A grid-based heavy rain forecasting method

Abstract

The application discloses a lattice rainstorm prediction method, and the method adopts circulation background field, U\V prediction field, rainfall prediction field, model adjustment field and data of a ground meteorological station, and constructs a training sample set; an XGBoost method is used to establish a prediction model based on a mapping relationship of a circulation background field of an EC model and prediction products and future 0-24h precipitation; the training sample set is trained through the prediction model, so that the prediction model can perform medium and short-term precipitation prediction. The prediction ability (rainstorm TS) of the method for precipitation above rainstorm is improved by more than 17% compared with the EC model prediction, the prediction precision of medium and short-term precipitation can be effectively improved, and the method has a good application prospect and provides more accurate prediction services for disaster prevention and reduction.

Description

Technical Field

[0001] This invention relates to the field of weather forecasting technology, and in particular to a gridded method for forecasting heavy rainfall. Background Technology

[0002] Against the backdrop of global warming, severe convective weather events triggering flash floods, urban flooding, and geological disasters have become increasingly common. Consequently, the demand for meteorological services from all sectors is growing, and the requirements are becoming more stringent. Simultaneously, with the increasing use of high spatiotemporal resolution numerical weather prediction models in operational forecasting, there is a growing expectation for meteorological departments to provide more accurate weather forecasts. However, due to factors such as initial field errors, output errors, and model stability, directly using the output results of numerical weather prediction models for forecasting results results in relatively low accuracy. The forecast level still falls significantly short of the actual needs for disaster prevention and mitigation in the event of major meteorological disasters, impacting meteorological disaster prevention and mitigation decision-making. Under these circumstances, meteorological departments urgently need to address the issue of correcting numerical model forecasts to improve their forecast accuracy.

[0003] Currently, numerical weather prediction products include multiple global numerical weather model prediction products from both domestic and international sources, as well as several regional model numerical weather prediction products from China. Abundant multi-source meteorological data provides us with more forecasting information for heavy precipitation correction forecasts. However, the massive amount of data sources also brings significant challenges to the correction forecasting work. How to effectively utilize numerical weather prediction products from different sources with different spatiotemporal resolutions and structures is a crucial aspect of tackling key technical challenges in objective heavy precipitation forecasting. Currently, operational numerical weather model precipitation correction forecasting methods mainly include two types: multi-model fusion techniques and statistical forecasting methods based on linear and nonlinear models. Their drawbacks are as follows:

[0004] (1) The gridded forecasting method based on multi-model fusion mainly uses a weighting method to weight each model or forecast member. This method only considers the previous forecast performance of each model member's precipitation forecast field, without considering the current circulation pattern field.

[0005] (2) Rainfall forecasting methods based on linear and nonlinear statistical methods. Theoretically, this method can incorporate various models of rainfall forecast fields and various physical quantity fields. However, based on the current literature, no scholar has used a large circulation background as a factor in the model as the input of the model. Secondly, since this method often models and forecasts based on stations, it cannot meet the current gridded requirements of precipitation forecasting due to the large number of grid points and the lack of corresponding real data on the grid points. Summary of the Invention

[0006] This invention provides a gridded heavy rainfall forecasting method to at least address the technical problem of over-reliance on multi-model fusion in current gridded forecasting techniques.

[0007] According to one aspect of the present invention, a gridded heavy rainfall forecasting method is provided, comprising:

[0008] Data from the background circulation field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations were used to construct a training sample set;

[0009] A forecasting model based on the EC model's background circulation field and its forecast products, and the mapping relationship between these fields and future 0–24 h precipitation, was established using the XGBoost method.

[0010] The training sample set is used to train the forecast model, enabling the forecast model to perform short- and medium-term precipitation forecasts.

[0011] Optionally, a training sample set is constructed using data from the circulation background field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations, specifically including:

[0012] Acquire the circulation background field, U / V forecast field of the set model, rainfall forecast field, model adjustment field, and real-time rainfall data from ground meteorological stations for the T-time forecast product.

[0013] Preprocessing was performed on the data from the background circulation field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological station data, respectively.

[0014] The preprocessed data is divided into multiple sample groups, each sample group xy includes several forecast factors x and a forecast object y;

[0015] By combining the aforementioned sample groups, a sample matrix is ​​constructed using spatiotemporal stacking and sample reconstruction techniques, thereby obtaining the training sample set.

[0016] Optionally, it also includes: based on the forecast model with the mapping relationship, using the rainfall forecast field of the CMA-SH9 model to perform sparse correction on the forecast results of the forecast model.

[0017] Optionally, the circulation background field is an EC model rainfall forecast field, and the UV forecast field adopts an EC model UV forecast field at 500hPa, 850hPa, and 925hPa.

[0018] Optionally, preprocessing the data of the circulation background field includes: determining whether there are troughs and shears in the 500hPa / 850hPa wind fields at the upper pressure level to be monitored within a set future time period; if both exist, it is recorded as 1, otherwise it is recorded as 0.

[0019] Optionally, preprocessing the U / V forecast field data at the start time T includes: for the EC wind field of the numerical model, calculating the wind direction and wind speed for each t3 hour when the start time of the numerical model is T-12h and the forecast lead time is t1 to t2h, and interpolating the calculation results to the station, where t1 is 12h, t2 is 36h, and t3 is 3h.

[0020] Optionally, the preprocessing of the rainfall forecast field data includes: preprocessing the rainfall forecast field data with a start time of T includes: calculating the cumulative precipitation for the next t1 to t2 hours based on the rainfall forecast fields of the EC model and CMA-SH9 model with a start time of T-12 hours, and interpolating the calculation results to the stations, denoted as ec24 and sh24 respectively; calculating the cumulative precipitation for the next 24-48 hours based on the rainfall forecast field data of the EC model with a start time of T-24 hours, and interpolating the calculation results to the stations, denoted as ec48 respectively.

[0021] According to another aspect of the present invention, a gridded rainstorm forecasting system is also provided, comprising:

[0022] The training sample module is used to construct a training sample set by using data from the circulation background field, U / V forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations.

[0023] The forecast model module is used to establish a forecast model based on the EC model circulation background field and its forecast products and the mapping relationship between the future 0-24h precipitation using the XGBoost method. The training sample set is trained through the forecast model so that the forecast model can make medium- and short-term precipitation forecasts.

[0024] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to perform the gridded rainstorm forecasting method described in any one of the above embodiments.

[0025] According to another aspect of the present invention, a processor is also provided, the processor being configured to run a program, wherein the program, when running, executes the gridded rainstorm forecasting method described in any of the preceding embodiments.

[0026] Compared with existing technologies, the present invention has the following advantages:

[0027] 1. In this embodiment of the invention, the method uses data from the circulation background field, U / V forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations to construct a training sample set. An XGBoost method is used to establish a forecasting model based on the mapping relationship between the EC model's circulation background field and its forecast products and future 0-24 hour precipitation. The training sample set is then used to train the forecasting model, enabling it to perform short- to medium-term precipitation forecasts. This method improves the prediction capability for heavy rainfall (TS) by more than 17% compared to the EC model forecast, effectively enhancing the accuracy of short- to medium-term precipitation forecasts. It has promising application prospects and provides more accurate forecasting services for disaster prevention and mitigation.

[0028] 2. This invention can use longer historical data (circulation background field, U / V forecast field, rainfall forecast field, model adjustment field and data from ground meteorological stations) for modeling, so that the model can better obtain the forecast performance of numerical models during training;

[0029] 3. Compared with existing numerical model interpretations based on statistical methods, this invention incorporates the circulation background field into the forecast factors, enabling the forecast model to learn the characteristics of precipitation systems in upper-level troughs and shear lines. At the same time, compared with traditional statistical forecasting methods that can only perform single-station modeling forecasts, the method of this invention can achieve gridded forecasting of station modeling. Attached Figure Description

[0030] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only one embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart of a gridded rainstorm forecasting method according to an embodiment of the present invention. Detailed Implementation

[0032] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0033] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0034] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0035] Example 1

[0036] According to an embodiment of the present invention, a gridded rainstorm forecasting method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0037] like Figure 1 This is a flowchart of a gridded rainstorm forecasting method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0038] Step S10: Use data from the circulation background field, U / V forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations to construct a training sample set;

[0039] As an optional embodiment, step S10 specifically includes:

[0040] Step S101: Obtain the following data: the start time of the forecast is T, the relevant circulation background field, the U / V forecast field of the set model, the rainfall forecast field, the model adjustment field, and the actual rainfall data from the ground meteorological station.

[0041] Specifically, the circulation background field is an upper-level trough and shear at 500hPa and 850hPa, and the U / V forecast field adopts the U / V forecast field of EC model at 500hPa, 850hPa and 925hPa. The rainfall forecast field adopts the rainfall forecast fields of EC model and CMA-SH9 model as correction references.

[0042] Step S102: Preprocess the data from the background circulation field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological station data, respectively. Specifically, this includes:

[0043] Step S1021: Preprocessing the data of the circulation background field includes: within a set future time period, determining whether there are troughs and shears in the 500hPa and 850hPa wind fields at the upper atmosphere that need to be monitored. If they exist simultaneously, it is recorded as 1; otherwise, it is recorded as 0.

[0044] Specifically, in the EC model, during the forecast period of 15 to 36 hours from T-12h, it is determined whether there is a trough in the 500hPa upper-level wind field and whether there is shear in the 850hPa and 925hPa upper-level wind fields. If there is, the result is 1; otherwise, the result is 0.

[0045] Step S1022: Preprocessing the data of the U / V forecast field includes: For forecasts with a start time of T (the time when forecasters produce 08:00 and 20:00 forecast products), calculating the wind direction and wind speed for each t3 hour when the EC model starts at T-12h and the forecast lead time is t1 to t2h, and interpolating the calculation results to the station, which are denoted as W respectively; where t1 is 12h, t2 is 36h, and t3 is 3h.

[0046] Specifically, the wind speed and direction at the station are calculated: using the 500hPa, 850hPa, and 925hPa EC model wind fields reported from T-12h, the wind direction and wind speed are calculated every 3 hours for a forecast lead time of 12 to 36 hours, and then interpolated (3rd order polynomial interpolation, the same below) to the station, denoted as W.

[0047] Step S1023: Preprocessing the rainfall forecast field data with a start time of T (the time when forecasters produce the 08:00 and 20:00 forecast products, the same below) includes: calculating the cumulative precipitation for the next 12 to 36 hours based on the rainfall forecast field data of the EC model and CMA-SH9 model with a start time of T-12 hours, and interpolating the calculation results to the stations, denoted as ec24 and sh24 respectively; calculating the cumulative precipitation for the next 24 to 48 hours based on the rainfall forecast field data of the EC model with a start time of T-24 hours, and interpolating the calculation results to the stations, denoted as ec48 respectively.

[0048] Step S1024, Preprocessing the data of the model adjustment field includes: subtracting the rainfall forecast field data ec48 of the EC model with a reporting time of T-24h from the rainfall forecast field data ec24 of the EC model with a reporting time of T-12h, and denoted as wc.

[0049] Step S103: The preprocessed data is divided into multiple sample groups. Each sample group xy includes several prediction factors x and one prediction object y, that is:

[0050] f={cx1,qb1,cx2,qb2,ec24,ec48,fs850,fx850,fs925,fx925,wc,lat,lon,y}

[0051] In the above formula, cx1 and qb1 are indicators of whether there is a trough shear line in the 12-36h (3h increments) at 500hPa, totaling 2 factors; cx2 and qb2 are indicators of whether there is a trough shear line in the 12-36h (3h increments) at 850hPa, totaling 2 factors; ec24, ec48, and wc are the cumulative precipitation forecasts for the same forecast period at the time of T-12h and T-24h respectively, and their differences, totaling 3 factors; fs850 and fx850 are the values ​​of precipitation at 850hPa wind speeds. For the field (EC model product reported from T-12h), wind speed and direction were calculated for the forecast fields from 12 to 36h in 3-hour increments, totaling 16 factors; fs925 and fx925 are wind fields (EC model product reported from T), wind speed and direction were calculated for the forecast fields from 15 to 36h in 3-hour increments, totaling 16 factors; (16 factors added for 500hPa); lat and lon are the latitude and longitude of the station where the sample group is located, respectively, totaling 2 factors; the total number of factors is 41. y represents the meteorological observation station, and for the reporting time T, the cumulative precipitation of the station in the 24 hours before T+24h.

[0052] Step S104: Combine the sample group and construct a sample matrix using spatiotemporal stacking and sample reconstruction techniques to obtain the training sample set.

[0053] As an optional embodiment, step S104 includes:

[0054] The sample groups are stacked sequentially according to time and space. That is, for each time sample t (t=1,2,...,n), the spatial station corresponding to that time point is first processed, and the data matrix (XY) is generated. t Stack them together, and further, combine all the data matrices (XY). t Stacking the samples over time t yields the original modeling sample dataset XYdata:

[0055]

[0056] in:

[0057]

[0058] m is the total number of stations in the study area, f t j Let t be a set of samples from the j-th (j≤m) station under time sample t (t≤n).

[0059] Based on the sample distribution ratio, a new training sample set is selected from the XYdata sample set as the training sample set for the model, that is, the training sample set is reconstructed, thus obtaining the training sample set.

[0060] Step S20: Use the XGBoost method to establish a forecast model based on the EC model circulation background field and its forecast products and the mapping relationship between the future 0-24h precipitation; where the forecast products are the forecast conclusions.

[0061] Step S30: Train the forecast model using the training sample set so that the forecast model can perform short- and medium-term precipitation forecasts.

[0062] As an optional embodiment, the gridded heavy rainfall forecasting method of the present invention further includes: based on the forecasting model with a mapping relationship, using the rainfall forecast field of the CMA-SH9 model to perform sparse correction on the forecasting results of the forecasting model.

[0063] Example 2

[0064] According to another aspect of the present invention, a gridded rainstorm forecasting system is also provided, comprising a brain training sample module and a forecasting model module, wherein...

[0065] The training sample module is used to construct a training sample set by using data from the circulation background field, U / V forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations.

[0066] The forecast model module is used to establish a forecast model based on the EC model circulation background field and its forecast products and the mapping relationship between the future 0-24h precipitation using the XGBoost method. The training sample set is trained through the forecast model so that the forecast model can make medium- and short-term precipitation forecasts.

[0067] This invention is not limited to the specific embodiments described above. The above are merely preferred embodiments of this invention and are not intended to limit the invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0068] Example 3

[0069] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to execute any of the above-described gridded rainstorm forecasting methods.

[0070] Optionally, in this embodiment, the computer-readable storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals, and the computer-readable storage medium includes a stored program.

[0071] Optionally, during program execution, the device containing the computer-readable storage medium performs the following functions: using data from the circulation background field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations, and constructing a training sample set; using the XGBoost method to establish a forecasting model based on the mapping relationship between the EC model circulation background field and its forecast products and the future 0-24h precipitation; and training the forecasting model with the training sample set, enabling the forecasting model to perform medium- and short-term precipitation forecasts.

[0072] Example 4

[0073] According to another aspect of the present invention, a processor is also provided for running a program, wherein the program executes any of the above-described gridded rainstorm forecasting methods.

[0074] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of a gridded rainstorm forecasting method.

[0075] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0076] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0077] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The system embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interface, and the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0078] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0080] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0081] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A gridded heavy rainfall forecasting method, characterized in that, include: Data from the background circulation field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations were used to construct a training sample set; specifically including: Acquire the circulation background field for producing T-time forecast products, the U / V forecast field for setting the model, the rainfall forecast field, the model adjustment field, and the actual rainfall data from ground meteorological stations; Preprocessing was performed on the data from the background circulation field, UV forecast field, rainfall forecast field, model adjustment field, and surface meteorological station data, respectively. The preprocessed data is divided into multiple sample groups, each group containing samples... Including several forecasting factors and a forecast object ; By combining the aforementioned sample groups, a sample matrix is ​​constructed using spatiotemporal stacking and sample reconstruction techniques, thereby obtaining the training sample set; A forecasting model based on the EC model circulation background field and its forecast products and the mapping relationship between future 0-24h precipitation was established using the XGBoost method. The training sample set is used to train the forecast model, enabling the forecast model to perform short- and medium-term precipitation forecasts.

2. The gridded rainstorm forecasting method according to claim 1, characterized in that, Also includes: Based on the forecast model with mapping relationship, the forecast results of the forecast model are corrected by eliminating gaps in the rainfall forecast field of CMA-SH9 model.

3. The gridded rainstorm forecasting method according to claim 1, characterized in that, The background circulation field is an EC model rainfall forecast field, and the UV forecast field adopts the EC model UV forecast field at 500hPa, 850hPa, and 925hPa.

4. The gridded rainstorm forecasting method according to claim 1, characterized in that, Preprocessing the data of the circulation background field includes: determining whether there are troughs and shears in the 500hPa and 850hPa wind fields at the upper atmosphere that need to be monitored within a set future time period. If both exist, it is recorded as 1; otherwise, it is recorded as 0.

5. The gridded rainstorm forecasting method according to claim 1, characterized in that, Preprocessing of U / V forecast field data with a forecast start time of T includes: for EC wind fields in numerical models, calculating wind direction and wind speed for each t3 hour when the numerical model starts time of T-12h and the forecast lead time is t1~t2h, and interpolating the calculation results to the stations; where t1 is 12h, t2 is 36h, and t3 is 3h.

6. The gridded rainstorm forecasting method according to claim 1, characterized in that, Preprocessing of rainfall forecast data with a start time of T includes: calculating the cumulative precipitation for the next t1 to t2 hours based on rainfall forecast data from the EC model and CMA-SH9 model with a start time of T-12 hours, and interpolating the calculation results to the stations, denoted as ec24 and sh24 respectively; calculating the cumulative precipitation for the next 24 to 48 hours based on rainfall forecast data from the EC model with a start time of T-24 hours, and interpolating the calculation results to the stations, denoted as ec48 respectively.

7. A gridded rainstorm forecasting system, characterized in that, The gridded heavy rainfall forecasting method according to any one of claims 1 to 6 includes: The training sample module is used to construct a training sample set by using data from the circulation background field, U / V forecast field, rainfall forecast field, model adjustment field, and surface meteorological stations. The forecast model module is used to establish a forecast model based on the mapping relationship between the EC model circulation background field and its forecast products and the future 0-24h precipitation using the XGBoost method. The training sample set is trained through the forecast model so that the forecast model can make medium- and short-term precipitation forecasts.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the gridded rainstorm forecasting method according to any one of claims 1 to 6.

9. A processor, characterized in that, The processor is used to run a program, wherein the program executes the gridded rainstorm forecasting method according to any one of claims 1 to 6.

Images