A method for short-time prediction of PM2.5 concentration changes based on GNSS tropospheric delay

Abstract

This invention discloses a method for predicting PM2.5 concentration changes based on GNSS tropospheric delay short-time (ZTD). First, a GNSS double-difference mode receiver receives GNSS satellite signals and calculates CORS station data to obtain the ZTD of each ground-based GNSS station within the observation area. Then, wavelet analysis is used to perform multi-scale signal reconstruction and abrupt change analysis on the ZTD. Wavelet analysis is also used to reconstruct meteorological and environmental data observed by the IGS stations. Finally, using the reconstructed PM2.5 concentration change as the dependent variable and the reconstructed ZTD, relative humidity, average wind speed, and NO2 concentration as independent variables, a multiple regression model for PM2.5 concentration is established using multiple regression analysis. This model is then used to predict PM2.5 concentration changes. This invention addresses the problem of uneven spatial distribution and limited number of meteorological stations, which hinders real-time monitoring and forecasting of haze, providing an important reference for haze weather forecasting.

Description

Technical Field

[0001] This invention relates to the field of haze monitoring technology, specifically to a method for short-time prediction of PM2.5 concentration changes based on GNSS tropospheric delay. Background Technology

[0002] The monitoring and forecasting of haze has become a crucial issue in meteorology. Although weather stations can obtain relatively accurate haze data, their uneven spatial distribution and limited number make real-time monitoring and forecasting difficult. Therefore, adopting appropriate methods for monitoring haze is urgently needed. With the rapid development of Global Navigation Satellite Systems (GNSS), the limitations of traditional water vapor detection technologies in terms of spatiotemporal resolution have been overcome. GNSS can provide large-scale, real-time, and high-precision water vapor distribution information for meteorological and disaster monitoring. Tropospheric delay inversion based on ground-based GNSS systems is being actively applied to haze monitoring and forecasting.

[0003] When haze occurs, the increase in particulate matter in the air leads to changes in water vapor content and humidity, which in turn causes changes in GNSS tropospheric delay. This makes it possible to use tropospheric delay to assist in monitoring and predicting haze weather changes. Currently, GNSS zenith tropospheric delay is still in its early stages in haze monitoring and forecasting. Researchers have only pointed out a strong correlation between zenith tropospheric delay (ZTD) and haze at a macroscopic level, mainly using traditional methods to analyze the relationship between ZTD and haze. While there are temporal and frequency spatial distribution characteristics between ZTD and haze (mainly the harmful PM2.5), the original sequences of both fluctuate significantly and are subject to noise interference. Traditional methods cannot detect the multi-scale characteristics of their evolution. Wavelet analysis, with its multi-resolution characteristics, can satisfy the analysis of the correlation and variation patterns between the two in the time and frequency domains, providing an important reference for haze weather forecasting. Summary of the Invention

[0004] The purpose of this invention is to provide a method for short-time prediction of PM2.5 concentration changes based on GNSS tropospheric delay, which can solve the problem that the current meteorological stations are spatially uneven and limited in number, making it difficult to achieve real-time monitoring and forecasting of haze.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A method for short-time prediction of PM2.5 mass concentration changes based on GNSS tropospheric delay includes the following steps:

[0007] Step 1: Receive GNSS satellite signals using a ground-based GNSS dual-difference mode receiver;

[0008] Step 2: Using the ground-based GNSS observation data from Step 1, calculate the CORS station data to obtain the zenith tropospheric delay (ZTD) of each ground-based GNSS station within the observation area, and simultaneously acquire meteorological and environmental data near the CORS station.

[0009] Step 3: Use the internal consistency accuracy and average deviation to check the accuracy of the zenith tropospheric delay ZTD obtained in Step 2. If the accuracy meets the requirements, proceed to Step 4; otherwise, return to Step 2 and recalculate.

[0010] Step 4: Use wavelet analysis to perform multi-scale, signal reconstruction, and abrupt change analysis on the zenith tropospheric delay (ZTD) obtained in Step 2;

[0011] Step 5: Reconstruct the meteorological and environmental data from the CORS station in Step 2 using wavelet analysis techniques;

[0012] Step 6: Using the reconstructed PM2.5 mass concentration change as the dependent variable, and the reconstructed ZTD, relative humidity, average wind speed, and NO2 concentration as independent variables, establish a multiple regression model for PM2.5 mass concentration using multiple regression analysis. Use this model to predict PM2.5 concentration changes.

[0013] Preferably, in step 2, the zenith tropospheric delay (ZTD) of each ground-based GNSS station within the observation area is obtained by using ground-based GNSS observation data, combined with observation data from distant IGS stations, and employing GAMIT software in a double-difference mode to solve the CORS station data.

[0014] Preferably, the IGS station observation data includes historical zenith tropospheric delay (ZTD), meteorological data and environmental data within the corresponding observation area, and the CORS station meteorological data and environmental data are sourced from CORS station data in various provinces, cities and other regions.

[0015] Preferably, the distance between the long-distance IGS station and the detection station is greater than 500km.

[0016] Preferably, in step 2, the Zenith tropospheric delay (ZTD) is obtained using ground-based GNSS observation data and precise ephemeris data for the corresponding time, employing the PPP mode.

[0017] Preferably, in steps 4 and 5, the db5 wavelet is used to perform sequence analysis on the meteorological and environmental data of ZTD and CORS stations, and the signal is reconstructed in the low-frequency coefficients of the fourth layer.

[0018] Preferably, in step 5, the meteorological data of the CORS station includes hourly relative humidity, average wind speed, and NO2 concentration within the observation area, while the environmental data includes hourly PM2.5 data within the observation area.

[0019] Preferably, the multiple regression model established in step 6 is as follows:

[0020] Let NO2 concentration be X1, relative humidity be X2, average wind speed be X3, ZTD data be X4, and PM2.5 mass concentration be Y. Using multiple regression analysis, the following multiple regression model for PM2.5 mass concentration is established:

[0021] Y = a0 + a1X1 + a2X2 + a3X3 + a4X4 + ε (1)

[0022] Wherein, parameters a0, a1, a2, a3, and a4 are the regression coefficients to be estimated, and ε is the random error (which roughly follows a normal distribution with a mean of 0). The above parameters are set according to actual usage requirements.

[0023] This invention utilizes the relationship between the zenith tropospheric delay (ZTD), a major error source in GNSS during haze, and haze. It employs standard orthogonal wavelet db5 analysis to reconstruct PM2.5, ZTD, and closely related parameters such as relative humidity, average wind speed, and NO2 using fourth-layer low-frequency coefficients. A multivariate regression model is then established using the reconstructed PM2.5, ZTD, relative humidity, average wind speed, and NO2 sequences to predict changes in PM2.5 mass concentration. Wavelet analysis effectively removes high-frequency noise and subtle disturbances, improving prediction accuracy. This invention addresses the problem of uneven spatial distribution and limited number of meteorological stations, hindering real-time monitoring and forecasting of haze, and provides an important reference for haze weather forecasting. Attached Figure Description

[0024] Figure 1 is a flowchart of Embodiment 1 of the present invention;

[0025] Figure 2 is a flowchart of Embodiment 2 of the present invention. Detailed Implementation

[0026] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0027] Example 1

[0028] As shown in Figure 1, this invention discloses a method for short-time prediction of PM2.5 concentration changes based on GNSS tropospheric delay, which includes the following steps:

[0029] Step 1: Receive GNSS satellite signals using a ground-based GNSS dual-difference mode receiver.

[0030] Step 2: Using the ground-based GNSS observation data from Step 1, combined with the observation data from the distant IGS station, the CORS station data is solved using the GAMIT software in a double-difference mode to obtain the zenith tropospheric delay (ZTD) of each ground-based GNSS station in the observation area. At the same time, meteorological and environmental data near the CORS station are also acquired.

[0031] The IGS station observation data includes historical zenith tropospheric delay (ZTD), meteorological data and environmental data within the corresponding observation area. The IGS station is more than 500 km away from the monitoring station. The meteorological and environmental data near the CORS station are from meteorological bureaus and environmental protection bureaus of various provinces and cities. The process of solving the CORS station data using the double-difference model is existing technology and will not be described in detail here.

[0032] When using the double-difference mode for data processing, common errors between different stations and different satellites can be eliminated. These common errors include receiver clock errors and satellite clock errors. At the same time, satellite orbit errors and tropospheric delay errors are also greatly reduced. Therefore, precise clock error information is not required, and the results can achieve high accuracy. Especially when the distance between stations is short, there is a strong correlation between the tropospheric delays between stations. In this case, only the relative tropospheric delay between stations can be calculated. Therefore, it is necessary to introduce several long-distance reference stations to reduce the correlation between stations. Typically, observation data from long-distance IGS stations are used to estimate the absolute tropospheric delay.

[0033] Step 3: Use the internal conformity accuracy (±1.3mm) and average deviation (7mm) to check the accuracy of the zenith tropospheric delay ZTD obtained in Step 2. If the accuracy meets the requirements, proceed to Step 4. If it does not meet the requirements, return to Step 2 and recalculate the zenith tropospheric delay ZTD for each station of the ground-based GNSS.

[0034] The process of verifying the accuracy of the Zenith Tropospheric Delay Data (ZTD) is existing technology and will not be described in detail here.

[0035] Step 4: Use wavelet analysis to perform multi-scale, signal reconstruction, and abrupt change analysis on the zenith tropospheric delay (ZTD) obtained in Step 2;

[0036] The zenith tropospheric delay (ZTD) of the original sequence and the original sequence of PM2.5, the main component of haze, fluctuate significantly and are subject to noise interference, making it impossible to detect the multi-scale characteristics of their evolution. Wavelet analysis, with its multi-resolution characteristics, can satisfy the analysis of their correlation and variation patterns in the time and frequency domains.

[0037] Reconstruction uses wavelet coefficients obtained from decomposition at multiple resolutions to synthesize the original signal from multi-scale wavelets, involving a denoising process. This is equivalent to filtering; selecting appropriate components for reconstruction can filter out other components. Alteration analysis extracts high-frequency coefficients from a specific layer after wavelet decomposition and determines the abrupt change points of ZTD by testing the singularity of these high-frequency coefficients. After considering the overall algorithm performance, the db5 wavelet is the optimal choice for ZTD and PM2.5 sequence analysis. The wavelet low-frequency signal is used to reconstruct the ZTD and PM2.5 sequences from the fourth layer of low-frequency coefficients.

[0038] Step 5: Reconstruct the meteorological and environmental data from the CORS station in Step 2 using wavelet analysis techniques;

[0039] CORS station meteorological data includes hourly relative humidity, average wind speed, and NO2 concentration within the observation area, while environmental data includes hourly PM2.5 data within the observation area. For provincial and municipal CORS stations, the meteorological and environmental data come from the meteorological bureau and the environmental protection bureau, respectively.

[0040] The process of reconstructing meteorological and environmental data is the same as step 4, which is existing technology and will not be described again.

[0041] Step 6: Using the reconstructed PM2.5 mass concentration change as the dependent variable, and the reconstructed ZTD, relative humidity, average wind speed, and NO2 concentration as independent variables, establish a multiple regression model for PM2.5 mass concentration using multiple regression analysis techniques, and use this model to predict PM2.5 concentration changes.

[0042] Let NO2 concentration be X1, relative humidity be X2, average wind speed be X3, ZTD data be X4, and PM2.5 mass concentration be Y. Using multiple regression analysis, the following multiple regression model for PM2.5 mass concentration is established:

[0043] Y = a0 + a1X1 + a2X2 + a3X3 + a4X4 + ε (1)

[0044] Wherein, parameters a0, a1, a2, a3, and a4 are the regression coefficients to be estimated, and ε is the random error (which roughly follows a normal distribution with a mean of 0). The above parameters are set according to actual usage requirements.

[0045] Example 2

[0046] like Figure 2 As shown, this embodiment is largely the same as embodiment one, except that in step 2, the Zenith tropospheric delay (ZTD) is obtained using PPP mode by combining ground-based GNSS observation data with the corresponding time precise ephemeris.

[0047] The advantages of choosing the PPP mode are that the observation stations are independent of each other, and the data processing results can directly obtain the absolute tropospheric delay over each observation station. However, its disadvantages are that in order to minimize the influence of other errors and meet the requirements of high-precision data processing, it is necessary to provide high-precision ephemeris and precise clock difference information of each satellite in advance, which increases the difficulty of real-time data processing.

[0048] The results of the regression model test and the PM2.5 concentration prediction curve demonstrate that the monitoring accuracy using this method is superior to traditional methods, especially when the number of days is shorter and there is a strong correlation between ZTD and PM2.5, the prediction results are better. This method has also achieved good results in predicting haze in Beijing and Changchun (selecting any 3-5 days).

[0049] This invention utilizes the multi-resolution characteristics of wavelet analysis to analyze the temporal correlation and variation patterns of ZTD and PM2.5 sequences, providing an important reference for the forecasting of haze weather. Wavelet analysis can eliminate high-frequency noise and subtle disturbances, thereby improving the accuracy of prediction.

Claims

1. A method for predicting the change of Beijing PM2.5 mass concentration based on GNSS troposphere delay short time prediction, characterized in that, Comprise the following steps in turn: Step 1, using ground-based GNSS double difference mode processing receiver received GNSS satellite signals; Step 2, using the observation data of ground-based GNSS in step 1, combined with the precise ephemeris of corresponding time, using PPP mode to obtain zenith tropospheric delay ZTD; At the same time, obtain the meteorological data and environmental protection data near the CORS station; Step 3, using the precision of internal compliance and average deviation to test the precision of zenith tropospheric delay ZTD obtained in step 2, if the precision meets the requirements, then enter step 4, if it does not meet the requirements, then return to step 2 to recalculate; Step 4, using db5 wavelet analysis technology to carry out multi-scale, signal reconstruction and mutation analysis on the zenith tropospheric delay ZTD obtained in step 2, and signal reconstruction is carried out in the 4th layer low frequency coefficient; Step 5, using db5 wavelet analysis technology to reconstruct the meteorological data and environmental protection data of CORS station in step 2, signal reconstruction is carried out in the 4th layer low frequency coefficient; The meteorological data of CORS station is the relative humidity, average wind speed and NO2 concentration in the observation area every hour, and the environmental protection data is the PM2.5 data in the observation area every hour; Step 6, taking the change of PM2.5 mass concentration after reconstruction as the dependent variable, taking the reconstructed ZTD, relative humidity, average wind speed and NO2 concentration as the independent variable, using multiple regression analysis technology, establishing a multiple regression model of PM2.5 mass concentration, using the model to predict the concentration change of PM2.5; The multiple regression model is as follows: NO2 concentration is recorded as X 1. Relative humidity is X 2. The average wind speed is X 3. ZTD data is X 4. PM2.5 mass concentration is Y , using multiple regression analysis techniques, the multiple regression model of PM2.5 mass concentration is as follows: Y = a0+ a1 X 1 + a2 X 2 + a3 X 3 + a4 X 4 + ε (1) Wherein, the parameters a0, a1, a2, a3, a4 are regression coefficients to be estimated, ε is a random error, and the above parameters are set according to actual use requirements.

Images