Rainfall event identification method and device

Abstract

The invention discloses a rainfall event identification method and device, and the method comprises the steps: obtaining a plurality of rainfall data points, and enabling each rainfall data point to comprise a corresponding relation among a time point, a spatial position and a rainfall amount; clustering the plurality of rainfall data points based on a time search radius, a space search radius and a minimum inclusion rainfall to obtain at least one rainfall event, the time distance between different rainfall data points in each rainfall event being less than or equal to the time search radius; the space position distance between different rainfall data points in each rainfall event is smaller than or equal to the space search radius, and the sum of the rainfall amount of all the rainfall data points contained in each rainfall event is larger than or equal to the minimum contained rainfall amount.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to a method and apparatus for identifying rainfall events. Background Technology

[0002] A rainfall event is a continuous rainfall process both in time and space. Currently, when identifying rainfall events, all rainfall data at a specific station is typically identified, and the data is classified into rainfall events using certain rules. For example, if the cumulative rainfall over a continuous period exceeds a threshold, the rainfall data during that period is identified as a rainfall event.

[0003] When using the above methods, it is difficult to characterize the spatial extent of rainfall events, which affects the accuracy of identifying rainfall events. Summary of the Invention

[0004] This application provides a method and apparatus for identifying rainfall events, which can improve the accuracy of rainfall event identification. The technical solution is as follows.

[0005] Firstly, a method for identifying rainfall events is provided, the method comprising:

[0006] Acquire multiple rainfall data points, each of which includes the correspondence between time point, spatial location, and precipitation amount;

[0007] Based on the time search radius, spatial search radius, and minimum contained rainfall, the multiple rainfall data points are clustered to obtain at least one rainfall event. The time distance between different rainfall data points in each rainfall event is less than or equal to the time search radius, the spatial distance between different rainfall data points in each rainfall event is less than or equal to the spatial search radius, and the total precipitation of all rainfall data points included in each rainfall event is greater than or equal to the minimum contained rainfall.

[0008] In some implementations, the plurality of rainfall data points includes a first rainfall data point, the at least one rainfall event includes a first rainfall event, the temporal search radius includes a first temporal radius, the spatial search radius includes a first spatial radius, the minimum included rainfall amount includes a first rainfall amount, and the clustering of the plurality of rainfall data points based on the temporal search radius, the spatial search radius, and the minimum included rainfall amount includes:

[0009] Based on the time point and spatial location of each rainfall data point, multiple nearby rainfall points are determined from the multiple rainfall data points. The time distance between each nearby rainfall point and the first rainfall data point is less than or equal to the first time radius, and the spatial location distance is less than or equal to the first spatial radius.

[0010] In response to the fact that the sum of precipitation at multiple nearby rainfall points of the first rainfall data point is greater than or equal to the first rainfall amount, the multiple nearby rainfall points and the first rainfall data point are classified into the first rainfall event.

[0011] In some implementations, the at least one rainfall event further includes a second rainfall event, the second rainfall event having a smaller time scale and a smaller spatial scale than the first rainfall event; the time search radius further includes a second time radius, the second time radius being smaller than the first time radius; the spatial search radius further includes a second spatial radius, the second spatial radius being smaller than the first spatial radius; the minimum included rainfall amount further includes a second rainfall amount, the second rainfall amount being smaller than the first rainfall amount; and the clustering of the plurality of rainfall data points based on the time search radius, the spatial search radius, and the minimum included rainfall amount includes:

[0012] Among the multiple rainfall data points, the rainfall data points other than the first rainfall event are identified as the first unclassified points. The first unclassified points include the second rainfall data points, and the sum of the precipitation of multiple nearby rainfall points of the second rainfall data point is less than the first rainfall amount.

[0013] Based on the time point and spatial location of each rainfall data point in the first unclassified points, multiple nearby rainfall points are determined from the first unclassified points. The time distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second time radius, and the spatial location distance is less than or equal to the second spatial radius.

[0014] In response to the fact that the sum of precipitation at multiple nearby rainfall points of the second rainfall data point is greater than or equal to the second rainfall amount, the multiple nearby rainfall points and the second rainfall data point are classified into the second rainfall event.

[0015] In some embodiments, the at least one rainfall event further includes a third rainfall event, which has a smaller time scale and a smaller spatial scale than the second rainfall event. The time search radius further includes a third time radius, which is smaller than the second time radius. The spatial search radius further includes a third spatial radius, which is smaller than the second spatial radius. The minimum included rainfall amount also includes a third rainfall amount, which is smaller than the second rainfall amount. The clustering of the plurality of rainfall data points based on the time search radius, spatial search radius, and minimum included rainfall amount includes:

[0016] Rainfall data points other than the second rainfall event in the first unclassified points are identified as second unclassified points. The second unclassified points include third rainfall data points, and the sum of precipitation of multiple nearby rainfall points of the third rainfall data point is less than the second rainfall amount.

[0017] Based on the time point and spatial location of each rainfall data point in the second unclassified points, multiple nearby rainfall points are determined from the second unclassified points. The time distance between each nearby rainfall point and the third rainfall data point is less than or equal to the third time radius, and the spatial location distance is less than or equal to the third spatial radius.

[0018] In response to the fact that the sum of precipitation at multiple nearby rainfall points of the third rainfall data point is greater than or equal to the third rainfall amount, the multiple nearby rainfall points and the third rainfall data point are classified into the third rainfall event.

[0019] In some implementations, before clustering the plurality of rainfall data points based on the time search radius, spatial search radius, and minimum contained rainfall, the method further includes:

[0020] The precipitation amount of each of the multiple precipitation data points is compared with the precipitation threshold, and the precipitation data points with daily precipitation less than the precipitation threshold are removed from the multiple precipitation data points.

[0021] In some implementations, before clustering the plurality of rainfall data points based on the time search radius, spatial search radius, and minimum contained rainfall, the method further includes:

[0022] The ratio between the temporal search radius and the spatial search radius is obtained and used as the spatiotemporal scale conversion coefficient.

[0023] The time point of each of the multiple rainfall data points is multiplied by the spatiotemporal scale conversion coefficient so that the resolution of the multiple rainfall data points in the time dimension is the same as the resolution in the spatial dimension.

[0024] In some implementations, after clustering the plurality of rainfall data points based on the time search radius, spatial search radius, and minimum contained rainfall to obtain at least one rainfall event, the method further includes:

[0025] The at least one rainfall event is input into the flood catastrophic model, and the expected loss of the at least one rainfall event is output by the flood catastrophic model.

[0026] Secondly, a device for identifying rainfall events is provided, the device comprising:

[0027] The acquisition unit is used to acquire multiple rainfall data points, each of which includes the correspondence between time point, spatial location, and precipitation amount.

[0028] A clustering unit is used to cluster the multiple rainfall data points based on the time search radius, the spatial search radius, and the minimum included rainfall amount to obtain at least one rainfall event. The time distance between different rainfall data points in each rainfall event is less than or equal to the time search radius, the spatial distance between different rainfall data points in each rainfall event is less than or equal to the spatial search radius, and the total precipitation of all rainfall data points included in each rainfall event is greater than or equal to the minimum included rainfall amount.

[0029] In some implementations, the plurality of rainfall data points includes a first rainfall data point, the at least one rainfall event includes a first rainfall event, the time search radius includes a first time radius, the spatial search radius includes a first spatial radius, the minimum included rainfall amount includes a first rainfall amount, and the clustering unit is used to determine a plurality of nearby rainfall points of the first rainfall data point from the plurality of rainfall data points based on the time point and spatial location of each rainfall data point, wherein the time distance between each nearby rainfall point and the first rainfall data point is less than or equal to the first time radius, and the spatial location distance is less than or equal to the first spatial radius; in response to the sum of the rainfall amounts of the plurality of nearby rainfall points of the first rainfall data point being greater than or equal to the first rainfall amount, the plurality of nearby rainfall points and the first rainfall data point are assigned to the first rainfall event.

[0030] In some embodiments, the at least one rainfall event further includes a second rainfall event, the second rainfall event having a smaller time scale and a smaller spatial scale than the first rainfall event; the time search radius further includes a second time radius, the second time radius being smaller than the first time radius; the spatial search radius further includes a second spatial radius, the second spatial radius being smaller than the first spatial radius; the minimum included rainfall amount further includes a second rainfall amount, the second rainfall amount being smaller than the first rainfall amount; the clustering unit is used to determine rainfall data points other than the first rainfall event among the plurality of rainfall data points as first unclassified points, the first unclassified points including the second rainfall data points, the sum of precipitation of a plurality of nearby rainfall points of the second rainfall data point being less than the first rainfall amount; based on the time point and spatial location of each rainfall data point in the first unclassified points, a plurality of nearby rainfall points of the second rainfall data point are determined from the first unclassified points, the time distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second time radius, and the spatial location distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second spatial radius; in response to the sum of precipitation of the plurality of nearby rainfall points of the second rainfall data point being greater than or equal to the second rainfall amount, the plurality of nearby rainfall points and the second rainfall data point are classified into the second rainfall event.

[0031] In some embodiments, the at least one rainfall event further includes a third rainfall event, the third rainfall event having a smaller time scale and a smaller spatial scale than the second rainfall event; the time search radius further includes a third time radius, the third time radius being smaller than the second time radius; the spatial search radius further includes a third spatial radius, the third spatial radius being smaller than the second spatial radius; the minimum included rainfall amount further includes a third rainfall amount, the third rainfall amount being smaller than the second rainfall amount; the clustering unit is used to determine rainfall data points other than the second rainfall event in the first unclassified points as second unclassified points, the second unclassified points including the first... Three rainfall data points are identified, wherein the sum of precipitation from multiple nearby rainfall points of the third rainfall data point is less than the second rainfall amount; based on the time point and spatial location of each rainfall data point in the second unclassified points, multiple nearby rainfall points of the third rainfall data point are determined from the second unclassified points, wherein the time distance between each nearby rainfall point and the third rainfall data point is less than or equal to the third time radius, and the spatial distance is less than or equal to the third spatial radius; in response to the sum of precipitation from multiple nearby rainfall points of the third rainfall data point being greater than or equal to the third rainfall amount, the multiple nearby rainfall points and the third rainfall data point are classified into the third rainfall event.

[0032] In some embodiments, the apparatus further includes:

[0033] The processing unit is used to compare the precipitation amount of each of the plurality of precipitation data points with a precipitation threshold, and to remove precipitation data points whose daily precipitation amount is less than the precipitation threshold from the plurality of precipitation data points.

[0034] In some embodiments, the apparatus further includes:

[0035] The processing unit is used to obtain the ratio between the time search radius and the spatial search radius as a spatiotemporal scale conversion coefficient; and to multiply the time point of each of the multiple rainfall data points by the spatiotemporal scale conversion coefficient so that the resolution of the multiple rainfall data points in the time dimension is the same as the resolution in the spatial dimension.

[0036] In some embodiments, the apparatus further includes:

[0037] A processing unit is configured to input the at least one rainfall event into a flood catastrophic model and output the expected loss of the at least one rainfall event through the flood catastrophic model.

[0038] Thirdly, a computer device is provided, the computer device comprising: a processor coupled to a memory, the memory storing at least one computer program instruction, the at least one computer program instruction being loaded and executed by the processor to enable the computer device to implement the method described in the first aspect or any optional method of the first aspect.

[0039] Fourthly, a computer-readable storage medium is provided, the storage medium storing at least one instruction that, when executed on a computer, causes the computer to perform the method described in the first aspect or any alternative method of the first aspect.

[0040] Fifthly, a computer program product is provided, the computer program product comprising one or more computer program instructions, which, when loaded and run by a computer, cause the computer to perform the method described in the first aspect or any optional method of the first aspect.

[0041] In a sixth aspect, a chip is provided, the chip including programmable logic circuitry and / or program instructions, which, when the chip is running, are used to implement the method as described in the first aspect or any alternative method of the first aspect.

[0042] The method provided in this embodiment introduces a time search radius and a spatial search radius to cluster rainfall data points. Rainfall data points whose time distance is less than the time search radius, whose spatial location distance is less than the spatial search radius, and whose total precipitation is greater than the minimum precipitation amount are classified as rainfall events. This is equivalent to clustering in both time and spatial dimensions at the same time, while also considering the spatiotemporal continuity of rainfall events. This allows the method to characterize the spatiotemporal range of rainfall, thereby improving the accuracy of identifying rainfall events.

[0043] Furthermore, considering that the different amounts of precipitation at each point in the precipitation data represent the importance of that point, the clustering results are influenced by minimizing the total amount of precipitation, which in turn affects the minimum amount of precipitation included, thus enabling a more accurate identification of the precipitation center. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of a DBSCAN algorithm provided in an embodiment of this application;

[0045] Figure 2 This is a flowchart of a rainfall event identification method provided in an embodiment of this application;

[0046] Figure 3 This is a schematic diagram of a DBSCAN spatiotemporal clustering method provided in an embodiment of this application;

[0047] Figure 4 This is a flowchart of a multi-level clustering method that considers differences in rainfall scale, provided in an embodiment of this application;

[0048] Figure 5 This is a schematic diagram of the structure of a rainfall event identification device provided in an embodiment of this application;

[0049] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0051] The following explains some terms and concepts involved in the embodiments of this application.

[0052] Flood disaster model: This refers to a model used in the insurance industry that uses Monte Carlo simulation to simulate expected losses and losses with different return periods by constructing a long-term set of random events. It is the mainstream method for quantifying catastrophic risks in the insurance field.

[0053] Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is a clustering algorithm based on spatial density. Unlike K-means and hierarchical clustering methods, DBSCAN defines a cluster as the largest set of density-connected points, enabling it to divide regions with sufficiently high density into clusters and discover clusters of arbitrary shapes in noisy spatial databases.

[0054] The following are some definitions in DBSCAN.

[0055] E-neighborhood: The region within a radius of E of a given object is called the E-neighborhood of that object.

[0056] Core object: If the number of sample points in the neighborhood of a given object is greater than or equal to MinPts, then the object is called a core object.

[0057] Direct density reachability: For a sample set D, if a sample point q is within the E neighborhood of p and p is the core object, then object q is directly density reachable from object p.

[0058] Density reachability: For a sample set D, given a series of sample points p1, p2, ..., pn, p = p1, q = pn, if object pi is directly density reachable from pi-1, then object q is density reachable from object p.

[0059] Density connectivity: If there exists a point o in a sample set D such that objects o are density-reachable from both objects p and q, then p and q are density-connected.

[0060] It can be observed that density reachability is a transitive closure of direct density reachability, and this relationship is asymmetric. Density connectivity is a symmetric relationship. The purpose of DBSCAN is to find the largest set of density-connected objects.

[0061] For example, if radius E=3, MinPts=3, point p has points {m,p,p1,p2,o} in its E-neighborhood, point m has points {m,q,p,m1,m2} in its E-neighborhood, point q has points {q,m} in its E-neighborhood, point o has points {o,p,s} in its E-neighborhood, and point s has points {o,s,s1} in its E-neighborhood.

[0062] Then the core objects are p, m, o, s (q is not a core object because the number of points in its E-neighborhood is equal to 2, which is less than MinPts = 3); point m is directly density-reachable from point p because m is within the E-neighborhood of p and p is a core object; point q is density-reachable from point p because point q is directly density-reachable from point m and point m is directly density-reachable from point p; point q and point s are density-connected because point q is density-reachable from point p and s is density-reachable from point p.

[0063] DBScan requires two parameters: the scan radius (eps) and the minimum number of points (minPts).

[0064] In the process of adopting the DBScan algorithm, start with an unvisited point at random and find all nearby points whose distance from it is within eps (including eps). If the number of nearby points ≥ minPts, the current point and its nearby points form a cluster, and the starting point is marked as visited. Then recursively, process all unvisited points within the cluster in the same way to expand the cluster. If the number of nearby points < minPts, the point is temporarily marked as a noise point. If the cluster is fully expanded, that is, all points within the cluster are marked as visited, then use the same algorithm to process unvisited points.

[0065] Exemplarily, please refer to the appendix Figure 1 , appendix Figure 1 is a schematic diagram of a DBSCAN algorithm provided by an embodiment of the present application. As Figure 1 shown, Figure 1 point A in the figure represents a core object, Figure 1 the red points in the figure represent points that have a density-reachable relationship with the core object, Figure 1 the yellow points B and C in the figure represent points within the neighborhood of the core object but the number of nearby points < minPts, Figure 1 the blue point N in the figure represents a point that is not within the neighborhood of the core object and the number of nearby points < minPts, and the blue point N is a noise point.

[0066] In flood catastrophic models, the identification of rainstorm events is crucial, affecting not only the event distribution of the random event set but also significantly influencing the model's computational complexity. In catastrophic models applied to insurance and reinsurance, many insurance clauses operate on events, such as single-event limits and deductibles; the triggering of catastrophic overpayment is directly influenced by the event definition. Furthermore, since most annual rainfall in China does not result in flooding, only a small portion of extreme rainfall causes disasters. Identifying these extreme rainfall events can significantly reduce the simulation duration required by flood models. However, rainfall events vary greatly in temporal and spatial scales, ranging from basin-scale and month-long extreme rainfall events to short-duration heavy rainfall events affecting only a single city for less than a day. Therefore, designing a method that accurately identifies the spatiotemporal correlations of rainstorms while efficiently meeting the requirements of flood catastrophic models for classifying millennia-long rainfall events is essential.

[0067] The classification of rainfall events is commonly used in the meteorological field, but current methods with specific rules can often only be applied to individual stations and cannot take into account the spatial correlation between multiple stations. This makes it difficult to identify whether events classified by different stations belong to the same event. More precise events (such as those defined in the top ten annual disaster events) are mostly classified after the fact, without specific rules, and cannot be classified and defined in advance.

[0068] In the field of catastrophic disaster models, flood events are often modeled using small watersheds as units. This method solves the problem of spatial correlation to some extent, but it also has three problems:

[0069] (1) The average within the computational unit reduces the spatial resolution of rainfall events and weakens the ability to identify small-scale extreme events.

[0070] (2) Rainfall events often cross watersheds and are not distributed along watershed boundaries, making it impossible to identify large-scale precipitation events that cross watersheds.

[0071] (3) Within the same watershed, rainfall is divided according to specific time intervals, making it difficult to consider the events formed by the movement of the rainfall system in time and space. That is, there may be two different events in the same watershed within the same time period, and rainfall events in the same watershed between two time periods may be the same event but are actually divided into two events.

[0072] Compared with the prior art, the embodiments of this application improve the spatial resolution of event recognition and the limitation of watershed boundaries. At the same time, they solve the limitation of dividing and merging time and space separately, and can identify the situation of rainfall events moving in space over time, which is more in line with reality.

[0073] ① Existing methods for classifying rainfall events are often based on a single site, which fails to depict their spatial extent.

[0074] The embodiments of this application take into account the spatiotemporal continuity of rainfall, which can complete the characterization of the spatiotemporal range of rainfall.

[0075] ② The existing definitions of rainfall events mostly determine whether they constitute an event based on their impact, and cannot serve as the basis for event definitions in catastrophic models.

[0076] This application's embodiments directly utilize rainfall data to generate events, without requiring subsequent disaster information.

[0077] The method flow of the embodiments of this application is illustrated below.

[0078] Appendix Figure 2 This is a flowchart illustrating a method for identifying rainfall events provided in an embodiment of this application. (Attached) Figure 2 The method shown includes the following steps S210 to S240.

[0079] Step S210: Obtain multiple rainfall data points, each rainfall data point including the correspondence between time point, spatial location and precipitation amount.

[0080] Step S220: Remove rainfall data points from multiple rainfall data points whose daily precipitation is less than the precipitation threshold.

[0081] For example, the daily precipitation of each rainfall data point in multiple rainfall data points is compared with a precipitation threshold. If the daily precipitation of a rainfall data point is greater than or equal to the precipitation threshold, the data point is retained. If the daily precipitation of a rainfall data point is less than the precipitation threshold, the data point is deleted. Considering that rainfall data points with excessively low daily precipitation are unlikely to form rainfall events with a high risk of flooding, removing rainfall data points with excessively low precipitation before clustering reduces the number of rainfall data points that need to be processed during the clustering process, thereby improving the speed of identifying rainfall events through clustering.

[0082] For example, the precipitation threshold is b millimeters. Considering that points with daily precipitation less than b millimeters are unlikely to cause urban flooding or flash floods, precipitation data points with daily precipitation less than b millimeters are removed during clustering, thereby improving clustering efficiency and more accurately distinguishing the core areas of precipitation events.

[0083] Step S230: Unify the time and spatial scales of multiple rainfall data points.

[0084] Considering that rainfall data points are clustered in three-dimensional space and time, it is necessary to unify the temporal and spatial distances. As an example, the ratio between the temporal search radius and the spatial search radius is obtained as the spatiotemporal scale conversion coefficient; the time point of each rainfall data point in the multiple rainfall data points is multiplied by the spatiotemporal scale conversion coefficient to make the resolution of the multiple rainfall data points in the temporal dimension the same as the resolution in the spatial dimension.

[0085] For example, the resolution of spatial distance is 0.1°, while the resolution of time is day. To unify the resolution of spatial distance and the resolution of time, an equivalent conversion is performed as follows.

[0086] Based on the definitions in DBscan, the conversion is performed by defining the limiting conditions for non-contiguous density between two precipitation events. Temporally, the limiting condition is defined as no daily precipitation exceeding b mm within *a* days for both events; therefore, the temporal search radius is considered to be *a / 2* = *c* days. Spatially, the limiting condition is defined as no daily precipitation exceeding b mm within a distance of *d*° (approximately *e* km); therefore, the spatial search radius is considered to be approximately *d / 2* = *f*°. Under these conditions, *c* days = *f*° in the cluster space. That is, multiplying *days* by *f* / / *c* serves as the coefficient for spatiotemporal scale conversion.

[0087] For example, the number of days between the time point of the rainfall data point and the starting time point is multiplied by the spatiotemporal scale conversion factor. For example, if the time point of the rainfall data point is the 2nd day, then 2 is multiplied by f / / c to obtain the rainfall data point after the time scale conversion.

[0088] Step S240: Based on the time search radius, spatial search radius, and minimum rainfall amount, cluster multiple rainfall data points to obtain at least one rainfall event.

[0089] A rainfall event is like a cluster obtained through clustering. Different rainfall data points in a rainfall event are like different points in a cluster. The temporal distance between different rainfall data points in each rainfall event is less than or equal to the temporal search radius, the spatial distance between different rainfall data points in each rainfall event is less than or equal to the spatial search radius, and the sum of precipitation for all rainfall data points included in each rainfall event is greater than or equal to the minimum included rainfall.

[0090] To distinguish between different rainfall events, we will use "first rainfall event" and "second rainfall event" to describe multiple different rainfall events. Similarly, to distinguish between different rainfall data points, we will use "first rainfall data point" and "second rainfall data point" to describe multiple different rainfall data points.

[0091] To facilitate understanding, the following explanation uses the example of searching for clustering events based on the first rainfall data point. For instance, multiple rainfall data points include the first rainfall data point, the rainfall event corresponding to the first rainfall data point is the first rainfall event, the temporal search radius includes the first temporal radius, the spatial search radius includes the first spatial radius, and the minimum included rainfall amount includes the first rainfall amount. During the clustering process for the first rainfall data point, based on the time point and spatial location of each rainfall data point, multiple nearby rainfall points are determined from the multiple rainfall data points. The temporal distance between each nearby rainfall point and the first rainfall data point is less than or equal to the first temporal radius, and the spatial distance is less than or equal to the first spatial radius. In response to the sum of the rainfall amounts of the multiple nearby rainfall points being greater than or equal to the first rainfall amount, the multiple nearby rainfall points and the first rainfall data point are assigned to the first rainfall event, and the rainfall center of the first rainfall event is determined as the first rainfall data point.

[0092] For example, iterating through multiple rainfall data points, except for the first rainfall data point, for each rainfall data point i, the temporal distance between the first and second rainfall data points is determined based on the time points of both the first and second rainfall data points. The spatial distance between the first and second rainfall data points is determined based on their spatial locations. The temporal distance between the first and second rainfall data points is compared to a first time radius, and the spatial distance between them is compared to a first spatial radius. If both the temporal distance and spatial distance are less than or equal to the first time radius, then rainfall data point i is identified as a nearby rainfall point. Finally, the precipitation amounts of all nearby rainfall points are summed to obtain the total precipitation amount of all nearby rainfall points of the first rainfall data point. The sum of precipitation from multiple nearby rainfall points is compared with the first rainfall amount. If the sum of precipitation from multiple nearby rainfall points is greater than or equal to the first rainfall amount, the multiple nearby rainfall points and the first rainfall data point are classified as the first rainfall event. If the sum of precipitation from multiple nearby rainfall points is less than the first rainfall amount, the first rainfall data point is identified as a noise point.

[0093] The aforementioned first rainfall event is equivalent to a set of rainfall data points that are density-reachable from the first rainfall data point, with the first rainfall data point as the core object. The first rainfall event can be represented as a sphere, with the first rainfall data point at its center. The spatial scale of the first rainfall event can be, for example, a watershed scale. For instance, the first rainfall event could be a watershed-scale, month-long extreme rainfall event.

[0094] As an example, density-based clustering (DBSCAN) is used to cluster multiple rainfall data points. In the DBSCAN algorithm, the scan radius (eps) is divided into two parameters: the temporal search radius and the spatial search radius. The minimum rainfall amount (minPts) is the last parameter that needs to be set. For example, with a temporal search radius of c days and a spatial search radius of f degrees, each search radius contains approximately N points.

[0095] By implementing the above method, and defining rainfall events simultaneously from both temporal and spatial dimensions, the spatial accuracy of event segmentation is improved compared to traditional methods, and the spatiotemporal correlation of precipitation events is more fully considered. Furthermore, the density-based clustering method (DBSCAN), compared to common partitioning clustering algorithms (K-means, Gaussian mixture, spectral clustering) and hierarchical clustering algorithms, has advantages such as clearly defined physical meanings of parameters, no need to predefine the number of clusters, and the cluster shape can adapt to various forms of rainfall bands.

[0096] Related technologies typically cluster rainfall events along a fixed dimension, such as spatial or temporal dimensions. However, the method provided in this embodiment introduces both temporal and spatial search radii to cluster rainfall data points. Rainfall data points whose temporal distance is less than the temporal search radius, whose spatial distance is less than the spatial search radius, and whose total precipitation is greater than the minimum included rainfall amount are classified as rainfall events. This is equivalent to simultaneous clustering along both temporal and spatial dimensions, while also considering the spatiotemporal continuity of rainfall events. This allows for a better depiction of the spatiotemporal range of rainfall, thus improving the accuracy of rainfall event identification.

[0097] Furthermore, considering that the different amounts of precipitation at each point in the precipitation data represent the importance of that point, the clustering results are influenced by minimizing the total amount of precipitation, which in turn affects the minimum amount of precipitation included, thus enabling a more accurate identification of the precipitation center.

[0098] Specifically, during clustering, the DBSCAN algorithm does not assign a value to each point itself, but only records the number of points. It discovers clusters based on the relationship between the number of points and the minimum Pts (minPts). In the scenario of rainfall event identification, the rainfall amount varies for each rainfall data point. By using the rainfall amount as the weight of each rainfall data point—for example, whenever a nearby point of the first rainfall data point is found, the rainfall amount of that nearby point is accumulated—the algorithm determines whether the first rainfall data point corresponds to a cluster (rainfall event) based on the relationship between the accumulated total rainfall amount and the minimum contained rainfall amount. This improves the accuracy of rainfall event identification, for example, helping to identify whether the first rainfall data point is the rainfall center of the first rainfall event.

[0099] The clustering algorithm described above helps to better segment rainfall events with large spatiotemporal scales. To more accurately capture rainfall events with smaller spatiotemporal scales, this application also introduces a secondary clustering process. The second clustering uses a smaller search radius and a stricter minPts condition than the first clustering to search for heavy rainfall events with medium spatiotemporal scales.

[0100] For example, at least one rainfall event also includes a second rainfall event, which has a smaller time scale and a smaller spatial scale than the first rainfall event. The time search radius also includes a second time radius, which is smaller than the first time radius. The spatial search radius also includes a second spatial radius, which is smaller than the first spatial radius. The minimum included rainfall amount also includes a second rainfall amount, which is smaller than the first rainfall amount.

[0101] For example, the second time radius is the product of the first time radius and the first ratio; the second spatial radius is the product of the first spatial radius and the second ratio; and the second rainfall is the product of the first rainfall and the third ratio. All three ratios are less than 1. Alternatively, the second time radius can be the difference between the first time radius and the first difference; the second spatial radius can be the difference between the first spatial radius and the second difference; and the second rainfall can be the difference between the first rainfall and the third difference. All three differences are greater than 0. As an example, the second time radius could be half the first time radius, the third time radius could be half the second time radius, and the second rainfall could be half the first rainfall.

[0102] Rainfall data points other than the first rainfall event are identified as first unclassified points. The first unclassified points include the second rainfall data point. The sum of precipitation from multiple nearby rainfall points of the second rainfall data point is less than the first rainfall amount. Based on the time point and spatial location of each rainfall data point in the first unclassified points, multiple nearby rainfall points of the second rainfall data point are identified from the first unclassified points. The time distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second time radius, and the spatial distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second spatial radius. In response to the sum of precipitation from multiple nearby rainfall points of the second rainfall data point being greater than or equal to the second rainfall amount, the multiple nearby rainfall points and the second rainfall data point are classified into the second rainfall event, and the rainfall center of the second rainfall event is determined as the second rainfall data point.

[0103] For example, please refer to the appendix. Figure 3 , attached Figure 3 This is a schematic diagram of a DBSCAN spatiotemporal clustering method provided in an embodiment of this application. Figure 3The blue dots represent rainfall data points from the first rainfall event, the red dots represent rainfall data points from the second rainfall event, and the black dots represent noise points (unclassified points). For example... Figure 3 As shown, the radius of the first rainfall event in the time dimension (Z-axis) (first time radius) is greater than the radius of the second rainfall event in the time dimension (second time radius), and the radius of the first rainfall event in the longitude and latitude dimensions (first spatial radius) is greater than the radius of the second rainfall event in the longitude and latitude dimensions (second spatial radius).

[0104] Considering that using only one clustering method to obtain clustered events makes it difficult to set the minimum rainfall coverage, setting it too high can lead to difficulty in finding consecutive rainfall events, resulting in scattered rainfall events. Setting it too low can cause all rainfall data points to be connected, making segmentation difficult. Using multiple clustering methods helps reduce the difficulty of setting the minimum rainfall coverage and identifies various types of rainfall events.

[0105] Optionally, in order to more accurately capture rainfall events with smaller spatiotemporal scales based on the secondary clustering, embodiments of this application also introduce a tertiary clustering process. The third clustering uses a smaller search radius and a more stringent minPts condition than the second clustering to search for short-duration heavy rainfall events with smaller spatiotemporal scales.

[0106] For example, at least one rainfall event also includes a third rainfall event, which has a smaller time scale and a smaller spatial scale than the second rainfall event. The time search radius also includes a third time radius, which is smaller than the second time radius. The spatial search radius also includes a third spatial radius, which is smaller than the second spatial radius. The minimum included rainfall also includes a third rainfall amount, which is smaller than the second rainfall amount.

[0107] For example, the third time radius is the product of the second time radius and the fourth ratio; the third spatial radius is the product of the second spatial radius and the fifth ratio; and the third rainfall is the product of the second rainfall and the sixth ratio. The fourth, fifth, and sixth ratios are all less than 1. Alternatively, the third time radius can be the difference between the second and fourth time radius differences; the third spatial radius can be the difference between the second and fifth spatial radius differences; and the third rainfall can be the difference between the second and sixth rainfall differences. The first, second, and third differences are all greater than 0. As an example, the third time radius is half of the second time radius, and the third rainfall is half of the second rainfall.

[0108] In the process of classifying the third rainfall event, the rainfall data points in the first unclassified points other than the second rainfall event are identified as the second unclassified points. The second unclassified points include the third rainfall data point. The sum of precipitation from multiple nearby rainfall points of the third rainfall data point is less than the second rainfall amount. Based on the time point and spatial location of each rainfall data point in the second unclassified points, multiple nearby rainfall points of the third rainfall data point are identified from the second unclassified points. The time distance between each nearby rainfall point and the third rainfall data point is less than or equal to the third time radius, and the spatial distance is less than or equal to the third spatial radius. In response to the sum of precipitation from multiple nearby rainfall points of the third rainfall data point being greater than or equal to the third rainfall amount, the multiple nearby rainfall points and the third rainfall data point are classified into the third rainfall event, and the rainfall center of the third rainfall event is identified as the third rainfall data point.

[0109] Since the three time radii (first, second, and third) have a decreasing numerical relationship, the three spatial radii (first, second, and third) have a decreasing numerical relationship, and the three minimum contained rainfall amounts (first, second, and third) have a decreasing numerical relationship, it is possible to sequentially classify rainfall events into large-scale, medium-scale, and small-scale events.

[0110] For a concrete example, please refer to the appendix. Figure 4 , attached Figure 4 This is a flowchart illustrating a multi-level clustering method considering differences in rainfall scales, as provided in an embodiment of this application. (See attached...) Figure 4 As shown, firstly, rainfall data points with a rainfall rate greater than 10 mm / h are selected. These data points are then clustered for the first time radius, the first spatial radius, and the first rainfall amount to obtain watershed-scale extreme rainfall events and noise points (first unclassified points). For the noise points (first unclassified points) in the first cluster, a second cluster is performed based on the second time radius, the second spatial radius, and the second rainfall amount to obtain large-scale rainfall events and noise points (second unclassified points). Finally, for the noise points (second unclassified points) in the second cluster, a third cluster is performed based on the third time radius, the third spatial radius, and the third rainfall amount to obtain localized heavy rainfall events and noise points (third unclassified points).

[0111] Multi-level clustering allows for more targeted parameter settings compared to single-level clustering. For example, the search radii used in the third, second, and first clusters are different, which helps to identify consecutive rainfall events of sufficient intensity.

[0112] In some implementations, after clustering to obtain at least one rainfall event, the at least one rainfall event is input into a flood catastrophic model, which outputs the expected loss for the at least one rainfall event. For example, the flood catastrophic model outputs the limit and deductible for each of the at least one rainfall event. Optionally, after outputting the expected loss for the at least one rainfall event based on the flood catastrophic model, the expected loss for the rainfall event is compared with a loss threshold. If the expected loss for the rainfall event is greater than or equal to the loss threshold, reinsurance payment is initiated.

[0113] Figure 5 This is a schematic diagram of the structure of a rainfall event identification device 500 provided in an embodiment of this application. The device 500 includes:

[0114] The acquisition unit 510 is used to acquire multiple rainfall data points, each rainfall data point including the correspondence between time point, spatial location and precipitation amount;

[0115] Clustering unit 520 is used to cluster the multiple rainfall data points based on the time search radius, the spatial search radius, and the minimum included rainfall amount to obtain at least one rainfall event. The time distance between different rainfall data points in each rainfall event is less than or equal to the time search radius, the spatial distance between different rainfall data points in each rainfall event is less than or equal to the spatial search radius, and the total precipitation of all rainfall data points included in each rainfall event is greater than or equal to the minimum included rainfall amount.

[0116] In some implementations, the plurality of rainfall data points includes a first rainfall data point, the at least one rainfall event includes a first rainfall event, the time search radius includes a first time radius, the spatial search radius includes a first spatial radius, the minimum included rainfall amount includes a first rainfall amount, and the clustering unit 520 is used to determine a plurality of nearby rainfall points of the first rainfall data point from the plurality of rainfall data points based on the time point and spatial location of each rainfall data point, wherein the time distance between each nearby rainfall point and the first rainfall data point is less than or equal to the first time radius, and the spatial location distance is less than or equal to the first spatial radius; in response to the sum of the rainfall amounts of the plurality of nearby rainfall points of the first rainfall data point being greater than or equal to the first rainfall amount, the plurality of nearby rainfall points and the first rainfall data point are assigned to the first rainfall event.

[0117] In some embodiments, the at least one rainfall event further includes a second rainfall event, the second rainfall event having a smaller time scale and a smaller spatial scale than the first rainfall event; the time search radius further includes a second time radius, the second time radius being smaller than the first time radius; the spatial search radius further includes a second spatial radius, the second spatial radius being smaller than the first spatial radius; the minimum included rainfall amount further includes a second rainfall amount, the second rainfall amount being smaller than the first rainfall amount; the clustering unit 520 is used to determine rainfall data points other than the first rainfall event among the plurality of rainfall data points as first unclassified points, the first unclassified points including the second rainfall data points, the sum of precipitation of a plurality of nearby rainfall points of the second rainfall data point being less than the first rainfall amount; based on the time point and spatial location of each rainfall data point in the first unclassified points, a plurality of nearby rainfall points of the second rainfall data point are determined from the first unclassified points, the time distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second time radius, and the spatial location distance is less than or equal to the second spatial radius; in response to the sum of precipitation of the plurality of nearby rainfall points of the second rainfall data point being greater than or equal to the second rainfall amount, the plurality of nearby rainfall points and the second rainfall data point are classified into the second rainfall event.

[0118] In some embodiments, the at least one rainfall event further includes a third rainfall event, which has a smaller time scale and a smaller spatial scale than the second rainfall event. The time search radius further includes a third time radius, which is smaller than the second time radius. The spatial search radius further includes a third spatial radius, which is smaller than the second spatial radius. The minimum included rainfall amount also includes a third rainfall amount, which is smaller than the second rainfall amount. The clustering unit 520 is used to identify rainfall data points other than the second rainfall event in the first unclassified points as second unclassified points. The second unclassified points include... The system includes a third rainfall data point, wherein the sum of precipitation from multiple nearby rainfall points of the third rainfall data point is less than the second rainfall amount; based on the time point and spatial location of each rainfall data point in the second unclassified points, multiple nearby rainfall points of the third rainfall data point are determined from the second unclassified points, wherein the time distance between each nearby rainfall point and the third rainfall data point is less than or equal to the third time radius, and the spatial distance is less than or equal to the third spatial radius; in response to the sum of precipitation from multiple nearby rainfall points of the third rainfall data point being greater than or equal to the third rainfall amount, the multiple nearby rainfall points and the third rainfall data point are classified into the third rainfall event.

[0119] In some embodiments, the device 500 further includes:

[0120] The processing unit 530 is used to compare the precipitation amount of each of the plurality of precipitation data points with the precipitation threshold, and to remove the precipitation data points whose daily precipitation amount is less than the precipitation threshold from the plurality of precipitation data points.

[0121] In some embodiments, the device 500 further includes:

[0122] The processing unit 530 is used to obtain the ratio between the time search radius and the spatial search radius as a spatiotemporal scale conversion coefficient; and to multiply the time point of each of the multiple rainfall data points by the spatiotemporal scale conversion coefficient so that the resolution of the multiple rainfall data points in the time dimension is the same as the resolution in the spatial dimension.

[0123] In some embodiments, the device 500 further includes:

[0124] Processing unit 530 is used to input the at least one rainfall event into a flood disaster model and output the expected loss of the at least one rainfall event through the flood disaster model.

[0125] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. The computer device 600 includes: a processor 601, which is coupled to a memory 602. The memory 602 stores at least one computer program instruction, which is loaded and executed by the processor 601 to enable the computer device 600 to perform... Figure 2 The method provided in the embodiments.

[0126] In some embodiments, a computer-readable storage medium is also provided, storing at least one instruction that, when executed on a computer, causes the computer to perform the above-described... Figure 2 The method provided in the embodiments.

[0127] In some embodiments, a computer program product is also provided, comprising one or more computer program instructions that, when loaded and executed by a computer, cause the computer to perform the aforementioned... Figure 2 The method provided in the embodiments.

[0128] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0129] A references B, which means that A is the same as B or A is a simple variation of B.

[0130] The terms "first" and "second," etc., used in the specification and claims of this application are used to distinguish different objects, not to describe a specific order of objects, and should not be construed as indicating or implying relative importance. For example, "first rainfall event" and "second rainfall event" are used to distinguish different rainfall events, not to describe a specific order of rainfall events, and should not be construed as the first rainfall event being more important than the second rainfall event.

[0131] The information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in the embodiments of this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions. For example, the rainfall data points involved in this application were all obtained with full authorization.

[0132] In this application, unless otherwise stated, "at least one" means one or more, and "multiple" means two or more. For example, multiple rainfall data points refer to two or more rainfall data points.

[0133] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, they can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0134] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for identifying rainfall events, characterized in that, The method includes: Acquire multiple rainfall data points, each of which includes the correspondence between time point, spatial location, and precipitation amount; Based on the time search radius, spatial search radius, and minimum contained rainfall, the multiple rainfall data points are clustered to obtain at least one rainfall event. The time distance between different rainfall data points in each rainfall event is less than or equal to the time search radius, the spatial distance between different rainfall data points in each rainfall event is less than or equal to the spatial search radius, and the total precipitation of all rainfall data points included in each rainfall event is greater than or equal to the minimum contained rainfall.

2. The method according to claim 1, characterized in that, The plurality of rainfall data points includes a first rainfall data point, the at least one rainfall event includes a first rainfall event, the time search radius includes a first time radius, the spatial search radius includes a first spatial radius, the minimum included rainfall amount includes a first rainfall amount, and the clustering of the plurality of rainfall data points based on the time search radius, the spatial search radius, and the minimum included rainfall amount includes: Based on the time point and spatial location of each rainfall data point, multiple nearby rainfall points are determined from the multiple rainfall data points. The time distance between each nearby rainfall point and the first rainfall data point is less than or equal to the first time radius, and the spatial location distance is less than or equal to the first spatial radius. In response to the fact that the sum of precipitation at multiple nearby rainfall points of the first rainfall data point is greater than or equal to the first rainfall amount, the multiple nearby rainfall points and the first rainfall data point are classified into the first rainfall event.

3. The method according to claim 2, characterized in that, The at least one rainfall event further includes a second rainfall event, which has a smaller time scale and a smaller spatial scale than the first rainfall event. The time search radius further includes a second time radius, which is smaller than the first time radius. The spatial search radius further includes a second spatial radius, which is smaller than the first spatial radius. The minimum included rainfall amount further includes a second rainfall amount, which is smaller than the first rainfall amount. The clustering of the plurality of rainfall data points based on the time search radius, spatial search radius, and minimum included rainfall amount includes: Among the multiple rainfall data points, the rainfall data points other than the first rainfall event are identified as the first unclassified points. The first unclassified points include the second rainfall data points, and the sum of the precipitation of multiple nearby rainfall points of the second rainfall data point is less than the first rainfall amount. Based on the time point and spatial location of each rainfall data point in the first unclassified points, multiple nearby rainfall points are determined from the first unclassified points. The time distance between each nearby rainfall point and the second rainfall data point is less than or equal to the second time radius, and the spatial location distance is less than or equal to the second spatial radius. In response to the fact that the sum of precipitation at multiple nearby rainfall points of the second rainfall data point is greater than or equal to the second rainfall amount, the multiple nearby rainfall points and the second rainfall data point are classified into the second rainfall event.

4. The method according to claim 3, characterized in that, The at least one rainfall event further includes a third rainfall event, which has a smaller time scale and a smaller spatial scale than the second rainfall event. The time search radius further includes a third time radius, which is smaller than the second time radius. The spatial search radius further includes a third spatial radius, which is smaller than the second spatial radius. The minimum included rainfall amount also includes a third rainfall amount, which is smaller than the second rainfall amount. The clustering of the multiple rainfall data points based on the time search radius, spatial search radius, and minimum included rainfall amount includes: Rainfall data points other than the second rainfall event in the first unclassified points are identified as second unclassified points. The second unclassified points include third rainfall data points, and the sum of precipitation of multiple nearby rainfall points of the third rainfall data point is less than the second rainfall amount. Based on the time point and spatial location of each rainfall data point in the second unclassified points, multiple nearby rainfall points are determined from the second unclassified points. The time distance between each nearby rainfall point and the third rainfall data point is less than or equal to the third time radius, and the spatial location distance is less than or equal to the third spatial radius. In response to the fact that the sum of precipitation at multiple nearby rainfall points of the third rainfall data point is greater than or equal to the third rainfall amount, the multiple nearby rainfall points and the third rainfall data point are classified into the third rainfall event.

5. The method according to claim 1, characterized in that, Before clustering the multiple rainfall data points based on the time search radius, spatial search radius, and minimum contained rainfall, the method further includes: The precipitation amount of each of the multiple precipitation data points is compared with the precipitation threshold, and the precipitation data points with daily precipitation less than the precipitation threshold are removed from the multiple precipitation data points.

6. The method according to claim 1, characterized in that, Before clustering the multiple rainfall data points based on the time search radius, spatial search radius, and minimum contained rainfall, the method further includes: The ratio between the temporal search radius and the spatial search radius is obtained and used as the spatiotemporal scale conversion coefficient. The time point of each of the multiple rainfall data points is multiplied by the spatiotemporal scale conversion coefficient so that the resolution of the multiple rainfall data points in the time dimension is the same as the resolution in the spatial dimension.

7. The method according to claim 1, characterized in that, After clustering the multiple rainfall data points based on the time search radius, spatial search radius, and minimum contained rainfall to obtain at least one rainfall event, the method further includes: The at least one rainfall event is input into the flood catastrophic model, and the expected loss of the at least one rainfall event is output by the flood catastrophic model.

8. A device for identifying rainfall events, characterized in that, The device includes: The acquisition unit is used to acquire multiple rainfall data points, each of which includes the correspondence between time point, spatial location, and precipitation amount. A clustering unit is used to cluster the multiple rainfall data points based on the time search radius, the spatial search radius, and the minimum included rainfall amount to obtain at least one rainfall event. The time distance between different rainfall data points in each rainfall event is less than or equal to the time search radius, the spatial distance between different rainfall data points in each rainfall event is less than or equal to the spatial search radius, and the total precipitation of all rainfall data points included in each rainfall event is greater than or equal to the minimum included rainfall amount.

9. A computer device, characterized in that, The computer device includes: a processor coupled to a memory, the memory storing at least one computer program instruction, the at least one computer program instruction being loaded and executed by the processor to enable the computer device to implement the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, which, when executed on a computer, causes the computer to perform the method as described in any one of claims 1-7.

Images