A method, system, device and medium for extended-term quantitative precipitation forecasting

Abstract

The present invention proposes a method, system, device and medium for quantitative precipitation forecasting for an extended period, and relates to the field of weather forecasting. The method comprises: based on the current forecast time validity of a certain grid point, taking the ensemble average of the real-time forecast as the object, in the re-forecast data sample library with the same forecast time validity and subjected to ensemble average processing, sorting according to the similarity between the ensemble average processed re-forecast data and the real-time forecast, taking the first predefined number of re-forecast data in the sorting as the model forecast similarity sample set; based on the model forecast similarity sample set, corresponding to the QPE data, obtaining the historical actual precipitation similarity set; based on the precipitation similarity set, using the probability matching average technology and the bilinear interpolation technology, calculating the extended period precipitation grid forecast product for a predefined number of days. The present invention conducts in-depth research on the problems of long extended period forecast time validity, small number of samples required for the same period, small number of extended period models, and large forecast uncertainty, and establishes a new generation of extended period precipitation forecast model technology.

Description

Technical Field

[0001] The present invention belongs to the field of weather forecasting, and particularly relates to a method, system, device and medium for extended-range quantitative precipitation forecasting. Background Art

[0002] Precipitation, as a major catastrophic weather affecting China, is an important inducement for serious flood disasters, geological disasters and urban waterlogging. The key to reducing precipitation and secondary disasters lies in improving the accuracy of precipitation forecasting and extending the forecasting time limit. At present, numerical models, as the basis of modern weather forecasting operations, play a very important role in precipitation forecasting. However, due to the large number of influencing factors for precipitation forecasting, it is considered one of the most difficult challenges for numerical model forecasting, and the improvement of forecasting skills is relatively slow. In addition, as the forecasting time limit extends, the uncertainty of the model's precipitation forecasting increases rapidly, and its forecasting reliability is further significantly reduced.

[0003] In recent years, in order to improve the long-term precipitation forecasting effect, weather forecasting centers in many countries around the world have been developing high-resolution numerical model systems and vigorously developing ensemble model technologies at the same time. The advantage of ensemble models lies in revealing the non-linear uncertainty factors of weather processes, which is particularly applicable to extended-range forecasting. The advantages of ensemble models also give rise to new difficulties. One of them is how forecasters can quickly and efficiently extract high-value information from the massive data of ensemble models, thereby improving precipitation forecasting skills. In view of this, at present, in the real-time quantitative precipitation forecasting service of the China Meteorological Administration, statistical post-processing technology products of various ensemble models have been established, such as probability matching average, quantile mapping, and also include machine learning technologies such as logistic regression, support vector machine, and Bayesian. In addition, there are also studies on the application of deep learning and other methods in precipitation forecasting, which currently have good effects in short-term / short-range forecasting. Some studies have classified the post-processing technologies of model forecasting and pointed out that different methods should be used for bias correction based on different correction purposes and application scenarios. In actual business applications, the above-mentioned correction products are mostly concentrated in the forecasting time limit of 0-10 days. For the forecasting time limit of 11-30 days, due to the increase in the uncertainty of model forecasting as the forecasting time limit extends, the forecasting difficulty further increases, and there is still a lack of mature and effective objective forecasting at present. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a technical solution for an extended-range quantitative precipitation forecasting method, system, device, and medium. For the 11-30-day extended-range precipitation grid forecasting service, using the real-time forecasting and reforecasting data of the European Centre (EC) ensemble model, multi-source integrated real-time grid products (QPE), etc. as basic data, and adopting the technical ideas of similar ensemble forecasting technology and probability matching technology, etc., a national extended-range grid precipitation forecasting correction method with daily updates, a time resolution of 24-hour intervals for 11-30 days, and a spatial resolution of 5 km is developed. To solve the above technical problems.

[0005] The first aspect of the present invention discloses an extended-range quantitative precipitation forecasting method, and the method includes:

[0006] Step S1: Based on the current forecasting time of a certain grid point, taking the ensemble average of the real-time forecasting as the object, in the reforecasting data sample library processed by ensemble average at the same forecasting time, sort according to the similarity between the reforecasting data processed by ensemble average and the real-time forecasting, and take the top predefined number of reforecasting data as the model forecasting similarity sample set;

[0007] Step S2: Based on the model forecasting similarity sample set, corresponding to the QPE data, obtain the historical real precipitation similarity set;

[0008] Step S3: Based on the historical real precipitation similarity set, adopt the probability matching average technology and bilinear interpolation technology to calculate and obtain the extended-range precipitation grid forecasting product for the predefined number of days.

[0009] According to the method of the first aspect of the present invention, in the step S1, the calculation of the similarity includes:

[0010]

[0011] Wherein, is the forecasting time; is the forecasting time in the reforecasting data, and its value is equal to the forecasting time in the real-time forecasting data ; is the ensemble average of the real-time forecasting for a certain point at a certain forecasting time; is the reforecasting data processed by ensemble average at the same point and the same time in the past; is the number of forecasting variables; is the weight coefficient; is the standard deviation of the sequence of reforecasting data processed by ensemble average at the same point; is the forecasting time step; to is the time window, for example, taking the current as the reference time, For the time length of sliding before and after the benchmark aging, that is, the sliding time window is to , where = 24 hours; is the predicted value of the i-th prediction variable, that is, the prediction time of the prediction; is the similar predicted value of the same i-th point in the past at the aging.

[0012] According to the method of the first aspect of the present invention, in the step S1, the predefined value is equal to 20;

[0013] The number of the prediction variables is 25 grid points in the range of (x - 2, x + 2, y - 2, y + 2) around the current grid point (x, y);

[0014] The weight coefficient is an equal weight;

[0015] The prediction time step is 12 hours;

[0016] The time window is from -24h to 24h.

[0017] According to the method of the first aspect of the present invention, in the step S3, the predefined number of days is 11 - 30 days.

[0018] The second aspect of the present invention discloses an extended - range quantitative precipitation forecasting system, and the system includes:

[0019] A first processing module, configured to, based on the current prediction time of a certain grid point, take the ensemble average value of the real - time prediction as an object, and in the re - prediction data sample library processed by ensemble average at the same prediction time, sort according to the similarity between the re - prediction data processed by ensemble average and the real - time prediction, and take the first predefined number of re - prediction data in the sorting as a pattern prediction similarity sample set;

[0020] A second processing module, configured to, based on the pattern prediction similarity sample set, obtain a historical observed precipitation similarity set corresponding to the QPE data;

[0021] A third processing module, configured to, based on the historical observed precipitation similarity set, calculate and obtain an extended - range precipitation grid forecast product for a predefined number of days by using the probability matching average technology and the bilinear interpolation technology.

[0022] According to the system of the second aspect of the present invention, the first processing module is specifically configured that the calculation of the similarity includes:

[0023]

[0024] where is the prediction time; is the forecast lead time in the reforecast data, and its value is equal to the forecast lead time in the real-time forecast data ; is the real-time forecast of the ensemble average processing for a certain point at a certain forecast lead time; is the reforecast data of the ensemble average processing for the same point and the same lead time in the past; is the number of forecast variables; is the weight coefficient; is the standard deviation of the sequence of reforecast data of the ensemble average processing for the same point; is the forecast time step; to is the time window. For example, taking the current as the reference lead time, is the sliding time length before and after the reference lead time, that is, the sliding time window is to , where = 24 hours; is the forecast value of the i-th forecast variable, that is, the forecast lead time; is the similar forecast value of the same i-th point and the same lead time in the past.

[0025] According to the system of the second aspect of the present invention, the first processing module is specifically configured such that the predefined value is equal to 20;

[0026] The number of forecast variables is 25 grid points in the range of (x - 2, x + 2, y - 2, y + 2) around the current grid point (x, y);

[0027] The weight coefficient is an equal weight;

[0028] The forecast time step is 12 hours;

[0029] The time window is from -24h to 24h.

[0030] According to the system of the second aspect of the present invention, the third processing module is specifically configured such that the predefined number of days is 11 - 30 days.

[0031] The third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor. When the processor executes a computer program stored in the memory, the steps in an extended - period quantitative precipitation forecasting method according to any one of the first aspects of the present disclosure are implemented.

[0032] A fourth aspect of the present invention discloses a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps in an extended-range quantitative precipitation forecasting method according to any one of the first aspects of the present disclosure are implemented.

[0033] In summary, the solution proposed by the present invention can deeply explore two problems, namely, the long extended-range forecasting period with relatively few required concurrent samples, and the few extended-range models with large forecasting uncertainties, and research and establish a new generation of extended-range precipitation forecasting model technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0035] Figure 1 FIG. is a flowchart of an extended-range quantitative precipitation forecasting method according to an embodiment of the present invention;

[0036] Figure 2 FIG. is a structural diagram of an extended-range quantitative precipitation forecasting system according to an embodiment of the present invention;

[0037] Figure 3 FIG. is a structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0039] The similar ensemble prediction technology is an ensemble prediction technology that selects the actual observations corresponding to the most similar historical model predictions through the comparison of the deterministic prediction with the similar characteristics of the model predictions in each historical period, and uses them as the ensemble members after error correction. This method was proposed by scientists such as NCAR Monache in 2010 for the point-based AnEn technique. It saves computing resources compared with traditional ensemble prediction, can avoid the uncertainty of EPS itself, and can provide estimates of spatial and temporal uncertainty without the need for post-processing. However, it requires a large amount of observational and historical prediction data. Depending on the prediction object, the required duration varies, but it is at least several months including one season, and is used to improve the short-term prediction effect.

[0040] To address the bottleneck problems faced in the construction of the extended-range precipitation prediction model: First, the extended-range prediction has a long time limit and requires fewer concurrent samples; second, there are few extended-range models and the prediction uncertainty is large. When constructing the extended-range similar ensemble prediction model, we followed the method and idea of the above similar ensemble prediction technology and made adaptive improvements, mainly reflected in: First, changing the deterministic prediction in the original method to the real-time prediction (ensemble mean) of the ensemble prediction, which filters out high-frequency information to a certain extent and reduces prediction uncertainty; second, changing the model predictions in each historical period to the reforecast data (ensemble mean) of the ensemble prediction, that is, on the one hand, a large number of concurrent historical samples are increased, and at the same time, due to the use of model historical data, it is beneficial to correct the systematic error of the model.

[0041] The first aspect of the present invention discloses an extended-range quantitative precipitation prediction method. Figure 1 As shown in the flowchart of an extended-range quantitative precipitation prediction method according to an embodiment of the present invention, Figure 1 as shown, the method includes:

[0042] Step S1: Based on the current prediction time limit of a certain grid point, taking the ensemble mean of the real-time prediction as the object, in the reforecast data sample library that has been processed by ensemble mean at the same prediction time limit, sort according to the similarity between the reforecast data processed by ensemble mean and the real-time prediction, and take the top predefined number of reforecast data as the similar sample set of the model prediction;

[0043] Step S2: Based on the similar sample set of the model prediction, corresponding to the QPE data, obtain the similar set of historical actual precipitation;

[0044] Step S3: Based on the similar set of historical actual precipitation, use the probability matching average technology and the bilinear interpolation technology to calculate and obtain the extended-range precipitation grid prediction product for the predefined number of days.

[0045] In step S1, based on the current forecast lead time of a certain grid point, taking the ensemble average of the real-time forecast as the object, in the re-forecast data sample library processed by ensemble averaging at the same forecast lead time, sort according to the similarity between the re-forecast data processed by ensemble averaging and the real-time forecast, and take the top predefined number of re-forecast data as the model forecast similarity sample set.

[0046] In some embodiments, in the step S1, the calculation of the similarity includes:

[0047]

[0048] Wherein, is the forecast lead time; is the forecast lead time in the re-forecast data, and its value is equal to the forecast lead time in the real-time forecast data ; is the ensemble-averaged real-time forecast for a certain point at a certain forecast lead time; is the re-forecast data processed by ensemble averaging at the same point and the same lead time in the past; is the number of forecast variables; is the weight coefficient; is the standard deviation of the sequence of re-forecast data processed by ensemble averaging at the same point; is the forecast time step; to is the time window. For example, taking the current as the reference lead time, is the sliding time length before and after the reference lead time, that is, the sample sliding time window in this project is to , where = 24 hours; is the forecast value of the i-th forecast variable, that is, the forecast lead time; is the similar forecast value at the same i-th point and the lead time in the past.

[0049] The predefined number is equal to 20;

[0050] The number of forecast variables is 25 grid points in the range of (x - 2, x + 2, y - 2, y + 2) around the current grid point (x, y);

[0051] The weight coefficient is an equal weight;

[0052] The forecast time step is 12 hours;

[0053] The time window is -24h to 24h.

[0054] The ensemble mean forecast (EC_M) calculates the mathematical mean of all members (E) of the ECMWF ensemble forecast using equal weights, which can reflect the overall trend of the results of all ensemble forecast members. The calculation formula is as follows:

[0055]

[0056] where, are the members of the ensemble forecast.

[0057] In step S3, based on the historical precipitation similarity set, using the probability matching average technique and the bilinear interpolation technique, an extended range precipitation grid forecast product for a predefined number of days is calculated.

[0058] In some embodiments, in the step S3, the predefined number of days is 11 - 30 days.

[0059] Probability matching average

[0060] Quantitative precipitation forecasts are rarely able to accurately predict the spatial distribution pattern of precipitation. Through ensemble mean calculation, the most likely precipitation center location can be indicated, but there is a magnitude bias in the precipitation ensemble mean. This is manifested in that the ensemble mean product smooths the precipitation distribution, that is, the strong precipitation values decrease while the range of small - magnitude rainfall expands.

[0061] To correct the precipitation magnitude bias of the ensemble mean product, the probability matching technique (Probabilitymatching) is used. The probability matching technique is used to fuse data sources with different spatio - temporal distributions. Usually, one data source has a better spatial distribution while another data has better accuracy. This technique is achieved by setting the probability distribution function (PDF) of the low - accuracy data to the PDF of the high - accuracy data. Practical examples include the fusion between radar and rain gauge observations, and precipitation estimation from polar - orbiting or geostationary satellites. In ensemble forecasting, this technique is used to combine the ensemble mean field with better spatial distribution and the ensemble member forecasts with better magnitude accuracy.

[0062] Select a certain region, arrange all the forecasts of n members in the region from largest to smallest, and then retain the forecast values at every n / 2 intervals; arrange the ensemble mean field from largest to smallest; match the sequence retained in the first step with the ensemble mean sequence, and the probability - matched ensemble mean product is obtained.

[0063] Experiment: May to September 2024 was selected as the test period, and the TS score test method commonly used in forecasting business was adopted to conduct forecast test and comparison between the revised product and the original model product. The results showed that the revised product has a significant improvement over the original model in the 11-30 day forecast period in each precipitation level (such as: ≥0.1 mm, ≥10 mm, ≥25 mm, ≥50 mm), especially in the heavy precipitation level (i.e. ≥25 mm, ≥50 mm), the TS score is more than 2 times higher than the original model.

[0064] In summary, the scheme proposed in this invention can address the two problems of long extended-term forecast validity and few samples required for the same period; and few extended-term models and large forecast uncertainty. It has carried out in-depth discussions and studied the establishment of a new generation of extended-term precipitation forecast model technology. It was finally determined that based on similar ensemble forecast technology, probability matching average, bilinear interpolation and other technologies, the model was constructed using ensemble model real-time forecast data, re-forecast data and actual data, and a national precipitation refined business forecast product with daily updates and a spatiotemporal resolution of 1 day and 5 kilometers was developed and generated. Among them, the study made adaptive improvements to the similar ensemble forecast technology, which is mainly reflected in: first, the original method based on deterministic forecast is changed to real-time forecast based on ensemble forecast (ensemble average), which filters high-frequency information to a certain extent and reduces forecast uncertainty; second, the model forecasts of each historical period are changed to ensemble forecast re-forecast data, that is, on the one hand, a large number of historical samples of the same period are added, and at the same time, due to the use of model historical data, it is conducive to correcting the model system error.

[0065] The second aspect of the present invention discloses an extended-period quantitative precipitation forecasting system. Figure 2 is a structural diagram of an extended period quantitative precipitation forecast system according to an embodiment of the present invention; Figure 2 As shown, the system 100 includes:

[0066] The first processing module 101 is configured to, based on the current forecast time effectiveness of a certain grid point, take the ensemble average of the real-time forecast as the object, sort the ensemble average processed re-forecast data in the re-forecast data sample library with the same forecast time effectiveness according to the similarity between the ensemble average processed re-forecast data and the real-time forecast, and take the first predefined value of the sorted re-forecast data as the model forecast similarity sample set;

[0067] The second processing module 102 is configured to obtain a historical actual precipitation similarity set based on the model forecast similarity sample set corresponding to the QPE data;

[0068] The third processing module 103 is configured to calculate the extended-range precipitation grid forecast product for a predefined number of days by using the probability matching average technology and the bilinear interpolation technology based on the historical actual precipitation similarity set.

[0069] For the system according to the second aspect of the present invention, the first processing module 101 is specifically configured such that the calculation of the similarity includes:

[0070]

[0071] wherein, is the forecast lead time; is the forecast lead time in the re-forecast data, and its value is equal to the forecast lead time in the real-time forecast data ; is the real-time forecast of the ensemble mean processing for a certain point at a certain forecast lead time; is the re-forecast data of the ensemble mean processing for the same point and the same lead time in the past; is the number of forecast variables; is the weight coefficient; is the standard deviation of the sequence of re-forecast data of the ensemble mean processing for the same point; is the forecast time step; to is the time window. For example, taking the current as the reference lead time, is the sliding time length before and after the reference lead time, that is, the sample sliding time window in this project is to , where = 24 hours; is the forecast value of the i-th forecast variable, that is, the forecast lead time; is the similar forecast value of the same i-th point and the lead time in the past.

[0072] The predefined value is equal to 20;

[0073] The number of forecast variables is 25 grid points in the range of (x - 2, x + 2, y - 2, y + 2) around the current grid point (x, y);

[0074] The weight coefficient is an equal weight;

[0075] The forecast time step is 12 hours;

[0076] The time window is from -24h to 24h.

[0077] The ensemble mean forecast (EC_M) calculates the mathematical average of all members (E) of the ECMWF ensemble forecast using equal weights, and can reflect the overall trend of the results of all ensemble forecast members. The calculation formula is as follows:

[0078]

[0079] wherein, For the members of the ensemble forecast.

[0080] For the system according to the second aspect of the present invention, the third processing module 103 is specifically configured such that the predefined number of days is 11 - 30 days.

[0081] Probability matching average

[0082] Quantitative precipitation forecasts are rarely able to accurately predict the spatial distribution pattern of precipitation. Through ensemble averaging calculation, it can indicate the location of the most likely precipitation center, but there is a magnitude bias in the precipitation ensemble average. This is manifested in that the ensemble average product smooths the precipitation distribution, that is, the strong precipitation value decreases while the range of small - magnitude rainfall expands.

[0083] To correct the precipitation magnitude bias of the ensemble average product, the probability matching technique (Probabilitymatching) is adopted. The probability matching technique is used to fuse data sources with different spatio - temporal distributions. Usually, one data source has a better spatial distribution while another data has better accuracy. This technique is achieved by setting the probability distribution function (PDF) of the low - accuracy data to the PDF of the high - accuracy data. Practical examples include the fusion between radar and rain gauge observations, and precipitation estimation from polar - orbiting or geostationary satellites. In ensemble forecasting, this technique is used to combine the ensemble average field with better spatial distribution and the ensemble member forecasts with better magnitude accuracy.

[0084] Select a certain area, arrange all the forecasts of n members in the area from large to small, and then retain the forecast values at every n / 2 intervals; arrange the ensemble average field from large to small; match the sequence retained in the first step with the ensemble average sequence, and the probability - matched ensemble average product is obtained.

[0085] The third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor. When the processor executes the computer program stored in the memory, it implements the steps in an extended - period quantitative precipitation forecasting method according to any one of the first aspects disclosed in the present invention.

[0086] Figure 3 It is a structural diagram of an electronic device according to an embodiment of the present invention, as Figure 3As shown, the electronic device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, near field communication (NFC), or other technologies. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the electronic device, or an external keyboard, touchpad, or mouse, etc.

[0087] Those skilled in the art can understand that Figure 3 the structure shown in is only a structural diagram of a part related to the technical solution of the present disclosure, and does not constitute a limitation on the electronic device to which the solution of the present application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0088] The fourth aspect of the present invention discloses a computer-readable storage medium. A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps in a method for extended-period quantitative precipitation forecasting according to any one of the first aspect of the present invention are implemented.

[0089] Please note that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combinations of these technical features do not conflict, they should all be considered as within the scope described in this specification. The above embodiments only represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for quantitative precipitation forecasting in the extended period, characterized in that, The method includes: Step S1: Based on the current forecast lead time of a certain grid point, taking the ensemble mean of the real-time forecast as the object, in the re-forecast data sample library processed by ensemble mean at the same forecast lead time, sort according to the similarity between the re-forecast data processed by ensemble mean and the real-time forecast, and take the top predefined number of re-forecast data as the model forecast similarity sample set; Step S2: Based on the model forecast similarity sample set, corresponding to the QPE data, obtain the historical observed precipitation similarity set; Step S3: Based on the historical observed precipitation similarity set, use the probability matching average technique and the bilinear interpolation technique to calculate and obtain the extended-range precipitation grid forecast product for a predefined number of days; In the said Step S1, the calculation of the similarity includes: ； Among them, is the forecast lead time; is the forecast lead time in the re-forecast data, and its value is equal to the forecast lead time in the real-time forecast data ; is the real-time forecast of the ensemble average processing for a certain point at a certain forecast lead time; is the re-forecast data of the ensemble average processing for the same point and the same lead time in the past; is the number of forecast variables; is the weight coefficient; is the standard deviation of the sequence of re-forecast data of the ensemble average processing for the same point; is the forecast time step; to is the time window. For example, taking the current as the reference lead time, is the sliding time length before and after the reference lead time, that is, the sliding time window is to , where = 24 hours; is the forecast value of the i-th forecast variable at the th forecast lead time; is the similar forecast value of the same i-th point and the th lead time in the past; In the said Step S1, the predefined number is equal to 20; The number of forecast variables is 25 grid points in the range of (x - 2, x + 2, y - 2, y + 2) around the current grid point (x, y); The weight coefficient is an equal weight; The forecast time step is 12 hours; The time window is from -24h to 24h; In the said Step S3, the predefined number of days is 11 - 30 days.

2. An extended period quantitative precipitation forecasting system, characterized in that, The system adopts the method described in Claim 1, and the system includes: The first processing module is configured to, based on the current forecast lead time of a certain grid point, take the ensemble mean of the real-time forecast as the object, in the re-forecast data sample library processed by ensemble mean at the same forecast lead time, sort according to the similarity between the re-forecast data processed by ensemble mean and the real-time forecast, and take the top predefined number of re-forecast data as the model forecast similarity sample set; The second processing module is configured to, based on the model forecast similarity sample set, corresponding to the QPE data, obtain the historical observed precipitation similarity set; The third processing module is configured to, based on the historical observed precipitation similarity set, use the probability matching average technique and the bilinear interpolation technique to calculate and obtain the extended-range precipitation grid forecast product for a predefined number of days; the predefined number of days is 11 - 30 days; The first processing module is specifically configured to, the calculation of the similarity includes: ； wherein, is the forecast lead time; is the forecast lead time in the reforecast data, and its value is equal to the forecast lead time in the real-time forecast data ; is the real-time forecast of the ensemble average processing for a certain point at a certain forecast lead time; is the reforecast data of the ensemble average processing for the same point and the same lead time in the past; is the number of forecast variables; is the weight coefficient; is the standard deviation of the reforecast data sequence of the ensemble average processing for the same point; is the forecast time step; to is the time window. For example, taking the current as the reference lead time, is the sliding time length before and after the reference lead time, that is, the sliding time window is to , where = 24 hours; is the forecast value of the i-th forecast variable at the i-th forecast lead time; is the similar forecast value at the same i-th point and the i-th lead time in the past; The first processing module is specifically configured to, the predefined number is equal to 20; The number of forecast variables is 25 grid points in the range of (x - 2, x + 2, y - 2, y + 2) around the current grid point (x, y); The weight coefficient is an equal weight; The forecast time step is 12 hours; The time window is from -24h to 24h.

3. An electronic device, characterized in that, The electronic device includes a memory and a processor. When the processor executes the computer program stored in the memory, it implements the steps in an extended-range quantitative precipitation forecasting method described in Claim 1.

4. A computer-readable storage medium, characterized in that, The computer program is stored on a computer-readable storage medium. When the computer program is executed by the processor, it implements the steps in an extended-range quantitative precipitation forecasting method described in Claim 1.

Images