A method for monitoring sandstorm weather throughout the day

Abstract

The present invention discloses a method for monitoring dust storms throughout the day, comprising the following steps: S1. Calculating a dust spectral index, S2. Extracting time-series features, S3. Constructing a visibility inversion model, and S4. Mapping dust intensity. The present invention utilizes time-series thermal infrared images from geostationary meteorological satellites, combined with spectral indices and brightness temperatures in each band, to construct a set of time-series features to characterize the dynamic characteristics of dust events. By coupling a bidirectional long-short-term memory network model, a model is constructed based on these time-series features, achieving the inversion of dust intensity, thereby achieving the effect of dynamic, high-precision monitoring of dust intensity throughout the day. In this process, the limitations of traditional spectral index methods and machine learning methods in fully utilizing time-series image information when inverting dust intensity are overcome, significantly improving the inversion accuracy of dust intensity throughout the day.

Description

Technical Field

[0001] The present invention relates to a sandstorm weather monitoring method, and in particular to a full-day sandstorm weather monitoring method. Background Art

[0002] Due to the high uncertainty and complexity of the spatiotemporal distribution of dust storms, traditional ground-based observation methods often fail to meet the needs of large-scale, real-time monitoring. Remote sensing technology, using sensors carried on satellite platforms, enables continuous monitoring of the formation, propagation, and decay of dust events at a macroscale. Therefore, remote sensing technology holds irreplaceable advantages in dust monitoring. Remote sensing-based dust monitoring satellite platforms can be categorized into two main types: polar-orbiting and geostationary. Polar-orbiting satellites orbit the Earth's poles, covering most of the globe and providing high-resolution spatial imagery. However, due to their limited re-entry periods, their temporal resolution is low, making it difficult to monitor the dynamic nature of dust storms. In contrast, geostationary satellites can continuously monitor fixed areas, greatly facilitating the real-time dynamic monitoring of dust events. The Himawari-8 satellite, in particular, provides high-precision data covering East Asia and the western Pacific at 10-minute intervals, providing a crucial data source for monitoring the occurrence, development, and propagation of dust storms. Developing all-day dust monitoring technology based on Himawari-8 satellite data is of great significance.

[0003] Considering the existing remote sensing-based dust monitoring technology, there are two main methods: spectral index-based methods and machine learning-based methods.

[0004] The spectral index method uses the unique reflectance characteristics of dust in the visible, near-infrared, and thermal infrared bands, which differ from those of other ground objects, to combine different bands to improve sensitivity to dust and reduce interference with background information, thereby enabling dust identification. This method has the advantages of simplicity and high computational efficiency, but it relies on spectral information in a limited number of bands and cannot describe complex nonlinear relationships. It has poor adaptability to different environmental backgrounds and is easily affected by external factors such as surface type.

[0005] Machine learning methods use information from all spectral bands of remote sensing imagery as input and rely on complex network structures to autonomously learn dust characteristics based on a large number of training samples. Dust monitoring is then performed based on the trained model. Commonly used methods include support vector machines, random forests, and convolutional neural networks. Machine learning methods not only fully utilize information from all spectral bands of remote sensing imagery but also can express complex nonlinear relationships from input to output, thereby enhancing the accuracy and robustness of dust monitoring.

[0006] However, existing research has primarily focused on binary classification, distinguishing between "dust" and "non-dust." In practice, the lack of dust intensity information makes it difficult to accurately implement dust mitigation measures. The limited research on dust intensity monitoring relies on imagery at the time of observation, using spectral indices or machine learning methods to establish a quantitative relationship between spectral information and dust intensity, thereby leveraging this relationship to detect dust intensity. However, these methods fail to utilize time-series imagery, resulting in limited accuracy in dust intensity monitoring. Prior to dust passage, the atmospheric spectral information undergoes a series of dynamic changes, and these temporal variations are crucial for improving the accuracy of dust identification. However, existing dust intensity monitoring methods are unable to effectively utilize this temporal spectral information, posing challenges in the accuracy of dust intensity monitoring.

[0007] This application is based on the time-series thermal infrared images of geostationary meteorological satellites. It uses spectral indices and brightness temperatures of each band to construct a set of time-series features to characterize this phenomenon, and couples a bidirectional long-short-term memory network to build a model based on the time-series features to achieve high-precision dynamic monitoring of sandstorm intensity throughout the day. Summary of the Invention

[0008] In order to solve the shortcomings of the above technologies, the present invention provides a method for monitoring sandstorm weather throughout the day.

[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for monitoring sandstorm weather all day long, comprising the following steps:

[0010] S1. Calculation of the dust spectral index: Cloud masking is performed based on hourly thermal infrared brightness temperature images from geostationary meteorological satellites. The hourly brightness temperature difference spectral index, three-band volcanic ash product index, brightness temperature adjusted dust index, and thermal infrared composite dust index are then calculated.

[0011] S2. Time series feature extraction, including the brightness temperature difference spectral index, three-band volcanic ash product index, brightness temperature adjusted dust index, thermal infrared integrated dust index, and the clear sky state value of brightness temperature in each band, the state value near the observation time, and the state value at the observation time;

[0012] S3. Construction of a visibility inversion model based on a bidirectional long short-term memory network model, using the temporal feature information in S2 as input;

[0013] S4. Dust intensity mapping.

[0014] Furthermore, the brightness temperature difference spectral index includes three typical calculation formulas as follows:

[0015] BTD 11-12 =BT 11.2 -BT 12.4 ;

[0016] BTD 3-12 =BT 3.9 -BT 11.2 ;

[0017] BTD 8-11 =BT 8.6 -BT 11.2 ;

[0018] Where, BTD 11-12 、BTD 3-12 、BTD 8-11 Represents three typical brightness temperature difference spectral indices, BT 3.9 , BT 8.6 , BT 11.2 , BT 12.4 They correspond to the 3.9μm, 8.6μm, 11.2μm and 12.4μm bands of geostationary meteorological satellites respectively.

[0019] Furthermore, the calculation formula of the three-band volcanic ash product index is as follows:

[0020] TVAP=60+10×(BT 12.4 -BT 11.2 )+3×(BT 3.9 -BT 11.2 ).

[0021] Furthermore, the calculation process of brightness temperature adjustment dust index is as follows:

[0022]

[0023] Where, BDI=(BT 3.9 -BT 12.4 ) 2 ×(BT 11.2 -BT 12.4 );

[0024] Where BADI is the brightness temperature adjusted dust index, BDI is the brightness temperature dust index, and BDI 0.95 Indicates the 95% quantile of BDI.

[0025] Furthermore, the calculation process of the thermal infrared comprehensive dust index is as follows:

[0026]

[0027] Furthermore, the clear sky status values ​​of the spectral index and brightness temperature of each band refer to the spectral index values ​​and brightness temperature values ​​calculated using the clear sky images synthesized from the images 7 days before the observation time. During the synthesis, it is mainly based on the 11.2μm band, and the images corresponding to the observation time of its maximum value in the 7 days are selected for synthesis of each band.

[0028] Furthermore, the state values ​​of the spectral index and brightness temperature of each band at the time of observation are calculated based on the images of each hour in the 30 hours before the observation time, and then the time series data of each index and brightness temperature of each band are calculated using the following formula:

[0029]

[0030] Where F(x,y) is the final spectral index value and brightness temperature value of each band of the pixel point (x,y) in the synthetic image; I i (x,y) represents the spectral index value or brightness temperature value of each band at pixel (x,y) of the image in the i-th hour; w i is the weight of the i-th hour;

[0031]

[0032] Where w i represents the weight of the image in the i-th hour, λ is a constant that controls the weight increment rate and is set to 0.5, n is the total number of hours and is set to 30, and the image at the moment farthest from the observation time is set to 1, and the image immediately adjacent to the observation time is set to 30.

[0033] Furthermore, the state values ​​of the spectral index and the brightness temperature of each band at the observation time refer to the spectral index value and the brightness temperature value of each band extracted based on the image at the observation time.

[0034] Furthermore, the visibility inversion model includes an input layer, a hidden layer, and an output layer;

[0035] The input layer corresponds to the input vector X at three different times t-2 、X t-1 、X t , X t-2 The clear sky state value corresponding to the brightness temperature of each band, X t-1 Corresponding to the state value at the time of observation, X t Corresponding state value at the observation moment;

[0036] The output layer corresponds to the visibility value Y at the observation time t t ;

[0037] The hidden layer consists of two sub-networks, which are responsible for processing the forward information flow and the reverse information flow respectively, so as to capture the forward feature information and the backward feature information of the input sequence.

[0038] Furthermore, the hourly visibility data is inverted based on the visibility inversion model in S3, and the sandstorm intensity is divided into five levels according to the hourly visibility data, namely, extremely strong sandstorm, strong sandstorm, sandstorm, floating dust / blowing sand, and no sandstorm. The hourly sandstorm intensity mapping is completed, and all-day sandstorm monitoring is realized.

[0039] A method for monitoring dust storms around the clock utilizes time-series thermal infrared imagery from geostationary meteorological satellites, combined with spectral indices and brightness temperatures across various bands, to construct a set of temporal features to characterize the dynamic characteristics of dust events. By coupling a bidirectional long-short-term memory network model, a model based on these temporal features is constructed to invert dust intensity, achieving dynamic, high-precision monitoring of dust intensity around the clock. This method overcomes the limitations of traditional spectral index and machine learning methods in fully utilizing temporal image information when inverting dust intensity, significantly improving the accuracy of inversion of dust intensity around the clock. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Flowchart of the present invention.

[0041] Figure 2 This is the framework of the visibility inversion model of the present invention.

[0042] Figure 3 This is the hourly dust intensity chart disclosed in Example 2. DETAILED DESCRIPTION

[0043] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Example 1

[0045] This embodiment is about the all-day sandstorm weather monitoring method, such as Figure 1 As shown, the following steps are included:

[0046] S1. Calculation of the dust spectral index: Cloud masking is performed based on the hourly thermal infrared brightness temperature images of the geostationary meteorological satellite, and then the hourly brightness temperature difference spectral index, three-band volcanic ash product index, brightness temperature adjusted dust index, and thermal infrared comprehensive dust index are calculated. It should be noted that the specific model of the geostationary meteorological satellite in this embodiment is preferably Himawari-8, and the hourly thermal infrared brightness temperature images are obtained using the AHI sensor of Himawari-8. The cloud masking method is a prior art method, specifically according to the method headed by author Yamamoto and published in the journal Geophysical Research Letters in 2016, Volume 43, Pages 5886–5894. It is not original to this application and is therefore not displayed or discussed.

[0047] Preferably, the brightness temperature difference spectral index (i.e., brightness temperature difference, BTD) specifically includes three typical brightness temperature difference spectral indices and the calculation formula is as follows:

[0048] BTD 11-12 =BT 11.2 -BT 12.4 ;

[0049] BTD 3-12 =BT 3.9 -BT 11.2 ;

[0050] BTD 8-11 =BT 8.6 -BT 11.2 ;

[0051] Where, BTD 11-12 、BTD 3-12 、BTD 8-11 Represents three typical brightness temperature difference spectral indices, BT 3.9 , BT 8.6 , BT 11.2 , BT 12.4 They correspond to the 3.9μm, 8.6μm, 11.2μm and 12.4μm bands of geostationary meteorological satellites respectively.

[0052] Preferably, the calculation formula of the three-band volcanic ash product index (i.e., Three-band volcanic ash product, TVAP) is as follows:

[0053] TVAP=60+10×(BT 12.4 -BT 11.2 )+3×(BT 3.9 -BT 11.2 ).

[0054] Preferably, the calculation process of the brightness temperature adjusted dust index (BADI) is as follows:

[0055]

[0056] Where, BDI=(BT 3.9 -BT 12.4 ) 2 ×(BT 11.2 -BT 12.4 );

[0057] Where BADI is the brightness temperature adjusted dust index as mentioned above, BDI is the brightness temperature dust index (BDI), and BDI 0.95 Indicates the 95% quantile of BDI.

[0058] Preferably, the calculation process of the Thermal Infrared Integrated Dust Index (TIIDI) is as follows:

[0059]

[0060] S2. Time series feature extraction, including the brightness temperature difference spectral index and the clear sky state value of each band, the state value near the observation time, and the state value at the observation time.

[0061] Preferably, the clear sky status values ​​of the spectral index and brightness temperature of each band refer to the spectral index values ​​and brightness temperature values ​​calculated using the clear sky image synthesized from the images 7 days before the observation time. During the synthesis, it is mainly based on the 11.2μm band, and the images corresponding to the observation time of its maximum value in the 7 days are selected for synthesis of each band.

[0062] Preferably, the state values ​​of the spectral index and the brightness temperature of each band at the time of observation near the observation time are calculated based on the images of each hour within 30 hours before the observation time, and then the time series data of each index and brightness temperature of each band are calculated using the following formula:

[0063]

[0064] Where F(x,y) is the final spectral index value and brightness temperature value of each band of the pixel point (x,y) in the synthetic image; I i (x,y) represents the spectral index value or brightness temperature value of each band at pixel (x,y) of the image in the i-th hour; w i is the weight of the i-th hour;

[0065]

[0066] Where w i represents the weight of the image in the i-th hour, λ is a constant that controls the weight increment rate and is set to 0.5, n is the total number of hours and is set to 30, and the image at the moment farthest from the observation time is set to 1, and the image immediately adjacent to the observation time is set to 30.

[0067] The observation time status values ​​of the spectral index and brightness temperature of each band refer to the spectral index value and brightness temperature value of each band extracted based on the image at the observation time.

[0068] S3. Construction of a visibility inversion model, based on a bidirectional long short-term memory network model, using the temporal feature information in S2 as input to build a visibility inversion model;

[0069] like Figure 2 As shown, the visibility inversion model includes input layer, hidden layer and output layer;

[0070] The input layer corresponds to the input vector X at three different times t-2 、X t-1 、X t , X t-2 The clear sky state value corresponding to the brightness temperature of each band, X t-1 Corresponding to the state value at the time of observation, X t Corresponding state value at the observation moment;

[0071] The hidden layer consists of two sub-networks, which are responsible for processing the forward information flow and the reverse information flow respectively, so as to capture the forward feature information and the backward feature information of the input sequence.

[0072] Preferably, each sub-network includes three LSTM basic modules, which process X t-2 、X t-1 、X t The input information is serially connected between them so that the information can be transmitted in time sequence. It should be noted that the LSTM basic module is well known and those skilled in the art can clearly understand its content, so it will not be explained in detail.

[0073] The output layer corresponds to the visibility value Y at the observation time t t , and the hidden layer processes X t The LSTM basic modules at each moment are connected.

[0074] S4. Dust intensity mapping uses the visibility inversion model in S3 to invert hourly visibility data, and based on the national standard "Dust Weather Grade (GB / T 20480-2017)", the dust intensity is divided into five levels: extremely strong sandstorm, strong sandstorm, sandstorm, floating dust / blowing sand, and no sandstorm. Hourly dust intensity mapping is completed to achieve all-day dust monitoring.

[0075] Example 2

[0076] Based on Example 1, Example 2 discloses the practical application of Example 1.

[0077] This example collects hourly thermal infrared band images of the Himawari-H8 satellite's AHI sensor over Inner Mongolia during the spring (March-May) from 2020 to 2022, as well as hourly visibility data and hourly wind speed data from 119 meteorological observation stations during the corresponding time period. The following steps are used to generate an hourly spatial distribution map of dust intensity over Inner Mongolia during 2020-2022. The specific steps are as follows:

[0078] A1. Referring to the method in step S1 of Example 1, the collected satellite imagery is first cloud masked according to the method described by author Yamamoto and published in the journal Geophysical Research Letters, Volume 43, Pages 5886–5894, 2016. Then, based on the formula disclosed in step S1 of Example 1, three typical brightness temperature difference spectral indices, a three-band volcanic ash product index, a brightness temperature adjusted dust index, and a thermal infrared integrated dust index are calculated.

[0079] A2. Extract the temporal characteristics of each observation moment by referring to the method of step S2 in Example 1, including the clear sky state value, the state value near the observation moment, and the state value at the observation moment of the above-mentioned spectral index and brightness temperature of each band.

[0080] A3. Extract the visibility values ​​observed at stations during the period of sandstorms in Inner Mongolia between 2020 and 2022. Randomly select 70% of the data as the training dataset and the remaining 30% as the validation dataset. Extract the temporal feature values ​​of the images corresponding to the ground-based visibility in the training dataset and use them as the training data set. Figure 2 The visibility inversion model shown is trained.

[0081] A4. Based on the trained visibility inversion model, use the time series features calculated in A2 as input to generate hourly visibility data products for Inner Mongolia in the spring of 2020-2022.

[0082] A5. Use the visibility data product generated in A4, based on the national standard "Dust Weather Grade (GB / T 20480-2017)", and combine it with the collected wind speed data to generate a dust intensity data product.

[0083] Based on the above steps, it is possible to monitor sandstorm weather all day long. The visibility inversion model constructed in step A4 is verified using the visibility data of the validation dataset, and the average relative error is only 23.83%. Figure 3 The hourly sandstorm intensity map shown during the sandstorm in Inner Mongolia on March 15, 2021, truly reflects the actual situation.

[0084] In summary, the method disclosed in Example 1 also has the potential to be extended to other geostationary satellite images, which is hereby explained.

[0085] This application discloses a method for monitoring dust storms around the clock. This method utilizes time-series thermal infrared imagery from geostationary meteorological satellites, combined with spectral indices and brightness temperatures in each band, to construct a set of time-series features to characterize the dynamic characteristics of dust events. By coupling a bidirectional long-short-term memory network model, a model based on these time-series features is constructed to invert dust intensity, achieving dynamic, high-precision monitoring of dust intensity around the clock. This process overcomes the limitations of traditional spectral index and machine learning methods in fully utilizing time-series imagery information when inverting dust intensity, significantly improving the accuracy of inversion of dust intensity around the clock.

[0086] The above embodiments are not limitations of the present invention, and the present invention is not limited to the above examples. Any changes, modifications, additions or substitutions made by technicians in this technical field within the scope of the technical solution of the present invention also fall within the scope of protection of the present invention.

Claims

1. A method for monitoring sandstorm weather throughout the day, characterized in that: The following steps are involved: S1. Calculation of the dust spectral index: Cloud masking is performed based on hourly thermal infrared brightness temperature images from geostationary meteorological satellites. The hourly brightness temperature difference spectral index, three-band volcanic ash product index, brightness temperature adjusted dust index, and thermal infrared composite dust index are then calculated. The brightness temperature difference spectral index includes three typical and respective calculation formulas are as follows: ； ； ； Where, 、 、 Represent three typical brightness temperature difference spectral indices, 、 、 、 These correspond to the 3.9µm, 8.6µm, 11.2µm, and 12.4µm bands of geostationary meteorological satellites; The calculation formula of the three-band volcanic ash product index is as follows: ； The calculation process of the brightness temperature adjusted dust index is as follows: ； in, ; Where, Adjust the dust index for brightness temperature, is the brightness temperature dust index, express 95% quantile of The calculation process of the thermal infrared comprehensive dust index is as follows: ， ； S2. Time series feature extraction, including the brightness temperature difference spectral index, three-band volcanic ash product index, brightness temperature adjusted dust index, thermal infrared integrated dust index, and the clear sky state value of brightness temperature in each band, the state value near the observation time, and the state value at the observation time; S3. Construction of a visibility inversion model based on a bidirectional long short-term memory network model, using the temporal feature information in S2 as input; S4. Dust intensity mapping.

2. The all-day sandstorm monitoring method according to claim 1, characterized in that: The clear sky status values ​​of the spectral index and brightness temperature of each band refer to the spectral index values ​​and brightness temperature values ​​calculated using the clear sky images synthesized using images from the 7 days before the observation time. During the synthesis, the 11.2μm band is mainly based on the observation time corresponding to the maximum value in the 7 days, and the images corresponding to the observation time at which the maximum value is found are selected for synthesis of each band.

3. The all-day sandstorm monitoring method according to claim 2, characterized in that: The state values ​​of the spectral index and brightness temperature of each band at the time of observation are calculated based on the images of each hour in the 30 hours before the observation time, and then the time series data of each index and brightness temperature of each band are calculated using the following formula: ， Where, Pixels in the synthetic image The final spectral index value and brightness temperature value of each band; Indicates the Hourly images in pixels spectral index value or brightness temperature value of each band; It is Hour weight; ， Where, Indicates the The weight of the hourly image, Is a constant that controls the weight increment rate and is set to 0.

5. The total number of hours is 30, and the image at the moment farthest from the observation time is 1, and the image immediately adjacent to the observation time is 30.

4. The all-day sandstorm monitoring method according to claim 3, characterized in that: The observation time state values ​​of the spectral index and the brightness temperature of each band refer to the spectral index value and the brightness temperature value of each band extracted based on the image at the observation time.

5. The all-day sandstorm monitoring method according to claim 1, characterized in that: The visibility inversion model includes an input layer, a hidden layer and an output layer; The input layer corresponds to the input vectors at three different times 、 、 , The clear sky state value corresponding to the brightness temperature of each band, Corresponding to the state value at the time of observation, Corresponding state value at the observation moment; The output layer corresponds to the observation time Visibility value ; The hidden layer includes two sub-networks, and the two sub-networks are responsible for processing the forward information flow and the reverse information flow respectively, so as to capture the forward feature information and the backward feature information of the input sequence.

6. The all-day sandstorm monitoring method according to claim 1, characterized in that: Based on the visibility inversion model in S3, the hourly visibility data is inverted. According to the hourly visibility data, the sandstorm intensity is divided into five levels, namely, extremely strong sandstorm, strong sandstorm, sandstorm, floating dust / blowing sand, and no sandstorm. The hourly sandstorm intensity mapping is completed to realize all-day sandstorm monitoring.

Images