A multi-source data fusion rain measuring radar high-precision rainfall inversion method

Abstract

The present application relates to the technical field of rainfall inversion, in particular to a multi-source data fusion rain measuring radar high-precision rainfall inversion method, comprising S1: inversion target definition and multi-source data selection; S2: multi-source data preprocessing and cloud layer influence prediction; S3: fusion algorithm selection and correction model construction; S4: history data assimilation; S5: dynamic weight optimization; The judgment logic can effectively distinguish the distorted data and the effective data shielded by the cloud layer, avoid blindly using the shielded precipitation particle detection data for inversion, reduce the appearance of false rainfall area or missed area, make the originally large-deviation precipitation particle information due to cloud layer reflection and absorption signal closer to the real situation, significantly improve the accuracy of rain intensity inversion, and the short-term historical data processed through the physical model and the long-term climate background field are integrated into the real-time inversion process, so that more smooth, stable and reasonable rainfall inversion results can be provided, and the accuracy of the short-term forecast of the rainfall product is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of rainfall inversion technology, and in particular to a high-precision rainfall inversion method for rainfall radar that integrates multi-source data. Background Technology

[0002] High-precision rainfall monitoring is a core foundation for hydrological forecasting, flash flood warnings, and water resource management. Traditional inversion methods based on single rainfall radars are limited by the inherent limitations of their physical observation principles, making it difficult to meet the growing application demands in terms of accuracy and reliability. Therefore, multi-source data fusion technology has emerged, aiming to construct high-precision rainfall inversion models by integrating heterogeneous information such as rainfall radar base data, ground rain gauges, satellite remote sensing, and numerical weather prediction, in order to correct systematic errors and compensate for insufficient spatiotemporal coverage. However, existing implementation schemes of this technical approach still face significant challenges in dealing with complex atmospheric structures and meeting the requirements for continuous temporal data.

[0003] However, during the implementation of the above technical solution, at least the following technical problems were discovered:

[0004] Regarding spatial inversion accuracy, existing fusion methods fail to adequately consider the interference of cloud layer structure, leading to distortion in the detection of the core precipitation particle layer. In the atmosphere, water droplets in low clouds (0-2 km) and ice crystals in high clouds (5-18 km) produce complex reflection, absorption, and secondary scattering effects on radar electromagnetic waves and satellite remote sensing signals passing through them, severely interfering with the accurate detection of precipitation particles located in the critical precipitation-forming layer at 2-6 km. This physical interference causes systematic overestimation or underestimation of precipitation intensity in radar inversion, blurring of precipitation zone boundaries, and even the generation of false precipitation signals or missed reports, directly affecting the accuracy of short-term heavy precipitation identification and flood forecasting. Although traditional methods have introduced multi-source data types (such as rain gauges, radar data, and meteorological cloud images) for correction, they usually remain at the statistical level of weighted averaging or simple regression, lacking in-depth utilization of data from raindrop spectrometers that can provide microphysical characteristics of particles, and failing to construct a physical mechanism-based error correction and complementarity model that can effectively remove cloud interference.

[0005] Regarding the stability of time-series products, existing methods are mostly limited to instantaneous fusion, resulting in inversion results lacking physical consistency and failing to support time-sensitive applications. Traditional fusion frameworks primarily process observational data within the current moment or a very short time window, with the output essentially being a series of independent time slices. This method ignores the continuous evolutionary patterns and historical climate statistical characteristics of the precipitation system itself, leading to two prominent problems: first, the inversion products are prone to unreasonable and drastic fluctuations in the time series, failing to smoothly reflect the true process of precipitation generation and dissipation; second, it cannot effectively suppress inversion fluctuations caused by random noise, instantaneous interference, or temporary missing data, resulting in a significant decrease in reliability when data quality is unstable. The root cause lies in the fact that while existing processes emphasize analyzing the characteristics of multi-source rainfall data, establishing a unified standardized framework, and striving to solve the problems of asynchronous data acquisition and spatial coverage differences, when constructing fusion inversion algorithms, whether using weighted methods or simple machine learning models, there is a lack of explicit constraints and assimilation of the time dimension. This results in products that cannot meet the application requirements of hydrological model simulation and refined nowcasting, which have extremely high requirements for the continuity and temporal consistency of input data. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a high-precision rainfall inversion method for rain-measuring radar based on multi-source data fusion, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A high-precision rainfall inversion method based on multi-source data fusion from a rainfall radar includes the following steps:

[0009] S1: Definition of Inversion Target and Selection of Multi-Source Data: Clarify the spatiotemporal resolution and accuracy indicators of rainfall inversion, select three types of core data sources as multi-source data, and unify the time standard and spatial coordinate system of all data sources;

[0010] S2: Multi-source data preprocessing and cloud impact prediction: Noise reduction, outlier removal, and format conversion are performed on multi-source data; cloud parameters are extracted; and the attenuation coefficient k is determined based on radar wavelength. This is then analyzed using A... est =k×H calculates the signal attenuation;

[0011] S3: Selection of fusion algorithm and construction of correction model: Select the appropriate algorithm according to the terrain complexity and precipitation type, build an embedded cloud attenuation correction module, and embed the cloud attenuation correction module into the algorithm. In the scenario that needs correction, cloud thickness and attenuation amount are included in the model input.

[0012] S4: Historical Data Assimilation: Obtain the inversion results through the algorithm in S3, optimize and calibrate the current inversion results, and collect short-term and long-term historical data.

[0013] S5: Dynamic weight optimization: Weights are allocated based on current data, short-term historical data, and long-term historical data. The current real-time inversion result is the core, while short-term and long-term historical data are used as optimization aids.

[0014] In one possible implementation, in step S1, the multi-source data includes rain-measuring radar data and satellite remote sensing data. The rain-measuring radar data includes reflectivity factor Z and radial velocity V, and the satellite remote sensing data includes brightness temperature and microwave radiation data.

[0015] In one possible implementation, the formula for calculating the attenuation in step S2 is: A = k * L;

[0016] Where A is the total attenuation of the radar signal;

[0017] k is the specific attenuation coefficient of the cloud layer;

[0018] L represents the effective path length of radar waves through the cloud layer.

[0019] In one possible implementation, in step S2, the thickness H of the target cloud layer is obtained from radar data or satellite data. Based on the radar wavelength and a preliminary judgment of the cloud type (liquid / ice crystal), a typical value of an attenuation coefficient k is set, such as k being between 0.5 and 2.0 dB / km for S-band radar and liquid low clouds.

[0020] Let the estimated signal attenuation be A. est， The formula for calculating the estimated signal attenuation is: A est =k*H;

[0021] Threshold A was set after calibration based on radar and local climate characteristics. threshold :

[0022] If A est threshold This indicates that the cloud layer is thin or has weak attenuation characteristics, and its impact on the radar signal is negligible. No correction is needed, and the original radar echo data can be used directly for rainfall inversion.

[0023] If A est >=A threshold This indicates that the cloud layer is thick or has strong attenuation characteristics, resulting in significant attenuation of the radar signal. Correction is required, and the original echo data should be sent to S3.

[0024] In one possible implementation, the reflectivity factor of the observation received by the radar antenna after cloud attenuation is set as Z. obs ;

[0025] ​The true reflectivity factor of precipitation particles that the radar should receive under conditions of no cloud cover will be set to Z. true ;

[0026] Let Z be the actual total attenuation of the signal in the propagation path. true ;

[0027] Z true The calculation formula is set as: Z true =Z obs +A true ;

[0028] Among them, A true Since it is unknown, the estimated attenuation A is used. est Replace A true ;

[0029] Let the logical formula for deduction be: Z true_est =Z obs +A est Z true_est This is the estimated true reflectivity factor after attenuation correction.

[0030] In one possible implementation, A is used again. est =k*H calculates the estimated signal attenuation, at which point A est For the correction amount, use the correction formula Z. true_est =Z obs +A est Calculate the corrected reflectivity factor, and then apply the corrected Z... true_est Substituting the ZR relationship, we finally obtain the corrected rainfall intensity R, which is closer to the true value. est .

[0031] In one possible implementation, during historical data assimilation, short-term historical data—that is, the historical inversion results obtained at the same time the previous day—is denoted as R. historical_short This data must be consistent with the current R est Similarly, the high-precision product generated through steps S1 to S3 of this method ensures consistency between the data source and the processing logic; long-term historical data: that is, acquiring historical inversion datasets from the past 3 to 5 years that fall within the same climatic season as the current date, and generating a climatological mean precipitation field, denoted as R, by averaging this dataset. clim .

[0032] In one possible implementation, during historical data assimilation, strict consistency preprocessing is required when using both short-term and long-term historical data. This includes resampling all historical data to a spatiotemporal resolution and spatial coordinate system consistent with the current inversion target, ensuring comparability and fusion between the data.

[0033] In one possible implementation, in dynamic weight optimization, three weight coefficients are specifically defined: W current R represents the current real-time inversion result est The weight of W is the most important, reflecting the importance of real-time detection information; its base value is set between 0.6 and 0.8. short R represents short-term historical data historical_short The weighting of W reflects the persistence of weather patterns, with a base value set between 0.1 and 0.3; seasonal Represents seasonality R clim The weights, reflecting climate patterns, are set between 0.05 and 0.1; these three weighting coefficients must sum to 1, i.e., W current +W short +W seasonal =1.

[0034] In one possible implementation, dynamic weight optimization involves dynamic fine-tuning based on real-time observations, and comparing the current inversion result R in real time. est and short-term historical data R historical_short The system automatically adjusts the weighting of the deviation between the current inversion result and the rain gauge observations, as well as the deviation between short-term historical data and the rain gauge observations. If the short-term historical data matches the actual situation better, and this situation persists for several periods, the system will automatically adjust the weighting of W. short Increase W by 0.05 and decrease it accordingly. current This will allow the continuous laws of history to play a greater calibrating role.

[0035] Beneficial effects compared to existing technologies:

[0036] 1. This solution addresses the detection challenges caused by low and high cloud cover. The judgment logic effectively distinguishes between distorted and valid data obscured by clouds, preventing the blind use of obscured precipitation particle detection data for inversion and reducing the occurrence of false rainfall areas or missed areas. The calculation logic specifically corrects distorted data, making precipitation particle information, which was previously significantly biased due to cloud reflection and signal absorption, more closely reflect reality, thus significantly improving the accuracy of rainfall intensity inversion. These two logics work together to improve the depiction of precipitation spatial distribution, making the presentation of precipitation range and intensity more realistic, and enhance the continuity of rainfall temporal changes. This provides more reliable data support for applications such as short-term precipitation warnings and flood forecasts, greatly improving the overall accuracy and practical value of multi-source data fusion rainfall inversion.

[0037] 2. In this scheme, short-term historical data processed by the physical model and long-term climate background field are integrated into the real-time inversion process. The dynamic weight allocation mechanism ensures the dominant position of real-time observation data. At the same time, by utilizing the continuous patterns of historical data and climate statistical characteristics, the inversion fluctuations caused by random noise or instantaneous interference in the data are effectively smoothed out. The scheme accurately reflects the instantaneous weather conditions and conforms to the continuous evolution of weather. It can provide smoother, more stable and reasonable rainfall inversion results, which greatly enhances the accuracy of rainfall products in short-term forecasts.

[0038] 3. This scheme deeply explores the correlation value of particle microphysical characteristics and constructs an error correction and complementarity model based on physical mechanisms. It breaks through the limitations of traditional statistical weighted averaging or simple regression. This scheme extracts cloud parameters, determines the specific attenuation coefficient, and accurately calculates the signal attenuation by combining radar wavelength and cloud type. It removes cloud interference from the perspective of physical principles rather than relying on surface data statistical correlation. It fully leverages the role of particle microphysical related information contained in multi-source data, making rainfall intensity inversion not only based on data appearance fusion but also more in line with the laws of atmospheric physical processes. This further improves the accuracy of spatial inversion, makes rainfall intensity estimation more realistic, and effectively avoids the problem of systematic overestimation or underestimation.

[0039] 4. In this scheme, an explicit constraint and assimilation mechanism in the time dimension is established to enhance the physical coherence of time-series products. Through rigorous historical data consistency preprocessing, the scheme integrates short-term historical data processed with the same high-standard procedures and long-term historical data representing the climate background field into the real-time inversion process. By using dynamic weight optimization to form a constraint relationship in the time dimension, it avoids the fragmentation of inversion results caused by traditional instantaneous fusion. Furthermore, it effectively constrains the inversion results on the time series through the continuous evolution law of the precipitation system and the climate statistical characteristics contained in the historical data. This smooths out unreasonable and drastic fluctuations and suppresses inversion fluctuations caused by random noise, instantaneous interference, or temporary missing data. Even when the data quality is unstable, the reliability of the inversion results can be guaranteed by relying on historical data, meeting the high requirements of hydrological model simulation and refined nowcasting for the continuity and temporal consistency of input data.

[0040] 5. In this scheme, the accuracy of the rainfall zone boundary delineation is optimized to reduce false precipitation and missed reports. Through the threshold judgment logic of cloud impact prediction in step S2, the degree of cloud interference is accurately distinguished. For scenarios that do not require correction, high-quality original data is directly used. For scenarios that require correction, the correction model in step S3 is used to specifically compensate for signal attenuation loss. This reduces the problem of blurred rainfall zone boundaries caused by cloud reflection, absorption, and secondary scattering from the source. Compared with the general treatment of cloud interference by traditional methods, this scheme achieves refined control of cloud interference, making the definition of rainfall zone boundaries clearer and more accurate. It significantly reduces the probability of false precipitation signals and missed precipitation reports, provides a more accurate basis for regional division for short-term heavy precipitation identification, and further improves the accuracy of flood forecasting. Attached Figure Description

[0041] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0042] Figure 1 This is a schematic diagram of the overall process structure of the present invention. Detailed Implementation

[0043] Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention can also be implemented in various different forms, and therefore the present invention is not limited to the embodiments described below.

[0044] The technical solution in this application embodiment is to solve the problems mentioned in the background art, and the overall idea is as follows:

[0045] Example:

[0046] Please refer to Figure 1As shown in the figure, this embodiment introduces a high-precision rainfall inversion method for rain-measuring radar by multi-source data fusion, including the following steps:

[0047] S1: Definition of Inversion Target and Selection of Multi-Source Data: Determine the spatiotemporal resolution of rainfall inversion, such as 1 hour / time, 1km×1km grid, and accuracy indicators, such as RMSE≤1.0mm, R²≥0.8. Select three types of core data sources as multi-source data, with rain-measuring radar data as the main data source, satellite remote sensing data to fill coverage blind spots, brightness temperature data and microwave radiation data collected, and ground rain gauge data as the benchmark calibration basis. Unify the time standard and spatial coordinate system of all data sources to ensure spatiotemporal benchmark consistency and lay the foundation for subsequent fusion and correction.

[0048] S2: Multi-source data preprocessing and cloud impact prediction: Noise reduction, outlier removal, and format conversion are performed on multi-source data. Radar data is freed from ground clutter and attenuation. Satellite data undergoes projection conversion and quality screening. Rain gauge data is interpolated to a continuous grid. Cloud parameters are extracted, such as the thickness of low clouds (0.1-2.0 km) and high clouds (6.0-18.0 km). The attenuation coefficient k is determined based on the radar wavelength. This is then analyzed using A... est =k×H calculates the signal attenuation, and sets a threshold A. threshold If the value is 2.0dB, determine whether data correction is needed; if the value is below the threshold, use the original data directly.

[0049] S3: Algorithm Selection and Model Construction: Based on terrain complexity and precipitation type, select an appropriate algorithm (traditional mathematical algorithm, intelligent algorithm, or optimal interpolation algorithm). Embed a cloud attenuation correction module into the component and the algorithm. In scenarios requiring correction, incorporate cloud thickness and attenuation into the model input, using Z... true_est =Z obs +A est Inversely calculate the true reflectivity factor; use the original data directly without needing to correct the scene.

[0050] In step S1, the multi-source data includes rain-measuring radar data and satellite remote sensing data. The rain-measuring radar data includes reflectivity factor Z and radial velocity V, while the satellite remote sensing data includes brightness temperature and microwave radiation data.

[0051] In step S2, the formula for calculating the attenuation is: A = k * L;

[0052] Where A is the total attenuation of the radar signal, which is the core indicator used for judgment;

[0053] k is the specific attenuation coefficient of the cloud layer, which describes the degree to which radar waves are attenuated by the cloud layer per unit distance. k depends on the radar wavelength (the S-band is less affected by attenuation, the C-band is more affected, and the X-band is most affected), the size and concentration of cloud droplets (the larger and denser the cloud droplets, the larger the k value), and the phase of the cloud layer (clouds composed of liquid water (such as low clouds) are attenuated more severely than clouds composed of ice crystals (such as high clouds).

[0054] L represents the effective path length of radar waves through the cloud layer.

[0055] In step S2, the thickness H of the target cloud layer is obtained from radar data or satellite data. Based on the radar wavelength and the preliminary judgment of the cloud type (liquid / ice crystal), a typical value of the attenuation coefficient k is set. For example, for S-band radar and liquid low clouds, k is between 0.5-2.0 dB / km.

[0056] Let the estimated signal attenuation be A. est， The formula for calculating the estimated signal attenuation is: A est =k*H;

[0057] Threshold A was set after calibration based on radar and local climate characteristics. threshold :

[0058] If A est threshold This indicates that the cloud layer is thin or has weak attenuation characteristics, and its impact on the radar signal is negligible. No correction is needed, and the original radar echo data can be used directly for rainfall inversion.

[0059] If A est >=A threshold This indicates that the cloud layer is thick or has strong attenuation characteristics, resulting in significant attenuation of the radar signal. Correction is required, and the original echo data should be sent to S3.

[0060] Let Z be the reflectivity factor of the radar antenna after attenuation by clouds. obs ;

[0061] The true reflectivity factor of precipitation particles that the radar should receive under conditions of no cloud cover will be set to Z. true ;

[0062] Let Z be the actual total attenuation of the signal in the propagation path. true ;

[0063] Z true The calculation formula is set as: Z true =Z obs +A true ;

[0064] Among them, A true ​Since it is unknown, the estimated attenuation A is used. est Replace A true ;

[0065] Let the logical formula for deduction be: Z true_est =Z obs +A est Z true_est This is the estimated true reflectivity factor after attenuation correction.

[0066] Use A again est =k*H calculates the estimated signal attenuation. At this point, A... est For the correction amount, use the correction formula Z. true_est =Z obs +A est Calculate the corrected reflectivity factor, and then apply the corrected Z... true_est Substituting the ZR relationship (Z = reflectivity factor Z detected by radar) into an empirical formula to calculate the true ground rainfall intensity R, is the core step in the process of rain-measuring radar moving from "cloud measurement" to "rain measurement" (e.g., R = a * Z^b, where a and b are localized empirical coefficients). This ultimately yields a corrected rainfall intensity R that is closer to the true value. est .

[0067] S4: Historical Data Assimilation: Using the algorithm in S3, the inversion results are obtained and optimized to better reflect the continuous evolution of weather patterns. This involves collecting and preparing two types of historical data sources:

[0068] 1. Short-term historical data: This refers to obtaining the historical inversion results for the same time the previous day, denoted as R. historical_short This data must be consistent with the current R est Similarly, the high-precision products generated through steps S1 to S3 of this method ensure the consistency between the data source and the processing logic.

[0069] 2. Long-term historical data: This involves acquiring historical inversion datasets from the past 3 to 5 years that fall within the same climatic season as the current date. By averaging this dataset, a climatological mean precipitation field, denoted as R, is generated to characterize the background climate field. clim The processing formula is: R clim (i,j)=(1 / N)×Σ[R year_k [i,j)], where k=1 to N, R clim (i,j) represents the value of the climatological mean precipitation field at the grid point in the i-th row and j-th column, where N is the total number of years involved in the calculation.

[0070] R year_k(i,j) represents the rainfall intensity value at the same grid point (i,j) in the same period of year k;

[0071] When using short-term and long-term historical data, it is necessary to strictly perform consistent preprocessing, including resampling all historical data to a spatiotemporal resolution and spatial coordinate system consistent with the current inversion target, to ensure the comparability and fusion of data.

[0072] S5: Dynamic Weight Optimization: Weights are assigned based on current data, short-term historical data, and long-term historical data. The current real-time inversion result is the core, with short-term and long-term historical data used as optimization aids. Three weight coefficients are specifically defined: W current R represents the current real-time inversion result est The weight of W is the most important, reflecting the importance of real-time detection information; its base value is usually set between 0.6 and 0.8. short R represents short-term historical data historical_short The weight of W reflects the persistence of weather patterns, and its base value is usually set between 0.1 and 0.3; seasonal Represents seasonality R clim The weights, reflecting climate patterns, are set between 0.05 and 0.1; these three weighting coefficients must sum to 1, i.e., W current +W short +W seasonal =1;

[0073] To enhance intelligence and adaptability, these weights can be dynamically fine-tuned based on real-time observations. This is achieved by comparing the current inversion results R in real time. est and short-term historical data R historical_short The deviation between the current inversion result and the measured values ​​from ground rain gauges is automatically adjusted in terms of weight allocation. Specifically, the deviation between the current inversion result and the rain gauge observations, as well as the deviation between short-term historical data and the rain gauge observations, are calculated. If it is found that the short-term historical data matches the actual situation better, and this situation persists for several periods, W will be automatically adjusted. short Increase W by 0.05 and decrease it accordingly. current This allows historical patterns to play a greater calibrating role. This dynamic adjustment mechanism ensures adaptability to changes in data characteristics under different weather conditions; after the weights are determined, the final fusion calculation is performed; the specific calculation formula is: R est_final =W current ×R est +W short ×R historical_short +W seasonal ×R clim , where R est_finalThe final output is a high-precision rainfall intensity product optimized from historical data. This is the final result, which makes the rainfall intensity estimation in areas with greater uncertainty, such as cloud edges and radar coverage blind spots, more reasonable and reliable. It can provide a smoother and more natural spatiotemporal transition, effectively avoiding unreasonable and drastic changes in the precipitation field. When real-time observation data is temporarily missing or of poor quality, it can also rely on historical data to provide relatively reliable estimates, ensuring the continuity and availability of the inversion product.

[0074] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A high-precision rainfall inversion method using multi-source data fusion from a rainfall radar, characterized in that, Includes the following steps: S1: Definition of Inversion Target and Selection of Multi-Source Data: Clarify the spatiotemporal resolution and accuracy indicators of rainfall inversion, select three types of core data sources as multi-source data, unify the time standard and spatial coordinate system of all data sources, and the multi-source data include rainfall radar data and satellite remote sensing data. Rainfall radar data includes reflectivity factor Z and radial velocity V, and satellite remote sensing data includes brightness temperature and microwave radiation data. S2: Multi-source data preprocessing and cloud impact prediction: Denoise, outlier removal and format conversion are performed on multi-source data, cloud parameters are extracted, attenuation coefficient k is determined in combination with radar wavelength, and signal attenuation is calculated. S3: Selection of fusion algorithm and construction of correction model: Select the appropriate algorithm according to the terrain complexity and precipitation type, build an embedded cloud attenuation correction module, and embed the cloud attenuation correction module into the algorithm. In the scenario that needs correction, cloud thickness and attenuation amount are included in the model input. S4: Historical Data Assimilation: Obtain the inversion results through the algorithm in S3, optimize and calibrate the current inversion results, and collect short-term and long-term historical data. S5: Dynamic weight optimization: The weights are then allocated based on the current data, short-term historical data, and long-term historical data, with the current real-time inversion results as the core and the short-term and long-term historical data as optimization aids. In step S2, the formula for calculating the attenuation is: A = k * L; Where A is the total attenuation of the radar signal; k is the specific attenuation coefficient of the cloud layer, which describes the degree to which radar waves are attenuated by the cloud layer per unit distance. k depends on the radar wavelength, the size and concentration of cloud droplets, and the phase of the cloud layer. L is the effective path length of the radar wave through the cloud layer; Obtain the thickness H of the target cloud layer from radar or satellite data, and make a preliminary judgment based on the radar wavelength and cloud type. The cloud type includes liquid clouds and ice crystal clouds. Set a typical value for an attenuation coefficient k. Let the estimated signal attenuation be A. est， The formula for calculating the estimated signal attenuation is: A est =k*H; H represents the thickness of the target cloud layer; Threshold A was set after calibration based on radar and local climate characteristics. threshold : If A est threshold This indicates that the cloud layer is thin or has weak attenuation characteristics, and its impact on the radar signal is negligible. No correction is needed, and the original radar echo data can be used directly for rainfall inversion.​ If A est >=A threshold This indicates that the cloud layer is thick or has strong attenuation characteristics, and the radar signal is significantly attenuated, requiring correction. The original echo data is then sent to S3. Let Z be the reflectivity factor of the radar antenna after attenuation by clouds. obs ; The true reflectivity factor of precipitation particles that the radar should receive under conditions of no cloud cover will be set to Z. true ; Let A be the actual total attenuation of the signal in the propagation path. true ; Z true The calculation formula is set as: Z true =Z obs +A true ; Among them, A true Since it is unknown, the estimated attenuation A is used. est Replace A true ; Let the logical formula for deduction be: Z true_est =Z obs +A est Z true_est This is the estimated true reflectivity factor after attenuation correction.

2. The high-precision rainfall inversion method for rain-measuring radar based on multi-source data fusion as described in claim 1, characterized in that, Use A again est =k*H calculates the estimated signal attenuation, at which point A est For the correction amount, use the correction formula Z. true_est =Z obs +A est Calculate the corrected reflectivity factor, and then apply the corrected Z... true_est Substituting the ZR relationship, we finally obtain the corrected rainfall intensity R, which is closer to the true value. est .

3. The high-precision rainfall inversion method for rain-measuring radar based on multi-source data fusion as described in claim 2, characterized in that, In historical data assimilation, short-term historical data refers to the historical inversion results obtained at the same time the previous day, denoted as R. historical_short This data must be consistent with the current R est Similarly, it is generated through steps S1 to S3 of this method to ensure the consistency of the data source and processing logic; Long-term historical data: This refers to obtaining historical inversion datasets from the past 3 to 5 years that fall within the same climatic season as the current date. By averaging this dataset, a climatological mean precipitation field representing the climatic background field is generated, denoted as R. clim .

4. The high-precision rainfall inversion method for rain-measuring radar based on multi-source data fusion as described in claim 3, characterized in that, In historical data assimilation, consistent preprocessing is required when using short-term and long-term historical data. This includes resampling all historical data to a spatiotemporal resolution and spatial coordinate system consistent with the current inversion target to ensure comparability and fusion between data.

5. The high-precision rainfall inversion method for rain-measuring radar based on multi-source data fusion as described in claim 3, characterized in that, In dynamic weight optimization, three weight coefficients are specifically defined: W current R represents the current real-time inversion result est The weights are set between 0.6 and 0.8; W short R represents short-term historical data historical_short The weight of W reflects the persistence of weather, and its value is set between 0.1 and 0.3; seasonal Represents seasonality R clim The weights, reflecting climate patterns, are set between 0.05 and 0.1; these three weight coefficients must satisfy the relationship that their sum is 1.

6. The high-precision rainfall inversion method for rain-measuring radar based on multi-source data fusion as described in claim 5, characterized in that, In dynamic weight optimization, dynamic fine-tuning is performed based on real-time observations, and the current inversion result R is compared in real time. est and short-term historical data R historical_short The system automatically adjusts the weight allocation to calculate the deviation between the current inversion result and the rain gauge observation, as well as the deviation between short-term historical data and the rain gauge observation.

Images