Sandstorm weather area forecasting method and system

Abstract

The present invention discloses a method and system for predicting the fallout area of sandstorms. The forecasting method is implemented based on a distributed sandstorm monitoring sensor network system, including constructing source area monitoring equipment, transmission path monitoring equipment, and potential fallout area monitoring equipment composed of PM10 and PM2.5 sensors, meteorological sensors, and optical visibility sensors; edge computing nodes collect and preprocess monitored parameter data in real time, establish a sandstorm source area-transmission-fallout area correlation model, and calculate a sandstorm diffusion index and a source area activity index; based on the calculation results of the correlation model, an improved space-time vector autoregressive model is used to predict the sandstorm transmission path and fallout area range, and generate a forecast product including the fallout area range, sandstorm intensity level, and arrival time; the present invention can significantly improve the accuracy and timeliness of sandstorm fallout area forecasts, and provide scientific decision-making support for sandstorm disaster prevention.

Description

Technical Field

[0001] The present invention belongs to the technical field of sandstorm area forecasting, and in particular relates to a sandstorm area forecasting method and system. Background Art

[0002] Dust storm is a common natural disaster. Traditional sandstorm forecasts mainly rely on meteorological satellite remote sensing monitoring and numerical weather forecast models. However, these methods often have problems such as low spatial resolution, poor timeliness, and inability to accurately predict the scope and intensity of the impact area, making it difficult to meet the needs of refined forecasts.

[0003] Existing sandstorm forecasting technologies are mainly based on physical models and statistical methods. Although physical models take into account complex atmospheric dynamic processes, they have large computational complexity, complex parameter adjustment, and lack effective characterization of local characteristics. Although statistical methods are simple to calculate, they have limited accuracy, difficulty in handling nonlinear relationships, and insufficient ability to predict new sandstorm events. In addition, existing technologies lack full-chain monitoring and correlation analysis of sandstorm source areas, transmission paths, and potential landing areas. The real-time nature of data collection and the degree of automation of preprocessing are insufficient, resulting in short forecast timeliness and low accuracy. Especially under complex meteorological conditions and the interaction of multi-source sandstorms, the error in the landing area forecast is large, making it difficult to provide timely and effective decision-making support for disaster prevention and mitigation.

[0004] Therefore, there is an urgent need to build a method and system that can accurately predict the location of sandstorm weather. Summary of the Invention

[0005] The present invention achieves the purpose of accurately predicting the dust transmission path, landing area range, intensity level and arrival time by constructing a distributed dust monitoring sensor network system, establishing a dust source area-transmission-falling area correlation model and an improved space-time vector autoregressive model.

[0006] In a first aspect, the present invention provides a method for predicting sandstorm areas, which is implemented based on a distributed sandstorm monitoring sensor network system and includes the following steps:

[0007] Step S1: Construct a distributed sand and dust monitoring sensor network system, which includes source area monitoring equipment composed of PM10, PM2.5 sensors, meteorological sensors and optical visibility sensors, transmission path monitoring equipment and potential landing area monitoring equipment.

[0008] Step S2: The edge computing node collects and pre-processes the monitored parameter data in real time, wherein the parameter data include PM10, PM2.5 concentration, wind speed, wind direction, humidity, air pressure and visibility parameters; taking the parameter data into consideration, a dust source area-transmission-falling area correlation model is established to calculate the dust diffusion index and the dust source area activity index.

[0009] Step S3: Based on the calculation results of the correlation model, the improved space-time vector autoregressive model is used to predict the dust transmission path and landing area, and generate a dust landing area forecast product, which includes the landing area range, dust intensity level and arrival time.

[0010] Furthermore, the edge computing preprocessing method includes: data time series smoothing processing.

[0011] The exponentially weighted moving average (EWMA) algorithm is used to smooth the data time series. ;in, for The smoothed value at time, for The original value of the moment, for The smoothed value at time, is the smoothing coefficient, and the smoothing coefficient ranges from 0.1 to 0.3.

[0012] Furthermore, the dust source-transmission-falling area correlation model is expressed as follows:

[0013] ;in, is the dust diffusion index; is the PM10 concentration; is the wind speed; For wind direction; is the direction of the target area relative to the monitoring point; is the relative humidity; is the pressure gradient; For visibility.

[0014] 、 、 、 、 is the weight coefficient, and each weight coefficient is determined according to the square ratio of its correlation coefficient with the historical dust event landing area: ,in, is the correlation coefficient between each factor and the landing area of historical sandstorm events.

[0015] Using the Soil Moisture Index and vegetation cover index Constructing a dust source area activity index :

[0016] ;in, is the normalized soil moisture index, is the normalized vegetation cover index; is the wind erosion index.

[0017] when hour, ;

[0018] when , ;

[0019] in, is the wind speed threshold for dust-raising, which is 6m / s.

[0020] Furthermore, based on the calculation results of the correlation model, the improved space-time vector autoregressive model is used to predict the dust transmission path and the range of the dust landing area. The dust transmission path is predicted by combining the SDI index and the DSAI index to calculate the transmission probability:

[0021] ,in, From the source area To the landing area The transmission probability, For historically source areas To the landing area The number of sandstorm events, Source area The total number of sandstorm events, Source area The dust diffusion index, is the average dust diffusion index of all regions, Source area The dust source area activity index, is the average dust source activity index of all regions.

[0022] Furthermore, the set of predicted points for the impact zone is obtained based on the dust feature vector. The calculation of the dust feature vector takes into account the SDI index and real-time monitoring parameter data. Its mathematical expression is:

[0023] ;in, For location In time The dust characteristic vector includes: SDI index, PM10 concentration and diffusion rate, The corresponding spatial location points constitute the set of predicted landing points , is the total number of predicted points in the landing area.

[0024] For location In time The dust feature vector of ; For location In time Meteorological characteristic vectors, including wind speed, wind direction and humidity; For location Neighborhood location In time The dust feature vector of ; For location The spatial neighborhood set of ; is the time autoregressive matrix, combined with the extracted local features, and estimated by the least squares method:

[0025] ; Represents the maximum time index in the dataset; is the exogenous variable influence matrix, which is estimated by the least squares method:

[0026] ; is the spatial correlation matrix, and its initial value is determined by the transmission probability Determine and update through iterative optimization; is the random error term; is the time autoregressive order, ranging from 1 to 3; is the lag order of the exogenous variable, ranging from 1 to 2.

[0027] Furthermore, the spatial correlation matrix Initial value of Expressed as:

[0028] ;in, From the position To its neighboring location transmission probability.

[0029] Spatial correlation matrix The iterative optimization update method uses the least squares method: ;in, For the The spatial correlation matrix of the iteration, is the learning rate, ranging from 0.01 to 0.05, is the loss function, defined as: ,in, is the observed dust feature vector, is the dust feature vector predicted by the model, Represents the Euclidean norm of a vector.

[0030] Furthermore, the dust fall area range in the sand and dust fall area forecast product is represented by probability contour lines and calculated using the SDI index using the kernel density estimation method: ;in, For location The probability density of dust impact is As the kernel function, the Gaussian kernel function is used: , The bandwidth parameter is 1.5 times the average site spacing.

[0031] Furthermore, the dust intensity level in the dust fall area forecast product is divided into four levels according to the SDI index and the PM10 concentration in the dust feature vector predicted by the model:

[0032] Mild level meets: Or PM10 concentration is ;

[0033] Moderate level satisfaction: Or PM10 concentration is ;

[0034] Severe level meets: Or PM10 concentration is ;

[0035] Severity level meets: Or PM10 concentration is .

[0036] Furthermore, the arrival time in the dust fall area forecast product is based on the transmission path and transmission probability. Calculate the arrival time of dust The mathematical expression is: ,in, is the arrival time, D is the distance, is the effective transmission speed of dust, is the transmission probability influence coefficient, which is set to 0.5. is the historical average transmission probability, The calculation formula is:

[0037] ,in, is the average wind speed along the transmission path, is the average elevation difference of the transmission path.

[0038] In a second aspect, the present invention provides a sandstorm weather area forecasting system for executing the sandstorm weather area forecasting method of the first aspect; the forecasting system includes: a distributed sandstorm monitoring sensor network, a data acquisition and processing module, and a sandstorm area forecast product generation module.

[0039] The distributed sand and dust monitoring sensor network includes source area monitoring equipment, transmission path monitoring equipment and potential landing area monitoring equipment composed of PM10, PM2.5 sensors, meteorological sensors and optical visibility sensors.

[0040] The data acquisition module collects and pre-processes the monitored parameter data in real time through the edge computing node. The parameter data include PM10, PM2.5 concentration, wind speed, wind direction, humidity, air pressure and visibility parameters; taking into account the parameter data, a sand and dust source area-transmission-falling area correlation model is established to calculate the sand and dust diffusion index and the sand and dust source area activity index.

[0041] The sand and dust fall area forecast product generation module uses an improved space-time vector autoregressive model to predict the sand and dust transmission path and fall area range based on the calculation results of the correlation model, and generates a sand and dust fall area forecast product. The sand and dust fall area forecast product includes the fall area range, sand and dust intensity level and arrival time.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention realizes real-time monitoring of the entire chain of sand and dust source areas, transmission paths and potential landing areas by constructing a distributed sand and dust monitoring sensor network system; uses edge computing nodes to efficiently preprocess multi-source parameter data; establishes a sand and dust source area-transmission-landing area correlation model to accurately calculate the sand and dust diffusion index and source area activity index; uses an improved space-time vector autoregressive model, combined with historical sand and dust event data and real-time monitoring parameters, to achieve accurate forecasts of sand and dust transmission paths, landing area ranges, intensity levels and arrival times, greatly improving the accuracy and timeliness of sand and dust landing area forecasts, and providing a scientific basis and decision-making support for the prevention and response to sand and dust disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of the sandstorm weather forecasting method of the present invention;

[0045] Figure 2 The figure is a schematic diagram of the composition of the sandstorm weather area forecasting system of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only part of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] It should be noted that in the embodiments, the distributed dust monitoring sensor network system is constructed using a multi-layered layout strategy. Source zone monitoring equipment is primarily deployed in desert edges and Gobi regions, transmission path monitoring equipment is located along major wind paths, and potential impact zone monitoring equipment is deployed in sensitive areas such as urban clusters and agricultural regions. All sensors utilize low-power designs and solar-powered power solutions to ensure continuous and stable operation in harsh environments. Monitoring data is transmitted in real time to a data processing center via 4G / 5G networks or satellite communications, enabling 24 / 7 uninterrupted monitoring.

[0048] Example 1

[0049] like Figure 1 As shown in the first aspect, the present invention provides a method for predicting the location of sandstorms. The method is implemented based on a distributed sandstorm monitoring sensor network system and includes the following steps:

[0050] Step S1: Construct a distributed sand and dust monitoring sensor network system, which includes source area monitoring equipment composed of PM10, PM2.5 sensors, meteorological sensors and optical visibility sensors, transmission path monitoring equipment and potential landing area monitoring equipment.

[0051] Edge computing preprocessing methods include: data outlier correction and determination, data time series smoothing, and local feature extraction. Edge computing preprocessing is performed using edge servers deployed near the monitoring equipment, reducing data transmission latency and the computational burden on central servers. Outlier correction technology can effectively identify data anomalies caused by sensor failure, calibration deviation, or transient interference. For example, if the PM10 concentration at a monitoring station suddenly increases from 200μg / m³ to 1500μg / m³ in a short period of time, while surrounding stations show no significant change, the system will automatically identify and correct the outlier.

[0052] When the deviation between the monitored value and the historical mean value for the same period exceeds three standard deviations, it is determined to be an outlier and corrected. The outlier is replaced by the average of at least four data points before and after the outlier. The expression for correcting the outlier is: ;in, is the abnormal coefficient, To monitor the value, The historical average for the same period. is the standard deviation, when , it is determined to be an abnormal value.

[0053] The exponentially weighted moving average (EWMA) algorithm is used to smooth the data time series. ;in, for The smoothed value at time, for The original value of the moment, for The smoothed value at time, is the smoothing coefficient, and the smoothing coefficient ranges from 0.1 to 0.3. Dynamic adjustment based on the fluctuation characteristics of different parameters, such as wind speed and other parameters with large fluctuations The value is set to be small (about 0.1), while relatively stable parameters such as PM10 concentration The value is set to a larger value (about 0.3) to balance the response speed and anti-interference ability.

[0054] The advantage of the EWMA algorithm is that it can preserve the time series characteristics of the data while effectively filtering out short-term noise interference. In practical applications, the initial smoothing value S0 is usually taken as the historical average value of the same period to reduce the fluctuation in the algorithm startup phase. The choice of the smoothing coefficient λ directly affects the algorithm performance. A smaller A value (such as 0.1) makes the smooth curve smoother, has strong anti-interference ability but slow response; a larger value A lower value (such as 0.3) makes the smooth curve closer to the original data, making the response more sensitive but less able to resist interference. For key parameters such as PM10, the system will dynamically adjust according to weather conditions. Values, for example, during a sandstorm warning period, The value will be increased appropriately to improve the system response speed.

[0055] Local features are extracted from the smoothed data time series, including PM10 / PM2.5 ratio, concentration change rate and wind field divergence calculation, to provide input data for the space-time vector autoregression model.

[0056] Step S2: The edge computing node collects and pre-processes the monitored parameter data in real time, wherein the parameter data include PM10, PM2.5 concentration, wind speed, wind direction, humidity, air pressure and visibility parameters; taking the parameter data into consideration, a dust source area-transmission-falling area correlation model is established to calculate the dust diffusion index and the dust source area activity index.

[0057] The expression of the dust source area-transmission-falling area correlation model is:

[0058] ;in, is the dust diffusion index; is the PM10 concentration; is the wind speed; For wind direction; is the direction of the target area relative to the monitoring point; is the relative humidity; is the pressure gradient; For visibility.

[0059] PM10 concentration directly reflects the content of sand and dust particles in the air; wind speed and wind direction Represents the effective wind force component to the target area; relative humidity term Reflects the degree of dryness. The lower the humidity, the more conducive it is to dust transmission. The pressure gradient It represents the atmospheric driving force, and the greater the gradient, the stronger the airflow; the visibility item VIS reflects the transparency of the atmosphere and is negatively correlated with the concentration of sand and dust; taking a city in a certain area as an example, when the SDI index of a monitoring station 500km away from this area exceeds 150 and lasts for more than 3 hours, the system will issue a sand and dust transmission warning, prompting relevant departments to make preventive preparations.

[0060] 、 、 、 、 is the weight coefficient, and each weight coefficient is determined according to the square ratio of its correlation coefficient with the historical dust event landing area: ,in, The correlation coefficients between each factor and the dust event areas in history were calculated by analyzing the historical data of 240 typical dust events in the past 10 years. The study found that the correlation coefficient between PM10 concentration and fall area is about 0.78, the correlation coefficient between effective wind component is about 0.65, the correlation coefficient between relative humidity is about -0.42, the correlation coefficient between pressure gradient is about 0.38, and the correlation coefficient between visibility is about -0.56.

[0061] The weight coefficients calculated based on this are: , , , , .

[0062] Using the Soil Moisture Index and vegetation cover index Constructing a dust source area activity index :

[0063] ;in, is the normalized soil moisture index, is the normalized vegetation cover index; is the wind erosion index.

[0064] The Normalized Soil Moisture Index (SWI) and the Vegetation Cover Index (VCI) are calculated from satellite remote sensing data (such as the SMAP soil moisture product and MODIS-NDVI data), respectively, and are seasonally corrected. When the soil is dry (SWI close to 0), vegetation is sparse (VCI close to 0), and wind speeds exceed the dust emission threshold, the DSAI value increases significantly, indicating that the source area has a high dust release potential. For example, in one region, the SWI typically has a value of 0.15 and the VCI typically has a value of 0.10 in spring. When wind speeds reach 8 m / s, the calculated WEI is 0.36 and the DSAI is approximately 0.27, indicating a highly active state and becoming a significant dust source.

[0065] when hour, ;

[0066] when , ;

[0067] in, The wind speed threshold for dust blowing is 6 m / s. When the wind speed is lower than this value, the wind force is insufficient to overcome the adhesion and gravity between soil particles and cannot cause dust blowing. It is applicable to typical sandy soil areas.

[0068] This threshold can be adjusted based on factors such as soil type, particle size distribution and surface roughness. For example, it can be lowered to 5m / s in fine sand areas and increased to 7m / s in coarse sand or gravel areas.

[0069] Step S3: Based on the calculation results of the correlation model, the improved space-time vector autoregressive model is used to predict the dust transmission path and landing area, and generate a dust landing area forecast product, which includes the landing area range, dust intensity level and arrival time.

[0070] According to the calculation results of the correlation model, the improved space-time vector autoregressive model is used to predict the dust transmission path and the range of the dust landing area. The dust transmission path is predicted by combining the SDI index and the DSAI index to calculate the transmission probability:

[0071] ,in, From the source area To the landing area The transmission probability, For historically source areas To the landing area The number of sandstorm events, Source area The total number of sandstorm events, Source area The dust diffusion index, is the average dust diffusion index of all regions, Source area The dust source area activity index, is the average dust source activity index of all regions.

[0072] The set of predicted points for the impact zone is obtained based on the dust feature vector. The calculation of the dust feature vector takes into account the SDI index and real-time monitoring parameter data. Its mathematical expression is:

[0073] ;in, For location In time The dust characteristic vector includes: SDI index, PM10 concentration and diffusion rate, The corresponding spatial location points constitute the set of predicted landing points , is the total number of predicted points in the landing area.

[0074] For location In time The dust feature vector of For location In time Meteorological characteristic vectors, including wind speed, wind direction and humidity; For location Neighborhood location In time The dust feature vector of For location The spatial neighborhood set of ; is the time autoregressive matrix, combined with the extracted local features, and estimated by the least squares method:

[0075] ; Represents the maximum time index in the dataset; is the exogenous variable influence matrix, which is estimated by the least squares method:

[0076] ; is the spatial correlation matrix, and its initial value is determined by the transmission probability Determine and update through iterative optimization; is the random error term; is the time autoregressive order, ranging from 1 to 3; is the lag order of the exogenous variable, ranging from 1 to 2.

[0077] Spatial correlation matrix Initial value of Expressed as:

[0078] ;in, From the position To its neighboring location transmission probability.

[0079] Spatial correlation matrix The iterative optimization update method uses the least squares method: ;in, For the The spatial correlation matrix of the iteration, is the learning rate, ranging from 0.01 to 0.05, is the loss function, defined as: ,in, is the observed dust feature vector, is the dust feature vector predicted by the model, Represents the Euclidean norm of a vector.

[0080] The gradient calculation formula is: ; When the iterative loss function change is less than the preset threshold (value is 0.001) or reaches the maximum number of iterations (value is 50), the iterative update is stopped.

[0081] When a new dust event is detected, the system will automatically trigger the matrix re-optimization process to adapt to the latest dust transmission characteristics.

[0082] The dust fall area forecast product uses probability contour lines to represent the fall area range, and is calculated using the SDI index using the kernel density estimation method: ;in, For location The probability density of dust impact is As the kernel function, the Gaussian kernel function is used: , is the bandwidth parameter, which is 1.5 times the average site spacing; the system generates four probability contour lines of 30%, 50%, 70% and 90%, representing areas that may, relatively likely, very likely and extremely likely be affected by sandstorms, respectively, providing a graded reference for disaster prevention and mitigation decisions.

[0083] The dust intensity level in the dust fall area forecast product is divided into four levels based on the SDI index and the PM10 concentration in the dust feature vector predicted by the model:

[0084] Mild level meets: Or PM10 concentration is ;

[0085] Moderate level satisfaction: Or PM10 concentration is ;

[0086] Severe level meets: Or PM10 concentration is ;

[0087] Severity level meets: Or PM10 concentration is .

[0088] Taking into account the varying sensitivities to dust in different regions, the dust intensity classification takes into account two key indicators: the SDI index and PM10 concentration. As long as one indicator reaches the corresponding threshold, the dust intensity is classified into the corresponding level. For example, in agricultural areas, even if PM10 concentrations are not particularly high, a high SDI index indicates that the dust has strong diffusion capacity and may still have a significant impact on crops. The four levels are mainly based on the degree of impact of dust on human health, visibility, transportation, and the ecological environment: light dust mainly affects sensitive groups; moderate dust causes discomfort to the general public, with visibility reduced to 2-5 kilometers; heavy dust significantly affects outdoor activities, with visibility reduced to 1-2 kilometers; and severe dust may lead to work and school suspensions, visibility below 1 kilometer, and may cause damage to infrastructure.

[0089] The arrival time in the dust fall area forecast product is based on the transmission path and transmission probability. Calculate the arrival time of dust The mathematical expression is: ,in, is the arrival time, D is the distance, is the effective transmission speed of dust, is the transmission probability influence coefficient, which is set to 0.5. is the historical average transmission probability, The calculation formula is:

[0090] ,in, is the average wind speed along the transmission path (km / h), is the average elevation difference of the transmission path (m).

[0091] Example 2

[0092] like Figure 2 The figure shows a schematic diagram of the composition of the sandstorm weather area forecasting system of the present invention, which includes: a distributed sandstorm monitoring sensor network, a data acquisition and processing module, and a sandstorm area forecast product generation module.

[0093] The distributed sand and dust monitoring sensor network includes source area monitoring equipment, transmission path monitoring equipment and potential landing area monitoring equipment composed of PM10, PM2.5 sensors, meteorological sensors and optical visibility sensors.

[0094] The distributed sand and dust monitoring sensor network consists of hundreds of monitoring stations, covering major sand and dust source areas, key transmission paths and important potential landing areas.

[0095] This system adopts a layered architecture design, and the data flow relationship between each module is as follows:

[0096] 1) Perception Layer: This layer consists of a distributed dust monitoring sensor network that collects data every five minutes and transmits it in real time to edge computing nodes via 4G / 5G networks or satellite communications. The system employs an adaptive sampling strategy, automatically increasing the sampling frequency to once per minute during dust events or warning periods, improving data timeliness.

[0097] 2) Edge Layer: This layer, comprised of edge servers deployed in each monitoring area, is responsible for receiving, preprocessing, and performing preliminary calculations on raw data. Edge nodes perform anomaly detection, smoothing, and local feature extraction on the data, then transmit the results to the cloud server every 10 minutes. The edge layer has a 72-hour data cache to ensure data is not lost in the event of a network outage.

[0098] 3) Cloud Layer: This layer, comprised of a high-performance computing cluster, is responsible for correlation model calculations, spatiotemporal autoregressive model training, and forecast product generation. The cloud layer updates the dust dispersion index and source activity index calculations hourly, and updates the transmission path and impact zone forecasts every three hours.

[0099] 4) Application Layer: Comprised of a visualization platform and a multi-channel warning push system, it provides sandstorm forecast products and warning services to a wide range of users. The application layer supports interactive user queries and can adjust the forecast display range, time step, and level of detail as needed. The overall system response time indicators are as follows: from monitoring data acquisition to edge preprocessing, the average time is no more than 30 seconds; from preprocessed data upload to the cloud to correlation model calculation, the average time is no more than 2 minutes; and from correlation model calculation to the generation of the impact zone forecast product, the average time is no more than 5 minutes. Under normal circumstances, the system's end-to-end response time does not exceed 10 minutes, meeting the timeliness requirements of sandstorm disaster warnings.

[0100] The system also incorporates a comprehensive fault-tolerance mechanism: if some monitoring equipment fails, the system automatically initiates a data interpolation algorithm to estimate missing values based on spatial correlation and historical data. If an edge computing node fails, an adjacent node automatically takes over its data processing tasks. If a partial failure occurs in the cloud service, the system switches to a backup server cluster to ensure business continuity. This multi-level fault-tolerance design enables stable 24 / 7 operation with an availability exceeding 99.9%.

[0101] The data acquisition module uses edge computing nodes to collect and preprocess monitored parameter data in real time. These parameter data include PM10 and PM2.5 concentrations, wind speed and direction, humidity, air pressure, and visibility. This data is then used to establish a dust source-transmission-fallout correlation model, which is used to calculate the dust diffusion index and the dust source activity index. The data acquisition and processing module utilizes a cloud-edge collaborative architecture, with edge computing nodes responsible for real-time data preprocessing and preliminary calculations, and cloud servers responsible for model training and complex computational tasks.

[0102] The sand and dust fall area forecast product generation module uses an improved space-time vector autoregressive model to predict the sand and dust transmission path and fall area range based on the calculation results of the correlation model, and generates a sand and dust fall area forecast product. The sand and dust fall area forecast product includes the fall area range, sand and dust intensity level and arrival time.

[0103] The dust fall forecast product generation module integrates Geographic Information System (GIS) technology, enabling the production of a variety of visualization products, including 2D and 3D plots and dynamic evolution animations, to meet the needs of diverse users. The system also provides multiple alert push services, including text messages, app notifications, emails, and specialized meteorological information terminals, ensuring timely delivery of warning information to decision-makers at all levels and the public.

[0104] To verify the technical effect of this invention, the research team conducted actual measurements on a typical sandstorm weather process that occurred from March 20 to 22, 2024.

[0105] This dust storm originated in a Gobi desert region and traveled through region A to region B. The system first detected anomalies in the source region at 6:00 PM on March 19th, with the SDI index reaching 185 and the DSAI index reaching 0.32. The system immediately initiated the forecast process. Using an improved space-time vector autoregressive model, the dust transmission probability (TP(region A, region B)) was calculated to be 0.87. The predicted impact zone covered region B, with a moderate to severe intensity level. It was expected to reach region B between 8:00 and 12:00 PM on March 21st.

[0106] Compared with traditional numerical models, the forecast results of the system of the present invention show significant advantages in the following aspects:

[0107] Forecast lead time: The traditional model only gave an effective forecast at 8:00 on March 20, but this system issued an early warning 24 hours in advance, buying valuable time for disaster prevention and mitigation;

[0108] Accuracy of impact zone: The predicted impact zone overlaps with the actual impact area by 87%, while the traditional model only overlaps by 62%.

[0109] Intensity forecast: The system predicted that the peak PM10 concentration in area B would be between 500 and 700 μg / m³. The actual observed value was 628 μg / m³, with an error of less than 10%. The traditional model predicted a value of 850 μg / m³, with an error of more than 35%.

[0110] Arrival time: This system predicts that the sandstorm will arrive in area B at around 10:00 on March 21. The actual time when the impact of the sandstorm was observed to intensify was 9:30 on March 21, with an error of only 30 minutes. The traditional model predicted that it would arrive at 14:00 on March 21, with an error of 4.5 hours.

[0111] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting sandstorm areas, characterized in that: The forecasting method is implemented based on a distributed dust monitoring sensor network system and includes the following steps: Step S1: Constructing a distributed dust monitoring sensor network system, the system including source area monitoring equipment, transmission path monitoring equipment, and potential fallout area monitoring equipment, which are composed of PM10 and PM2.5 sensors, meteorological sensors, and optical visibility sensors; Step S2: Using edge computing nodes to collect and pre-process monitored parameter data in real time, the parameter data includes PM10 and PM2.5 concentrations, wind speed, wind direction, humidity, air pressure, and visibility parameters; comprehensively considering the parameter data, a dust source area-transmission-fall area correlation model is established to calculate the dust diffusion index and the dust source area activity index; Step S3: Based on the calculation results of the correlation model, the improved space-time vector autoregressive model is used to predict the dust transmission path and landing area, and generate a dust landing area forecast product, which includes the landing area, dust intensity level and arrival time; According to the calculation results of the correlation model, the improved space-time vector autoregressive model is used to predict the dust transmission path and the range of the falling area, wherein the dust transmission path is predicted by combining the SDI index and the DSAI index to calculate the transmission probability: ,in, From the source area To the landing area The transmission probability, For historically source areas To the landing area The number of sandstorm events, Source area The total number of sandstorm events, Source area The dust diffusion index, is the average dust diffusion index of all regions, Source area The dust source area activity index, is the average dust source activity index of all regions; The set of predicted points for the impact zone is obtained based on the dust feature vector. The calculation of the dust feature vector takes into account the SDI index and real-time monitoring parameter data. Its mathematical expression is: ;in, For location In time The dust characteristic vector includes: SDI index, PM10 concentration and diffusion rate, The corresponding spatial location points constitute the set of predicted landing points , is the total number of predicted points in the landing area; For location In time The dust feature vector of For location In time Meteorological characteristic vectors, including wind speed, wind direction and humidity; For location Neighborhood location In time The dust feature vector of For location The spatial neighborhood set of ; is the time autoregressive matrix, combined with the extracted local features, and estimated by the least squares method: ; Represents the maximum time index in the dataset; is the exogenous variable influence matrix, which is estimated by the least squares method: ; is the spatial correlation matrix, and its initial value is determined by the transmission probability Determine and update through iterative optimization; is the random error term; is the time autoregressive order, ranging from 1 to 3; is the lag order of the exogenous variable, ranging from 1 to 2; Spatial correlation matrix Initial value of Expressed as: ;in, From the position To its neighboring location The transmission probability of Spatial correlation matrix The iterative optimization update method uses the least squares method: ;in, For the The spatial correlation matrix of the iteration, is the learning rate, ranging from 0.01 to 0.05, is the loss function, defined as: ,in, is the observed dust feature vector, is the dust feature vector predicted by the model, represents the Euclidean norm of a vector; The dust fall area forecast product uses probability contour lines to represent the fall area range, and is calculated using the SDI index using the kernel density estimation method: ;in, For location The probability density of dust impact is As the kernel function, the Gaussian kernel function is used: , The bandwidth parameter is 1.5 times the average site spacing.

2. The sandstorm weather forecasting method according to claim 1, characterized in that: Edge computing preprocessing methods include: data time series smoothing; The exponentially weighted moving average (EWMA) algorithm is used to smooth the data time series. ;in, for The smoothed value at time, for The original value of the moment, for The smoothed value at time, is the smoothing coefficient, and the smoothing coefficient ranges from 0.1 to 0.

3.

3. The method for predicting sandstorm areas according to claim 2, characterized in that: The expression of the dust source area-transmission-falling area correlation model is: ;in, is the dust diffusion index; is the PM10 concentration; is the wind speed; For wind direction; is the direction of the target area relative to the monitoring point; is the relative humidity; is the pressure gradient; For visibility; 、 、 、 、 is the weight coefficient, and each weight coefficient is determined according to the square ratio of its correlation coefficient with the historical dust event landing area: ,in, is the correlation coefficient between each factor and the location of historical dust events; Using the Soil Moisture Index and vegetation cover index Constructing a dust source area activity index : ;in, is the normalized soil moisture index, is the normalized vegetation cover index; is the wind erosion index; when hour, ; when , ; in, is the wind speed threshold for dust-raising, which is 6m / s.

4. The method for predicting sandstorm areas according to claim 3, characterized in that: The dust intensity level in the dust fall area forecast product is divided into four levels based on the SDI index and the PM10 concentration in the dust feature vector predicted by the model: Mild level meets: Or PM10 concentration is ; Moderate level satisfies: Or PM10 concentration is ; Severe level meets: Or PM10 concentration is ; Severity level meets: Or PM10 concentration is .

5. The sandstorm weather forecasting method according to claim 4, characterized in that: The arrival time in the dust fall area forecast product is based on the transmission path and transmission probability. Calculate the arrival time of dust The mathematical expression is: ,in, is the arrival time, D is the distance, is the effective transmission speed of dust, is the transmission probability influence coefficient, which is set to 0.

5. is the historical average transmission probability, The calculation formula is: ,in, is the average wind speed along the transmission path, is the average elevation difference of the transmission path.

6. A sandstorm forecasting system for executing the sandstorm forecasting method according to any one of claims 1 to 5, characterized in that: The forecast system includes: a distributed dust monitoring sensor network, a data acquisition and processing module, and a dust fall area forecast product generation module; The distributed dust monitoring sensor network includes source area monitoring equipment, transmission path monitoring equipment and potential fallout area monitoring equipment composed of PM10, PM2.5 sensors, meteorological sensors and optical visibility sensors; The data acquisition and processing module collects and pre-processes the monitored parameter data in real time through the edge computing node. The parameter data includes PM10 and PM2.5 concentrations, wind speed, wind direction, humidity, air pressure, and visibility parameters. Taking the parameter data into consideration, a dust source area-transmission-fall area correlation model is established to calculate the dust diffusion index and the dust source area activity index. The sand and dust fall area forecast product generation module uses an improved space-time vector autoregressive model to predict the sand and dust transmission path and fall area range based on the calculation results of the correlation model, and generates a sand and dust fall area forecast product. The sand and dust fall area forecast product includes the fall area range, sand and dust intensity level and arrival time.

Images