Raindrop size distribution retrieval method considering environmental factors
By using dual-polarization radar data and machine learning methods, an environmentally driven raindrop spectrum inversion system was constructed, which solved the problem that the impact of environmental factors was not considered in the existing technology and achieved high-precision precipitation monitoring and forecasting.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGZHOU INST OF GEOGRAPHY GUANGDONG ACAD OF SCI
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-02
AI Technical Summary
Existing raindrop spectrum inversion methods fail to effectively consider the influence of various environmental factors, resulting in insufficient accuracy in precipitation monitoring and forecasting, especially with large errors under complex terrain or specific weather systems.
The raindrop spectrum was initially inverted using dual-polarization radar data. A random forest classification model was trained by combining Gamma distribution and K-means clustering. An IF-THEN-style classification rule base was constructed, taking into account the influence of environmental factors on the raindrop spectrum.
It improves the inversion accuracy of precipitation parameters such as rainfall rate, and enables adaptive adjustment and refined monitoring and forecasting of changes in raindrop spectral morphology.
Smart Images
Figure CN122132980A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of raindrop spectrum prediction technology, and in particular to a raindrop spectrum inversion method that comprehensively considers environmental factors. Background Technology
[0002] Raindrop spectrum (DSD) represents the distribution of the number of raindrops of different sizes per unit volume of air. It is a core parameter for quantitatively describing the microphysical processes of precipitation. Inaccurate DSD simulations can lead to serious errors in precipitation rate estimation. Currently, the mathematical description of raindrop spectrum mainly relies on parameterized distribution models. The most commonly used are the MP (Marshall-Palmer) exponential distribution and the more general three-parameter Gamma distribution. The parameters of these models need to be inverted through observation. Gamma distribution models can characterize the raindrop spectrum using a few parameters (such as the intercept parameter N0, the slope parameter Λ, and the shape parameter μ). The changes in these parameters reflect the microphysical differences of different precipitation events. However, in practical applications, especially in radar quantitative precipitation estimation and model parameterization, there is a significant limitation: the construction of raindrop spectrum distribution models is too simplistic in its consideration of the environmental control factors of the raindrop spectrum distribution parameters, and the adjustment methods are coarse.
[0003] Current technologies primarily rely on classifying precipitation based on its origin, empirically adjusting raindrop spectrum parameters by categorizing it into stratiform and convective cloud precipitation. For example, when applying the MP distribution, different NO coefficients are used for stratiform and convective cloud precipitation; when establishing radar ZR relationships, Z=200R1.6 is chosen for stratiform cloud precipitation, while Z=300R1.4 is used for convective cloud precipitation. While this dichotomy is simple, it greatly simplifies the complexity of the real atmosphere. In fact, besides precipitation type, numerous observations and theoretical studies have confirmed that various environmental factors, such as temperature, humidity, wind speed, topography, latitude, land-sea location, and aerosol background, all have crucial influences on raindrop spectra. Temperature and humidity significantly trim the droplet ends of the raindrop spectrum and alter its overall shape by affecting the subcloud evaporation process; topography, through dynamic lifting, can enhance collision and merging processes, producing a wider raindrop spectrum that more closely resembles convective characteristics; vertical wind shear affects the dynamic structure within clouds, thus indirectly influencing microphysical processes. These environmental factors have a clear impact on raindrop spectra, but existing inversion schemes lack a systematic and quantitative method to incorporate these factors into the inversion process, leading to significant inversion errors under complex terrain or specific weather systems. Furthermore, inversion models are typically static and cannot adaptively adjust to the actual physical causes of precipitation, making it difficult to achieve refined inversion tailored to specific scenarios. The importance of each factor also varies across different regions; current operational paradigms generalize complex and continuously changing environmental influences into a single category. This results in biases in the description of the true morphology of raindrop spectra, becoming a significant source of uncertainty in radar quantitative precipitation estimation and errors in numerical model simulations of precipitation.
[0004] There is an urgent need in this field to develop a new technical solution that breaks through the constraints of the traditional "stratification-convection" dichotomy, and can clearly and quantitatively reveal the relative contribution and influence mechanism of various environmental factors in the region on the formation of different raindrop spectral modes. Based on this, multiple typical modes of raindrop spectra in the region can be automatically discovered, thereby providing technical support for achieving more accurate precipitation monitoring and forecasting based on physical processes. Summary of the Invention
[0005] This invention provides a raindrop spectrum inversion method that comprehensively considers environmental factors. By refining the classification of rainfall events, it reduces the impact of multiple factors on the inversion of key raindrop spectrum parameters, reduces systematic errors caused by the mismatch between the model and the microphysical characteristics of precipitation, and improves the inversion accuracy of precipitation parameters such as rainfall rate.
[0006] The first aspect of this invention provides a raindrop spectrum inversion method that comprehensively considers environmental factors, comprising the following steps: Based on the rainfall dataset, obtain the dual-polarization radar base data for the corresponding region; Based on dual-polarization radar data, the raindrop spectrum is initially inverted using the Gamma distribution. The extracted feature vectors are used to construct a dataset, and unsupervised classification is performed using the K-means clustering algorithm to generate raindrop spectrum classification labels; environmental factor datasets are obtained, and environmental feature matrices are constructed. Using environmental features as input and raindrop spectrum classification labels as prediction targets, a random forest classification model is trained to generate the dominant factors affecting the changes in raindrop spectrum morphology and the classification thresholds corresponding to the dominant factors. Construct an IF-THEN-style classification rule base based on dominant factors and classification thresholds; The environmental data corresponding to the rainfall to be predicted is used as input to the classification rule base. The rainfall to be predicted is classified into the preset raindrop spectrum classification label for precipitation monitoring and forecasting.
[0007] Furthermore, the initial inversion of the raindrop spectrum based on dual-polarization radar-based data using the Gamma distribution includes the following steps: The horizontal reflectivity factor, differential reflectivity, and differential propagation phase shift in the dual-polarization radar base data are obtained, and the dual-polarization radar base data is preprocessed. Set up a three-parameter Gamma distribution; construct a radar observation operator; Based on the radar observation operator and the optimal estimation algorithm, the Gamma distribution parameters describing the raindrop spectrum are solved; Set normalization parameters, and characterize the raindrop spectrum as a normalized eigenvector based on the normalization parameters.
[0008] Furthermore, the normalized parameters include the mass-weighted average diameter and the standardized intercept parameter; the mass-weighted average diameter is calculated using the following formula: in, The mass-weighted average diameter. For shape parameters, This is the slope parameter.
[0009] The standardized intercept parameter is calculated using the following formula: in, To standardize the intercept parameter, Liquid water content, This is the density of water.
[0010] Furthermore, the step of constructing a dataset from the extracted feature vectors and performing unsupervised classification using the K-means clustering algorithm to generate raindrop spectral classification labels includes the following steps: Obtain the feature vectors extracted in the initial inversion and construct a dataset from these feature vectors; Unsupervised classification of the dataset was performed using the K-means clustering algorithm. The optimal number of clusters is determined based on the elbow rule or the profile coefficient method, and the raindrop spectrum is divided into multiple categories with different microphysical characteristics, which serve as raindrop spectrum classification labels.
[0011] Furthermore, the environmental feature matrix includes features such as thermal features, humidity features, dynamic features, topographic features, and weather causes.
[0012] Furthermore, the step of using environmental features as input, raindrop spectrum classification labels as prediction targets, training a random forest classification model, and generating the dominant factors influencing raindrop spectrum morphology changes and the corresponding classification thresholds for these dominant factors includes the following steps: Using the environmental feature matrix as the input variable and the obtained raindrop spectrum category labels as the prediction target, a random forest classification model is trained to generate a feature importance ranking corresponding to the features. Based on the importance of features, environmental factors that have a significant impact on the morphological changes of raindrop spectra are selected. Based on the statistical distribution, physical meaning, and natural boundary points of environmental factors in different raindrop spectral categories, classification thresholds are set for environmental factors.
[0013] Furthermore, the classification rule base constructs IF-THEN-style classification logic based on the dominant factors and their corresponding classification thresholds; For each classification result corresponding to a raindrop spectral category, the optimized raindrop spectral distribution mode for that category is calculated, forming a set of classification and inversion mapping rules that can be executed automatically.
[0014] A second aspect of the present invention provides a raindrop spectrum inversion system that comprehensively considers environmental factors, including a first processing unit for acquiring dual-polarization radar base data of a corresponding region based on a rainfall dataset; The second processing unit is used to perform initial inversion of raindrop spectrum based on dual-polarization radar base data through Gamma distribution; The third processing unit is used to construct a dataset from the extracted feature vectors, perform unsupervised classification using the K-means clustering algorithm, generate raindrop spectrum classification labels, acquire environmental factor datasets, and construct an environmental feature matrix. The fourth processing unit is used to take environmental features as input, raindrop spectrum classification labels as prediction targets, train a random forest classification model, and generate the dominant factors affecting the changes in raindrop spectrum morphology and the classification thresholds corresponding to the dominant factors. The fifth processing unit is used to construct a classification rule base in IF-THEN form based on the dominant factors and classification thresholds; The sixth processing unit is used to take the environmental data corresponding to the rainfall to be predicted as input to the classification rule base, classify the rainfall to be predicted into the preset raindrop spectrum classification label, and carry out precipitation monitoring and forecasting.
[0015] A third aspect of the present invention provides a computer device, comprising: Memory, transceiver, processor, and bus system; The memory is used to store programs; The processor is used to execute the program in the memory, including executing the raindrop spectrum inversion method that comprehensively considers environmental factors as described above. The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.
[0016] A fourth aspect of the present invention provides a readable storage medium storing computer-readable instructions, which, when executed by a processor, implement the steps of the raindrop spectrum inversion method that comprehensively considers environmental factors as described above.
[0017] As can be seen from the above technical solutions, the present invention has the following advantages: This invention acquires dual-polarization radar-based data and performs initial raindrop spectrum inversion using Gamma distribution. This refines the classification of rainfall events, reducing the impact of multiple factors on the inversion of key raindrop spectrum parameters. Based on the unsupervised classification method of K-means clustering, it can objectively discover multiple microphysical modes of raindrop spectrum actually existing in the study area from historical data. This allows for matching more targeted inversion models for different types of precipitation, thereby reducing systematic errors caused by the mismatch between the model and the microphysical characteristics of precipitation, and improving the inversion accuracy of key precipitation parameters such as rainfall rate.
[0018] Secondly, by integrating multi-dimensional environmental data and using a random forest model for feature importance analysis and threshold extraction, this invention can clearly identify and quantify the dominant role of different environmental factors in the formation and evolution of specific raindrop spectrum morphology. This strengthens the characterization and understanding of the physical mechanism of raindrop spectrum changes and provides a clear, physically meaningful, and repeatable objective standard for raindrop spectrum classification based on environmental conditions. This allows the inversion scheme to be adaptively adjusted to closely follow the actual physical background of precipitation occurrence.
[0019] Finally, this invention solidifies complex machine learning models and multi-factor relationships into concise, transparent, and automatically executable business logic. For new observational data, only easily obtainable environmental parameters need to be input, and the rule base can automatically determine its category and call the corresponding optimization inversion scheme, eliminating the need for repetitive and complex modeling calculations, thus improving the method's engineering practicality and business potential.
[0020] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from an examination of the following, or may be learned from the practice of the invention. Attached Figure Description
[0021] Figure 1 The method flowchart provided by the present invention. Detailed Implementation
[0022] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0023] Example 1 The implementation method in this embodiment can be implemented in a system, on a server, or on a terminal; no specific limitation is made. The method in this application will be described below from the perspective of system implementation. As shown in the figure, a raindrop spectrum inversion method that comprehensively considers environmental factors includes the following steps: Based on the rainfall dataset, obtain the dual-polarization radar base data for the corresponding region; Based on dual-polarization radar data, the raindrop spectrum is initially inverted using the Gamma distribution. The extracted feature vectors are used to construct a dataset, and unsupervised classification is performed using the K-means clustering algorithm to generate raindrop spectrum classification labels; environmental factor datasets are obtained, and environmental feature matrices are constructed. Using environmental features as input and raindrop spectrum classification labels as prediction targets, a random forest classification model is trained to generate the dominant factors affecting the changes in raindrop spectrum morphology and the classification thresholds corresponding to the dominant factors. Construct an IF-THEN-style classification rule base based on dominant factors and classification thresholds; The environmental data corresponding to the rainfall to be predicted is used as input to the classification rule base. The rainfall to be predicted is classified into the preset raindrop spectrum classification label for precipitation monitoring and forecasting.
[0024] The dual-polarization radar base data includes horizontal reflectivity factor, differential reflectivity, and differential propagation phase shift, containing information on raindrop size, shape, and concentration. The dual-polarization radar base data undergoes quality control and preprocessing, including clutter removal, attenuation correction, noise filtering, and system bias correction.
[0025] Next, a corresponding radar observation operator is constructed based on the three-parameter Gamma distribution, establishing a physical connection between radar observations and the microphysical parameters of the raindrop spectrum, including intercept, shape, and slope parameters. The expression for the Gamma distribution is: in, Where is the raindrop diameter, and is the droplet concentration. For the intercept parameter, For shape parameters, The slope parameter is used. These parameters are solved using an optimal estimation algorithm, and the normalized features are further calculated, thus completing the initial transformation from radar signal to normalized raindrop spectral features.
[0026] After obtaining the raindrop spectrum characteristics of a large number of historical rainfall samples, unsupervised clustering algorithms such as K-means are used to analyze the feature parameter space and divide the raindrop spectrum into several physical categories with significant differences, avoiding the problem of simple dichotomy of stratification / convection based solely on experience in traditional methods.
[0027] By collecting multi-source environmental data such as temperature, humidity, wind field, topography, and weather patterns synchronized with radar observations, and constructing a feature matrix, machine learning models such as random forests were used to quantitatively diagnose the environmental factors that play a dominant role in the formation of different raindrop spectral types, and association rules between environmental conditions and raindrop spectral morphology were established. Based on these rules, an IF-THEN-style classification rule base was constructed.
[0028] For new rainfall events, by inputting real-time and analyzed environmental field data, the raindrop spectrum category to which it belongs can be quickly determined by querying a pre-built rule base; then, the inversion scheme optimized for that category can be invoked for final calculation.
[0029] Furthermore, based on the best inversion practices corresponding to the classified raindrop spectrum categories, more refined and customized inversion schemes can be adopted to adjust the distribution function model for each category and establish a specific ZR relationship (reflectivity factor-rainfall intensity relationship) or relationship (differential propagation phase shift rate-rainfall intensity relationship). This improves the accuracy of regional heavy precipitation monitoring and forecasting.
[0030] Example 2 The difference between this embodiment and Embodiment 1 is that the initial inversion of the raindrop spectrum based on dual-polarization radar base data through the Gamma distribution includes the following steps: The horizontal reflectivity factor, differential reflectivity, and differential propagation phase shift in the dual-polarization radar base data are obtained, and the dual-polarization radar base data is preprocessed. Set up a three-parameter Gamma distribution; construct a radar observation operator; Based on the radar observation operator and the optimal estimation algorithm, the Gamma distribution parameters describing the raindrop spectrum are solved; Set normalization parameters, and characterize the raindrop spectrum as a normalized eigenvector based on the normalization parameters.
[0031] The expression for the three-parameter Gamma distribution is: in, The diameter of the raindrop. This refers to the concentration of the droplet spectrum. For the intercept parameter, For shape parameters, This is the slope parameter.
[0032] The radar observation operator includes a horizontal reflectivity factor, differential reflectivity, and differential propagation phase shift rate; the horizontal reflectivity factor is calculated using the following formula: The differential reflectance is calculated using the following formula: The differential propagation phase shift rate is calculated using the following formula: in, It is the backscattering cross section. Forward scattering amplitude, The dielectric constant factor of water, This refers to the radar wavelength.
[0033] Next, the cost function is defined, and the optimal estimation method is used to solve for the Gamma distribution parameters. The expression for the cost function is: in, This is the radar measured vector. Based on the current Estimated radar observation vector, Let the observation error covariance matrix be denoted as . The cost function is minimized using a numerical optimization algorithm. Solve for the optimal Parameter group.
[0034] The normalized parameters include the mass-weighted average diameter and the standardized intercept parameter; the mass-weighted average diameter is calculated using the following formula: in, The mass-weighted average diameter. For shape parameters, This is the slope parameter.
[0035] The standardized intercept parameter is calculated using the following formula: in, To standardize the intercept parameter, Liquid water content, The density is given by the water. The raindrop spectrum of each radar sampling volume is characterized as a feature vector. .
[0036] Example 3 The difference between this embodiment and Embodiment 2 is that the step of constructing a dataset from the extracted feature vectors and performing unsupervised classification using the K-means clustering algorithm to generate raindrop spectrum classification labels includes the following steps: Obtain the feature vectors extracted in the initial inversion and construct a dataset from these feature vectors; Unsupervised classification of the dataset was performed using the K-means clustering algorithm. The optimal number of clusters is determined based on the elbow rule or the profile coefficient method, and the raindrop spectrum is divided into multiple categories with different microphysical characteristics, which serve as raindrop spectrum classification labels.
[0037] First, obtain the feature vectors of all samples. This forms a multidimensional dataset. Next, the K-means clustering algorithm is applied to perform unsupervised classification on this dataset. By iteratively optimizing the feature vectors, the samples are divided into K clusters. The elbow rule or silhouette coefficient method is used for evaluation to minimize the feature space distance between samples within the same cluster and maximize the distance between different clusters. Through the above steps, various raindrop spectra that have historically occurred in the study area are systematically divided into several significantly different microphysical categories. Each category corresponds to a typical raindrop spectrum morphology. The inherent grouping patterns of the raindrop spectrum microphysical structure are mined from a large amount of historical data, thus eliminating the dependence on prior classification knowledge.
[0038] Example 4 The difference between this embodiment and Embodiment 3 is that the environmental feature matrix includes thermal features, humidity features, dynamic features, topographic features, and weather phenomena. The features of the environmental feature matrix can be limited as follows: The thermal characteristics can be described as follows: (850 hPa temperature) (The temperature difference between 850 hPa and 500 hPa, i.e., temperature stratification), is used to characterize the thermal state and stratification stability of the atmosphere; Humidity characteristics can be described as follows: (950 hPa relative humidity, i.e., lower-level humidity) (Vertical humidity gradient) is used to describe the water vapor conditions in the lower atmosphere and the humidity profile of the entire layer; The dynamic characteristics can be described as follows: (850hPa and 500hPa vertical wind shear) is used to reflect the influence of the environmental wind field on cloud organization and development dynamics; Topographic features: (altitude), (Estimated rate of topographic uplift), used to quantify the dynamic forcing effect of topography on precipitation; Causes of weather: (Convective available potential energy), and perform weather classification (such as cold front, warm front, etc.) to describe the synoptic-scale background and convective potential of precipitation occurrence.
[0039] Example 5 The difference between this embodiment and embodiment four is that the step of using environmental features as input, raindrop spectrum classification labels as prediction targets, training a random forest classification model, and generating the dominant factors affecting raindrop spectrum morphology changes and the corresponding classification thresholds for the dominant factors includes the following steps: Using the environmental feature matrix as the input variable and the obtained raindrop spectrum category labels as the prediction target, a random forest classification model is trained to generate a feature importance ranking corresponding to the features. Based on the importance of features, environmental factors that have a significant impact on the morphological changes of raindrop spectra are selected. Based on the statistical distribution, physical meaning, and natural boundary points of environmental factors in different raindrop spectral categories, classification thresholds are set for environmental factors.
[0040] First, a random forest classification model is trained using the environmental feature matrix as the input variable and the raindrop spectrum category label as the target variable. The random forest model constructs and integrates numerous decision trees to predict the classification of the environmental feature matrix, calculating the importance score of each environmental feature in the classification decision, and thus deriving a feature importance ranking. This feature importance ranking objectively identifies the dominant environmental factors that contribute most to distinguishing different raindrop spectrum categories, such as identifying low-level humidity or vertical wind shear as key discriminant factors. The top 2-3 environmental factors in the feature importance ranking are extracted from the trained random forest model and identified as the dominant factors influencing the morphological changes of raindrop spectra in that region.
[0041] After identifying the dominant factors, the specific numerical distribution of these factors in different raindrop spectral categories is analyzed, and their known physical meanings are considered, such as high CAPE typically corresponding to strong convection potential. This allows for the determination of critical environmental parameter values that most effectively distinguish each category. For each dominant factor, its statistical distribution in different raindrop spectral categories is analyzed, and one or more classification thresholds are set for each factor, taking into account its physical meaning and natural data boundaries.
[0042] Example 6 The difference between this embodiment and embodiment five is that the classification rule base constructs classification logic in the form of IF-THEN based on the dominant factors and their corresponding classification thresholds; For each classification result corresponding to a raindrop spectral category, the optimized raindrop spectral distribution mode for that category is calculated, forming a set of classification and inversion mapping rules that can be executed automatically.
[0043] Constructing an IF-THEN rule base allows us to express the rules as: IF (a dominant environmental factor satisfies a certain threshold condition), THEN (the sample belongs to a specific raindrop spectrum category).
[0044] For example, the following rules can be constructed: Rule 1: IF THEN "Dry Orographic Rain" category; Rule 2: IF THEN "Wet stratified rain" type; Rule 3: IF THEN "Strong Convection" category.
[0045] In practical business applications, for rainfall events to be inverted, only real-time analysis or forecast environmental field data needs to be input. The system substitutes these environmental data into the rule base for logical judgment to determine the raindrop spectrum category. Then, it calls the optimized inversion algorithm module preset for that category to perform the final calculation and obtain a more refined raindrop spectrum inversion mode for each category.
[0046] Example 7 A raindrop spectrum inversion system that comprehensively considers environmental factors includes a first processing unit for acquiring dual-polarization radar base data of the corresponding region based on a rainfall dataset. The second processing unit is used to perform initial inversion of raindrop spectrum based on dual-polarization radar base data through Gamma distribution; The third processing unit is used to construct a dataset from the extracted feature vectors, perform unsupervised classification using the K-means clustering algorithm, generate raindrop spectrum classification labels, acquire environmental factor datasets, and construct an environmental feature matrix. The fourth processing unit is used to take environmental features as input, raindrop spectrum classification labels as prediction targets, train a random forest classification model, and generate the dominant factors affecting the changes in raindrop spectrum morphology and the classification thresholds corresponding to the dominant factors. The fifth processing unit is used to construct a classification rule base in IF-THEN form based on the dominant factors and classification thresholds; The sixth processing unit is used to take the environmental data corresponding to the rainfall to be predicted as input to the classification rule base, classify the rainfall to be predicted into the preset raindrop spectrum classification label, and carry out precipitation monitoring and forecasting.
[0047] Example 8 A computer device, comprising: Memory, transceiver, processor, and bus system; The memory is used to store programs; The processor is used to execute the program in the memory, including executing the raindrop spectrum inversion method that comprehensively considers environmental factors as described above. The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.
[0048] Example 9 A readable storage medium storing computer-readable instructions, characterized in that, when executed by a processor, the computer-readable instructions implement the steps of the raindrop spectrum inversion method that comprehensively considers environmental factors as described above.
[0049] In summary, this invention acquires dual-polarization radar-based data and performs initial raindrop spectrum inversion using Gamma distribution, refining the classification of rainfall events and reducing the impact of multiple factors on the inversion of key raindrop spectrum parameters. Based on the unsupervised classification method of K-means clustering, it can objectively discover multiple microphysical modes actually existing in the raindrop spectrum within the study area from historical data, matching more targeted inversion models for different types of precipitation. This reduces systematic errors caused by the mismatch between the model and the microphysical characteristics of precipitation, and improves the inversion accuracy of key precipitation parameters such as rainfall rate.
[0050] Secondly, by integrating multi-dimensional environmental data and using a random forest model for feature importance analysis and threshold extraction, this invention can clearly identify and quantify the dominant role of different environmental factors in the formation and evolution of specific raindrop spectrum morphology. This strengthens the characterization and understanding of the physical mechanism of raindrop spectrum changes and provides a clear, physically meaningful, and repeatable objective standard for raindrop spectrum classification based on environmental conditions. This allows the inversion scheme to be adaptively adjusted to closely follow the actual physical background of precipitation occurrence.
[0051] Finally, this invention solidifies complex machine learning models and multi-factor relationships into concise, transparent, and automatically executable business logic. For new observational data, only easily obtainable environmental parameters need to be input, and the rule base can automatically determine its category and call the corresponding optimization inversion scheme, eliminating the need for repetitive and complex modeling calculations, thus improving the method's engineering practicality and business potential.
[0052] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.
[0053] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.
[0054] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0055] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A raindrop spectrum inversion method that comprehensively considers environmental factors, characterized in that, Includes the following steps: Based on the rainfall dataset, obtain the dual-polarization radar base data for the corresponding region; Based on dual-polarization radar data, the raindrop spectrum is initially inverted using the Gamma distribution. The extracted feature vectors are used to construct a dataset, and unsupervised classification is performed using the K-means clustering algorithm to generate raindrop spectrum classification labels. Obtain an environmental factor dataset and construct an environmental feature matrix; Using environmental features as input and raindrop spectrum classification labels as prediction targets, a random forest classification model is trained to generate the dominant factors affecting the changes in raindrop spectrum morphology and the classification thresholds corresponding to the dominant factors. Construct an IF-THEN-style classification rule base based on dominant factors and classification thresholds; The environmental data corresponding to the rainfall to be predicted is used as input to the classification rule base. The rainfall to be predicted is classified into the preset raindrop spectrum classification label for precipitation monitoring and forecasting.
2. The raindrop spectrum inversion method that comprehensively considers environmental factors according to claim 1, characterized in that, The initial inversion of raindrop spectra based on dual-polarization radar data using Gamma distribution includes the following steps: The horizontal reflectivity factor, differential reflectivity, and differential propagation phase shift in the dual-polarization radar base data are obtained, and the dual-polarization radar base data is preprocessed. Set up a three-parameter Gamma distribution; construct a radar observation operator; Based on the radar observation operator and the optimal estimation algorithm, the Gamma distribution parameters describing the raindrop spectrum are solved; Set normalization parameters, and characterize the raindrop spectrum as a normalized eigenvector based on the normalization parameters.
3. The raindrop spectrum inversion method that comprehensively considers environmental factors according to claim 1, characterized in that, The normalized parameters include the mass-weighted average diameter and the standardized intercept parameter; the mass-weighted average diameter is calculated using the following formula: in, The mass-weighted average diameter. For shape parameters, This is the slope parameter. The standardized intercept parameter is calculated using the following formula: in, To standardize the intercept parameter, Liquid water content, This is the density of water.
4. The raindrop spectrum inversion method that comprehensively considers environmental factors according to claim 1, characterized in that, The process of constructing a dataset from the extracted feature vectors and performing unsupervised classification using the K-means clustering algorithm to generate raindrop spectral classification labels includes the following steps: Obtain the feature vectors extracted in the initial inversion and construct a dataset from these feature vectors; Unsupervised classification of the dataset was performed using the K-means clustering algorithm. The optimal number of clusters is determined based on the elbow rule or the profile coefficient method, and the raindrop spectrum is divided into multiple categories with different microphysical characteristics, which serve as raindrop spectrum classification labels.
5. The raindrop spectrum inversion method that comprehensively considers environmental factors according to claim 1, characterized in that, The environmental feature matrix includes thermal features, humidity features, dynamic features, topographic features, and weather causes.
6. The raindrop spectrum inversion method considering environmental factors according to claim 1, characterized in that, The process of using environmental features as input, raindrop spectral classification labels as prediction targets, training a random forest classification model, and generating dominant factors influencing raindrop spectral morphology changes and corresponding classification thresholds for these dominant factors includes the following steps: Using the environmental feature matrix as the input variable and the obtained raindrop spectrum category labels as the prediction target, a random forest classification model is trained to generate a feature importance ranking corresponding to the features. Based on the importance of features, environmental factors that have a significant impact on the morphological changes of raindrop spectra are selected. Based on the statistical distribution, physical meaning, and natural boundary points of environmental factors in different raindrop spectral categories, classification thresholds are set for environmental factors.
7. The raindrop spectrum inversion method that comprehensively considers environmental factors according to claim 1, characterized in that, The classification rule base constructs IF-THEN-style classification logic based on the dominant factors and their corresponding classification thresholds; For each classification result corresponding to a raindrop spectral category, the optimized raindrop spectral distribution mode for that category is calculated, forming a set of classification and inversion mapping rules that can be executed automatically.
8. A raindrop spectrum inversion system that comprehensively considers environmental factors, characterized in that, It includes a first processing unit, used to acquire dual-polarization radar base data for the corresponding area based on the rainfall dataset; The second processing unit is used to perform initial inversion of raindrop spectrum based on dual-polarization radar base data through Gamma distribution; The third processing unit is used to construct a dataset from the extracted feature vectors, perform unsupervised classification using the K-means clustering algorithm, and generate raindrop spectrum classification labels. Obtain an environmental factor dataset and construct an environmental feature matrix; The fourth processing unit is used to take environmental features as input, raindrop spectrum classification labels as prediction targets, train a random forest classification model, and generate the dominant factors affecting the changes in raindrop spectrum morphology and the classification thresholds corresponding to the dominant factors. The fifth processing unit is used to construct a classification rule base in IF-THEN form based on the dominant factors and classification thresholds; The sixth processing unit is used to take the environmental data corresponding to the rainfall to be predicted as input to the classification rule base, classify the rainfall to be predicted into the preset raindrop spectrum classification label, and carry out precipitation monitoring and forecasting.
9. A computer device, characterized in that, include: Memory, transceiver, processor, and bus system; The memory is used to store programs; The processor is used to execute the program in the memory, including executing the raindrop spectrum inversion method that comprehensively considers environmental factors as described in any one of claims 1 to 7; The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.
10. A readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, they implement the steps of the raindrop spectrum inversion method that comprehensively considers environmental factors as described in any one of claims 1 to 7.