the similar pixel according to the spatio-temporal index of the similar pixel and the delay embedding distance of the current prediction pixel as: ； Among them, is the weight after normalization, represents the delayed embedding distance between the selected pixels and the current predicted pixel; The estimated value of each predicted pixel is calculated according to the weight as follows: ； Among them, represents the estimated value of radar reflectivity predicted only based on the numerical relationship between the current candidate factor and the radar reflectivity; represents the radar reflectivity values corresponding to similar pixels under the spatio-temporal index; After completing the traversal of the predicted pixels, calculate the Pearson correlation coefficient between the estimated value and the true value of the predicted pixels, and calculate the significance level corresponding to the correlation coefficient; Perform a significance test according to the significance level corresponding to the correlation coefficient. If the significance test is passed, there is a causal relationship between the current candidate factor and the radar reflectivity; otherwise, there is no causal relationship between the current candidate factor and the radar reflectivity.

6. The severe convective weather forecasting method based on spatio-temporal cross-convergence mapping according to claim 1, characterized in that The process of calculating the strength of the causal relationship includes: Traverse the current candidate factors with a window of a preset size, and randomly select m windows of a corresponding size from the current candidate factors. Evaluate each window using spatio-temporal cross-convergence mapping to obtain m the mean value of the correlation coefficients of the correlation coefficients; the window of the preset size is three-dimensional, including all time segments and part of the longitude and latitude grid. The shape of the window is represented in the form of L×L×T , where L×L represents the geographical space range, and T represents all time segments; Construct a statistic of the student distribution for the mean of the correlation coefficients for a significance test; the statistic of the student distribution is: ； Among them, represents the statistic of student distribution, represents the mean of correlation coefficients, is the number of randomly selected prediction pixels; Iteratively select the window and perform the significance test in a binary search manner until the minimum window when the mean of the correlation coefficient passes the significance test is found, and use the L value as the strength of the causal relationship.

7. A severe convective weather forecasting device based on spatio-temporal cross-convergence mapping, characterized in that, The device includes: A data preprocessing module, configured to obtain radar reflectivity data and meteorological element data at the beginning of convection and perform preprocessing; A candidate factor set construction module, configured to perform a correlation analysis on the preprocessed radar reflectivity data and meteorological element data, screen out candidate factors according to the significance level, and construct a candidate factor set; A causal relationship judgment and severe convection forecasting factor determination module, configured to randomly select predicted pixels in the radar reflectivity, and calculate whether there is a causal relationship between each candidate factor pixel and the radar reflectivity respectively by using the spatio-temporal cross-convergent mapping algorithm according to the delayed embedding of each candidate factor pixel and the predicted pixel; if there is a causal relationship, further calculate the strength of the causal relationship, and use the candidate factor as a severe convection forecasting factor; A severe convection weather forecasting module, configured to use all the severe convection forecasting factors for severe convection weather forecasting to obtain a severe convection weather forecasting result.

Images