A method and apparatus for inverting the effective radius of particles at the bottom of a water cloud.

Abstract

This invention proposes a method and apparatus for inverting the effective radius of cloud-based particles at the bottom of water clouds. The method includes: constructing a training dataset based on a first effective radius of cloud-based particles, first observation data, and first auxiliary meteorological data; obtaining target input parameters based on the contribution of all training parameters; inputting the target input parameters into multiple models; evaluating each model with the first effective radius of cloud-based particles as the prediction target; selecting a cloud-based particle effective radius inversion model based on the evaluation results; processing second observation data, second auxiliary meteorological data, and rainfall data to obtain a multidimensional feature dataset; performing cloud detection and cloud phase detection sequentially on the multidimensional feature dataset; selecting water cloud observation data based on the detection results; extracting features from the water cloud observation data; inputting the features of the water cloud observation data into the cloud-based particle effective radius inversion model to obtain the cloud-based particle effective radius inversion result corresponding to the water cloud observation data.

Description

Technical Field

[0001] This application relates to the field of cloud remote sensing technology, and more specifically, to a method and apparatus for inverting the effective radius of particles at the bottom of water clouds. Background Technology

[0002] Water clouds cover approximately 20-30% of the Earth's surface and play a crucial role in the Earth's radiation balance. The effective radius of cloud particles is one of the key parameters representing the radiative characteristics of water clouds. It not only significantly impacts the simulation of cloud-climate feedback but also holds great importance for the study of aerosol-cloud-precipitation interactions. The method of determining cloud parameters using satellite observation data is known as cloud remote sensing inversion.

[0003] Currently, remote sensing inversion of cloud particle effective radius based on passive satellite observation data is often performed using the dual-spectral reflectance method. This method is based on two bands that are sensitive to optical thickness and particle size: visible light (0.8 μm) and near-infrared (1.6, 2.1 or 3.7 μm). The measured reflectance in the visible and near-infrared bands is compared with the reflectance in a pre-calculated lookup table, and the cloud optical thickness and cloud particle effective radius are obtained simultaneously through optimal estimation.

[0004] However, since the particle size in actual clouds varies significantly with altitude, the traditional effective radius obtained by existing methods using average information from the cloud top or the entire layer is based on the assumption of vertical uniformity of the cloud and the premise that the sensitivity of cloud bottom information is weak only through direct observation by passive satellites, resulting in low accuracy of the effective radius. Summary of the Invention

[0005] This application provides a method and apparatus for inverting the effective radius of cloud-based particles at the cloud base. It ranks the contribution of a series of relevant data from active satellite observations and selects target input parameters based on this ranking for evaluation using various models to obtain an effective radius inversion model. It also performs cloud detection and cloud phase detection using a series of relevant data from passive satellite observations, selecting water cloud observation data to input into the effective radius inversion model to obtain the inversion result. Therefore, it no longer relies on the assumption of vertical uniformity of clouds, and improves the accuracy of the predicted effective radius by using both active and passive satellite observations to invert the effective radius of cloud-based particles.

[0006] In a first aspect, embodiments of this application provide a method for inverting the effective radius of particles at the bottom of a water cloud, the method comprising: Step S1: Obtain the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data from active satellite observations, and construct a training dataset based on the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data; Step S2: Obtain the contribution of all training parameters in the cloud bottom particle effective radius inversion from the training dataset, and obtain the target input parameters based on the contribution of all training parameters; wherein, the target input parameters represent the training parameters used to construct the cloud bottom particle effective radius inversion model; Step S3: Input the target input parameters into a variety of preset models, and evaluate each model with the first effective radius of the cloud bottom particle as the prediction target to obtain the evaluation results of each model. Based on the evaluation results, select the model that meets the requirements as the cloud bottom particle effective radius inversion model. Step S4: Obtain the second observation data, second auxiliary meteorological data, and rainfall data of the passive satellite corresponding to the target input parameters, and process the second observation data, the second auxiliary meteorological data, and the rainfall data to obtain a multidimensional feature dataset; Step S5: Perform cloud detection and cloud phase detection on the multidimensional feature dataset in sequence, and select water cloud observation data according to the detection results. Perform feature extraction on the water cloud observation data, and input the features of the water cloud observation data into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data; wherein, the water cloud observation data represents pixels with clouds and water clouds.

[0007] In one possible implementation, step S1, acquiring the effective radius of cloud-bottom particles, first observation data, and first auxiliary meteorological data from active satellite observations, includes: The system acquires satellite data of various target products from active satellites and filters the satellite data of the various target products based on a variety of preset filtering conditions to obtain satellite data that meets the various filtering conditions; wherein, the satellite data includes observation data and first auxiliary meteorological data; the various filtering conditions include at least cloud type conditions, temperature conditions, precipitation conditions, and valid data conditions; The index position of the satellite data is determined, and the satellite data required for each target product is extracted based on the index position; wherein, the satellite data of the first target product includes at least the average geometric radius of the droplets; the effective radius of the cloud-bottom particles is calculated based on the average geometric radius of the droplets in the first target product, and the effective radius of the cloud-bottom particles replaces the average geometric radius of the droplets in the satellite data and all parameters are merged to obtain the training dataset.

[0008] In one possible implementation, the filtering of satellite data for the various target products based on preset multiple filtering conditions includes: The cloud type in the satellite data is determined. When the cloud type is a single-layer cloud, stratocumulus cloud, stratus cloud, or water cloud, the satellite data is determined to meet the cloud type condition. The location temperature in the satellite data is determined, and when the location temperature is greater than a preset first temperature threshold, the satellite data is determined to meet the temperature condition. The precipitation information in the satellite data is determined. When the precipitation information indicates no precipitation, the satellite data is determined to meet the precipitation conditions. The geometric radius value of liquid water in the satellite data is determined. If there are at least a target number of consecutive valid geometric radius values ​​of liquid water, then the satellite data is determined to meet the valid data condition.

[0009] In one possible implementation, in step S2, obtaining the contribution of all training parameters in the cloud bottom particle effective radius inversion and obtaining the target input parameters based on the contribution of all training parameters includes: All training parameters in the training dataset are input into the basic random forest model for contribution evaluation, and the contribution of all training parameters in the cloud bottom particle effective radius inversion is obtained. Rank the contributions of all training parameters and select the target training parameters that meet the contribution requirements. The target training parameters and all training parameters are respectively input into the random forest model to re-invert the effective radius of the cloud bottom particles, and the second effective radius of the cloud bottom particles is obtained. Based on the effective radius of the first cloud bottom particle and the effective radius of the second cloud bottom particle, the target training parameters and the estimation accuracy of each of the training parameters in the random forest model are determined, and the target input parameters are selected based on the number of training parameters and the estimation accuracy; wherein, the estimation accuracy characterizes the estimation accuracy of the random forest model for the effective radius of the cloud bottom particle.

[0010] In one possible implementation, step S3, evaluating each model and obtaining the evaluation result for each model, includes: The evaluation index value corresponding to each model is calculated based on the preset multi-dimensional evaluation index; wherein, the evaluation index includes at least the coefficient of determination, root mean square error and mean absolute error; The evaluation index values ​​corresponding to each model are visualized using a density scatter plot, and the evaluation results of each model are obtained based on the density scatter plot.

[0011] In one possible implementation, in step S4, the processing of the second observation data, the second auxiliary meteorological data, and the rainfall data to obtain a multidimensional feature dataset includes: Determine the target time window and target spatial grid for spatiotemporal matching, and perform spatiotemporal matching on the second observation data, the second auxiliary meteorological data, and the rainfall data based on the target time window and the target spatial grid; After identifying and removing invalid data from the second observation data, the second auxiliary meteorological data, and the rainfall data, the data are fused to obtain a multidimensional feature dataset.

[0012] In one possible implementation, in step S5, the process of sequentially performing cloud detection and cloud phase detection on the multidimensional feature dataset, and then filtering out water cloud observation data based on the detection results, includes: Perform cloud detection on the multidimensional feature dataset to determine whether the multidimensional feature dataset contains clouds; If the multidimensional feature dataset contains clouds, then cloud phase detection is performed on the multidimensional feature dataset to determine the cloud phase of the multidimensional feature dataset; wherein, the cloud phase includes at least water clouds and ice clouds; If the cloud phase of the multidimensional feature dataset is water cloud, then the water cloud observation data in the multidimensional feature dataset are selected.

[0013] Secondly, embodiments of this application also provide a device for inverting the effective radius of particles at the bottom of a water cloud, the device comprising: The module is used to acquire the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data from active satellite observations, and to construct a training dataset based on the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data. The first acquisition module is used to acquire the contribution of all training parameters in the cloud bottom particle effective radius inversion of the training dataset, and to acquire the target input parameters based on the contribution of all training parameters; wherein, the target input parameters represent the training parameters used to construct the cloud bottom particle effective radius inversion model; The second acquisition module is used to input the target input parameters into a variety of preset models, and evaluate each model with the first effective radius of the cloud bottom particle as the prediction target, obtain the evaluation result of each model, and select the model that meets the requirements as the cloud bottom particle effective radius inversion model based on the evaluation result. The third acquisition module is used to acquire the second observation data, the second auxiliary meteorological data, and the rainfall data of the passive satellite corresponding to the target input parameters, and to process the second observation data, the second auxiliary meteorological data, and the rainfall data to obtain a multidimensional feature dataset; The fourth acquisition module is used to sequentially perform cloud detection and cloud phase detection on the multidimensional feature dataset, and filter out water cloud observation data according to the detection results. Based on the water cloud observation data, feature extraction is performed, and the features of the water cloud observation data are input into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data; wherein, the water cloud observation data represents pixels with clouds and water clouds.

[0014] In one possible implementation, the building module is specifically used for: The system acquires satellite data of various target products from active satellites and filters the satellite data of the various target products based on a variety of preset filtering conditions to obtain satellite data that meets the various filtering conditions; wherein, the satellite data includes observation data and first auxiliary meteorological data; the various filtering conditions include at least cloud type conditions, temperature conditions, precipitation conditions, and valid data conditions; The index position of the satellite data is determined, and the satellite data required for each target product is extracted based on the index position; wherein, the satellite data of the first target product includes at least the average geometric radius of the droplets; the effective radius of the cloud-bottom particles is calculated based on the average geometric radius of the droplets in the first target product, and the effective radius of the cloud-bottom particles replaces the average geometric radius of the droplets in the satellite data and all parameters are merged to obtain the training dataset.

[0015] In one possible implementation, the building module is specifically used for: The cloud type in the satellite data is determined. When the cloud type is a single-layer cloud, stratocumulus cloud, stratus cloud, or water cloud, the satellite data is determined to meet the cloud type condition. The location temperature in the satellite data is determined, and when the location temperature is greater than a preset first temperature threshold, the satellite data is determined to meet the temperature condition. The precipitation information in the satellite data is determined. When the precipitation information indicates no precipitation, the satellite data is determined to meet the precipitation conditions. The geometric radius value of liquid water in the satellite data is determined. If there are at least a target number of consecutive valid geometric radius values ​​of liquid water, then the satellite data is determined to meet the valid data condition.

[0016] In one possible implementation, the first acquisition module is specifically used for: All training parameters in the training dataset are input into the basic random forest model for contribution evaluation, and the contribution of all training parameters in the cloud bottom particle effective radius inversion is obtained. Rank the contributions of all training parameters and select the target training parameters that meet the contribution requirements. The target training parameters and all training parameters are respectively input into the random forest model to re-invert the effective radius of the cloud bottom particles, and the second effective radius of the cloud bottom particles is obtained. Based on the effective radius of the first cloud bottom particle and the effective radius of the second cloud bottom particle, the target training parameters and the estimation accuracy of each of the training parameters in the random forest model are determined, and the target input parameters are selected based on the number of training parameters and the estimation accuracy; wherein, the estimation accuracy characterizes the estimation accuracy of the random forest model for the effective radius of the cloud bottom particle.

[0017] In one possible implementation, the second acquisition module is specifically used for: The evaluation index value corresponding to each model is calculated based on the preset multi-dimensional evaluation index; wherein, the evaluation index includes at least the coefficient of determination, root mean square error and mean absolute error; The evaluation index values ​​corresponding to each model are visualized using a density scatter plot, and the evaluation results of each model are obtained based on the density scatter plot.

[0018] In one possible implementation, the third acquisition module is specifically used for: Determine the target time window and target spatial grid for spatiotemporal matching, and perform spatiotemporal matching on the second observation data, the second auxiliary meteorological data, and the rainfall data based on the target time window and the target spatial grid; After identifying and removing invalid data from the second observation data, the second auxiliary meteorological data, and the rainfall data, the data are fused to obtain a multidimensional feature dataset.

[0019] In one possible implementation, the fourth acquisition module is specifically used for: Perform cloud detection on the multidimensional feature dataset to determine whether the multidimensional feature dataset contains clouds; If the multidimensional feature dataset contains clouds, then cloud phase detection is performed on the multidimensional feature dataset to determine the cloud phase of the multidimensional feature dataset; wherein, the cloud phase includes at least water clouds and ice clouds; If the cloud phase of the multidimensional feature dataset is water cloud, then the water cloud observation data in the multidimensional feature dataset are selected.

[0020] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the water cloud bottom particle effective radius inversion method as described in any of the first aspects.

[0021] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the steps of the water cloud bottom particle effective radius inversion method described in any one of the first aspects.

[0022] This application provides the aforementioned method and apparatus for inverting the effective radius of cloud-based particles at the cloud base. It acquires the first effective radius of cloud-based particles from active satellite observations, first observation data, and first auxiliary meteorological data. A training dataset is constructed based on the first effective radius of cloud-based particles, the first observation data, and the first auxiliary meteorological data. The contribution of all training parameters in the training dataset to the inversion of the effective radius of cloud-based particles is obtained. Based on the contribution of all training parameters, a target input parameter is obtained. The target input parameter is then input into several preset models. Using the first effective radius of cloud-based particles as the prediction target, each model is evaluated, and the evaluation result for each model is obtained. Based on the evaluation results, a model that meets the requirements is selected as the cloud base particle effective radius inversion model. Second observation data, second auxiliary meteorological data, and precipitation data of passive satellites corresponding to the target input parameters are obtained. The second observation data, second auxiliary meteorological data, and precipitation data are processed to obtain a multi-dimensional feature dataset. Cloud detection and cloud phase detection are performed on the multi-dimensional feature dataset in sequence. Water cloud observation data are selected according to the detection results. Feature extraction is performed on the water cloud observation data. The features of the water cloud observation data are input into the cloud base particle effective radius inversion model to obtain the cloud base particle effective radius inversion result corresponding to the water cloud observation data. This application ranks the contribution of a series of relevant data from active satellite observations and selects target input parameters based on the contribution ranking to evaluate multiple models, thus obtaining a cloud base particle effective radius inversion model. It also performs cloud detection and cloud phase detection using a series of relevant data from passive satellite observations, and selects water cloud observation data to input into the cloud base particle effective radius inversion model to obtain the cloud base particle effective radius inversion result. Therefore, it no longer relies on the assumption of vertical uniformity of clouds, and improves the accuracy of the predicted cloud base particle effective radius by using both active and passive satellite observations to assist in the inversion. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1A flowchart of the first method for inverting the effective radius of particles at the bottom of a water cloud, as provided in an embodiment of this application, is shown. Figure 2 A schematic diagram of the overall process for inverting the effective radius of particles at the bottom of a water cloud is shown. Figure 3 This illustration shows a schematic diagram of the structure of a water cloud bottom particle effective radius inversion device provided in an embodiment of this application; Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0026] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0027] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0028] Given that water clouds cover approximately 20-30% of the Earth's surface and play a crucial role in the Earth's radiation balance, the effective radius of cloud particles is a key parameter representing the radiative characteristics of water clouds. It significantly impacts not only the simulation of cloud-climate feedback but also the study of aerosol-cloud-precipitation interactions. The method of determining cloud parameters using satellite observation data is known as cloud remote sensing inversion.

[0029] Currently, remote sensing inversion of cloud particle effective radius based on passive satellite observation data is often performed using the dual-spectral reflectance method. This method is based on two bands that are sensitive to optical thickness and particle size: visible light (0.8 μm) and near-infrared (1.6, 2.1 or 3.7 μm). The measured reflectance in the visible and near-infrared bands is compared with the reflectance in a pre-calculated lookup table, and the cloud optical thickness and cloud particle effective radius are obtained simultaneously through optimal estimation.

[0030] However, since the particle size in actual clouds varies significantly with altitude, the traditional effective radius obtained by existing methods using average information from the cloud top or the entire layer is based on the assumption of vertical uniformity of the cloud and the premise that the sensitivity of cloud bottom information is weak only through direct observation by passive satellites, resulting in low accuracy of the effective radius.

[0031] Based on this, this application provides a method and apparatus for inverting the effective radius of cloud-based particles at the bottom of water clouds. It ranks the contribution of a series of relevant data from active satellite observations and selects target input parameters for evaluation using this ranking, resulting in a cloud-based particle effective radius inversion model. It also performs cloud detection and cloud phase detection using a series of relevant data from passive satellite observations, selecting water cloud observation data to input into the cloud-based particle effective radius inversion model to obtain the inversion result. Therefore, it no longer relies on the assumption of vertical uniformity of clouds, but uses both active and passive satellite observations to assist in the inversion of cloud-based particle effective radius, improving the accuracy of the predicted effective radius of cloud-based particles.

[0032] like Figure 1 As shown in the embodiments of this application, a method for inverting the effective radius of particles at the bottom of a water cloud is provided, the method comprising: S101. Obtain the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data from active satellite observations, and construct a training dataset based on the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data.

[0033] S102. Obtain the contribution of all training parameters in the training dataset to the inversion of the effective radius of cloud-bottom particles, and obtain the target input parameters based on the contribution of all training parameters.

[0034] S103. Input the target input parameters into the preset multiple models respectively, and evaluate each model with the first cloud bottom particle effective radius as the prediction target, obtain the evaluation results of each model, and select the model that meets the requirements as the cloud bottom particle effective radius inversion model based on the evaluation results.

[0035] S104. Obtain the second observation data, second auxiliary meteorological data, and rainfall data of the passive satellite corresponding to the target input parameters, and process the second observation data, second auxiliary meteorological data, and rainfall data to obtain a multidimensional feature dataset.

[0036] S105. Perform cloud detection and cloud phase detection on the multidimensional feature dataset in sequence, and select water cloud observation data according to the detection results. Perform feature extraction based on the water cloud observation data, and input the features of the water cloud observation data into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data.

[0037] In the aforementioned method for inverting the effective radius of cloud-based particles at the bottom of water clouds, a series of relevant data from active satellite observations are used to rank their contributions. Target input parameters are then selected based on this ranking and evaluated using various models to obtain the effective radius inversion model for cloud-based particles. Conversely, a series of relevant data from passive satellite observations are used for cloud detection and cloud phase detection. Water cloud observation data is then selected and input into the effective radius inversion model to obtain the inversion result. Therefore, the method no longer relies on the assumption of vertical uniformity of clouds. By using both active and passive satellite observations to assist in the inversion of the effective radius of cloud-based particles, the accuracy of the predicted effective radius of cloud-based particles is improved.

[0038] The exemplary embodiments described above will be explained below: S101. Obtain the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data from active satellite observations, and construct a training dataset based on the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data.

[0039] In this embodiment, the first effective radius of the cloud base particle is the actual effective radius of the cloud base particle; the first observation data is the observation data from the active satellite. A training dataset is constructed using the first effective radius of the cloud base particle, the first observation data, and the first auxiliary meteorological data for subsequent processing. For example, as... Figure 2 As shown, active satellite observation data, auxiliary meteorological data, and the effective radius of cloud-bottom particles observed by active satellite are obtained. These three data represent the first observation data, the first auxiliary meteorological data, and the first effective radius of cloud-bottom particles, respectively.

[0040] The first observation data includes cloud parameters and rainfall parameters, and the training dataset includes the effective radius of the first cloud base particle and training parameters. Specifically, the training dataset includes at least the effective radius of the first cloud base particle, the total liquid water path, the cloud liquid water path, the precipitation liquid water path, the layer height, the cloud base height, the cloud top height, and meteorological field parameters. The meteorological field parameters include at least wind speed, temperature, humidity, and pressure.

[0041] In some implementations, when acquiring the effective radius of cloud-bottom particles, first observation data, and first auxiliary meteorological data from active satellite observations, satellite data of multiple target products from the active satellite are acquired, and the satellite data of multiple target products are filtered based on multiple preset filtering conditions to obtain satellite data that meets the multiple filtering conditions; the index position of the satellite data is determined, and the satellite data required for each target product is extracted based on the index position; the effective radius of cloud-bottom particles is calculated based on the average geometric radius of droplets in the first target product, and the effective radius of cloud-bottom particles replaces the average geometric radius of droplets in the satellite data, and all parameters are merged to obtain a training dataset.

[0042] The satellite data includes observational data and primary auxiliary meteorological data; multiple screening conditions include at least cloud type, temperature, precipitation, and valid data conditions; the satellite data of the first target product includes at least the average geometric radius of droplets; all target parameters include at least three types: cloud parameters, rainfall parameters, and primary auxiliary meteorological data; the satellite data of the first target product includes at least the average geometric radius of droplets, total liquid water path, cloud liquid water path, precipitation liquid water path, and layer height; the satellite data of the second target product includes at least cloud base height and cloud top height; the satellite data of the second target product includes at least meteorological field parameters, which include at least wind speed, temperature, humidity, and pressure.

[0043] Optionally, when filtering satellite data of multiple target products based on preset multiple filtering conditions, the cloud type in the satellite data is determined. If the cloud type is a single-layer cloud, stratocumulus cloud, stratus cloud, or water cloud, the satellite data is determined to meet the cloud type condition. The location temperature in the satellite data is determined. If the location temperature is greater than a preset first temperature threshold, the satellite data is determined to meet the temperature condition. The precipitation information in the satellite data is determined. If the precipitation information is no precipitation, the satellite data is determined to meet the precipitation condition. The geometric radius value of liquid water in the satellite data is determined. If there are at least a target number of consecutive valid geometric radius values ​​of liquid water, the satellite data is determined to meet the valid data condition.

[0044] Specifically, for example, according to the data acquisition method provided on the CloudSat official website, download the three target products related to CloudSat / CPR: 2B-CLDCLASS-LIDAR, 2B-CWC-RVOD, and ECMWF-AUX. Filter the satellite data of these three products, selecting satellite data that simultaneously meets the following four filtering conditions: 1) Cloud type condition: single-layer cloud, stratocumulus or stratus cloud, water cloud; 2) Temperature condition: temperature at the corresponding location greater than 273.15K; 3) Precipitation condition: no precipitation; 4) Valid data condition: at least three consecutive valid liquid water geometric radius values. Record the index position of these satellite data that meet the filtering conditions in the file. Based on the index location, extract the parameters (i.e., satellite data) required for each product: including the average geometric radius of droplets, total liquid water path, cloud liquid water path, precipitation liquid water path, and corresponding layer height in 2B-CWC-RVOD; cloud base height and cloud top height in 2B-CLDCLASS-LIDAR; and meteorological field parameters (wind speed, temperature, humidity, pressure, etc.) in ECMWF-AUX.

[0045] Continuing on, the average geometric radius of the droplets in 2B-CWC-RVOD is calculated according to the following formula (1). Converted to effective radius of droplets (particles) This refers to the effective radius of the first cloud-bottom particle. After extraction, all parameters, namely the effective radius of the first cloud-bottom particle, are replaced by the average geometric radius of the droplet and combined with all other parameters to form the training data. (1) in, This represents an empirical parameter, which can be 0.38.

[0046] S102. Obtain the contribution of all training parameters in the training dataset to the inversion of the effective radius of cloud-bottom particles, and obtain the target input parameters based on the contribution of all training parameters.

[0047] In this embodiment, the target input parameter represents the training parameters used to construct the cloud base particle effective radius inversion model, i.e., the input scheme of the training parameters; the training parameters include at least cloud base height, cloud top height, total liquid water path, cloud liquid water path, precipitation liquid water path, and meteorological field parameters; the contribution of all training parameters in the cloud base particle effective radius inversion is obtained, and the target input parameter is obtained based on the contribution of all training parameters for subsequent processing.

[0048] In some implementations, all training parameters in the training dataset are input into a basic random forest model for contribution evaluation, obtaining the contribution of each training parameter in the inversion of the effective radius of the cloud bottom particle. The contributions of all training parameters are ranked, and a target training parameter that meets the contribution requirements is selected. The target training parameter and all training parameters are then input into the random forest model to re-invert the effective radius of the cloud bottom particle, obtaining a second effective radius. Based on the first and second effective radii of the cloud bottom particle, the estimation accuracy of the target training parameter and all training parameters in the random forest model is determined, and the target input parameter is selected based on the number of training parameters and the estimation accuracy. Here, the estimation accuracy characterizes the accuracy of the random forest model in estimating the effective radius of the cloud bottom particle.

[0049] Specifically, all training parameters are input into a basic random forest model. The contribution evaluation module in the model outputs the contribution of each training parameter to the inversion of the effective radius of cloud-bottom particles, and ranks the contributions. The top 5, 7, 10, and 20 target training parameters, along with all other training parameters, are selected and input into the random forest model to re-invert the effective radius of cloud-bottom particles, resulting in a second effective radius. The optimal input parameter scheme, i.e., the target input parameters, is selected by considering both the number of input variables (5, 7, 10, 20, and all) and the estimation accuracy.

[0050] S103. Input the target input parameters into the preset multiple models respectively, and evaluate each model with the first cloud bottom particle effective radius as the prediction target, obtain the evaluation results of each model, and select the model that meets the requirements as the cloud bottom particle effective radius inversion model based on the evaluation results.

[0051] In this embodiment, multiple models include a random forest model, a distributed gradient boosting model, a histogram gradient boosting model, and a deep neural network (DNN) model. The target input parameters from step S102 are input into the random forest model, the distributed gradient boosting model, the histogram gradient boosting model, and the deep neural network model, respectively. The first effective radius of the cloud bottom particle in step S101 is used as the prediction target. Each model is evaluated, and based on the evaluation results, the model that meets the requirements is selected as the cloud bottom particle effective radius inversion model. For example, such as... Figure 2 As shown.

[0052] Optionally, when evaluating each model and obtaining the evaluation result for each model, the evaluation index value corresponding to each model is calculated based on the preset multi-dimensional evaluation index; wherein, the evaluation index includes at least the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE); the evaluation index value corresponding to each model is visualized through a density scatter plot, and the evaluation result of each model is obtained based on the density scatter plot.

[0053] Specifically, all models use the target input parameters as input features, with the effective radius of the first cloud-bottom particle as the prediction target. For data partitioning, a uniform 70% training set and 30% test set ratio can be used. For deep neural network models, the training dataset can be further divided into 60% training, 20% validation, and 20% test. Model evaluation can employ 3-fold cross-validation, using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE) as evaluation metrics to obtain the evaluation results for each model.

[0054] S104. Obtain the second observation data, second auxiliary meteorological data, and rainfall data of the passive satellite corresponding to the target input parameters, and process the second observation data, second auxiliary meteorological data, and rainfall data to obtain a multidimensional feature dataset.

[0055] In this embodiment, the second observation data includes secondary cloud parameters and observation geometry information; the first auxiliary meteorological data and precipitation data are spatiotemporally matched with the second observation data; the multidimensional feature dataset includes cloud physical parameters, observation geometry, environmental meteorology and precipitation characteristics; the passive satellite second observation data, the second auxiliary meteorological data and precipitation data are processed to obtain the multidimensional feature dataset for subsequent processing.

[0056] Optionally, the target time window and target spatial grid for spatiotemporal matching are determined, and the second observation data, the second auxiliary meteorological data, and the rainfall data are spatiotemporally matched based on the target time window and the target spatial grid; invalid data in the second observation data, the second auxiliary meteorological data, and the rainfall data are identified and removed, and then fused to obtain a multidimensional feature dataset.

[0057] Specifically, passive satellite (e.g., POLDER, MODIS) observation data corresponding to the target input parameters are acquired, namely secondary cloud parameter products (e.g., cloud top height, cloud base height, cloud optical thickness, cloud water path, cloud droplet effective radius, and cloud phase, etc.) and observational geometric information (e.g., solar zenith angle, satellite zenith angle, relative azimuth angle, and scattering angle, etc.). Simultaneously, spatiotemporally matched secondary auxiliary meteorological data (e.g., wind speed, temperature, humidity, pressure, etc.) and precipitation data (e.g., microwave radiometer precipitation data) are acquired. Through rigorous spatiotemporal matching (15-minute time window and 0.1°×0.1° spatial grid) and quality control, the above multi-source data are integrated into a multi-dimensional feature dataset containing cloud physical parameters, observational geometry, environmental meteorology, and precipitation characteristics. This provides a high-quality, high-confidence input foundation for the subsequent construction of the cloud base particle effective radius inversion model.

[0058] S105. Perform cloud detection and cloud phase detection on the multidimensional feature dataset in sequence, and select water cloud observation data according to the detection results. Perform feature extraction based on the water cloud observation data, and input the features of the water cloud observation data into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data.

[0059] In this embodiment, water cloud observation data represents pixels containing clouds that are water clouds. Cloud detection and cloud phase detection are sequentially performed on the multi-dimensional feature dataset, and water cloud observation data are selected based on the detection results. Feature extraction is performed on the water cloud observation data, and the features of the water cloud observation data are input into the cloud base particle effective radius inversion model to obtain the cloud base particle effective radius inversion result corresponding to the water cloud observation data. For example, as... Figure 2 As shown.

[0060] Optionally, when performing cloud detection and cloud phase detection sequentially on the multidimensional feature dataset, and filtering out water cloud observation data based on the detection results, cloud detection is performed on the multidimensional feature dataset to determine whether the dataset contains clouds. If the dataset contains clouds, cloud phase detection is then performed to determine the cloud phase. If the cloud phase of the dataset is water cloud, the water cloud observation data in the dataset is filtered out. The cloud phase includes at least water clouds and ice clouds.

[0061] Specifically, water cloud pixels are selected based on passive satellite cloud detection and cloud phase detection results to form the input dataset: the L1B product of the POLDER data records the cloud inspection results (cloud_indicator), where 0 represents clear sky, 100 represents cloudy sky, and 50 represents uncertainty about whether there are clouds; the parameters in the L2B-RGB product record the cloud phase (cloud_phase), where 0-99 represents water clouds, 100-199 represents ice clouds, 200-229 represents mixed phase clouds, 230-239 represents uncertain phase, 240 represents clear sky, and 250 represents no observation. Based on these two parameters, POLDER pixels that are both cloudy and water clouds (i.e., water cloud pixels / water cloud observation data) are selected simultaneously.

[0062] Continuing, the selected passive satellite water cloud observation data is input into the cloud base particle effective radius inversion model, and the output is the cloud base particle effective radius inversion result of water cloud pixels. That is, the water cloud observation data is feature extracted and input into the cloud base particle effective radius inversion model. The model performs pixel-by-pixel inversion of the cloud base particle effective radius of each water cloud pixel through forward propagation calculation. The model output is the predicted value of the cloud base particle effective radius corresponding to each water cloud pixel, in micrometers (μm).

[0063] This application provides a method for inverting the effective radius of cloud-base particles in water clouds. It obtains the first effective radius of cloud-base particles from active satellite observations, first observation data, and first auxiliary meteorological data. A training dataset is constructed based on the first effective radius of cloud-base particles, the first observation data, and the first auxiliary meteorological data. The contribution of all training parameters in the training dataset to the inversion of the effective radius of cloud-base particles is obtained. Based on the contribution of all training parameters, the target input parameters are obtained. These target input parameters are then input into several preset models. Using the first effective radius of cloud-base particles as the prediction target, each model is evaluated, and the evaluation result for each model is obtained. Based on the evaluation results, models that meet the requirements are selected as cloud base particle effective radius inversion models. Second observation data, second auxiliary meteorological data, and precipitation data from passive satellites corresponding to the target input parameters are obtained. The second observation data, second auxiliary meteorological data, and precipitation data are processed to obtain a multidimensional feature dataset. Cloud detection and cloud phase detection are performed on the multidimensional feature dataset in sequence. Water cloud observation data are selected based on the detection results. Feature extraction is performed on the water cloud observation data, and the features of the water cloud observation data are input into the cloud base particle effective radius inversion model to obtain the cloud base particle effective radius inversion result corresponding to the water cloud observation data. The method for inverting the effective radius of cloud-based particles at the bottom of water clouds in this application ranks the contribution of a series of relevant data from active satellite observations, selects target input parameters based on the contribution ranking, and evaluates various models to obtain an effective radius inversion model for cloud-based particles. It also performs cloud detection and cloud phase detection using a series of relevant data from passive satellite observations, and selects water cloud observation data to input into the effective radius inversion model to obtain the inversion result. Therefore, it no longer relies on the assumption of vertical uniformity of clouds, and improves the accuracy of the predicted effective radius of cloud-based particles by using both active and passive satellite observations to assist in the inversion.

[0064] Reference Figure 3 As shown, this application provides an embodiment of a device for inverting the effective radius of particles at the bottom of a water cloud. The device includes: The construction module 301 is used to obtain the first effective radius of the cloud bottom particle, the first observation data, and the first auxiliary meteorological data from active satellite observation, and to construct a training dataset based on the first effective radius of the cloud bottom particle, the first observation data, and the first auxiliary meteorological data.

[0065] The first acquisition module 302 is used to acquire the contribution of all training parameters in the cloud bottom particle effective radius inversion of the training dataset, and to acquire the target input parameters based on the contribution of all training parameters; wherein, the target input parameters represent the training parameters used to construct the cloud bottom particle effective radius inversion model.

[0066] The second acquisition module 303 is used to input the target input parameters into a variety of preset models, and evaluate each model with the first effective radius of the cloud bottom particle as the prediction target, obtain the evaluation result of each model, and select the model that meets the requirements as the cloud bottom particle effective radius inversion model based on the evaluation result.

[0067] The third acquisition module 304 is used to acquire the second observation data, second auxiliary meteorological data and rainfall data of the passive satellite corresponding to the target input parameters, and to process the second observation data, second auxiliary meteorological data and rainfall data to obtain a multidimensional feature dataset.

[0068] The fourth acquisition module 305 is used to perform cloud detection and cloud phase detection on the multidimensional feature dataset in sequence, and to filter out water cloud observation data according to the detection results. Based on the water cloud observation data, feature extraction is performed, and the features of the water cloud observation data are input into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data; wherein, the water cloud observation data represents the pixels with clouds and water clouds.

[0069] In one possible implementation, the building module is specifically used for: The system acquires satellite data of various target products from active satellites and filters the satellite data of various target products based on a variety of preset filtering conditions to obtain satellite data that meets the filtering conditions. Among them, the satellite data includes observation data and first auxiliary meteorological data. The various filtering conditions include at least cloud type conditions, temperature conditions, precipitation conditions, and valid data conditions. Determine the index position of the satellite data and extract the satellite data required for each target product based on the index position; wherein, the satellite data of the first target product includes at least the average geometric radius of the droplets; calculate the effective radius of the cloud-bottom particles based on the average geometric radius of the droplets in the first target product, replace the average geometric radius of the droplets in the satellite data with the effective radius of the cloud-bottom particles, and merge all parameters to obtain the training dataset.

[0070] In one possible implementation, the building module is specifically used for: Determine the cloud type in the satellite data. If the cloud type is a single-layer cloud, stratocumulus cloud, stratus cloud, or water cloud, then the satellite data is determined to meet the cloud type condition. The location temperature in the satellite data is determined. If the location temperature is greater than a preset first temperature threshold, the satellite data is determined to meet the temperature conditions. Determine precipitation information in satellite data; if the precipitation information indicates no precipitation, then the satellite data is determined to meet the precipitation conditions. The geometric radius of liquid water in satellite data is determined. If there are at least a target number of consecutive valid geometric radius values ​​of liquid water, then the satellite data is considered to meet the conditions for valid data.

[0071] In one possible implementation, the first acquisition module is specifically used for: All training parameters in the training dataset are input into the basic random forest model for contribution evaluation, and the contribution of all training parameters in the inversion of the effective radius of cloud bottom particles is obtained. Rank the contributions of all training parameters and select the target training parameters that meet the contribution requirements. The target training parameters and all training parameters are input into the random forest model to re-invert the effective radius of the cloud bottom particles, and the second effective radius of the cloud bottom particles is obtained. Based on the effective radius of the first and second cloud-bottom particles, the target training parameters and the estimation accuracy of each training parameter in the random forest model are determined, and the target input parameters are selected based on the number of training parameters and the estimation accuracy; whereby the estimation accuracy characterizes the estimation accuracy of the random forest model for the effective radius of the cloud-bottom particles.

[0072] In one possible implementation, the second acquisition module is specifically used for: The evaluation index value corresponding to each model is calculated based on the preset multi-dimensional evaluation index; among which, the evaluation index includes at least the coefficient of determination, root mean square error and mean absolute error; The evaluation index values ​​corresponding to each model are visualized using density scatter plots, and the evaluation results of each model are obtained based on the density scatter plots.

[0073] In one possible implementation, the third acquisition module is specifically used for: Determine the target time window and target spatial grid for spatiotemporal matching, and perform spatiotemporal matching on the second observation data, the second auxiliary meteorological data, and the rainfall data based on the target time window and target spatial grid; After identifying and removing invalid data from the second observation data, the second auxiliary meteorological data, and the rainfall data, the data are fused to obtain a multidimensional feature dataset.

[0074] In one possible implementation, the fourth acquisition module is specifically used for: Perform cloud detection on a multidimensional feature dataset to determine whether the multidimensional feature dataset contains clouds; If the multidimensional feature dataset contains clouds, then continue to perform cloud phase detection on the multidimensional feature dataset to determine the cloud phase of the multidimensional feature dataset; wherein, the cloud phase includes at least water clouds and ice clouds; If the cloud phase of the multidimensional feature dataset is water cloud, then the water cloud observation data in the multidimensional feature dataset will be selected.

[0075] This application provides the aforementioned device for inverting the effective radius of cloud-based particles at the cloud base. It acquires the first effective radius of cloud-based particles from active satellite observations, first observation data, and first auxiliary meteorological data. Based on these data, a training dataset is constructed. The contribution of all training parameters in the effective radius inversion is obtained. Target input parameters are then obtained based on these contributions. These target input parameters are input into various preset models. Using the first effective radius of cloud-based particles as the prediction target, each model is evaluated to obtain its evaluation result. Based on the evaluation results, models that meet the requirements are selected as cloud base particle effective radius inversion models. Second observation data, second auxiliary meteorological data, and precipitation data from passive satellites corresponding to the target input parameters are obtained. The second observation data, second auxiliary meteorological data, and precipitation data are processed to obtain a multidimensional feature dataset. Cloud detection and cloud phase detection are performed on the multidimensional feature dataset in sequence. Water cloud observation data are selected based on the detection results. Feature extraction is performed on the water cloud observation data, and the features of the water cloud observation data are input into the cloud base particle effective radius inversion model to obtain the cloud base particle effective radius inversion result corresponding to the water cloud observation data. The cloud-bottom particle effective radius inversion device of this application ranks the contribution of a series of relevant data from active satellite observations, and selects target input parameters for evaluation of multiple models to obtain the cloud-bottom particle effective radius inversion model. It also performs cloud detection and cloud phase detection using a series of relevant data from passive satellite observations, and selects water cloud observation data to input into the cloud-bottom particle effective radius inversion model to obtain the cloud-bottom particle effective radius inversion result. Therefore, it no longer relies on the assumption of vertical uniformity of clouds, and improves the accuracy of the predicted cloud-bottom particle effective radius by using both active and passive satellite observations to assist in the inversion of cloud-bottom particle effective radius.

[0076] like Figure 4 As shown in the embodiment of this application, an electronic device 400 includes a processor 401, a memory 402, and a bus. The memory 402 stores machine-readable instructions that can be executed by the processor 401. When the electronic device is running, the processor 401 communicates with the memory 402 via the bus. The processor 401 executes the machine-readable instructions to perform the steps of the above-described method for inverting the effective radius of particles at the bottom of water clouds.

[0077] Specifically, the memory 402 and processor 401 mentioned above can be general-purpose memory and processor, without any specific limitations. When the processor 401 runs the computer program stored in the memory 402, it can execute the above-mentioned method for inverting the effective radius of particles at the bottom of the water cloud.

[0078] Corresponding to the above-described method for inverting the effective radius of particles at the bottom of water clouds, this application also provides a computer-readable storage medium storing a computer program. When the computer program is run by a processor, it executes the steps of the above-described method for inverting the effective radius of particles at the bottom of water clouds.

[0079] The electronic devices and storage media provided in this application, in real time, rank the contribution of a series of relevant data from active satellite observations, and then select target input parameters through the contribution ranking to evaluate multiple models, thereby obtaining a cloud-base particle effective radius inversion model. A series of relevant data from passive satellite observations are used for cloud detection and cloud phase detection, and water cloud observation data is selected and input into the cloud-base particle effective radius inversion model to obtain the cloud-base particle effective radius inversion result. Therefore, it no longer relies on the assumption of vertical uniformity of clouds, and improves the accuracy of the predicted cloud-base particle effective radius by using both active and passive satellite observations to assist in the inversion of cloud-base particle effective radius.

[0080] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0081] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0082] In addition, the functional units in the various embodiments of this application 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.

[0083] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0084] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for inverting the effective radius of particles at the bottom of a water cloud, characterized in that, The method includes: Step S1: Obtain the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data from active satellite observations, and construct a training dataset based on the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data; Step S2: Obtain the contribution of all training parameters in the cloud bottom particle effective radius inversion from the training dataset, and obtain the target input parameters based on the contribution of all training parameters; wherein, the target input parameters represent the training parameters used to construct the cloud bottom particle effective radius inversion model; Step S3: Input the target input parameters into a variety of preset models, and evaluate each model with the first effective radius of the cloud bottom particle as the prediction target to obtain the evaluation results of each model. Based on the evaluation results, select the model that meets the requirements as the cloud bottom particle effective radius inversion model. Step S4: Obtain the second observation data, second auxiliary meteorological data, and rainfall data of the passive satellite corresponding to the target input parameters, and process the second observation data, the second auxiliary meteorological data, and the rainfall data to obtain a multidimensional feature dataset; Step S5: Perform cloud detection and cloud phase detection on the multidimensional feature dataset in sequence, and select water cloud observation data according to the detection results. Perform feature extraction on the water cloud observation data, and input the features of the water cloud observation data into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data; wherein, the water cloud observation data represents pixels with clouds and water clouds.

2. The method for inverting the effective radius of particles at the bottom of water clouds according to claim 1, characterized in that, In step S1, acquiring the effective radius of cloud-bottom particles, first observation data, and first auxiliary meteorological data from active satellite observations includes: The system acquires satellite data of various target products from active satellites and filters the satellite data of the various target products based on a variety of preset filtering conditions to obtain satellite data that meets the various filtering conditions; wherein, the satellite data includes observation data and first auxiliary meteorological data; the various filtering conditions include at least cloud type conditions, temperature conditions, precipitation conditions, and valid data conditions; The index position of the satellite data is determined, and the satellite data required for each target product is extracted based on the index position; wherein, the satellite data of the first target product includes at least the average geometric radius of the droplets; the effective radius of the cloud-bottom particles is calculated based on the average geometric radius of the droplets in the first target product, and the effective radius of the cloud-bottom particles replaces the average geometric radius of the droplets in the satellite data and all parameters are merged to obtain the training dataset.

3. The method for inverting the effective radius of particles at the bottom of water clouds according to claim 2, characterized in that, The filtering of satellite data for the various target products based on preset multiple filtering conditions includes: The cloud type in the satellite data is determined. When the cloud type is a single-layer cloud, stratocumulus cloud, stratus cloud, or water cloud, the satellite data is determined to meet the cloud type condition. The location temperature in the satellite data is determined, and when the location temperature is greater than a preset first temperature threshold, the satellite data is determined to meet the temperature condition. The precipitation information in the satellite data is determined. When the precipitation information indicates no precipitation, the satellite data is determined to meet the precipitation conditions. The geometric radius value of liquid water in the satellite data is determined. If there are at least a target number of consecutive valid geometric radius values ​​of liquid water, then the satellite data is determined to meet the valid data condition.

4. The method for inverting the effective radius of particles at the bottom of water clouds according to claim 1, characterized in that, In step S2, obtaining the contribution of all training parameters in the inversion of the effective radius of cloud-bottom particles, and obtaining the target input parameters based on the contribution of all training parameters, includes: All training parameters in the training dataset are input into the basic random forest model for contribution evaluation, and the contribution of all training parameters in the cloud bottom particle effective radius inversion is obtained. Rank the contributions of all training parameters and select the target training parameters that meet the contribution requirements. The target training parameters and all training parameters are respectively input into the random forest model to re-invert the effective radius of the cloud bottom particles, and the second effective radius of the cloud bottom particles is obtained. Based on the effective radius of the first cloud bottom particle and the effective radius of the second cloud bottom particle, the target training parameters and the estimation accuracy of each of the training parameters in the random forest model are determined, and the target input parameters are selected based on the number of training parameters and the estimation accuracy; wherein, the estimation accuracy characterizes the estimation accuracy of the random forest model for the effective radius of the cloud bottom particle.

5. The method for inverting the effective radius of particles at the bottom of water clouds according to claim 1, characterized in that, In step S3, evaluating each model and obtaining the evaluation result for each model includes: The evaluation index value corresponding to each model is calculated based on the preset multi-dimensional evaluation index; wherein, the evaluation index includes at least the coefficient of determination, root mean square error and mean absolute error; The evaluation index values ​​corresponding to each model are visualized using a density scatter plot, and the evaluation results of each model are obtained based on the density scatter plot.

6. The method for inverting the effective radius of particles at the bottom of water clouds according to claim 1, characterized in that, In step S4, the second observation data, the second auxiliary meteorological data, and the rainfall data are processed to obtain a multidimensional feature dataset, including: Determine the target time window and target spatial grid for spatiotemporal matching, and perform spatiotemporal matching on the second observation data, the second auxiliary meteorological data, and the rainfall data based on the target time window and the target spatial grid; After identifying and removing invalid data from the second observation data, the second auxiliary meteorological data, and the rainfall data, the data are fused to obtain a multidimensional feature dataset.

7. The method for inverting the effective radius of particles at the bottom of water clouds according to claim 1, characterized in that, In step S5, the process of sequentially performing cloud detection and cloud phase detection on the multidimensional feature dataset, and then filtering out water cloud observation data based on the detection results, includes: Perform cloud detection on the multidimensional feature dataset to determine whether the multidimensional feature dataset contains clouds; If the multidimensional feature dataset contains clouds, then cloud phase detection is performed on the multidimensional feature dataset to determine the cloud phase of the multidimensional feature dataset; wherein, the cloud phase includes at least water clouds and ice clouds; If the cloud phase of the multidimensional feature dataset is water cloud, then the water cloud observation data in the multidimensional feature dataset are selected.

8. A device for inverting the effective radius of particles at the bottom of a water cloud, characterized in that, The device includes: The module is used to acquire the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data from active satellite observations, and to construct a training dataset based on the first effective radius of cloud-bottom particles, the first observation data, and the first auxiliary meteorological data. The first acquisition module is used to acquire the contribution of all training parameters in the cloud bottom particle effective radius inversion of the training dataset, and to acquire the target input parameters based on the contribution of all training parameters; wherein, the target input parameters represent the training parameters used to construct the cloud bottom particle effective radius inversion model; The second acquisition module is used to input the target input parameters into a variety of preset models, and evaluate each model with the first effective radius of the cloud bottom particle as the prediction target, obtain the evaluation result of each model, and select the model that meets the requirements as the cloud bottom particle effective radius inversion model based on the evaluation result. The third acquisition module is used to acquire the second observation data, the second auxiliary meteorological data, and the rainfall data of the passive satellite corresponding to the target input parameters, and to process the second observation data, the second auxiliary meteorological data, and the rainfall data to obtain a multidimensional feature dataset; The fourth acquisition module is used to sequentially perform cloud detection and cloud phase detection on the multidimensional feature dataset, and filter out water cloud observation data according to the detection results. Based on the water cloud observation data, feature extraction is performed, and the features of the water cloud observation data are input into the cloud bottom particle effective radius inversion model to obtain the cloud bottom particle effective radius inversion result corresponding to the water cloud observation data; wherein, the water cloud observation data represents pixels with clouds and water clouds.

9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the water cloud bottom particle effective radius inversion method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method for inverting the effective radius of particles at the bottom of a water cloud as described in any one of claims 1 to 7.

Images