A method for interpreting precipitation forecast products from numerical models based on spatiotemporal local models

Through the method based on the local model of time and space, the numerical model precipitation forecast is optimized, and the interpretation model parameters are combined with observation data and historical forecast products, the uncertainty and regional differences of numerical model precipitation forecast are solved, and the accuracy of precipitation forecast is improved. and release level.

CN113988350BActive Publication Date: 2024-11-08HAINAN INSTITUTE OF METEOROLOGICAL SCIENCE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010733993.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-07-27
Publication Date
2024-11-08
Estimated Expiration
2040-07-27

AI Technical Summary

Technical Problem

There are uncertainties and regional differences in existing numerical mode precipitation forecasts, making it difficult to provide accurate precipitation forecasts in different regions and time periods, and multimodal integration methods usually assume that all samples are of the same importance and lack processing for local characteristics of space-time.

Method used

Using a method based on a spatiotemporal local model, the area to be released is divided into sub-regions, and the parameters of the release are optimized through interpolation and sample sets, combined with observation data and historical forecast products, and using weighted regression, spatiotemporal distance model and other technologies to carry out Accurate interpretation of precipitation drop zone and magnitude forecast.

Benefits of technology

It improves the accuracy and release level of numerical mode precipitation forecasts, can better adapt to time and space changes, and enhances the objectivity and reliability of forecast products.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FHA0000007269920000011
    Figure FHA0000007269920000011
  • Figure FHA0000007269920000021
    Figure FHA0000007269920000021
  • Figure FHA0000007269920000031
    Figure FHA0000007269920000031
Patent Text Reader

Abstract

The present invention discloses a method for distributing numerical model precipitation forecast products based on a spatiotemporal local model, comprising the following steps: determining an area to be distributed and dividing it into a number of sub-areas; obtaining numerical model precipitation forecast products in the forecast area to be distributed and interpolating them to a desired discrete grid; constructing a sample set; preprocessing and quality control of samples; optimizing parameters of the spatiotemporal local model for distribution of precipitation area forecast; for each sub-area, using optimized or set parameters to distribute the precipitation area forecast results of the forecast period to determine the precipitation area; optimizing parameters of the spatiotemporal local model for distribution of precipitation level forecast; for each sub-area, using optimized or set parameters to distribute the precipitation level forecast results in the above-mentioned area. The beneficial effects of the present invention are: using the spatiotemporal local model to more accurately model the relationship between the numerical model precipitation forecast products and the actual situation that change with time and space, and further improving the distributing level of the numerical model precipitation forecast products.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent grid forecasting in meteorology, and in particular to a method for interpreting and using numerical model precipitation forecast products based on a spatiotemporal local model. Background Art

[0002] Precipitation is one of the most important elements of weather and climate. Accurate forecasting of precipitation is one of the main contents of weather forecasting and climate forecasting. With the development of atmospheric science and other related disciplines, the spatial and temporal scale and timeliness of numerical models for precipitation forecasting have been continuously expanded, and the resolution and accuracy of precipitation forecasting have been continuously improved. At present, numerical model forecasts have become the basis for precipitation forecasting.

[0003] However, since numerical models can only approximate the atmosphere and atmospheric motion has chaotic characteristics, there is always uncertainty in the forecast of precipitation by numerical models. In addition, different countries have different model implementation plans, resulting in different forecast capabilities of different models for precipitation in different regions and at different time periods. Therefore, correcting the precipitation forecasted by numerical models or integrating the precipitation products forecasted by different numerical models has become an important means to improve the level of precipitation forecasting.

[0004] At present, linear regression, Bayesian model averaging, Kalman filtering, machine learning and other methods are the mainstream methods for correcting numerical model precipitation forecast products or integrating multi-model forecast products. However, when using the above methods to correct or integrate model products, some only use samples from a single grid point or station, some use all samples at the same time in a certain area, and some assume that all samples have the same importance.

[0005] In order to further improve the objective interpretation level of numerical model precipitation forecast products, the present invention proposes a numerical model precipitation forecast product interpretation method based on a spatiotemporal local model. Summary of the invention

[0006] The present invention is designed to solve the above technical problems and is a method for interpreting and using numerical model precipitation forecast products based on a spatiotemporal local model.

[0007] The technical solution adopted by the present invention to solve its technical problem is:

[0008] A method for interpreting a numerical model precipitation forecast product based on a spatiotemporal local model, first interpreting a numerical model precipitation area forecast result, and then interpreting a numerical model precipitation magnitude forecast result, the method comprising the following steps:

[0009] Step 1: Determine the area to be released and divide it into several sub-areas;

[0010] Step 2: Obtain the numerical model precipitation forecast product in the forecast expected area, and use the interpolation method to interpolate the numerical model precipitation forecast product to the required discrete grid;

[0011] Step 3: Obtain precipitation observation data and historical numerical model precipitation forecast products in the area to be released, build a precipitation (observation data, forecast product) sample set, and obtain a precipitation (observation indicator value, forecast product) sample set through conversion;

[0012] Step 4: Pre-process and quality control the samples in step 3 and store them for future use;

[0013] Step 5: For the precipitation area forecast, for each sub-area, select the corresponding interpretation model, weight function model, spatiotemporal distance model, evaluation index and sliding training period; use the corresponding sample set processed in step 4 to optimize the parameters of the spatiotemporal local model for the precipitation area forecast, i.e., the number of neighborhood samples N1, the spatiotemporal distance model parameters and the weight function bandwidth, or set the values ​​of the above parameters based on experience;

[0014] Step 6: For each sub-area, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 5 and the optimized or set spatiotemporal local model parameters for precipitation area forecast interpretation are used to interpret the numerical model precipitation area forecast in step 2 to determine the precipitation area;

[0015] Step 7: For the precipitation level forecast, for each sub-region, select the corresponding interpretation model, weight function model, spatiotemporal distance model, evaluation index and sliding training period, and use the corresponding sample set processed in step 4 to optimize the parameters of the spatiotemporal local model for precipitation level forecast interpretation, i.e., the number of neighborhood samples N2, the spatiotemporal distance model parameters and the weight function bandwidth, or set the values ​​of the above parameters based on experience;

[0016] Step 8: For each sub-region, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 7 and the optimized or set spatiotemporal local model parameters for precipitation level forecast interpretation are used to interpret the precipitation level in the precipitation area determined in step 6;

[0017] Step 9: Establish relevant business systems and repeat steps 2-8.

[0018] The method for interpreting and using numerical model precipitation forecast products based on a spatiotemporal local model, wherein the numerical model precipitation forecast product in step 2 is a deterministic forecast, ensemble average or a combination thereof derived from one or more models.

[0019] The method for interpreting precipitation forecast products of numerical models based on a spatiotemporal local model, the precipitation observation data in step 3 includes but is not limited to site precipitation observation data and multi-source data fusion precipitation actual data; the observation indicator value calculation method in step 3 is: when the observation value is greater than 0, the observation indicator value is 1, otherwise it is 0.

[0020] In the method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the preprocessing and quality control of samples in step 4 include but are not limited to eliminating redundant samples and abnormal samples.

[0021] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the interpretation models that can be used in step 5 include but are not limited to weighted logistic regression and its corresponding multi-scale model.

[0022] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the evaluation indicators that can be used in step 5 include but are not limited to hit rate, false alarm rate, TS score or a combination thereof.

[0023] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the weight function models that can be used in steps 5 and 7 include but are not limited to distance threshold model ①, Gaussian model ②, exponential model ③, and bisquare model ④.

[0024]

[0025]

[0026] Among them, d i For sample x i The space-time distance to the grid point x0 to be interpreted, r is the bandwidth of the weight function, W i (x i ) is the sample x i The weight of .

[0027] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the interpretation models that can be used in step 7 include but are not limited to weighted linear regression, weighted partial least squares regression, weighted conditional probability density function, weighted frequency matching method, weighted lasso regression and their corresponding multi-scale models.

[0028] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the evaluation indicators that can be used in step 7 include but are not limited to mean absolute error, root mean square error, CRPS score or a combination thereof.

[0029] The method for interpreting precipitation forecast products of numerical models based on a spatiotemporal local model, the spatiotemporal local model in steps 5 to 8 is specifically as follows: according to the selected sliding training period, for any grid point or sample point, a relationship between a precipitation observation indicator value and the forecast product is constructed using its surrounding N1 spatiotemporal samples and the corresponding interpretation model, weight function model, and spatiotemporal distance model, and the precipitation fall area forecast for the grid point is interpreted using this relationship; or, according to the selected sliding training period, for any grid point or sample point, a relationship between a precipitation observation value and a forecast value is constructed using its surrounding N2 spatiotemporal samples and the corresponding interpretation model, weight function model, and spatiotemporal distance model, and the precipitation magnitude forecast for the grid point is interpreted using this relationship.

[0030] In the method for interpreting precipitation forecast products of numerical models based on a spatiotemporal local model, the spatiotemporal distance model that can be used by the spatiotemporal local model in step 5 and step 7 includes but is not limited to formula ⑤.

[0031]

[0032] Where d is the space-time position x0 and x i The distance between (u0,v0,t0), (u i ,v i ,t i ) are the space-time positions x0 and x i The space-time coordinates of ; μ is the space-time distance model parameter, which is determined based on prior knowledge or cross-validation method.

[0033] The method for interpreting a numerical model precipitation forecast product based on a spatiotemporal local model, the business system in step 9 includes but is not limited to a data acquisition module, a numerical model precipitation forecast product interpolation module, a sample pair construction module, a sample quality control module, a fall area forecast interpretation parameter optimization and setting module, a fall area forecast interpretation module, a magnitude forecast interpretation parameter optimization and setting module, a magnitude forecast interpretation module or a combination thereof.

[0034] The samples described in the present invention all include corresponding unified reference time and space coordinates, which are not explicitly represented for simplicity.

[0035] The beneficial effect of the present invention is that the relationship between the numerical model precipitation forecast products and the actual situation that change with time and space is more accurately modeled by adopting the spatiotemporal local model, thereby further improving the interpretation level of the numerical model precipitation forecast products. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic diagram of space-time distance and space-time neighborhood. DETAILED DESCRIPTION

[0037] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0038] like Figure 1 As shown, the present invention provides a method for interpreting a numerical model precipitation forecast product based on a spatiotemporal local model, firstly interpreting the numerical model precipitation area forecast result, and then interpreting the numerical model precipitation magnitude forecast result, the method comprising the following steps:

[0039] Step 1: Determine the area to be released and divide it into several sub-areas based on factors such as the size of the area to be released and the weather and climate characteristics;

[0040] Step 2: Obtain the numerical model precipitation forecast product in the forecast expected area, and use the interpolation method to interpolate the numerical model precipitation forecast product to the required discrete grid;

[0041] Step 3: Obtain the precipitation observation data and historical numerical model precipitation forecast products in the area to be released, and use the interpolation method to interpolate the numerical model precipitation forecast products to the precipitation observation data position, construct a precipitation (observation data, forecast product) sample set, and obtain the precipitation (observation indicator value, forecast product) sample set through conversion;

[0042] Step 4: Pre-process and quality control the samples in step 3 and store them for future use;

[0043] Step 5: For the precipitation area forecast, for each sub-area, select the corresponding interpretation model, weight function model, spatiotemporal distance model, evaluation index and sliding training period; use the corresponding sample set processed in step 4 and the random search parameter optimization method to optimize the parameters of the spatiotemporal local model for the precipitation area forecast, i.e., the number of neighborhood samples N1, the spatiotemporal distance model parameters and the weight function bandwidth, or set the values ​​of the above parameters based on experience;

[0044] Step 6: For each sub-area, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 5 and the optimized or set spatiotemporal local model parameters for precipitation area forecast interpretation are used to interpret the numerical model precipitation area forecast in step 2 to determine the precipitation area;

[0045] Step 7: For the precipitation level forecast, for each sub-region, select the corresponding interpretation model, weight function model, spatiotemporal distance model, evaluation index and sliding training period, use the corresponding sample set processed in step 4 and the random search parameter optimization method to optimize the parameters of the spatiotemporal local model for precipitation level forecast interpretation, i.e., the number of neighborhood samples N2, the spatiotemporal distance model parameters and the weight function bandwidth, or set the values ​​of the above parameters based on experience;

[0046] Step 8: For each sub-region, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 7 and the optimized or set spatiotemporal local model parameters for precipitation level forecast interpretation are used to interpret the precipitation level in the precipitation area determined in step 6;

[0047] Step 9: Establish relevant business systems and repeat steps 2-8.

[0048] The method for interpreting and using numerical model precipitation forecast products based on a spatiotemporal local model, wherein the numerical model precipitation forecast product in step 2 is a deterministic forecast, ensemble average or a combination thereof derived from one or more models.

[0049] The method for interpreting precipitation forecast products of numerical models based on a spatiotemporal local model, the precipitation observation data in step 3 includes but is not limited to site precipitation observation data and multi-source data fusion precipitation actual data; the calculation method of the observation indicator value in step 3 is: when the observation value is greater than 0, the observation indicator value is 1, otherwise it is 0.

[0050] In the method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the preprocessing and quality control of samples in step 4 include but are not limited to eliminating redundant samples and abnormal samples.

[0051] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the interpretation models that can be used in step 5 include but are not limited to weighted logistic regression and its corresponding multi-scale model.

[0052] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the evaluation indicators that can be used in step 5 include but are not limited to hit rate, false alarm rate, TS score or a combination thereof.

[0053] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the weight function models that can be used in steps 5 and 7 include but are not limited to one or more of the distance threshold model ①, Gaussian model ②, exponential model ③, and bisquare model ④ (for multi-scale models).

[0054] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the interpretation models that can be used in step 7 include but are not limited to weighted linear regression, weighted partial least squares regression, weighted conditional probability density function, weighted frequency matching method, weighted lasso regression and their corresponding multi-scale models.

[0055] The method for interpreting precipitation forecast products of numerical models based on spatiotemporal local models, the evaluation indicators that can be used in step 7 include but are not limited to mean absolute error, root mean square error, CRPS score or a combination thereof.

[0056] The objective interpretation method of the numerical model forecast product based on the spatiotemporal local model, the parameter optimization methods that can be used in step 5 and step 7 include but are not limited to the grid search parameter optimization method and the random search parameter optimization method; taking the grid search parameter optimization method as an example: the possible intervals of all parameters in step 7 are discretized and combined; for each parameter combination, the selected interpretation model, weight function model, and spatiotemporal distance model are used to construct a relationship between precipitation observation data and forecast products for each sample point in the sub-area of ​​the sliding training period, and this relationship is used to interpret the numerical model precipitation forecast product of the sample point, and the interpretation results are stored for use; according to the interpretation results, the evaluation index of each parameter combination is calculated, and the optimal parameter combination is determined according to the evaluation index.

[0057] The method for interpreting precipitation forecast products of numerical models based on a spatiotemporal local model, the spatiotemporal local model in steps 5 to 8 is specifically as follows: according to the selected sliding training period, for any grid point or sample point, a relationship between a precipitation observation indicator value and the forecast product is constructed using its surrounding N1 spatiotemporal samples and the corresponding interpretation model, weight function model, and spatiotemporal distance model, and the precipitation fall area forecast for the grid point is interpreted using this relationship; or, according to the selected sliding training period, for any grid point or sample point, a relationship between a precipitation observation value and a forecast value is constructed using its surrounding N2 spatiotemporal samples and the corresponding interpretation model, weight function model, and spatiotemporal distance model, and the precipitation magnitude forecast for the grid point is interpreted using this relationship.

[0058] The method for interpreting precipitation forecast products of numerical models based on a spatiotemporal local model, the spatiotemporal distance models that can be used in step 5 and step 7 include but are not limited to formula ⑤.

[0059]

[0060] Where d is the space-time position x0 and x i The distance between (u0,v0,t0), (u i ,v i ,t i ) are the space-time positions x0 and x i The space-time coordinates of ; μ is the space-time distance model parameter, which is determined based on prior knowledge or cross-validation method.

[0061] The method for interpreting a numerical model precipitation forecast product based on a spatiotemporal local model, the business system in step 9 includes but is not limited to a data acquisition module, a numerical model precipitation forecast product interpolation module, a sample pair construction module, a sample quality control module, a fall area forecast interpretation parameter optimization and setting module, a fall area forecast interpretation module, a magnitude forecast interpretation parameter optimization and setting module, a magnitude forecast interpretation module or a combination thereof.

[0062] Example 1

[0063] Objectively interpret the Hainan Island fine grid (0.125°×0.125°) precipitation forecast product for the first day reported by the ECMWF model at 00:00 (UTC) every day.

[0064] Step 1: Divide Hainan Island into 18 sub-regions according to administrative divisions;

[0065] Step 2: Obtain the Hainan Island fine grid (0.125°×0.125°) precipitation forecast product for the first day reported by the ECMWF model at 00:00 (UTC) every day during the forecast period;

[0066] Step 3: Obtain the CMPAS China daily precipitation rapid fusion real-time analysis product (0.05°×0.05°); and use the inverse distance interpolation method to construct the (observation data, forecast product) sample set with the Hainan Island fine grid precipitation forecast product of the same period in history, and at the same time obtain the (observation indicator value, forecast product) sample set through conversion;

[0067] Step 4: Pre-process and quality control the samples in step 3 and store them for future use;

[0068] Step 5: For the precipitation area forecast, for each sub-area, the interpretation model uses weighted logistic regression, the weight function model uses the Gaussian model (Formula ②), the spatiotemporal distance model uses Formula ⑤, the evaluation index uses TS scoring, the sliding training period is set to 70 days, and the corresponding sample set processed in step 4 and the random search parameter optimization method are used to optimize the parameters of the spatiotemporal local model for the precipitation area forecast, namely the number of neighborhood samples N1, the spatiotemporal distance model parameter μ1 and the weight function bandwidth r1;

[0069] Step 6: For each sub-area, the interpretation model, weight function model, spatiotemporal distance model, sliding training period, and optimized spatiotemporal local model parameters of the precipitation area forecast interpretation selected in step 5 are used to interpret the precipitation area forecast in step 2 and determine the precipitation area;

[0070] Step 7: For each sub-region, the interpretation model uses univariate weighted linear regression, the weight function model uses the Gaussian model (Formula ②), the spatiotemporal distance model uses Formula ⑤, the evaluation index uses the root mean square error, the sliding training period is set to 70 days, and the corresponding sample set processed in Step 4 and the random search parameter optimization method are used to optimize the parameters of the spatiotemporal local model for precipitation level forecast interpretation, namely the number of neighborhood samples N2, the spatiotemporal distance model parameter μ2, and the weight function bandwidth r2;

[0071] Step 8: For each sub-region, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 7 and the optimized spatiotemporal local model parameters for precipitation level forecast interpretation are used to interpret the precipitation level in the precipitation area determined in step 6;

[0072] Step 9: Establish relevant business systems and repeat steps 2-8.

[0073] Example 2

[0074] Objectively interpret the Hainan Island fine grid (0.125°×0.125°) precipitation forecast product for the first day reported by the ECMWF model at 00:00 (UTC) every day.

[0075] Step 1: Divide Hainan Island into 18 sub-regions according to administrative divisions;

[0076] Step 2: Obtain the Hainan Island fine grid (0.125°×0.125°) precipitation forecast product for the first day reported by the ECMWF model at 00:00 (UTC) every day during the forecast period;

[0077] Step 3: Obtain the CMPAS China daily precipitation rapid fusion real-time analysis product (0.05°×0.05°); and use the inverse distance interpolation method and the Hainan Island fine grid precipitation forecast product of the same period in history to construct the precipitation (observation data, forecast product) sample set, and at the same time obtain the (observation indicator value, forecast product) sample set through conversion;

[0078] Step 4: Pre-process and quality control the samples in step 3 and store them for future use;

[0079] Step 5: For each sub-region, the interpretation model uses weighted logistic regression, the weight function model uses the Gaussian model (Formula ②), the spatiotemporal distance model uses Formula ⑤, the evaluation index uses TS scoring, the sliding training period is set to 70 days, and the corresponding sample set processed in Step 4 and the random search parameter optimization method are used to optimize the parameters of the spatiotemporal local model for precipitation forecast interpretation, namely the number of neighborhood samples N1, the spatiotemporal distance model parameter μ1, and the weight function bandwidth r1;

[0080] Step 6: For each sub-area, the interpretation model, weight function model, spatiotemporal distance model, sliding training period, and optimized spatiotemporal local model parameters of the precipitation area forecast interpretation selected in step 5 are used to interpret the precipitation area forecast in step 2 and determine the precipitation area;

[0081] Step 7: For each sub-region, the interpretation model uses the weighted conditional probability density function, the weight function model uses the Gaussian model (Formula ②), the spatiotemporal distance model uses Formula ⑤, the evaluation index uses the CRPS score, the sliding training period is set to 70 days, and the corresponding sample set processed in Step 4 and the random search parameter optimization method are used to optimize the parameters of the spatiotemporal local model for precipitation level forecast interpretation, namely the number of neighborhood samples N2, the spatiotemporal distance model parameter μ2, and the weight function bandwidth r2;

[0082] Step 8: For each sub-region, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 7 and the optimized spatiotemporal local model parameters for precipitation level forecast interpretation are used to interpret the precipitation level in the precipitation area determined in step 6;

[0083] Step 9: Establish relevant business systems and repeat steps 2-8.

Claims

1. A method for interpreting numerical model precipitation forecast products based on a spatiotemporal local model, first interpreting the numerical model precipitation area forecast result, and then interpreting the numerical model precipitation magnitude forecast result, characterized in that: The method comprises the following steps: Step 1: Determine the area to be released and divide it into several sub-areas; Step 2: Obtain the numerical model precipitation forecast product in the forecast expected area, and use the interpolation method to interpolate the numerical model precipitation forecast product to the required discrete grid; Step 3: Obtain precipitation observation data and historical numerical model precipitation forecast products in the area to be released, build a precipitation "observation data, forecast product" sample set, and obtain a precipitation "observation indicator value, forecast product" sample set through conversion; Step 4: Pre-process and quality control the samples in step 3 and store them for future use; Step 5: For the precipitation area forecast, for each sub-area, select the corresponding interpretation model, weight function model, spatiotemporal distance model, evaluation index and sliding training period; use the corresponding sample set processed in step 4 to optimize the parameters of the spatiotemporal local model for the precipitation area forecast, i.e., the number of neighborhood samples N1, the spatiotemporal distance model parameters and the weight function bandwidth, or set the values ​​of the above parameters based on experience; The interpretation model includes weighted logistic regression and its corresponding multi-scale model; The evaluation index includes hit rate, false alarm rate, TS score or a combination thereof; The weight function models include distance threshold model ①, Gaussian model ②, exponential model ③, and bi-square model ④; Among them, d i For sample x i The space-time distance to the grid point x0 to be interpreted, r is the bandwidth of the weight function, W i (x i ) is the sample x i The weight of Step 6: For each sub-area, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 5 and the optimized or set spatiotemporal local model parameters for precipitation area forecast interpretation are used to interpret the discrete grid numerical model precipitation area forecast in step 2 to determine the precipitation area; Step 7: For the precipitation level forecast, for each sub-region, select the corresponding interpretation model, weight function model, spatiotemporal distance model, evaluation index and sliding training period, and use the corresponding sample set processed in step 4 to optimize the parameters of the spatiotemporal local model for precipitation level forecast interpretation, i.e., the number of neighborhood samples N2, the spatiotemporal distance model parameters and the weight function bandwidth, or set the values ​​of the above parameters based on experience; The weight function models include distance threshold model ①, Gaussian model ②, exponential model ③, and bi-square model ④; Among them, d i For sample x i The space-time distance to the grid point x0 to be interpreted, r is the bandwidth of the weight function, W i (x i ) is the sample x i The weight of the method comprises weighted linear regression, weighted partial least squares regression, weighted conditional probability density function, weighted frequency matching method, weighted lasso regression and their corresponding multi-scale models; the evaluation index comprises mean absolute error, root mean square error, CRPS score or a combination thereof; Step 8: For each sub-region, the interpretation model, weight function model, spatiotemporal distance model and sliding training period selected in step 7 and the optimized or set spatiotemporal local model parameters for precipitation level forecast interpretation are used to interpret the precipitation level in the precipitation area determined in step 6; Step 9: Establish relevant business systems and repeat steps 2-8; The spatiotemporal local model in step 5 and step 6 is specifically as follows: according to the selected sliding training period, for any grid point or sample point, using the surrounding N1 spatiotemporal samples and the corresponding interpretation model, weight function model, and spatiotemporal distance model, a relationship between the precipitation observation indicator value and the forecast product is constructed, and the precipitation area forecast of the grid point is interpreted using this relationship; the spatiotemporal local model in step 7 and step 8 is specifically as follows: according to the selected sliding training period, for any grid point or sample point, using the surrounding N2 spatiotemporal samples and the corresponding interpretation model, weight function model, and spatiotemporal distance model, a relationship between the precipitation observation value and the forecast value is constructed, and the precipitation level forecast of the grid point is interpreted using this relationship; the spatiotemporal distance model used in step 5 and step 7 includes formula ⑤; Where d is the space-time position x0 and x i The distance between (u0,v0,t0), (u i ,v i ,t i ) are the space-time positions x0 and x i The space-time coordinates of ; μ is the space-time distance model parameter, which is determined based on prior knowledge or cross-validation method; The business system in step 9 includes a data acquisition module, a numerical model precipitation forecast product interpolation module, a sample pair construction module, a sample quality control module, a fall area forecast interpretation parameter optimization and setting module, a fall area forecast interpretation module, a magnitude forecast interpretation parameter optimization and setting module, and a magnitude forecast interpretation module.

2. The method for interpreting and using a numerical model precipitation forecast product based on a spatiotemporal local model according to claim 1, characterized in that: The numerical model precipitation forecast product in step 2 is a deterministic forecast and ensemble average derived from one or more models.

3. The method for interpreting and using a numerical model precipitation forecast product based on a spatiotemporal local model according to claim 1, characterized in that: The precipitation observation data in step 3 includes site precipitation observation data and multi-source data fusion precipitation real-time data; the calculation method of the observation indicator value in step 3 is: when the observation value is greater than 0, the observation indicator value is 1, otherwise it is 0.

4. The method for interpreting and using a numerical model precipitation forecast product based on a spatiotemporal local model according to claim 1, characterized in that: The sample preprocessing and quality control in step 4 includes eliminating redundant samples and abnormal samples.