A cloud top height quality inspection method, device, computing device and storage medium

Abstract

The present invention discloses a cloud top height quality inspection method, device, computing device and storage medium, belonging to the technical field of atmospheric detection and remote sensing, including: using lidar cloud profile data as the detection true value and imaging radiometer cloud product data as the data to be quality inspected. On the basis of the spatio-temporal matching of the two types of data, unqualified data is screened out through spatio-temporal uniformity tests, and for the qualified matching data, through layer-by-layer and step-by-step analysis, the total sample size, the correct detection, missed detection, false detection, correctly screened sample size and proportion are statistically calculated, as well as the correct identification sample size, wrong identification sample size and the corresponding cloud top height difference of each cloud type including cirrus clouds and lower layer clouds, so as to achieve complete and accurate cloud top height quality detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of atmospheric detection and remote sensing, and particularly relates to a cloud top height quality inspection method, device, computing device and storage medium. Background Art

[0002] As a basic parameter in the fields of meteorology and climatology, cloud top height has important scientific and application values in aspects such as short-term weather forecasting, climate prediction, and cloud radiative forcing. The distribution of clouds is uneven in time and space, and the change speed is affected by many aspects, which puts high requirements on the time, space resolution, and coverage of the instruments used to observe cloud top height. Spaceborne remote sensing technology provides a global perspective for cloud top height measurement. Among them, the imaging radiometer that can be carried on geostationary satellites and polar orbiting satellite platforms relies on a large observation range and a high temporal resolution observation mode, and has become the most stable source of cloud observation data at present. Since the imaging radiometer does not have the ability to directly detect cloud top height, it is necessary to rely on preprocessing steps such as cloud detection and cloud classification to obtain prior values, and then use a radiative transfer model to invert cloud top height. However, the imaging radiometer has poor observation ability for cirrus clouds, that is, thin clouds mainly composed of ice crystal particles existing at altitudes above 6 kilometers. This is because the large transmittance of cirrus clouds causes the observed radiation signal to be interfered by the radiation under the cloud. When only cirrus clouds exist in the actual scene, the cirrus clouds cannot be accurately detected. When multiple cloud layers including cirrus clouds exist in the actual scene, only the low-level clouds can be detected and classified. This leads to large errors in the cloud top height product data of cirrus clouds and related multi-layer clouds.

[0003] As one of the main observation targets of meteorological satellites in various countries, the current cloud top height product data all have operational verification methods. Taking my country's Fengyun-4 multi-channel scanning imaging radiometer (FY-4 / AGRI) as an example, the paper "Consistency Evaluation of the Accuracy of FY-4B and FY-4A Cloud Top Property Products" announced the accuracy evaluation results of the cloud top property products of Fengyun-4A (abbreviated as FY-4A) and B (abbreviated as FY-4B). The evaluation method is to use the cloud products of the US EOS / MODIS polar-orbiting meteorological satellite and Japan's second-generation geostationary meteorological satellite Himawari-8, which have partially overlapping fields of view and equivalent imaging performance, to verify the consistency of the accuracy of the cloud top products inverted by FY-4A and FY-4B. The conclusion is that the test results of the full sample show that the accuracy of the FY-4B cloud top parameter product is roughly equivalent to that of FY-4A, and has good consistency. Among them, the large difference in the multi-layer clouds of the B satellite is mainly due to the cloud classification misjudging cirrus clouds as multi-layer clouds. It can be seen that the misjudgment of the previous step caused by cirrus has a great impact on the cloud top height result. At the same time, although the Japanese and American imagers used in the test have certain differences in spectral channel design, calibration accuracy and radiation transmission mode used in the inversion algorithm, there is no obvious difference in the instrument and inversion mechanism. There are large differences in the detection and classification of cirrus clouds, which is a common problem of imaging radiometers, resulting in their inaccurate reference values ​​for cloud top height assessment.

[0004] In response to the above problems, a Chinese patent document with a publication number of CN116150151A discloses a cloud top height inspection method and system based on strict cloud phase matching, which includes obtaining a test source and a tested source to be matched in time and space and performing time and space matching; reading the time and space data of the test source and the tested source, searching for pixels with consistent cloud phases in the test source and the tested source, and calculating the distance between the pixels using the time and space data; calculating the time difference between the test source and the tested source using the time and space data, selecting pixels with consistent cloud phases within a preset time difference range and a preset distance range as matching data set output; using the matching data set to perform statistics on cloud top height products of the same phase, and generating statistical indicators for cloud top height inspection. This method strictly matches cloud phases, and its goal is to inspect cloud top heights on the premise that cloud classification is as consistent as possible, without evaluating the impact of passive remote sensing's failure to accurately detect and identify cirrus clouds on cirrus cloud heights and multi-layer cloud height inversion. Summary of the invention

[0005] In view of the above, in order to solve the technical problem in the existing cloud top height product quality inspection process that the subsequent impact assessment is incomplete due to the incorrect detection and classification of cirrus clouds, the present invention provides a cloud top height quality inspection method, device, computing device and storage medium, introducing spaceborne lidar as the inspection true value with higher sensitivity to cirrus clouds. By matching the observation data of spaceborne lidar and imaging radiometer, and using the lidar profile observation results with higher sensitivity to cirrus clouds, a step-by-step and layer-by-layer analysis of the errors existing in the cloud top height inversion based on the imaging radiometer observation data, including cloud detection and cloud classification, is carried out to achieve complete and accurate cloud top height quality detection.

[0006] To achieve the above invention objective, a cloud top height quality inspection method provided by an embodiment includes the following steps:

[0007] Obtain lidar cloud profile data as the detection true value, imaging radiometer cloud product data to be quality inspected, and auxiliary data;

[0008] Based on the spatio-temporal data in the auxiliary data, perform spatio-temporal matching of the imaging radiometer cloud product data relative to the lidar cloud profile data and determine the matching data;

[0009] According to the spatio-temporal information matching situation in the matching data, match the cloud detection and cloud classification results in the lidar cloud profile data with the cloud detection and cloud classification results in the imaging radiometer cloud product;

[0010] Perform a spatio-temporal homogeneity test on the matching data to screen out the data with unqualified spatio-temporal homogeneity, obtain the qualified matching data and count the total sample size;

[0011] Compare the qualified matching imaging radiometer cloud detection results with the lidar cloud detection results, and count the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples; Among the correctly detected samples, compare the qualified matching imaging radiometer cloud classification results with the lidar cloud classification results, and count the sample sizes of correct and incorrect classifications of each cloud type from the aspects of single-layer clouds and multi-layer clouds and in combination with cloud top height assessment as the judgment condition;

[0012] Compare the matching imaging radiometer cloud product data with the lidar cloud profile data, and calculate the cloud top height differences of each correctly classified and incorrectly classified single-layer cloud type, the cloud top height differences and layer average height differences of correctly classified multi-layer clouds, and the cloud top height differences, layer top height differences and layer average height differences of incorrectly classified multi-layer clouds;

[0013] Output the total sample size, the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples, the sample sizes of correct and incorrect classifications of each cloud type, and the corresponding cloud top height differences, layer top height differences and layer average height differences as the complete cloud top height quality inspection results.

[0014] Preferably, performing spatio-temporal matching of the imaging radiometer cloud product data with respect to the lidar cloud profile data based on spatio-temporal auxiliary data and determining the matching data, including:

[0015] Determining a time threshold and a space threshold, wherein the spatio-temporal auxiliary data includes measurement time and longitude and latitude, and the imaging radiometer cloud product data includes imaging radiometer cloud product samples measured at different times and longitudes and latitudes;

[0016] Based on the measurement time and longitude and latitude corresponding to each profile in the lidar cloud profile data, screening out the imaging radiometer cloud product samples with the closest spatial distance to each profile within each time threshold and space threshold to form the matching data.

[0017] Preferably, performing a spatio-temporal homogeneity test on the matching data to filter out the data with unqualified spatio-temporal homogeneity and obtain qualified matching data, including:

[0018] Performing a lidar profile spatio-temporal homogeneity test, a radiometer pixel internal homogeneity test, and a radiometer pixel peripheral homogeneity test on the matching data, and integrating the sample data that all three tests are qualified to obtain qualified matching data.

[0019] Preferably, the lidar profile spatio-temporal homogeneity test refers to testing the homogeneity between multiple basic profiles included in each lidar cloud profile, calculating the average deviation of the cloud top height and the difference in the number of levels of all basic profiles. When the number of levels changes or the average deviation of the cloud top height is greater than or equal to a given cloud top height deviation threshold, it is considered that the spatio-temporal homogeneity of all lidar cloud profile samples is unqualified and is screened out. When the number of levels is consistent and the average deviation of the cloud top height is less than the given cloud top height deviation threshold, it is considered that all lidar cloud profile samples are qualified;

[0020] The radiometer pixel internal homogeneity test refers to testing the homogeneity between multiple pixels included in each radiometer cloud product pixel, calculating the standard deviation of the brightness temperature in the infrared band between all pixels. When the standard deviation of the brightness temperature is greater than or equal to a given brightness temperature standard deviation threshold, it is considered that the spatio-temporal homogeneity of all radiometer cloud product samples is unqualified and is screened out. When the standard deviation of the brightness temperature is less than the given brightness temperature standard deviation threshold, it is considered that all radiometer cloud product samples are qualified;

[0021] The radiometer pixel perimeter uniformity test means that, based on time thresholds and spatial thresholds, the search time - space range is expanded, and all cloud product samples with longitude - latitude differences from each profile within the search time - space range are selected from the matching data. The average deviation between the cloud features of all cloud product samples and the average cloud feature is calculated, where the average cloud feature is the average value of the cloud features of all cloud product samples. When the average deviation is greater than or equal to the given cloud feature deviation threshold, it is considered that the time - space uniformity of all cloud product samples is unqualified and they are screened out. When the average deviation is less than the given cloud feature deviation threshold, it is considered that all cloud product samples are qualified. Among them, the cloud feature is a matrix composed of marking clouds as 1 and non - clouds as 0.

[0022] Preferably, the cloud top height deviation threshold between lidar profiles is 0 - 1 km, the standard deviation threshold of brightness temperature in the infrared band between radiometer pixels is 0 - 3 K, and the cloud feature deviation threshold between radiometer pixels is 0 - 0.25.

[0023] Preferably, the sample sizes of correct classification and misclassification of each cloud type are statistically counted by evaluating from two aspects of single - layer clouds and multi - layer clouds and combining with cloud top height as the judgment condition, including:

[0024] If the single - layer clouds identified by the lidar and the imaging radiometer are of the same type, it is recorded as a correctly classified single - layer cloud, where the single - layer cloud types include water clouds, ice clouds, and cirrus clouds;

[0025] If the lidar and the imaging radiometer both identify single - layer clouds but the types are inconsistent, and the cloud top height deviation < 3 km, it is recorded as a misclassified single - layer cloud, corresponding to the situation where the phase determination is incorrect but the layer identification is consistent in actual measurement;

[0026] If the lidar and the imaging radiometer both identify multi - layer clouds and the cloud top height deviation < 3 km, it is recorded as a correctly classified multi - layer cloud;

[0027] If the lidar and the imaging radiometer have inconsistent layer identifications, or both identify multi - layer clouds but the cloud top height deviation ≥ 3 km, it is recorded as a misclassified multi - layer cloud, corresponding to the situation where the imaging radiometer fails to identify cirrus clouds or misidentifies high - altitude aerosols as cirrus clouds in actual measurement;

[0028] Statistically count the sample sizes of correct classification and misclassification of each cloud type.

[0029] Preferably, the cloud top height difference refers to the difference between the cloud top height of the imaging radiometer and the cloud top height of the top - layer cloud in the lidar cloud profile.

[0030] The layer top height difference refers to the difference between the cloud top height of the imaging radiometer and the layer top height of the closest layer measured by the lidar cloud profile;

[0031] The height difference of the said layers refers to the difference between the cloud top height of the imaging radiometer and the average height of multiple layers in the lidar cloud profile.

[0032] To achieve the above-mentioned invention objective, the embodiment also provides a cloud top height quality inspection device, including:

[0033] A data acquisition module, which is used to acquire lidar cloud profile data as the detection true value, imaging radiometer cloud product data to be quality inspected, and auxiliary data;

[0034] A spatio-temporal matching module, which is used to perform spatio-temporal matching of the imaging radiometer cloud product data relative to the lidar cloud profile data based on the spatio-temporal data in the auxiliary data and determine the matching data;

[0035] A cloud detection and classification matching module, which is used to match the cloud detection and cloud classification results in the lidar cloud profile data and the cloud detection and cloud classification results in the imaging radiometer cloud product according to the spatio-temporal information matching situation in the matching data;

[0036] A homogeneity test module, which is used to perform spatio-temporal homogeneity tests on the matching data to screen out the data with unqualified spatio-temporal homogeneity, obtain qualified matching data and count the total sample size;

[0037] A cloud detection and classification result layer-by-layer comparison module, which is used to compare the cloud detection results of the qualified matching imaging radiometer with the lidar cloud detection results, count the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples, and in the samples of correct detections, compare the cloud classification results of the qualified matching imaging radiometer with the lidar cloud classification results, and count the sample sizes of correct and incorrect classifications of each cloud type from two aspects of single-layer clouds and multi-layer clouds and in combination with cloud top height evaluation as the judgment condition;

[0038] A difference calculation module, which is used to compare the matching imaging radiometer cloud product data with the lidar cloud profile data, calculate the cloud top height differences of each single-layer cloud type of correct and incorrect classifications, the cloud top height differences and layer average height differences of correctly classified multi-layer clouds, and the cloud top height differences, layer top height differences and layer average height differences of incorrectly classified multi-layer clouds;

[0039] A data output module, which is used to output the total sample size, the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples, the correct and incorrect classification sample sizes of each cloud type, and the corresponding cloud top height differences, layer top height differences and layer average height differences as the complete cloud top height quality inspection results.

[0040] To achieve the above-mentioned invention objectives, the embodiments further provide a computing device, including a memory and one or more processors. An executable code is stored in the memory. When the one or more processors execute the executable code, it is used to implement the above-mentioned cloud top height quality inspection method.

[0041] To achieve the above-mentioned invention objectives, the embodiments further provide a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements the above-mentioned cloud top height quality inspection method.

[0042] Compared with the prior art, the beneficial effects of the present invention at least include:

[0043] The cloud top height quality inspection method provided by the embodiments of the present invention gradually inspects the entire process link that affects the cloud top height of the radiation imager, including cloud detection and cloud classification, and maximally compares the cloud top height inverted by the radiation imager with the layers actually detected by the lidar. Especially relying on the high-precision identification of cirrus clouds by spaceborne lidar, it evaluates the missed detections, false detections, and misclassified samples caused by cirrus clouds and other layers, making the cloud top height inspection results more complete and detailed. Compared with the detection results under only the same cloud detection and cloud type, the effective cloud top height detection sample size has increased by 127%, and it makes a more accurate judgment on the inversion accuracy of the cloud top height, especially the cloud top height of multi-layer clouds, under different previous determination conditions, providing more comprehensive data support for subsequent product optimization. This method has broad application prospects in the fields of meteorological observation, climate research, and remote sensing data quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0045] Figure 1 is a flowchart of the cloud top height quality inspection method provided by an embodiment;

[0046] Figure 2 is a specific flowchart of the homogeneity test step in the cloud top height quality inspection method provided by an embodiment;

[0047] Figure 3 is a specific flowchart of the layer-by-layer comparison of cloud detection and classification results in the cloud top height quality inspection method provided by an embodiment;

[0048] Figure 4Schematic diagram of the observation orbit of spaceborne lidar (shown by the gray line) and the observation coverage of the imaging radiometer (shown within the red circle) provided by an embodiment;

[0049] Figure 5 Cross-sectional view of the cloud profile of spaceborne lidar provided by an embodiment, and schematic diagram of the comparison result with the cloud top height (shown by triangles) of the imaging radiometer;

[0050] Figure 6 Inspection result diagram of cloud top height provided by an embodiment;

[0051] Figure 7 Schematic structural diagram of the cloud top height quality inspection device provided by an embodiment. Detailed implementation manners

[0052] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation manners described herein are only used to explain the present invention and do not limit the protection scope of the present invention.

[0053] As Figure 1 shown, the cloud top height quality inspection method provided by the embodiment includes the following steps:

[0054] S1. Obtain data: Obtain lidar cloud profile data as the detection true value, its cloud determination result, the imaging radiometer cloud product data to be quality inspected, and auxiliary data.

[0055] In the embodiment, the lidar cloud profile data is detailed information about the vertical distribution of clouds obtained through lidar technology, specifically including layer identification, layer classification, and phase state classification results. Compared with other data, when used as the detection true value, this lidar cloud profile data is clearer about the cloud layer information, the corresponding cloud top height, cloud bottom height, and cloud thickness. The lidar classifies clouds through the scattering ability and polarization sensitivity of cloud particles to laser, and can also obtain the cloud phase state more accurately, that is, classify clouds into liquid-phase water clouds and ice-phase ice clouds according to the physical characteristics of clouds (such as temperature, water phase state, etc.). Based on these lidar cloud profile data, cloud determination results such as whether there are clouds, cloud types, and cloud top height can be determined.

[0056] The imaging radiometer cloud product data refers to various characteristic information about clouds obtained by the satellite imaging radiometer, specifically including cloud detection, cloud classification, and cloud top height data. Among them, the cloud top height data is a prior value obtained through previous steps such as cloud detection and cloud classification, and is obtained by inverting through the radiation transfer model. Compared with the cloud top height data obtained by the lidar, there are problems such as low accuracy of the cloud top height and easy missed detection for cirrus clouds.

[0057] The auxiliary data mainly includes the measurement time information and measurement space information of lidar and satellite imaging radiometers. The measurement space information is generally longitude and latitude information. In addition, it also includes the basic profiles and basic pixel information at Level 1 (calibrated raw data).

[0058] From the sample perspective, the imaging radiometer cloud product data includes imaging radiometer cloud product samples measured at different times and longitudes and latitudes. Each imaging radiometer cloud product sample is cloud detection, cloud classification, and cloud top height data within a fixed size range (e.g., 4km×4km) corresponding to each acquisition time point.

[0059] S2, spatio-temporal matching: Perform spatio-temporal matching of the imaging radiometer cloud product data relative to the lidar cloud profile data based on the spatio-temporal data in the auxiliary data and determine the matching data.

[0060] The necessary step for cloud top height quality inspection of the data to be inspected according to the detection truth value is spatio-temporal matching. Specifically, perform spatio-temporal matching of the imaging radiometer cloud product data with respect to the lidar cloud profile data using spatio-temporal auxiliary data. During specific spatio-temporal matching, first, the time threshold and space threshold are considered. The time threshold can be half of the time resolution of the geostationary satellite, e.g., 7.5 minutes. Considering the reduced resolution of the edge pixels of the geostationary satellite, the space threshold usually reaches about 1.5 times the central resolution (e.g., if the central resolution is 4km×4km, the edge resolution may be reduced to 6km×6km), and the threshold range should be appropriately expanded, e.g., 5km.

[0061] Then, based on the measurement time and longitude and latitude corresponding to each profile in the lidar cloud profile data, select the imaging radiometer cloud product sample with the closest spatial distance to each profile within each time threshold and space threshold range to form the matching data. Each profile in the matching data corresponds to an imaging radiometer cloud product sample, where the spatial distance is calculated using the Haversine formula, and the formula is:

[0062] (1)

[0063] where is the longitude and latitude of the lidar, is the longitude and latitude of the imaging radiometer cloud product sample, is and the difference, is and the difference, represents the radius of the earth, i.e., 6371km.

[0064] S3, Cloud Detection and Classification Matching: Match the cloud detection and cloud classification results in the lidar cloud profile data with those in the imaging radiometer cloud products according to the spatio-temporal information matching in the matching data.

[0065] Perform cloud classification on cloud samples. The obtained cloud classification results of the imaging radiometer include water clouds, ice clouds, cirrus clouds, and multi-layer clouds, where the ice cloud is a thick ice cloud that does not contain cirrus clouds; add the layer height of the lidar for determination. On the basis of directly comparing the cloud classification products of the imaging radiometer, add the comparison of other layers identified by the lidar profile to supplement the missed detection and misdetection of individual layers.

[0066] There are samples of imaging radiometer cloud products with qualified matching for each profile in the matching data, and there are cloud detection and cloud classification results in the lidar cloud profile data and cloud detection and cloud classification results in the imaging radiometer cloud products. In this way, the matching of cloud detection results and the matching of cloud classification results can be achieved according to the matching situation in the matching data.

[0067] S4, Homogeneity Test: Perform spatio-temporal homogeneity tests on the matching data to filter out the data with unqualified spatio-temporal homogeneity, obtain qualified matching data, and count the total sample size.

[0068] In actual detection, clouds may show uneven distribution and rapid changes at the edge of the cloud system. In this case, since the imaging radiometer samples at a certain time interval during detection, it often occurs that the unevenness and rapid changes of the cloud at the sampling time point of the imaging radiometer cause changes in the observation target, and some pixels contain both cloud pixels and non-cloud pixels, or single-layer cloud pixels and multi-layer cloud pixels at the same time. The same is true for the lidar, which further leads to inconsistent observation targets between the lidar cloud profile and the imaging radiometer cloud product. This situation is considered unqualified spatio-temporal homogeneity. Therefore, it is necessary to filter out the data in this situation. It can be filtered according to the deviation between the basic profiles of the lidar, the internal (inter-pixel) deviation and external (peripheral) deviation or standard deviation of the imaging radiometer pixels, as Figure 2 shown. The specific process is as follows:

[0069] Perform the spatio-temporal homogeneity test on the matched data, that is, test the homogeneity among multiple basic profiles included in each lidar cloud profile (for example, in a cloud profile with a spatial resolution of 5 km, the basic profile resolution is 333 m, that is, it contains 15 basic profiles). Calculate the average deviation of cloud top heights and the difference in the number of levels for all basic profiles. When the number of levels changes or the average deviation of cloud top heights is greater than or equal to the given cloud top height deviation threshold, it is considered that the spatio-temporal homogeneity of all lidar cloud profile samples is unqualified and they are screened out. When the number of levels is consistent and the average deviation of cloud top heights is less than the given cloud top height deviation threshold, it is considered that all lidar cloud profile samples are qualified. The cloud top height deviation threshold can be set to 1 km.

[0070] Perform the internal homogeneity test of radiometer pixels on the matched data, that is, test the homogeneity among multiple pixels included in each radiometer cloud product pixel (for example, in a pixel with a spatial resolution of 4 km × 4 km, the resolution of each pixel is 1 km × 1 km, that is, it contains 16 pixels). Calculate the standard deviation of brightness temperatures in the infrared band (such as 10 - 12 μm) among all pixels. When the standard deviation of brightness temperatures is greater than or equal to the given standard deviation threshold of brightness temperatures, it is considered that the spatio-temporal homogeneity of all radiometer cloud product samples is unqualified and they are screened out. When the standard deviation of brightness temperatures is less than the given standard deviation threshold of brightness temperatures, it is considered that all radiometer cloud product samples are qualified. The standard deviation threshold of brightness temperatures can be set to 3 K.

[0071] Perform the radiometer pixel perimeter homogeneity test on the imaging radiometer data with qualified internal pixel homogeneity, that is, expand the search spatio-temporal range based on time threshold and spatial threshold. Generally, choose 2 times the time threshold and 2 times the spatial threshold, that is, take the spatio-temporal range determined by the time threshold and spatial threshold as the center and expand 2 times outward as the search spatio-temporal range. Select all cloud product samples from the matched data whose longitude and latitude differences from each profile are within the search spatio-temporal range, and calculate the average deviation of cloud characteristics of all cloud product samples from the average cloud characteristics , where the average cloud characteristics are the average of the cloud characteristics of all cloud product samples; when the average deviation is greater than or equal to the given cloud characteristic deviation threshold, it is considered that the spatio-temporal homogeneity of all cloud product samples is unqualified and they are screened out. When the average deviation is less than the given cloud characteristic deviation threshold, it is considered that all cloud product samples are qualified, obtain the qualified matched data and count the total sample size;

[0072] (2)

[0073] where N represents the number of samples within the search spatio-temporal range, represents the cloud property characteristic of the i-th closest sample; It represents the average cloud characteristics of all cloud product samples within the range, where the cloud characteristics are a matrix composed of marking clouds as 1 and marking no clouds as 0, and the cloud characteristic deviation threshold can be set to 0.25.

[0074] Finally, by synthesizing the samples that all pass the three tests, qualified matching data is obtained.

[0075] S5. Layer-by-layer comparison of cloud detection and classification results: Compare the cloud detection results of the imaging radiometer with those of the lidar for the qualified matching, and count the sample sizes and proportions of correct detections, missed detections, false detections, and correct rejections; among the correctly detected samples, compare the cloud classification results of the imaging radiometer with those of the lidar for the qualified matching, and count the sample sizes of correct classification and misclassification for each cloud type by taking into account single-layer clouds and multi-layer clouds and combining the cloud top height evaluation as the judgment condition.

[0076] In the embodiment, after counting the samples and sample quantities of clouds and no clouds in the cloud detection results of the imaging radiometer, count the sample size and proportion of correct detections (True mask, TM), missed detections (False negative, FN), false detections (False positive, FP), and correct rejections (True screen, TS) in the cloud detection results of the imaging radiometer. Among them, a correct detection means that the lidar detects a cloud and the radiometer also detects the cloud; a missed detection means that the lidar detects a cloud but the radiometer does not detect the cloud; a false detection means that the lidar does not detect a cloud but the radiometer detects the cloud, and a correct rejection means that the lidar does not detect a cloud and the radiometer does not detect the cloud either.

[0077] In the embodiment, on the basis of comparing the cloud detection results, a comparison of cloud classification results is also carried out for the correctly detected samples. Specifically, compare the cloud classification results of the imaging radiometer with those of the lidar for the qualified matching, and count the sample sizes of correct classification (True classification, TC) and misclassification (False classification, FC) for each cloud type by taking into account single-layer clouds and multi-layer clouds and combining the cloud top height evaluation as the judgment condition. Among them, a correct classification means that the cloud type identified by the lidar is the same as the cloud type identified by the radiometer, and a misclassification means that the radiometer fails to identify the cloud type confirmed by the lidar. Specifically, it includes:

[0078] Comparing the number of layers in the misclassified samples, the number of layers that can be recognized by the imaging radiometer is divided into single-layer and multi-layer. The lidar can recognize any number of layers until the signal energy is completely attenuated. If both the imaging radiometer and the lidar are single-layer clouds or both are multi-layer clouds (the number of lidar layers ≥ 2), and the cloud top height difference < 3 km, then even if the classification of the cloud phase property is incorrect, it is still judged as correctly recognized in terms of layers. If the cloud top height difference > 3 km, it is considered that the imaging radiometer has missed or misdetected the high-altitude cloud layer and is judged as misclassified. If the number of layers is inconsistent, it is judged as misclassified. The sample sizes of correct classification and misclassification for each cloud type are counted.

[0079] More specifically, as Figure 3 shown, from the two aspects of single-layer clouds and multi-layer clouds and combined with the cloud top height assessment as the judgment condition, the sample sizes of correct classification and misclassification for each cloud type are counted, including:

[0080] If the single-layer clouds recognized by the lidar and the imaging radiometer are of the same type, it is recorded as correctly classified single-layer clouds, where the single-layer cloud types include water clouds, ice clouds, and cirrus clouds;

[0081] If the lidar and the imaging radiometer both recognize single-layer clouds but the types are inconsistent, and the cloud top height deviation < 3 km, it is recorded as misclassified single-layer clouds, corresponding to the situation where the phase determination is incorrect but the layer recognition is consistent in actual measurement;

[0082] If the lidar and the imaging radiometer both recognize multi-layer clouds and the cloud top height deviation < 3 km, it is recorded as correctly classified multi-layer clouds;

[0083] If the lidar and the imaging radiometer have inconsistent layer recognition, or both recognize multi-layer clouds but the cloud top height deviation ≥ 3 km, it is recorded as misclassified multi-layer clouds, corresponding to the situation where the imaging radiometer fails to recognize cirrus clouds or misidentifies high-altitude aerosols as cirrus clouds in actual measurement;

[0084] The sample sizes of correct classification and misclassification for each cloud type are counted.

[0085] S6. Calculate the cloud top height difference: Compare the matched imaging radiometer cloud product data with the lidar cloud profile data, and calculate the cloud top height differences of each correctly classified and misclassified single-layer cloud type, the cloud top height differences and average layer height differences of the correctly classified multi-layer clouds, and the cloud top height differences, layer top height differences, and average layer height differences of the misclassified multi-layer clouds.

[0086] In the embodiments, for the correctly classified samples and misclassified samples of each cloud type, the cloud top height data included in the cloud product samples and the cloud top height data included in the corresponding radar cloud profile are used to calculate the cloud height difference (CTD) between the two cloud tops. The CTD refers to the difference between the cloud top height measured by the imaging radiometer and the cloud top height of the top layer cloud in the lidar cloud profile. The calculation formula is as follows:

[0087] (3)

[0088] where x represents each cloud type and each type of correctly / incorrectly identified sample, is the number of samples for each cloud type, represents the cloud top height measured by the radiometer for the x-th cloud type, represents the cloud top height measured by the lidar for the x-th cloud type.

[0089] For multi-layer clouds, due to the mixed effect of radiation signals, the limitation of instrument resolution, and the limitation of the inversion algorithm in the radiometric imager, the inverted cloud top height may be inaccurate and closer to a certain height between the two layers of clouds. At the same time, the prior value of the radiometric imager will cause the same phenomenon for the pixels misjudged as multi-layer clouds. In addition, the imaging radiometer has certain limitations in identifying cirrus clouds. Generally, it cannot identify optically thin cirrus clouds. Therefore, since multi-layer clouds contain cirrus clouds, the imaging radiometer will largely identify multi-layer clouds as water clouds or thick ice clouds.

[0090] The lidar cloud profile data has complete cloud layer information, that is, the cloud top height and cloud bottom height of each layer of cloud. Based on this characteristic of the lidar cloud profile data, for the correctly classified multi-layer clouds, the cloud top height difference and the average layer height difference are calculated. For the misclassified multi-layer clouds, the cloud top height difference, the layer top height difference, and the average layer height difference are calculated. The difference is that the cloud top height difference compares with the cloud top height of the top layer cloud in the lidar cloud profile, the layer top height difference compares with the height of the layer closest to the imaging radiometer in the lidar cloud profile, and the average layer height difference compares with the average value of the heights of multiple layers in the lidar cloud profile.

[0091] The layer top height difference refers to the difference between the cloud top height of the imaging radiometer and the layer top height of the layer closest to the one measured by the lidar cloud profile, that is, find the layer closest to the radiometric imager in the cloud profile data and use its layer top height as and substitute it into formula (3) to calculate the layer top height difference of multi-layer clouds .

[0092] The difference in average layer height refers to the difference between the cloud top height of the imaging radiometer and the average height of multiple layers in the lidar cloud profile. That is, the average cloud height is calculated based on the cloud top heights of multiple layers in the cloud profile data, and the average cloud height is used as , and substitute it into formula (3) to calculate the difference in average layer height of multi-layer clouds .

[0093] S7, output the total sample size, the sample sizes and proportions of correct detection, missed detection, false detection, and correct rejection, the correct and incorrect classification sample sizes of each cloud type, and the corresponding cloud top height difference, layer top height difference, and average layer height difference as the complete cloud top height quality inspection results.

[0094] Through steps S4 - S7, the total sample size for quality inspection of the imaging radiometer cloud product data is obtained, including the sample sizes and proportions of correct detection, missed detection, false detection, and correct rejection, as well as the correct and incorrect classification sample sizes of each cloud type and the corresponding cloud top height difference, layer top height difference, and average layer height difference, to obtain the complete cloud top height quality inspection results.

[0095] The embodiment also provides an experimental example based on the above cloud top height quality inspection method, as Figure 4 and Figure 5 shown. Specifically, the cloud profile of the DQ-1 / ACDL lidar, the cloud product of the FY-4 / AGRI multi-channel scanning imaging radiometer of Fengyun-4, and auxiliary data from January 1, 2023 to January 16, 2023 are obtained. The spatial resolution of the DQ-1 / ACDL cloud profile used is 5 km along the orbit, the spatial resolution of the FY-4 / AGRI is 4 km, and the time resolution is 15 minutes. The set time threshold is 7.5 minutes, the spatial threshold is 5 km, and the expanded search spatio-temporal range is set to a time difference of 15 minutes and a spatial difference of 10 km. After processing the above data through steps S2 and S3, on the basis of spatio-temporal matching, the DQ-1 / ACDL cloud profile and the actual observation target of the FY-4 / AGRI cloud product are inconsistent due to the uneven distribution of clouds and the rapid changes at the edges of cloud systems, which are excluded through spatio-temporal homogeneity tests. Among the 564,789 pairs of matching data obtained, the pass rate of the lidar is 93.1%, the pass rate of the radiometric imager is 79.4%, and the total pass rate is 75.9%.

[0096] For the qualified matching data, through steps S4 and S5, the obtained qualified sample size is 428,878. Among the qualified samples, the sample size of correct detection (TM + TS) is 348,920, accounting for 89.0%. The total number of samples (FP) where the radiometer detected clouds but the lidar did not is 14,604, accounting for 3.4%. The total number of samples (FN) where the radiometer did not detect clouds but the lidar did is 32,731, accounting for 7.6%. Among the FN samples, the total number of samples where the lidar detected cirrus clouds is 29,177, with a relative proportion reaching 89.1%, which is the main source of cloud detection errors in the radiometer, as shown in Table 1:

[0097] Table 1 Sample Sizes of Detection Results

[0098]

[0099] Meanwhile, the values in parentheses represent the detection results of all samples without spatio-temporal homogeneity testing, and the correct detection proportion is 83.5%. It can be seen that the spatio-temporal homogeneity testing improves the correct detection proportion when removing samples with lower reliability.

[0100] Among the above-mentioned correctly detected samples, cloud classification was also carried out through step S5, and the specific classification results are shown in Table 2:

[0101] Table 2 Cloud Classification Results

[0102]

[0103] It can be seen that except for the case of misidentifying single-layer ice clouds as water clouds, the main misclassification of clouds by the radiometer occurs in the multi-layer cloud scenario, mainly due to the radiometer missing the upper-layer cirrus clouds. The values in parentheses are the comparison results of multi-layer cloud classification without introducing the layer height of the lidar, and some cases of missing the upper-layer cirrus clouds in the multi-layer cloud scenario (i.e., the actual number of cloud layers ≥ 3) are lost.

[0104] Based on the cloud classification, the height difference between single-layer and multi-layer clouds was also calculated through step S6. Before the calculation, the comparison results of the cloud profile of the spaceborne lidar and the cloud top height of the imaging radiometer are as Figure 6 shown. Compared with the detection results only under the condition of consistent cloud detection and cloud type, the sample size of effective cloud top height detection increased by 127%. The specific calculation results are shown in Tables 3 and 4:

[0105] Table 3 Calculation Results of Cloud Top Height Difference

[0106]

[0107] It can be seen that, on the premise that the radiometer correctly identifies cirrus clouds, the error in the cirrus cloud top height is not large, and the main error comes from the misclassification cases. However, the cirrus cloud height retrieved by the radiometer slightly exceeds the average cloud top height of lidar (basically lower than the cloud top height measured by lidar for other cloud types), indicating that there may be overcorrection for cirrus clouds.

[0108] Table 4 Calculation results of the differences in the cloud top heights and the average differences in layers of multi-layer clouds

[0109]

[0110] The calculation results of the height differences of multi-layer clouds show that there are great differences when directly comparing the cloud top heights of misclassified multi-layer clouds, which is usually due to lidar detecting cirrus clouds with extremely high heights (above 10 km). Below the influence of this layer of cirrus clouds, the radiometer overestimates the cloud height of misclassified clouds compared with the corresponding layers of lidar (usually the second layer). If the average heights of similar layers are compared, the cirrus clouds are underestimated, while other cloud types are overestimated, which is in line with the expected results based on cloud optical properties.

[0111] Based on the same inventive concept, as Figure 7 shown, the embodiment also provides a cloud top height quality inspection device, including: a data acquisition module 71, a spatio-temporal matching module 72, a cloud detection and classification matching module 73, a uniformity test module 74, a layer-by-layer comparison module 75 for cloud detection and classification results, a difference calculation module 76, and a data output module 77.

[0112] Among them, the data acquisition module 71 is used to acquire lidar cloud profile data as the detection truth value, imaging radiometer cloud product data to be quality inspected, and auxiliary data; the spatio-temporal matching module 72 is used to perform spatio-temporal matching of the imaging radiometer cloud product data relative to the lidar cloud profile data based on the spatio-temporal data in the auxiliary data and determine the matching data; the cloud detection and classification matching module 73 is used to match the cloud detection and cloud classification results in the lidar cloud profile data and the cloud detection and cloud classification results in the imaging radiometer cloud product according to the spatio-temporal information matching situation in the matching data; the homogeneity test module 74 is used to perform spatio-temporal homogeneity tests on the matching data to screen out the data with unqualified spatio-temporal homogeneity, obtain qualified matching data and count the total sample size; the cloud detection and classification result layer-by-layer comparison module 75 is used to compare the imaging radiometer cloud detection results of the qualified matching with the lidar cloud detection results, count the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples, and in the samples of correct detections, compare the imaging radiometer cloud classification results of the qualified matching with the lidar cloud classification results, and count the sample sizes of correct and incorrect classifications of each cloud type from two aspects of single-layer clouds and multi-layer clouds and in combination with the cloud top height evaluation as the determination condition; the difference calculation module 76 is used to compare the matching imaging radiometer cloud product data with the lidar cloud profile data, calculate the cloud top height differences of each single-layer cloud type of correct and incorrect classifications, the cloud top height differences and layer average height differences of correctly classified multi-layer clouds, and the cloud top height differences, layer top height differences, and layer average height differences of incorrectly classified multi-layer clouds; the data output module 77 is used to output the total sample size, the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples, the sample sizes of correct and incorrect classifications of each cloud type, and the corresponding cloud top height differences, layer top height differences, and layer average height differences as the complete cloud top height quality inspection results.

[0113] It should be noted that when the cloud top height quality inspection device provided in the above embodiment performs cloud top height detection, the above function modules should be used as examples for illustration. The above functions can be assigned to different function modules according to needs, that is, the internal structure of the terminal or server is divided into different function modules to complete all or part of the functions described above. In addition, the cloud top height quality inspection device provided in the above embodiment and the cloud top height quality inspection method embodiment belong to the same concept. For the specific implementation process, please refer to the cloud top height quality inspection method embodiment, which will not be elaborated here.

[0114] The embodiment also provides a computing device, including a memory and one or more processors. The memory stores executable code. When the one or more processors execute the executable code, they are used to implement the above cloud top height quality inspection method, specifically including the following steps:

[0115] S1, Data acquisition: Acquire lidar cloud profile data and its cloud determination results as the detection truth values, imaging radiometer cloud product data to be quality inspected, and auxiliary data;

[0116] S2, Spatiotemporal matching: Perform spatiotemporal matching of the imaging radiometer cloud product data relative to the lidar cloud profile data based on the spatiotemporal data in the auxiliary data and determine the matching data;

[0117] S3, Cloud detection and classification matching: Match the cloud detection and cloud classification results in the lidar cloud profile data with the cloud detection and cloud classification results in the imaging radiometer cloud products according to the spatiotemporal information matching situation in the matching data;

[0118] S4, Homogeneity test: Perform a spatiotemporal homogeneity test on the matching data to screen out the data with unqualified spatiotemporal homogeneity, obtain the qualified matching data and count the total sample size;

[0119] S5, Layer-by-layer comparison of cloud detection and classification results: Compare the cloud detection results of the qualified matching imaging radiometer with the lidar cloud detection results, and count the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples; Among the samples with correct detections, compare the cloud classification results of the qualified matching imaging radiometer with the lidar cloud classification results, and count the sample sizes of correct and incorrect classifications of each cloud type from two aspects of single-layer clouds and multi-layer clouds and combined with the cloud top height evaluation as the judgment condition;

[0120] S6, Calculate the cloud top height difference: Compare the matching imaging radiometer cloud product data with the lidar cloud profile data, and calculate the cloud top height differences of each single-layer cloud type with correct and incorrect classifications, the cloud top height differences and layer average height differences of correctly classified multi-layer clouds, and the cloud top height differences, layer top height differences, and layer average height differences of incorrectly classified multi-layer clouds;

[0121] S7, Data output: Output the total sample size, the sample sizes and proportions of correct detections, missed detections, false detections, and correctly screened samples, the sample sizes of correct and incorrect classifications of each cloud type, and the corresponding cloud top height differences, layer top height differences, and layer average height differences as the complete cloud top height quality inspection results.

[0122] The computing device provided by the embodiment, at the hardware level, in addition to including a processor and a memory, further includes an internal bus, a network interface, a memory, and other hardware required for other services. The memory is a non-volatile memory, and the processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the cloud top height quality inspection method described in the above steps S1 - S7. Of course, in addition to the software implementation method, the present invention does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, and can also be hardware or logic devices.

[0123] Based on the same inventive concept, the embodiment also provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, the above cloud top height quality inspection method is implemented, specifically including the following steps:

[0124] S1, Obtain data: Obtain lidar cloud profile data and its cloud determination result as the detection truth value, imaging radiometer cloud product data to be quality inspected, and auxiliary data;

[0125] S2, Spatiotemporal matching: Based on the spatiotemporal data in the auxiliary data, perform spatiotemporal matching of the imaging radiometer cloud product data relative to the lidar cloud profile data and determine the matching data;

[0126] S3, Cloud detection and classification matching: According to the spatiotemporal information matching situation in the matching data, match the cloud detection and cloud classification results in the lidar cloud profile data with the cloud detection and cloud classification results in the imaging radiometer cloud product;

[0127] S4, Homogeneity test: Perform a spatiotemporal homogeneity test on the matching data to screen out the data with unqualified spatiotemporal homogeneity, obtain the qualified matching data and count the total sample size;

[0128] S5, Layer-by-layer comparison of cloud detection and classification results: Compare the cloud detection results of the qualified matching imaging radiometer with the lidar cloud detection results, and count the sample size and proportion of correct detections, missed detections, false detections, and correctly screened samples; Among the correctly detected samples, compare the cloud classification results of the qualified matching imaging radiometer with the lidar cloud classification results, and count the sample size of correct classification and misclassification of each cloud type from two aspects of single-layer clouds and multi-layer clouds and in combination with cloud top height evaluation as the determination condition;

[0129] S6, Calculate the cloud top height difference: Compare the matching imaging radiometer cloud product data with the lidar cloud profile data, and calculate the cloud top height difference of each single-layer cloud type of correct classification and misclassification, the cloud top height difference and layer average height difference of correctly classified multi-layer clouds, and the cloud top height difference, layer top height difference, and layer average height difference of misclassified multi-layer clouds;

[0130] S7, Data Output: Output the total sample size, the number and proportion of correctly detected, missed detected, misdetected, and correctly screened samples, the number of correctly and incorrectly classified samples for each cloud type, and the corresponding differences in cloud top height, layer top height, and average layer height as the complete cloud top height quality inspection results.

[0131] In the embodiments, the computer-readable medium includes permanent and non-permanent, removable and non-removable media and can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data.

[0132] The above-described specific embodiments have detailed the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, supplements, equivalent replacements, etc. made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cloud top height quality inspection method, characterized in that: The following steps are involved: Obtain lidar cloud profile data as the detection truth value, imaging radiometer cloud product data to be inspected, and auxiliary data; Based on the spatiotemporal data in the auxiliary data, the imaging radiometer cloud product data is temporally and spatially matched with the lidar cloud profile data and the matching data is determined; Match the cloud detection and cloud classification results in the lidar cloud profile data with the cloud detection and cloud classification results in the imaging radiometer cloud products according to the matching of the spatiotemporal information in the matching data; Perform a spatiotemporal homogeneity test on the matching data to filter out data that do not meet the spatiotemporal homogeneity requirements, obtain qualified matching data, and calculate the total sample size; Compare the qualified imaging radiometer cloud detection results with the lidar cloud detection results, and count the number and proportion of samples that are correctly detected, missed, falsely detected, and correctly screened out; among the correctly detected samples, compare the qualified imaging radiometer cloud classification results with the lidar cloud classification results, and count the number of correctly classified and incorrectly classified samples of each cloud type from the perspectives of single-layer clouds and multi-layer clouds, combined with cloud top height assessment as the judgment condition; Compare the imaging radiometer cloud product data and the lidar cloud profile data, and calculate the cloud top height difference of each single-layer cloud type that is correctly classified and incorrectly classified, the cloud top height difference and layer average height difference of correctly classified multi-layer clouds, and the cloud top height difference, layer top height difference and layer average height difference of incorrectly classified multi-layer clouds; Output the total sample size, the sample size and proportion of correct detection, missed detection, false detection, and correct screening, the sample size of correct and incorrect classification for each cloud type, and the corresponding cloud top height difference, layer top height difference, and layer average height difference as the complete cloud top height quality inspection result.

2. The cloud top height quality inspection method according to claim 1, characterized in that: Based on the spatiotemporal auxiliary data, the imaging radiometer cloud product data is temporally and spatially matched with the lidar cloud profile data and the matching data is determined, including: Determine the time threshold and space threshold. The time and space auxiliary data include the measurement time and longitude and latitude. The imaging radiometer cloud product data include the imaging radiometer cloud product samples measured at different times and longitudes and latitudes. Based on the measurement time and longitude and latitude corresponding to each profile in the lidar cloud profile data, the imaging radiometer cloud product samples with the closest spatial distance to each profile within each time threshold and spatial threshold are screened out to form matching data.

3. The cloud top height quality inspection method according to claim 2, characterized in that: Perform a temporal and spatial homogeneity test on the matching data to filter out data that does not meet the temporal and spatial homogeneity requirements and obtain qualified matching data, including: The matching data are tested for the spatiotemporal uniformity of the lidar profile, the internal uniformity of the radiometer pixels, and the peripheral uniformity of the radiometer pixels. The qualified matching data are obtained by combining the sample data that pass all three tests.

4. The cloud top height quality inspection method according to claim 3 is characterized in that: The lidar profile spatiotemporal uniformity test refers to testing the uniformity of multiple basic profiles contained in each lidar cloud profile, calculating the average deviation of the cloud top height and the difference in the number of layers of all basic profiles. When the number of layers changes or the average deviation of the cloud top height is greater than or equal to a given cloud top height deviation threshold, the spatiotemporal uniformity of all lidar cloud profile samples is considered unqualified and screened out. When the number of layers is consistent and the average deviation of the cloud top height is less than the given cloud top height deviation threshold, all lidar cloud profile samples are considered qualified. The radiometer pixel internal uniformity test refers to testing the uniformity between multiple pixels contained in each radiometer cloud product pixel, calculating the standard deviation of infrared band brightness temperature between all pixels, and when the brightness temperature standard deviation is greater than or equal to a given brightness temperature standard deviation threshold, the temporal and spatial uniformity of all radiometer cloud product samples are considered unqualified and are screened out; when the brightness temperature standard deviation is less than the given brightness temperature standard deviation threshold, all radiometer cloud product samples are considered qualified; The radiometer pixel peripheral uniformity test refers to expanding the search time and space range based on the time threshold and the space threshold, selecting all cloud product samples whose latitude and longitude differences with each profile are within the search time and space range from the matching data, and calculating the average deviation between the cloud characteristics of all cloud product samples and the average cloud characteristics, wherein the average cloud characteristic is the average value of the cloud characteristics of all cloud product samples. When the average deviation is greater than or equal to the given cloud characteristic deviation threshold, the time and space uniformity of all cloud product samples is considered unqualified and are screened out. When the average deviation is less than the given cloud characteristic deviation threshold, all cloud product samples are considered qualified, wherein the cloud characteristic is a matrix composed of marking cloud as 1 and marking no cloud as 0.

5. The cloud top height quality inspection method according to claim 4, characterized in that: The cloud top height deviation threshold between lidar profiles is 0-1 km, the brightness temperature standard deviation threshold of the infrared band between radiometer pixels is 0-3 K, and the cloud feature deviation threshold between radiometer pixels is 0-0.

25.

6. The cloud top height quality inspection method according to claim 1, characterized in that: The number of samples correctly classified and misclassified for each cloud type is counted from both single-layer and multi-layer clouds, combined with cloud top height assessment as the judgment criterion, including: If the single-layer clouds identified by the lidar and the imaging radiometer are of the same type, they are recorded as correctly classified single-layer clouds, where single-layer cloud types include water clouds, ice clouds, and cirrus clouds; If both the lidar and the imaging radiometer identified single-layer clouds but the types were inconsistent, and the cloud top height deviation was <3 km, it was recorded as misclassified single-layer clouds, corresponding to the situation in which the phase state judgment was wrong in the actual measurement but the layer identification was consistent; If both the lidar and the imaging radiometer identified multi-layer clouds and the cloud top height deviation was <3 km, it was recorded as correctly classified multi-layer clouds; If the lidar and imaging radiometer do not agree on the layer recognition, or both recognize multi-layer clouds but the cloud top height deviation is ≥3km, it is recorded as misclassified multi-layer clouds, which corresponds to the situation in which the imaging radiometer fails to recognize cirrus clouds or misidentifies high-altitude aerosols as cirrus clouds in actual measurements; Count the number of correctly classified and incorrectly classified samples for each cloud type.

7. The cloud top height quality inspection method according to claim 1, characterized in that: The cloud top height difference refers to the difference between the cloud top height of the imaging radiometer and the cloud top height of the top cloud of the lidar cloud profile. The layer top height difference refers to the difference between the cloud top height of the imaging radiometer and the layer top height of the closest layer measured by the lidar cloud profile; The layer average height difference refers to the difference between the cloud top height of the imaging radiometer and the average height of multiple layers in the lidar cloud profile.

8. A cloud top height quality inspection device, characterized in that: include: A data acquisition module, which is used to acquire the lidar cloud profile data as the detection true value, the imaging radiometer cloud product data to be inspected, and the auxiliary data; A time-space matching module, which is used to perform time-space matching of the imaging radiometer cloud product data relative to the lidar cloud profile data based on the time-space data in the auxiliary data and determine the matching data; The cloud detection and classification matching module is used to match the cloud detection and cloud classification results in the lidar cloud profile data with the cloud detection and cloud classification results in the imaging radiometer cloud product according to the matching of the spatiotemporal information in the matching data; A uniformity test module is used to perform a temporal and spatial uniformity test on the matching data to filter out data that does not meet the temporal and spatial uniformity requirements, obtain qualified matching data, and calculate the total sample size; The cloud detection and classification result layer-by-layer comparison module is used to compare the qualified imaging radiometer cloud detection results with the lidar cloud detection results, and to count the number and proportion of samples that are correctly detected, missed, falsely detected, and correctly screened out. Among the correctly detected samples, the qualified imaging radiometer cloud classification results and the lidar cloud classification results are compared, and the number of correctly classified and incorrectly classified samples of each cloud type is counted from the perspectives of single-layer clouds and multi-layer clouds, combined with cloud top height assessment as the judgment condition; A difference calculation module is used to compare the matching imaging radiometer cloud product data with the lidar cloud profile data, calculate the cloud top height difference of each single-layer cloud type that is correctly classified and incorrectly classified, the cloud top height difference and layer average height difference of correctly classified multi-layer clouds, and the cloud top height difference, layer top height difference and layer average height difference of incorrectly classified multi-layer clouds; The data output module is used to output the total sample size, the sample size and proportion of correct detection, missed detection, false detection, and correct screening, the sample size of each cloud type that is correctly and incorrectly classified, and the corresponding cloud top height difference, layer top height difference, and layer average height difference as the complete cloud top height quality inspection results.

9. A computing device comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that: When the one or more processors execute the executable code, they are used to implement the cloud top height quality inspection method described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the cloud top height quality inspection method described in any one of claims 1-7 is implemented.

Images