A multi-band weather radar fusion-based minute-level rainfall rolling spatiotemporal correction method

Abstract

This invention discloses a method for minute-level rolling spatiotemporal correction of precipitation based on multi-band weather radar fusion, relating to the fields of meteorological observation and short-term nowcasting. The method includes: acquiring multi-band weather radar base data and automatic rain gauge observation data, and preprocessing them; generating a minute-level quantitative precipitation estimation field based on the preprocessed multi-band weather radar base data; extrapolating to generate a minute-level quantitative precipitation forecast field; constructing station residuals and generating a spatial error field; using the spatial error field to spatially correct the minute-level quantitative precipitation forecast field to obtain an intermediate forecast field; extracting precipitation dynamic evolution characteristics to determine time-rolling correction coefficients, and performing time correction on the intermediate forecast field to obtain a corrected minute-level quantitative precipitation forecast field. By generating a precipitation estimation field through multi-band radar fusion, and then combining this with station observations to construct a spatial error field and time correction coefficients, it achieves accurate spatial and temporal correction of minute-level extrapolated forecasts.

Description

Technical Field

[0001] This invention relates to the field of meteorological detection and short-term nowcasting, and more specifically to a minute-level rolling spatiotemporal correction method for rainfall based on multi-band weather radar fusion. Background Technology

[0002] Short-term quantitative precipitation forecasting is a core support for emergency response operations such as urban flooding and prevention of mountain torrents and geological disasters. In current operations, weather radar is the main detection method. It generates quantitative precipitation estimates (QPE) through rainfall intensity inversion, and then generates quantitative precipitation forecasts (QPF) for future periods based on echo extrapolation technology, providing decision-making departments with high-frequency rainfall trend information.

[0003] However, existing single-band radars have inherent limitations in detection capabilities: although the S-band has a wide coverage and weak attenuation, its spatial resolution is low, making it difficult to characterize the fine structure of the core area of ​​strong convection; the X-band has high resolution and fast update speed, but it is easily affected by rain attenuation, which reduces the reliability of long-distance detection. Simple splicing or fixed weight fusion cannot give full play to the complementary advantages of multi-bands, resulting in systematic biases in fused QPE in complex terrain and areas of heavy precipitation, which in turn affects the basic quality of subsequent extrapolation forecasts.

[0004] In the extrapolation forecasting stage, mainstream methods (such as optical flow and TREC) only rely on historical echo motion vectors for advection propagation and lack the ability to respond to the generation, dissipation and intensity abrupt changes of precipitation systems. Therefore, the extrapolated QPF generally has an overall overestimation or underestimation bias, and the error shows significant locality. Affected by factors such as radar blockage and insufficient attenuation correction, the direction and magnitude of the bias vary greatly in different regions. Traditional global constant scaling or simple interpolation correction is difficult to characterize this non-uniform spatial error structure.

[0005] In the time dimension, the error of extrapolation forecasts accumulates rapidly with the increase of lead time and is closely related to the rainfall intensity level: the deviation is relatively stable during weak precipitation processes, while extrapolation under severe convective weather is prone to sudden intensity lag or location shift, resulting in both false alarms and missed alarms. Existing methods often only perform static corrections on spatial errors or simply shift the overall intensity, lacking a dynamic adaptive mechanism for forecast lead time and unable to simultaneously optimize spatial location and intensity deviations within the effective window of short-term operations.

[0006] Therefore, how to design a minute-level rolling spatiotemporal correction method for multi-band weather radar fusion that can collaboratively handle spatial and temporal errors, possess dynamic adaptive capabilities, and meet operational stability requirements, thereby improving the hit rate and reliability of short-term forecast products in complex precipitation processes, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a minute-level rolling spatiotemporal correction method for precipitation based on multi-band weather radar fusion. It aims to refine the minute-level precipitation forecast generated by multi-band radar extrapolation, so as to solve the problems of difficulty in characterizing local spatial errors, inability to dynamically respond to accumulated time deviations, and instability of the correction process during operation, so that the final output forecast is closer to the actual precipitation situation in terms of spatial location and intensity values.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for minute-level rolling spatiotemporal correction of precipitation based on multi-band weather radar fusion includes the following steps: S1. Acquire multi-band weather radar base data and automatic rain gauge observation data, and perform preprocessing; S2. Based on the preprocessed multi-band weather radar base data, generate a minute-level quantitative precipitation estimation field; S3. Based on the minute-level quantitative precipitation estimation field, extrapolate to generate a minute-level quantitative precipitation forecast field; S4. Compare the preprocessed automatic rain gauge observation data with the minute-level quantitative precipitation estimation field, construct the station residuals and generate the spatial error field; S5. Use the spatial error field to spatially correct the minute-level quantitative precipitation forecast field to obtain the intermediate forecast field; S6. Extract the dynamic evolution characteristics of rainfall to determine the time rolling correction coefficient, and use the time rolling correction coefficient to perform time correction on the intermediate forecast field to obtain the corrected minute-level quantitative precipitation forecast field.

[0010] Preferably, S2 includes: Rainfall intensity was inverted from S-band and X-band radar base data respectively to obtain their respective rainfall intensity fields, which were then mapped onto the same Cartesian grid to generate the S-band gridded rainfall intensity field. and X-band grid rain intensity field ; For each grid point, S-band fusion weights are constructed based on radar coverage quality, range, and attenuation factor. Weighting of X-band ; The minute-level quantitative precipitation estimation field was calculated using a weighted average:

[0011] Where i and j are the row and column indices of the grid, and t represents the forecast time.

[0012] Preferably, S3 includes: Based on the start time Minute-level quantitative precipitation estimation field and its adjacent time points Motion estimation is performed to obtain two-dimensional motion information. ; For each lead time Based on two-dimensional motion information Minute-level quantitative precipitation estimation field at the current moment Advection extrapolation was performed to obtain minute-level quantitative precipitation forecast fields. .

[0013] Preferably, in step S4, constructing the site residual includes: At the forecast time t, the minute-level quantitative precipitation estimation field will be... The location of automatic rain gauge k was obtained through spatial interpolation sampling. ,get , where k is the site index; Observe the rainfall intensity at the station With the Perform difference calculations to obtain the station residuals. .

[0014] Preferably, in step S4, generating the spatial error field includes: For any grid point (i,j), effective stations are selected within a preset spatial influence radius h, centered on its spatial location, to form a local station set N(i,j,t); Calculate the spatial weight w(k;i,j,t) of each station k in the local station set w(k;i,j,t) relative to the grid point (i,j):

[0015] in, Let p be the distance from station k to grid point (i,j), p be the distance exponent, and δ be a zero-prevention decimal to prevent the denominator from being zero. The spatial error field E(i,j,t) of the grid point (i,j) is obtained by weighting the station residuals using the spatial weight w(k;i,j,t).

[0016] Preferably, S5 includes: For spatial error field By applying stability constraints, the spatial error field after applying the constraints is obtained. ; constrained spatial error field With minute-level quantitative precipitation forecast field Add them together to get the intermediate forecast field. .

[0017] Preferably, the stability constraints include: Rain area masking constraint: If grid point (i,j) is below the preset threshold θ in both the minute-level quantitative precipitation estimation field and the minute-level quantitative precipitation forecast field, and there is no rain in its neighborhood, then let E(i,j,t)=0; Quantile limiting constraint: Calculate the q-quantile C(t) of the absolute values ​​of the residuals of all available stations at the current forecast time, and introduce a robust limiting coefficient α to restrict the value of the spatial error field E(i,j,t) to the interval [missing information]. Inside; Missed Rainfall Compensation Limit Constraint: When the minute-level quantitative precipitation forecast field is less than the preset missed rainfall threshold at grid point (i,j)... When the spatial error field E(i,j,t) is positive, limited compensation is only allowed within the rain area, and the compensation range must not exceed the preset upper limit for missed rain reporting. .

[0018] Preferably, in step S6, extracting rainfall dynamic evolution characteristics to determine the time-rolling correction coefficient includes: The rainfall is divided into multiple rainfall intensity levels based on the cumulative rainfall within a preset time period. Based on the current rainfall intensity level and the dynamic evolution characteristics of rainfall, the initial time correction coefficient is determined through a preset mapping relationship. ; The initial time correction coefficient is adjusted using a lead decay mechanism to obtain the time correction coefficient. .

[0019] Preferably, the lead decay mechanism reduces the time rolling correction coefficient. With lead time As it increases, it gradually converges towards 1:

[0020] in, This is the time scale for lead decay.

[0021] Preferably, in step S6, the corrected minute-level quantitative precipitation forecast field is represented as follows:

[0022] in, For time-rolling correction factor, For intermediate forecast field, This is the preset physical upper limit of the final output rainfall intensity.

[0023] As can be seen from the above technical solution, compared with the prior art, the technical solution of the present invention has the following beneficial effects: 1. This method generates a high spatiotemporal resolution minute-level quantitative precipitation estimation field through the fusion processing of multi-band radar data, and then extrapolates the initial forecast field based on this. Subsequently, automatic rain gauge observation data is introduced to construct station residuals and generate a spatial error field, thereby realizing the spatial correction of the initial forecast field. Finally, an overall scheme of fusion estimation, extrapolation forecast, station correction, and rolling output is formed, ensuring that minute-level precipitation forecasts have a solid observational foundation and a traceable optimization path.

[0024] 2. By interpolating discrete station residuals into a continuous gridded spatial error field through local weighting, the local error structure of precipitation can be effectively characterized. At the same time, through multiple stability constraints such as rain area masking, quantile limiting, and upper limit of missed rainfall compensation, false corrections and abnormal spikes in rainless areas are suppressed, so that the correction process can still maintain the continuity of spatial distribution in complex scenarios with sparse stations or discrete errors.

[0025] 3. By extracting dynamic evolution characteristics such as the range of rainfall impact, regional average rainfall intensity, and effective rainfall intensity, and combining them with the rainfall intensity classification strategy, a time correction coefficient determination strategy driven by the evolution trend of the precipitation system itself was constructed. At the same time, a lead time decay mechanism was introduced to make the correction intensity naturally weaken as the forecast lead time increases, and operational constraints were imposed by coefficient limit ranges, so as to realize dynamic and adaptive correction of systematic overestimation or underestimation problems in minute-level extrapolation forecasts. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0027] Figure 1 A flowchart of a minute-level rolling spatiotemporal correction method for multi-band weather radar fusion provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of rainfall intensity distribution observed by an automatic rain gauge, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the rainfall intensity distribution in the multi-band radar precipitation forecast before correction, provided in an embodiment of the present invention. Figure 4 A schematic diagram of the corrected multi-band radar precipitation forecast rainfall intensity distribution provided for an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] like Figure 1 As shown, this embodiment provides a minute-level rolling spatiotemporal correction method for rainfall based on multi-band weather radar fusion, including the following steps: S1. Acquire multi-band weather radar base data and automatic rain gauge observation data, and perform preprocessing; S2. Based on the preprocessed multi-band weather radar base data, generate a minute-level quantitative precipitation estimation field; S3. Based on the minute-level quantitative precipitation estimation field, extrapolate to generate a minute-level quantitative precipitation forecast field; S4. Compare the preprocessed automatic rain gauge observation data with the minute-level quantitative precipitation estimation field, construct the station residuals and generate the spatial error field; S5. Use the spatial error field to spatially correct the minute-level quantitative precipitation forecast field to obtain the intermediate forecast field; S6. Extract the dynamic evolution characteristics of rainfall to determine the time rolling correction coefficient, and use the time rolling correction coefficient to perform time correction on the intermediate forecast field to obtain the corrected minute-level quantitative precipitation forecast field.

[0030] This method transforms discrete station deviations into continuous gridded corrections through a spatial error field, enabling precise corrections to be applied to different local areas. The time correction coefficients, driven by rainfall dynamics, allow the correction intensity to adaptively adjust with the evolution of precipitation events. Multiple stability constraints effectively suppress false corrections in rainless areas and amplification of abnormal extreme values. This makes the spatial distribution of extrapolated forecasts closer to actual rainfall distribution, meeting the continuity and reliability requirements of operational needs.

[0031] The following provides further explanation of each step in the above method; In this embodiment, S1, multi-band weather radar base data and automatic rain gauge observation data are acquired and preprocessed; In actual business operations, the system triggers on a rolling basis with a 6-minute cycle, and the start time of each round of reporting is... Accurate to the second, it simultaneously acquires S-band and X-band radar base data from the past hour, as well as 5-minute cumulative rainfall observations from automatic rain gauges during the corresponding time period; To standardize the units of measurement, the 5-minute rainfall at the stations is converted into hourly rainfall intensity. The radar data needs to undergo clutter suppression, outlier removal, and necessary attenuation correction. The station data is then filtered to remove missing data, negative values, and abnormal records that exceed physical thresholds. The latitude and longitude information is also checked for legality. Through time alignment, all data are unified into the same forecast time system to ensure the accuracy of subsequent comparisons.

[0032] In this embodiment S2, based on preprocessed multi-band weather radar base data, a minute-level quantitative precipitation estimation field is generated; including: Rainfall intensity inversion is performed on S-band and X-band radar base data separately. The respective rainfall intensity fields can be obtained using the ZR relation or dual polarization inversion algorithm, and then uniformly mapped onto the same Cartesian grid to generate the S-band gridded rainfall intensity field. and X-band grid rain intensity field ; For each grid point, S-band fusion weights are constructed based on radar coverage quality, range, and attenuation factor. Weighting of X-band ; The minute-level quantitative precipitation estimation field was calculated using a weighted average:

[0033] Where i and j are the row and column indices of the grid, and t represents the forecast time.

[0034] It leverages the complementary advantages of stable S-band coverage and high X-band resolution to generate minute-level quantitative precipitation estimation fields that retain the fine structure of heavy precipitation areas while maintaining continuity in distant regions, providing high-quality initial estimates for subsequent extrapolation and error analysis.

[0035] In this embodiment S3, based on the minute-level quantitative precipitation estimation field, an extrapolation is performed to generate a minute-level quantitative precipitation forecast field; including: Based on the start time Minute-level quantitative precipitation estimation field and its adjacent time points Motion estimation is performed to obtain two-dimensional motion information. ; Specifically, it can be based on the start time of the report. and the QPE sequence from the previous few time points, such as -6min -12 min, the TREC algorithm was used to estimate the two-dimensional motion information of the precipitation system. If motion estimation fails, then roll back to use the motion information from the previous round or adopt the zero displacement assumption; For each lead time Based on two-dimensional motion information Minute-level quantitative precipitation estimation field at the current moment Advection extrapolation was performed to obtain minute-level quantitative precipitation forecast fields. .

[0036] This step generates multi-time extrapolation forecasts based on the continuity of precipitation movement. However, since the formation and dissipation evolution of precipitation systems are not considered, the forecast results have systematic biases and spatial errors, which need to be corrected by further ground observations.

[0037] In this embodiment, S4, the preprocessed automatic rain gauge observation data is compared with the minute-level quantitative precipitation estimation field to construct the station residuals and generate a spatial error field. The construction of site residuals includes: At the forecast time t, the minute-level quantitative precipitation estimation field will be... The location of automatic rain gauge k was obtained through spatial interpolation sampling. ,get , where k is the site index; Observe the rainfall intensity at the station With the Perform difference calculations to obtain the station residuals. .

[0038] Furthermore, generating the spatial error field includes: For any grid point (i,j), effective stations are selected within a preset spatial influence radius h, centered on its spatial location, to form a local station set N(i,j,t); Calculate the spatial weight w(k;i,j,t) of each station k in the local station set w(k;i,j,t) relative to the grid point (i,j):

[0039] in, Let p be the distance from station k to grid point (i,j), p be the distance exponent, and δ be a zero-prevention decimal to prevent the denominator from being zero. The spatial error field E(i,j,t) of the station residuals is obtained by weighting the spatial weights w(k;i,j,t) using the spatial weights w(k;i,j,t). This step transforms discrete station deviations into a continuous gridded error field, accurately characterizing the spatial deviation distribution of QPE relative to actual rainfall, and providing a basis for subsequent local fine-grained corrections.

[0040] In this embodiment S5, the spatial error field is used to spatially correct the minute-level quantitative precipitation forecast field to obtain the intermediate forecast field; including: For spatial error field By applying stability constraints, the spatial error field after applying the constraints is obtained. ; constrained spatial error field With minute-level quantitative precipitation forecast field Add them together to get the intermediate forecast field. .

[0041] Furthermore, stability constraints include: Rain area masking constraint: If grid point (i,j) is below the preset threshold θ in both the minute-level quantitative precipitation estimation field and the minute-level quantitative precipitation forecast field, and there is no rain in its neighborhood, then let E(i,j,t)=0; Quantile limiting constraint: Calculate the q-quantile C(t) of the absolute values ​​of the residuals of all available stations at the current forecast time, and introduce a robust limiting coefficient α to restrict the value of the spatial error field E(i,j,t) to the interval [missing information]. Inside; Missed Rainfall Compensation Limit Constraint: When the minute-level quantitative precipitation forecast field is less than the preset missed rainfall threshold at grid point (i,j)... When the spatial error field E(i,j,t) is positive, limited compensation is only allowed within the rain area, and the compensation range must not exceed the preset upper limit for missed rain reporting. .

[0042] Through additive correction and robust constraints, the minute-level quantitative precipitation forecast field was effectively corrected. This reduces local spatial bias and avoids false precipitation and extreme anomalous values, providing a more spatially accurate intermediate field for time correction.

[0043] In this embodiment, S6, the dynamic evolution characteristics of rainfall are extracted to determine the time rolling correction coefficient, and the time rolling correction coefficient is used to correct the intermediate forecast field in time to obtain the corrected minute-level quantitative precipitation forecast field. Among them, the extraction of rainfall dynamic evolution characteristics to determine the time-rolling correction coefficient includes: Based on the cumulative rainfall over a preset time period, the rainfall is divided into multiple rainfall intensity levels (such as 0-10, 10-20, 20-40, 40-80, >80 mm / h). Based on the current rainfall intensity level and the dynamic evolution characteristics of rainfall, the initial time correction coefficient is determined through a preset mapping relationship. ; The initial time correction coefficient is adjusted using a lead decay mechanism to obtain the time correction coefficient. .

[0044] Furthermore, determining the preset mapping relationship includes: Collect minute-level quantitative precipitation estimation fields, automatic rain gauge observation data, and minute-level quantitative precipitation forecast fields for historical periods to form a historical sample set. Each sample corresponds to a reporting start time and multiple lead times. For each historical sample, based on the minute-level quantitative precipitation estimation field and station observations in the sample, the time-by-time residuals of each station are calculated, and the dynamic evolution characteristics of rainfall in the sample within the preset rolling period are extracted: the trend of rainfall impact range change S, the trend of regional average rainfall intensity change A, and the trend of effective average rainfall intensity change R. The sample is divided according to the rainfall intensity level. Within each rainfall intensity level, all candidate time correction coefficients are iterated. The optimal correction coefficient for the sample under the rainfall intensity level is determined with the goal of minimizing the root mean square error between the forecast and the observation after correction for all stations under that level. Historical samples were grouped according to rainfall intensity level. Multiple regression fitting was performed on the rainfall dynamic evolution characteristics S, A, and R of all samples in each group and the corresponding optimal correction coefficients to establish a mapping function from S, A, and R to the time correction coefficients for each rainfall intensity level.

[0045] Furthermore, the lead decay mechanism increases the time rolling correction factor. With lead time As it increases, it gradually converges towards 1:

[0046] in, This is the time scale for lead decay.

[0047] Furthermore, it also includes the calculated time-rolling correction factor. Trim the data to ensure it always remains within the preset business safety zone. , Within this embodiment, [ , [0.6, 1.4].

[0048] Furthermore, the revised minute-level quantitative precipitation forecast field is expressed as follows:

[0049] in, For time-rolling correction factor, For intermediate forecast field, This is the preset physical upper limit of the final output rainfall intensity.

[0050] This step uses dynamic feature-driven coefficient adjustment to enable the correction intensity to adapt to the evolution of precipitation processes, while also taking into account the increase in forecast lead time, ultimately outputting a corrected forecast field with reasonable intensity and spatiotemporal continuity.

[0051] Furthermore, the method also includes: modifying the minute-level quantitative precipitation forecast field at the station location before correction. Compared with the revised minute-level quantitative precipitation forecast field The test and evaluation were carried out, and the hit rate (TS), bias (BIAS), and root mean square error (RMSE) were calculated under multiple rainfall intensity thresholds. Based on the evaluation results, the relevant constraint parameters or correction coefficients were continuously optimized.

[0052] In this embodiment, the minute-level rainfall rolling spatiotemporal correction method based on multi-band weather radar fusion achieves systematic optimization of minute-level rainfall forecasts through multi-band radar fusion and spatiotemporal rolling correction strategies. The technical advantages of this method are fully demonstrated in actual heavy rainfall forecasting scenarios. At the spatial correction level, such as Figure 2 The station observations shown indicate that a center of heavy precipitation with wind speeds of 50–250 mm / h occurred near 28.1°–28.3°N and 112.8°E; Figure 3 As shown, the multi-band radar forecast before correction was significantly weaker than expected in this area, failing to capture heavy precipitation of more than 50 mm / h. To address this issue, steps S4-S5 construct a spatial error field, transforming the observation bias of discrete stations into a continuous gridded correction. Combined with multiple stability constraints such as rain area masking and quantile limiting, the false corrections in rainless areas and the amplification of abnormal extreme values ​​are effectively suppressed, enabling the correction to be accurately applied to areas of heavy precipitation. Furthermore, at the time correction level, step S6 extracts the dynamic evolution characteristics of rainfall and determines the time rolling correction coefficient by combining it with the lead decay mechanism, so that the correction intensity can be adaptively adjusted according to the evolution trend of precipitation process; such as Figure 4 As shown, after correction using the method of this embodiment, the forecast field accurately reproduced the center of heavy precipitation of 50-250 mm / h in the vicinity of 28.1°-28.3°N and 112.8°E, and the overall rainfall intensity distribution was consistent with... Figure 1 The observations from the stations are highly consistent, significantly improving the ability to pinpoint areas of heavy rainfall; By refining the local error field and adaptively adjusting the rolling correction coefficient over time, the extrapolation forecast's spatial distribution area is made closer to the actual rainfall distribution, and the intensity evolution is more in line with the formation and dissipation patterns of the precipitation system. Under the premise of maintaining the continuity and stability of operational operations, it effectively improves the accuracy and reliability of minute-level rainfall forecasts.

[0053] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0054] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion, characterized in that, Includes the following steps: S1. Acquire multi-band weather radar base data and automatic rain gauge observation data, and perform preprocessing; S2. Based on the preprocessed multi-band weather radar base data, generate a minute-level quantitative precipitation estimation field; S3. Based on the minute-level quantitative precipitation estimation field, extrapolate to generate a minute-level quantitative precipitation forecast field; S4. Compare the preprocessed automatic rain gauge observation data with the minute-level quantitative precipitation estimation field, construct the station residuals and generate the spatial error field; S5. Use the spatial error field to spatially correct the minute-level quantitative precipitation forecast field to obtain the intermediate forecast field; S6. Extract the dynamic evolution characteristics of rainfall to determine the time rolling correction coefficient, and use the time rolling correction coefficient to perform time correction on the intermediate forecast field to obtain the corrected minute-level quantitative precipitation forecast field.

2. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, S2 includes: Rainfall intensity was inverted from S-band and X-band radar base data respectively to obtain their respective rainfall intensity fields, which were then mapped onto the same Cartesian grid to generate the S-band gridded rainfall intensity field. and X-band grid rain intensity field ; For each grid point, S-band fusion weights are constructed based on radar coverage quality, range, and attenuation factor. Weighting of X-band ; The minute-level quantitative precipitation estimation field was calculated using a weighted average: Where i and j are the row and column indices of the grid, and t represents the forecast time.

3. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, S3 includes: Based on the start time Minute-level quantitative precipitation estimation field and its adjacent time points Motion estimation is performed to obtain two-dimensional motion information. ; For each lead time Based on two-dimensional motion information Minute-level quantitative precipitation estimation field at the current moment Advection extrapolation was performed to obtain minute-level quantitative precipitation forecast fields. .

4. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, In S4, constructing the site residuals includes: At the forecast time t, the minute-level quantitative precipitation estimation field will be... The location of automatic rain gauge k was obtained through spatial interpolation sampling. ,get , where k is the site index; Observe the rainfall intensity at the station With the Perform difference calculations to obtain the station residuals. .

5. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, In step S4, generating the spatial error field includes: For any grid point (i,j), effective stations are selected within a preset spatial influence radius h, centered on its spatial location, to form a local station set N(i,j,t); Calculate the spatial weight w(k;i,j,t) of each station k in the local station set w(k;i,j,t) relative to the grid point (i,j): in, Let p be the distance from station k to grid point (i,j), p be the distance exponent, and δ be a zero-prevention decimal to prevent the denominator from being zero. The spatial error field E(i,j,t) of the grid point (i,j) is obtained by weighting the station residuals using the spatial weight w(k;i,j,t).

6. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, S5 includes: For spatial error field By applying stability constraints, the spatial error field after applying the constraints is obtained. ; constrained spatial error field With minute-level quantitative precipitation forecast field Add them together to get the intermediate forecast field. .

7. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 6, characterized in that, The stability constraints include: Rain area masking constraint: If grid point (i,j) is below the preset threshold θ in both the minute-level quantitative precipitation estimation field and the minute-level quantitative precipitation forecast field, and there is no rain in its neighborhood, then let E(i,j,t)=0; Quantile limiting constraint: Calculate the q-quantile C(t) of the absolute values ​​of the residuals of all available stations at the current forecast time, and introduce a robust limiting coefficient α to restrict the value of the spatial error field E(i,j,t) to the interval [missing information]. Inside; Missed Rainfall Compensation Limit Constraint: When the minute-level quantitative precipitation forecast field is less than the preset missed rainfall threshold at grid point (i,j)... When the spatial error field E(i,j,t) is positive, limited compensation is only allowed within the rain area, and the compensation range must not exceed the preset upper limit for missed rain reporting. .

8. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, In step S6, extracting the dynamic evolution characteristics of rainfall to determine the time-rolling correction coefficient includes: The rainfall is divided into multiple rainfall intensity levels based on the cumulative rainfall within a preset time period. Based on the current rainfall intensity level and the dynamic evolution characteristics of rainfall, the initial time correction coefficient is determined through a preset mapping relationship. ; The initial time correction coefficient is adjusted using a lead decay mechanism to obtain the time correction coefficient. .

9. A method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 8, characterized in that, The advance decay mechanism adjusts the time rolling correction factor. With lead time As it increases, it gradually converges towards 1: in, This is the time scale for lead decay.

10. The method for minute-level rolling spatiotemporal correction of rainfall based on multi-band weather radar fusion according to claim 1, characterized in that, In S6, the corrected minute-level quantitative precipitation forecast field is represented as follows: in, For time-rolling correction factor, For intermediate forecast field, This is the preset physical upper limit of the final output rainfall intensity.

Images