CN118053077A - Method for detecting cloud-aerosol by using FY-4A full-disc data at all days - Google Patents
Method for detecting cloud-aerosol by using FY-4A full-disc data at all days Download PDFInfo
- Publication number
- CN118053077A CN118053077A CN202311703249.4A CN202311703249A CN118053077A CN 118053077 A CN118053077 A CN 118053077A CN 202311703249 A CN202311703249 A CN 202311703249A CN 118053077 A CN118053077 A CN 118053077A
- Authority
- CN
- China
- Prior art keywords
- aerosol
- cloud
- data
- model
- type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000000443 aerosol Substances 0.000 title claims abstract description 129
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 6
- 238000013145 classification model Methods 0.000 claims description 32
- 238000012549 training Methods 0.000 claims description 15
- 239000000428 dust Substances 0.000 claims description 12
- 239000000779 smoke Substances 0.000 claims description 8
- 238000004140 cleaning Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 239000000853 adhesive Substances 0.000 claims description 4
- 230000001070 adhesive effect Effects 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract description 5
- 230000002159 abnormal effect Effects 0.000 abstract description 2
- 239000000047 product Substances 0.000 description 15
- 238000007637 random forest analysis Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000001816 cooling Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 238000001429 visible spectrum Methods 0.000 description 2
- 241001085205 Prenanthella exigua Species 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 239000005427 atmospheric aerosol Substances 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a method for detecting cloud-aerosol by using FY-4A full-disc data at all days, which comprises the following steps: marking CALIPSO track pixels according to atmospheric characteristics and aerosol type data of CALIPSO VFM products; projecting the FY-4A TOAR data set into a plane coordinate system, adjusting the spatial resolution of ERA-5 data to be 4km which is the same as that of FY-4A through a nearest neighbor interpolation method, and combining the FY-4A TOAR and ERA-5 data to generate a grid sample data set; in the space-time matching of 4km x 4km grid sample data and CALIPSO on-orbit data, selecting the time difference to be within 30 minutes, and matching each CALIPSO track pixel with a grid pixel with the nearest longitude and latitude to generate an on-orbit sample data set; constructing a sub-model based on the nonlinear hybrid model; and inputting the grid sample data set of the whole disc area into each submodel for detection, and obtaining the whole disc prediction data. The method can integrate the advantages of a plurality of machine learning models, can effectively reduce the risk of overfitting, and has certain robustness to abnormal values.
Description
Technical Field
The invention belongs to the technical field of atmospheric pollution monitoring, and particularly relates to a method for detecting cloud-aerosol by using FY-4A full-disc data in all days.
Background
Atmospheric aerosols are composed mainly of solid or liquid particles, which not only reflect and absorb solar radiation, affecting the radiation balance in the atmosphere, but also directly involve the health of humans and the long-term stability of the terrestrial climate system. The cloud may reflect solar radiation, reducing the sun and temperature of the earth's surface, thereby cooling the earth's surface. This cooling effect helps to maintain the energy balance of the earth. Cloud and aerosol are important components of the earth's atmosphere that have important effects on the earth's climate system, air quality and hydrologic cycle. Therefore, intensive research into cloud-aerosols is necessary. However, although there have been many studies on cloud-aerosol identification, accurate identification of cloud-aerosol at night is still a challenging task in the current field of atmospheric science.
At present, ground-based remote sensing and satellite remote sensing data are commonly used in aerosol related research. Compared with the daytime, the night aerosol recognition is more difficult, and the night aerosol is difficult to directly observe due to the lack of a visible light wave band, so that the conventional aerosol recognition method is greatly influenced. In addition, the characteristics of the same aerosol in daytime and nighttime conditions may be greatly different, and the environmental conditions, meteorological factors and other factors may be affected by day-night alternation. These changes in aerosols can have a major impact on urban planning, environmental management and disaster warning. Therefore, the importance of night recognition work of aerosols is not neglected.
On the other hand, similar to aerosol recognition, many challenges still remain in the current night time cloud recognition work. The illumination of the sun provides a natural light source that makes the cloud layer relatively easy to observe and identify in the visible spectrum. However, at night, the lack of solar light sources can only rely on artificial light sources (such as night illumination of cities) or active light sources (such as satellite-borne lidar or infrared sensors) to observe clouds, which makes night cloud identification more complex. Like aerosols, night clouds and daytime clouds may have different physical properties. Daytime clouds are typically illuminated by sunlight, appear bright white or gray, and reflect light in the visible spectrum. Night clouds then typically appear darker in appearance, primarily by emitting or reflecting light from the illuminated source, which makes the night clouds more difficult to identify and distinguish. Today, students recognize night clouds from bright temperature data in the infrared band.
The wind cloud No. A star (FY-4A) is the first star of the second generation of stationary meteorological satellites in China, and is successfully launched in 2016. The satellite can cover the whole world of China and is provided with a multichannel scanning imaging radiometer (AGRI). The FY-4A satellite can be imaged 205 times per day on a full disk, including 40 images of the east Asia region. This frequent imaging frequency provides adequate data support. The FY-4A satellite is capable of providing data products with higher time resolution than the polar orbit satellite. In addition, in recent years, as various machine learning algorithms are gradually applied to satellite remote sensing aerosol related researches, such as Random Forest (RF), support vector machine (SVR), convolutional Neural Network (CNN), and the like, related works such as cloud identification and aerosol classification have been successfully realized. This shows that the machine learning model has unique advantages in solving the challenging problem in the cloud-aerosol remote sensing field. In addition, the full disk TOAR (Top of Atmosphere Reflectance) dataset of the FY-4A satellite has wider coverage and high time resolution, and provides more reliable and effective data support for cloud-aerosol identification work.
However, most aerosol classification researches based on remote sensing satellite data are subjected to cloud screening treatment, and an aerosol detection result cannot be obtained in a cloud coverage area, so that obvious adverse effects are caused on the generation of fully covered aerosol classification products.
Disclosure of Invention
In order to overcome the defects and shortcomings of the prior art, the invention aims to provide a method for classifying cloud-aerosol by using an FY-4A TOAR dataset, which uses CALIPSO VFM product to mark cloud and aerosol types, mainly relies on TOAR infrared band data of FY-4A as model input and is assisted by surface information data to generate a full disc night cloud-aerosol classified product with high space-time resolution.
The invention is realized in that a method for detecting cloud-aerosol by using FY-4A full-disc data at all days comprises the following steps:
S1, marking CALIPSO track pixels according to atmospheric characteristics and aerosol type data of CALIPSO VFM products, wherein marking comprises marking atmospheric characteristic classification labels, marking aerosol type labels and marking cloud type labels;
s2, projecting the FY-4A TOAR data set into a plane coordinate system, adjusting the spatial resolution of ERA-5 data to be 4km which is the same as that of FY-4A through a nearest neighbor interpolation method, and combining the FY-4A TOAR data and the ERA-5 data to generate a grid sample data set;
S3, in space-time matching of 4km x 4km grid sample data and CALIPSO on-orbit data, selecting a time difference within 30 minutes, and matching each CALIPSO track pixel with a grid pixel with the nearest longitude and latitude to generate an on-orbit sample dataset;
s4, constructing a sub-model based on the nonlinear hybrid model, wherein the sub-model comprises an atmospheric feature classification model, an aerosol classification model and a cloud classification model;
S5, inputting the grid sample data set of the whole disc region into each submodel for detection, and obtaining complete whole disc prediction data, wherein the complete whole disc prediction data comprises an atmospheric feature classification result, an aerosol classification result and a cloud classification result.
Preferably, in step S1, in the atmospheric feature classification label, the atmospheric feature type data ("enclosed" and "Tropospheric aerosol") in Feature Classification Flags are selected, each track pixel is marked, and all samples are marked as four types of pixels, namely, a clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer.
Preferably, in step S1, in the Aerosol type tag, aerosol type data (durt, polluted continental, etc.) in Aerosol sub-type is used to mark Aerosol types in an "Aerosol-cloud/Aerosol mixed layer of cloudless area", the Aerosol types of cloudless area and cloudy area are classified into "clean marine Aerosol-polluted marine Aerosol-Dust Aerosol-clean continental Aerosol-artificial related Aerosol-lifting smoke", the artificial related Aerosol including polluted continental Aerosol and polluted Dust Aerosol ".
Preferably, in step S1, in the Cloud type tag, cloud type data (low overcast (transparent), low overcast (opaque), etc.) in Cloud sub-type is used to mark the Cloud type in "Cloud region-Cloud/aerosol mix layer" and the Cloud type is divided into "transparent low Cloud-opaque low Cloud-high Cloud-curly Cloud-deep convective Cloud".
Preferably, in step S4, the process of constructing the nonlinear hybrid model includes the following steps:
Training four different types of basic classification models (ET, RF, XGB, LGT) on a training set, and for each basic classification model, generating a prediction result by using data in the training set;
and using the prediction result as the meta-feature and the real label of the training set for meta-model training.
Preferably, the meta-model is an extreme tree for combining the predictions of the base model to obtain the final classification result.
Preferably, in step S5, the detecting includes the steps of:
acquiring a clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer through an atmospheric characteristic classification model;
Detecting an area of an aerosol layer-cloud/aerosol mixed layer (acquired by an atmospheric feature classification model) based on an aerosol classification model, and classifying the aerosol of the area into the following categories: "cleaning marine aerosol-pollution marine aerosol-dust aerosol-cleaning continental aerosol-artificial related aerosol-lifting smoke";
Detecting a region of a 'cloud layer-cloud/aerosol mixed layer' (acquired by an atmospheric feature classification model) based on a cloud classification model, and classifying the cloud layer of the region into: "transparent low cloud-low cloud opaque-high cloud-coiled cloud-deep convective cloud".
Compared with the defects and shortcomings of the prior art, the invention has the following beneficial effects:
(1) Traditional experience threshold algorithm often consumes a large amount of calculation power in the process of acquiring the threshold value, is easily influenced by factors such as geographic environment and the like, and has limited research area. Compared with a single model, the method has higher performance, can effectively reduce the risk of over-fitting, and has certain robustness to abnormal values.
(2) Compared with the daytime, the cloud-aerosol identification work at night has larger element differences of meteorological elements, environmental changes and the like, and is more challenging. According to the invention, FY-4A TOAR infrared band data is mainly used, only surface type data is used as supplement, and meteorological element data is less used, so that the influence of meteorological field change on model performance is reduced to a certain extent, the coverage range of TOAR dataset is wide, the time resolution is high, the time resolution of a product generated based on the data can reach 15min, and the spatial resolution is 4km.
(3) Currently, identifying cloud-aerosols within a cloud/aerosol mixing region remains a challenging task in the field of atmospheric science. The invention not only realizes the accurate classification of the pure cloud layer and the pure aerosol layer, but also successfully identifies various cloud types and aerosol types of the cloud/aerosol mixed layer.
Drawings
FIG. 1 is a schematic construction diagram of a nonlinear hybrid model;
FIG. 2 is a flow chart of model identification;
FIG. 3 is an atmospheric signature classification result (UTC: 2019.3.1:15:00);
FIG. 4 is an aerosol classification result (UTC: 2019.3.1:15:00);
FIG. 5 is a cloud classification result (UTC: 2019.3.1:15:00);
In fig. 3-5, the bold line is CALIPSO satellite travel track lines,
FIG. 6 is CALIPSO VFM product results (UTC: 2019.3.1 14:53); wherein (a) atmospheric signature type , 0 = invalid (bad or missing data), 1 = clear air, 2 = cloud, 3 = tropospheric aerosol, 4 = stratospheric aerosol, 5 = surface, 6 = subsurface, 7 = no signal;（b） tropospheric aerosol subtype ,0 = not determined, 1 = clean marine, 2 = dust, 3 = polluted continental/smoke, 4 = clean continental, 5 = polluted dust, 6 = elevated smoke, 7 = dusty marine;（c） Yun Ya ,0 = not determined, 1 = low overcast（transparent）, 2 = low overcast（opaque）, 3 = transition stratocumulus, 4 = broken cumulus, 5 = altocumulus (transparent) , 6 = cirrus (transparent) , 7 = deep convective ,8 = no signal.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the invention discloses a method for detecting cloud-aerosol by using FY-4A full-disc data at all days, which comprises the following steps:
s1, as shown in FIG. 1, marking CALIPSO track pixels according to the atmospheric characteristics, aerosol type and cloud type data of CALIPSO VFM products, wherein marking comprises marking atmospheric characteristic classification labels, marking aerosol type labels and marking cloud type labels.
In an embodiment of the invention, CALIPSO track pixels are marked using atmospheric characteristics, aerosol type and cloud type data for CALIPSO VFM products (vertical height range of choice: -0.5km-20.2 km). It should be noted that the excessive thickness of the cloud layer may cause the signal received by the CALIPSO sensor to be attenuated, and a missing value may exist, which cannot be determined whether aerosol exists under a non-clear air condition, and the missing data is deleted.
In step S1, the process of marking includes the steps of:
s11, labeling atmospheric feature classification labels
And marking each track pixel by selecting 'closed' and 'Tropospheric aerosol' data in the atmospheric characteristic type data, and finally marking all track pixel samples as four types of pixels, namely 'clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer'.
S12, labeling aerosol type labels
Five data of the aerosol type data, such as best, polluted continental and Polluted Dust, are used for marking the aerosol type in an aerosol-cloud/aerosol mixed layer of a cloud-free area, and the aerosol type of the area of the aerosol-cloud/aerosol mixed layer of the cloud-free area is marked as clean ocean aerosol-polluted ocean aerosol-sand aerosol-clean continental aerosol-artificial related aerosol-lifting smoke.
S13, labeling cloud type labels
Seven types of data, low overcast, transition stratocumulus, deep convective, among the cloud type data, are used to mark the cloud type of the "cloud layer region-cloud/aerosol mixed layer" region, labeled as "low cloud layer (transparent) -low cloud layer (opaque) -high cloud layer-coiled cloud-deep convective cloud".
S2, projecting the FY-4A TOAR dataset into a plane coordinate system, and adjusting the spatial resolution of ERA-5 data to be 4km which is the same as that of FY-4A through a nearest neighbor interpolation method.
In step S2, FY-4A TOAR data and ERA-5 data are adjusted to uniform resolution (4 km) by nearest neighbor interpolation, and combined into a grid sample dataset.
S3, in space-time matching of 4km x 4km grid sample data and CALIPSO on-orbit data, selecting a time difference within 30 minutes, and matching each CALIPSO track pixel with a grid pixel with the nearest longitude and latitude to generate an on-orbit sample data set.
In step S3, since the track of the on-orbit data is a curve on the two-dimensional plane CALIPSO, it is necessary to match the grid sample data and the satellite orbit to the same size orbit data in order to match the sample data distributed in the grid. Thus, in the space-time matching of grid sample data (4 km x4 km) and CALIPSO on-track data (VFM product: resolution 5 km), the time difference is chosen to be within 30 minutes, each CALIPSO track pixel matches one of the nearest grid pixels in longitude and latitude, and an on-track sample data set is generated (as shown in equation 1).
（1）
Wherein Input represents an on-orbit dataset matched with CALIPSO tracks, i represents the position of the grid, j represents time, m represents channel number, and h represents vertical height (100 hpa/500hpa/850 hpa). The model inputs include: TOAR dataset, BTD (obtained by subtracting different bright temperature channels in TOAR), time (Time: 0-365), longitude (LON), latitude (LAT), LC (Land cover).
S4, constructing a sub-model based on the nonlinear hybrid model, wherein the sub-model comprises an atmospheric feature classification model, an aerosol classification model and a cloud classification model.
In step S4, in combination with the advantages of the extreme tree, the random forest, the XGB model and the LGT model algorithm, the present invention constructs a nonlinear hybrid model (fig. 1 shows a nonlinear hybrid model composition), which has better performance. The specific operation is as follows: four different types of base classification models are trained on the training set, for each of which data in the training set is used to generate their predictions. These predictions will be used as new features for the next meta-model training. The generated meta-feature and the real label of the training set are used for training the meta-model, the extreme tree is selected as the meta-model, the task of the meta-model is to combine the prediction results of the basic model to obtain the final classification result, and the comparison result of the nonlinear hybrid model and the single machine learning model is shown in the table 1.
Table 1 comparison of various machine learning models and nonlinear hybrid model accuracy
| Model | Atmospheric feature classification | Aerosol classification | Cloud classification |
| ET | 0.805 | 0.854 | 0.786 |
| RF | 0.806 | 0.857 | 0.783 |
| XGB | 0.796 | 0.853 | 0.784 |
| LGT | 0.799 | 0.851 | 0.782 |
| Nonlinear hybrid model | 0.817 | 0.869 | 0.809 |
Based on the nonlinear hybrid model, as shown in fig. 2, the study built a total of three sub-models: atmospheric feature classification model-aerosol classification model-cloud classification model. The three sub-model inputs are all shown in the formula (1), and the input variables are slightly adjusted according to the difference between the atmospheric characteristic classification model, the aerosol classification model and the cloud classification model.
S5, inputting the grid sample data set of the whole disc area into each submodel for detection.
In step S5, a grid sample data set (feature variable is consistent with model input) of the entire disc region is input into the model for detection, and a recognition result of the entire disc region is obtained. The specific process comprises the following steps:
s51, acquiring a clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer through an atmospheric characteristic classification model.
The result of the atmospheric feature classification model is shown in fig. 3, and the "clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer" is obtained. Because the aerosol content is higher at night, the pure condition area is fewer, the aerosol is concentrated in the south-north two-stage mode, and the cloud/aerosol mixed layer area is the most.
The "cloud/aerosol hybrid layer" is known to contain cloud and aerosol layers, which are not easily distinguished, and different models are used to treat this area in the embodiments of the present invention.
Firstly, detecting an area (acquired by an atmospheric feature classification model) of an aerosol adhesive layer-cloud/aerosol mixed layer based on an aerosol classification model, and classifying the aerosol adhesive of the area into the following categories: "cleaning marine aerosol-polluting marine aerosol-dust aerosol-cleaning continental aerosol-artificial related aerosol-lifting smoke". As fig. 4 shows the aerosol classification results, the most clean marine aerosols were identified, most concentrated in the marine area. The most artificial related aerosols in land areas, the most dust aerosols are concentrated in the dust source areas, and the polluted marine aerosols are concentrated in the areas near the boundary of land and sea.
Secondly, detecting a 'cloud layer-cloud/aerosol mixed layer' region (acquired by an atmospheric feature classification model) based on a cloud classification model, and classifying the cloud layer of the region into: "low cloud (transparent) -low cloud (opaque) -high cloud-roll-deep convective cloud". Fig. 5 shows the cloud classification result, and the volume cloud is concentrated in the area near the equator and the highest in the ratio. The low cloud layer cloud is widely distributed and mostly transparent low cloud.
S52, comparing the data products of aerosol subtypes and cloud subtypes in CALIPSO VFM products with the model identification result to verify and refine the classification result. Fig. 6 (a), (b) and (c) show the atmospheric characteristic type, aerosol subtype and Yun Ya type of CALIPSO VFM products respectively, the time of selection is UTC: 2019.3.1:14, 53, and the bold line in fig. 3-5 corresponds to the CALIPSO satellite motion track of the present time. It can be seen that the cloud-aerosol distribution in the VFM product is approximately consistent with the model recognition results of the present invention. The comprehensive method is combined with remote sensing products to evaluate, so that the complexity of a cloud/aerosol mixed layer can be better understood, and the identification and distinguishing capability of the invention on the atmospheric components can be improved.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (7)
1. A method for detecting cloud-aerosols using full-disc data of FY-4A at full day, the method comprising the steps of:
S1, marking CALIPSO track pixels according to atmospheric characteristics and aerosol type data of CALIPSO VFM products, wherein marking comprises marking atmospheric characteristic classification labels, marking aerosol type labels and marking cloud type labels;
s2, projecting the FY-4A TOAR data set into a plane coordinate system, adjusting the spatial resolution of ERA-5 data to be 4km which is the same as that of FY-4A through a nearest neighbor interpolation method, and combining the FY-4A TOAR data and the ERA-5 data to generate a grid sample data set;
S3, in space-time matching of 4km x 4km grid sample data and CALIPSO on-orbit data, selecting a time difference within 30 minutes, and matching each CALIPSO track pixel with a grid pixel with the nearest longitude and latitude to generate an on-orbit sample dataset;
s4, constructing a sub-model based on the nonlinear hybrid model, wherein the sub-model comprises an atmospheric feature classification model, an aerosol classification model and a cloud classification model;
S5, inputting the grid sample data set of the whole disc region into each submodel for detection, and obtaining complete whole disc prediction data, wherein the complete whole disc prediction data comprises an atmospheric feature classification result, an aerosol classification result and a cloud classification result.
2. The method as claimed in claim 1, wherein in step S1, the atmospheric feature type data in Feature Classification Flags is selected in the atmospheric feature classification label, each track pixel is marked, and all samples are marked as four types of pixels, namely, a clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer.
3. The method according to claim 1, wherein in step S1, the Aerosol type in the Aerosol type label is marked with Aerosol type data in Aerosol sub-type in an "cloudless area Aerosol-cloud/Aerosol mixed layer", the cloudless area and the cloudy area Aerosol types are classified into "clean marine Aerosol-polluted marine Aerosol-dust Aerosol-clean continental Aerosol-human related Aerosol-lifting smoke", the human related Aerosol comprises polluted continental Aerosol and polluted dust Aerosol ".
4. The method of claim 1, wherein in step S1, cloud type data in Cloud sub-type, low overcast is used to mark the Cloud type in "Cloud region-Cloud/aerosol mix layer" and the Cloud type is divided into "transparent low Cloud-opaque low Cloud-high Cloud-volume Cloud-deep convective Cloud" in the Cloud type label.
5. The method according to claim 1, wherein in step S4, the process of constructing the nonlinear hybrid model comprises the steps of:
Training four different types of basic classification models ET, RF, XGB, LGT on the training set, and for each basic classification model, generating a prediction result by using data in the training set;
and using the prediction result as the meta-feature and the real label of the training set for meta-model training.
6. The method of claim 5, wherein the meta-model is an extreme tree for combining the predictions of the base model to obtain a final classification result.
7. The method according to claim 1, wherein in step S5, the detecting comprises the steps of:
acquiring a clear sky-cloud layer-aerosol layer-cloud/aerosol mixed layer through an atmospheric characteristic classification model;
Detecting an area of an aerosol adhesive layer-cloud/aerosol mixed layer based on an aerosol classification model, and classifying the aerosol adhesive of the area into the following categories: "cleaning marine aerosol-pollution marine aerosol-dust aerosol-cleaning continental aerosol-artificial related aerosol-lifting smoke";
Detecting a cloud layer-cloud/aerosol mixed layer region based on a cloud classification model, and classifying the cloud layer of the region into: "transparent low cloud-low cloud opaque-high cloud-coiled cloud-deep convective cloud".
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311703249.4A CN118053077A (en) | 2023-12-12 | 2023-12-12 | Method for detecting cloud-aerosol by using FY-4A full-disc data at all days |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311703249.4A CN118053077A (en) | 2023-12-12 | 2023-12-12 | Method for detecting cloud-aerosol by using FY-4A full-disc data at all days |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118053077A true CN118053077A (en) | 2024-05-17 |
Family
ID=91052810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311703249.4A Pending CN118053077A (en) | 2023-12-12 | 2023-12-12 | Method for detecting cloud-aerosol by using FY-4A full-disc data at all days |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118053077A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190057171A1 (en) * | 2017-02-10 | 2019-02-21 | Guangxi Institute of Meteorlogical Disaster- Reducing Research | Novel nonlinear method for area-wide near surface air temperature precision retrieval |
| CN110942111A (en) * | 2019-12-31 | 2020-03-31 | 北京弘象科技有限公司 | Method and device for identifying strong convection cloud cluster |
| CN112464920A (en) * | 2021-02-01 | 2021-03-09 | 中国人民解放军国防科技大学 | FY-3D infrared hyperspectral cloud detection method based on extreme random tree |
| CN112966710A (en) * | 2021-02-01 | 2021-06-15 | 中国人民解放军国防科技大学 | FY-3D infrared hyperspectral cloud detection method based on linear discriminant analysis |
| CN114863169A (en) * | 2022-04-27 | 2022-08-05 | 电子科技大学 | Image classification method combining parallel ensemble learning and federal learning |
| CN116338625A (en) * | 2023-02-20 | 2023-06-27 | 武汉大学 | Multi-scale classification method for cloud and aerosol and computer readable medium |
| CN116698688A (en) * | 2023-04-20 | 2023-09-05 | 兰州大学 | Method for estimating concentration of atmospheric particulates based on double-star of cloud number 4 |
| CN116756560A (en) * | 2023-05-15 | 2023-09-15 | 上海期智研究院 | Dust aerosol recognition method, system, model training method, medium and equipment |
-
2023
- 2023-12-12 CN CN202311703249.4A patent/CN118053077A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190057171A1 (en) * | 2017-02-10 | 2019-02-21 | Guangxi Institute of Meteorlogical Disaster- Reducing Research | Novel nonlinear method for area-wide near surface air temperature precision retrieval |
| CN110942111A (en) * | 2019-12-31 | 2020-03-31 | 北京弘象科技有限公司 | Method and device for identifying strong convection cloud cluster |
| CN112464920A (en) * | 2021-02-01 | 2021-03-09 | 中国人民解放军国防科技大学 | FY-3D infrared hyperspectral cloud detection method based on extreme random tree |
| CN112966710A (en) * | 2021-02-01 | 2021-06-15 | 中国人民解放军国防科技大学 | FY-3D infrared hyperspectral cloud detection method based on linear discriminant analysis |
| CN114863169A (en) * | 2022-04-27 | 2022-08-05 | 电子科技大学 | Image classification method combining parallel ensemble learning and federal learning |
| CN116338625A (en) * | 2023-02-20 | 2023-06-27 | 武汉大学 | Multi-scale classification method for cloud and aerosol and computer readable medium |
| CN116698688A (en) * | 2023-04-20 | 2023-09-05 | 兰州大学 | Method for estimating concentration of atmospheric particulates based on double-star of cloud number 4 |
| CN116756560A (en) * | 2023-05-15 | 2023-09-15 | 上海期智研究院 | Dust aerosol recognition method, system, model training method, medium and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Derrien et al. | MSG/SEVIRI cloud mask and type from SAFNWC | |
| Lee et al. | Impact of the smoke aerosol from Russian forest fires on the atmospheric environment over Korea during May 2003 | |
| CN105988146B (en) | A kind of application data processing method of satellite-borne microwave radiometer | |
| Zhang et al. | Shortwave direct radiative effects of above-cloud aerosols over global oceans derived from 8 years of CALIOP and MODIS observations | |
| van der Linden et al. | Satellite‐based climatology of low‐level continental clouds in southern West Africa during the summer monsoon season | |
| Pangaud et al. | Assimilation of AIRS radiances affected by mid-to low-level clouds | |
| Li et al. | Simulated convective invigoration processes at trade wind cumulus cold pool boundaries | |
| Kotchenova et al. | Lidar remote sensing for modeling gross primary production of deciduous forests | |
| Xue et al. | Shortwave radiation calculation for forest plots using airborne LiDAR data and computer graphics | |
| Hart et al. | Building a tropical–extratropical cloud band metbot | |
| CN107016095A (en) | Climate change comprehensive estimation method based on multi-source carbon number evidence | |
| Carvajal et al. | Relationship between atmospheric corrections and training-site strategy with respect to accuracy of greenhouse detection process from very high resolution imagery | |
| Misra et al. | Validation of version 5.1 MODIS aerosol optical depth (Deep blue algorithm and dark target approach) over a semi-arid location in Western India | |
| Khan et al. | Dust plume formation in the free troposphere and aerosol size distribution during the Saharan Mineral Dust Experiment in North Africa | |
| WO2024021225A1 (en) | High-resolution true-color visible light model generation method, high-resolution true-color visible light model inversion method, and system | |
| Zhang et al. | Sunlight-absorbing aerosol amplifies the seasonal cycle in low-cloud fraction over the southeast Atlantic | |
| CN114005048A (en) | Multi-temporal data-based land cover change and thermal environment influence research method | |
| Sreekanth et al. | Spatio-temporal variations in columnar aerosol optical properties over Bay of Bengal: Signatures of elevated dust | |
| Chappell et al. | Satellites reveal Earth's seasonally shifting dust emission sources | |
| Mijani et al. | Exploring the effect of COVID-19 pandemic lockdowns on urban cooling: A tale of three cities | |
| Alonso-Montesinos et al. | Impact of a Saharan dust intrusion over southern Spain on DNI estimation with sky cameras | |
| CN110175638B (en) | Raise dust source monitoring method | |
| CN115546658B (en) | Night cloud detection method combining quality improvement and CNN improvement of data set | |
| CN118053077A (en) | Method for detecting cloud-aerosol by using FY-4A full-disc data at all days | |
| CN115950832A (en) | Method for inverting near-surface nitrogen dioxide concentration based on satellite data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |