A method for predicting long-term change of Antarctic ozone based on polar vortex morphology constraint

Abstract

The application discloses a kind of based on polar vortex form constraint Antarctic ozone long-term change prediction method, comprising: using artificial intelligence method, based on preprocessed reanalysis data and satellite data, the missing part in ozone station observation data is filled, and complete ozone data sequence is obtained;According to the preprocessed reanalysis data, the climate average state of the polar vortex boundary on 530K isentropic surface is determined;Based on the climate average state of polar vortex form, the complete ozone data sequence is screened, and effective ozone observation data are obtained;Based on effective ozone observation data, a nonlinear regression model is constructed to predict the long-term trend of Antarctic ozone change.The application fills in missing data by artificial intelligence, screens effective observation according to polar vortex form, and constructs nonlinear regression model, which significantly improves the accuracy and reliability of Antarctic ozone long-term change prediction.

Description

Technical Field

[0001] This invention belongs to the field of Antarctic ozone technology, and particularly relates to a method for predicting long-term changes in Antarctic ozone based on polar vortex morphology constraints. Background Technology

[0002] The ozone layer effectively absorbs harmful ultraviolet radiation from the sun, acting as a protective shield for life on Earth. However, since the 1960s and 70s, industrial emissions of substances such as chlorofluorocarbons (CFCs) have severely damaged the ozone layer, particularly in Antarctica. Although the international community adopted relevant treaties in the 1980s to strengthen the control of ozone-depleting substances, signs of recovery only began to appear at the beginning of this century. When the ozone layer will recover to its pre-depletion levels has become a major concern for the international community.

[0003] Field observation data from Antarctic stations are primary sources for studying long-term changes in Antarctic ozone, and their accuracy is higher compared to other data. However, the harsh observation conditions in Antarctica and the often unstable instruments result in gaps in data collection, leading to inconsistencies and a lack of data continuity. Filling in these gaps using scientific methods would significantly improve the completeness of the data from Antarctic stations.

[0004] Secondly, the distribution of Antarctic ozone is closely related to the morphology of the stratospheric polar vortex. The strong westerly jet stream at the boundary of the polar vortex hinders ozone transport to the pole, resulting in a much lower total ozone column inside the vortex compared to the outside. The dynamically changing polar vortex causes Antarctic stations to be sometimes inside and sometimes outside the vortex, causing significant fluctuations in their observational data. Such raw observation sequences require validity screening before they can be used for long-term ozone variation analysis; otherwise, false analytical results may be obtained.

[0005] Statistical modeling based on regression equations is a common method for predicting ozone recovery. However, the linear form of the regression equation, or unreasonable selection of independent and dependent variables, has prevented previous studies from yielding ideal predictive results. Therefore, further improvements are needed to enhance the accuracy of the models. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a method for predicting long-term Antarctic ozone variations based on polar vortex morphology constraints, thereby resolving the issues present in the existing technologies.

[0007] To achieve the above objectives, this invention provides a method for predicting long-term Antarctic ozone variations based on polar vortex morphology constraints, comprising: Preprocessing of multi-source ozone data; the multi-source ozone data includes ozone observation data from Antarctic stations, reanalysis data, and satellite data; Using artificial intelligence methods, the missing parts of ozone station observation data were filled in based on preprocessed reanalysis data and satellite data to obtain a complete ozone data sequence; Based on the preprocessed reanalysis data, the average climate state of the polar vortex boundary on the 530K isentropic surface was determined. Based on the climate average state, the complete ozone data sequence was screened, and the observation data when the station was located outside the polar vortex were removed to obtain effective ozone observation data. A nonlinear regression model was constructed based on the effective ozone observation data, and the long-term trend of Antarctic ozone variation was predicted based on the nonlinear regression model.

[0008] Optionally, preprocessing of ozone observation data, reanalysis data, and satellite data from Antarctic stations may include: converting the units of total ozone column to Dobson units and converting the units of geopotential vorticity to geopotential vorticity units.

[0009] Optionally, the process of obtaining a complete ozone data sequence includes: Based on the latitude and longitude of the Antarctic station, the time series of total ozone column and geopotential vorticity at the corresponding location are extracted from the preprocessed multi-source ozone data. Linear interpolation was performed on the ozone observation data, and correlation analysis was used to retain parameter sequences with significant correlations. A decision tree model was then constructed to filter out all feature input parameters. After standardizing all the selected feature input parameters, they are input into the long short-term memory neural network model for training. The model output is then de-standardized to generate a complete time series of total ozone column. The complete ozone column total time series is matched with the ozone observation data to fill in the missing parts and obtain a complete ozone data series.

[0010] Optionally, the process of determining the climatological mean state of the polar vortex boundary on the 530K isentropic surface based on preprocessed reanalysis data includes: Interpolate the geopotential vorticity data on the isobaric surface to the 530K isentropic surface; perform dimensional correction on the interpolated geopotential vorticity data; calculate the equivalent latitude of each contour line in the corrected geopotential vorticity field; determine the daily polar vortex boundary based on the gradient of geopotential vorticity with the equivalent latitude; calculate the average value of the daily geopotential vorticity value of the polar vortex boundary over historical periods as the climate average state.

[0011] Optionally, the process of obtaining effective ozone observation data includes: comparing daily geopotential vorticity values ​​with the climate average state to generate a binary sequence that identifies the station's position relative to the polar vortex; and based on the binary sequence, performing masking processing on the complete ozone data sequence to retain data where the station is located inside the polar vortex.

[0012] Optionally, a nonlinear regression model is constructed using the effective ozone observation data as the dependent variable and the stratospheric equivalent chlorine and polar stratospheric cloud generation potential as independent variables; the future data of the stratospheric equivalent chlorine and polar stratospheric cloud generation potential are substituted into the nonlinear regression model to predict the long-term trend of Antarctic ozone.

[0013] Optionally, the stratospheric equivalent chlorine is a time series calculated based on an emission inventory of ozone-depleting substances.

[0014] Optionally, the polar stratospheric cloud generation potential is a time series calculated based on stratospheric temperature and potential vorticity data, as shown in the following formula: ; in, The volume of polar stratospheric clouds. Let be the volume of the polar vortex.

[0015] Optionally, the nonlinear regression model is: ; This is the monthly average time series of the total ozone column from the valid ozone observation data. For EESC time series, For PFP time series, to For regression coefficients, This is the error term.

[0016] Compared with the prior art, the present invention has the following advantages and technical effects: This invention scientifically fills in missing data from Antarctic ozone stations by introducing artificial intelligence methods, effectively solving the problem of incomplete observational data sequences from Antarctic stations. Secondly, this invention combines polar vortex morphology to screen the filled station data. Before and after screening, the average total ozone column difference across multiple stations was only 29 ± 13.5 DU, fully ensuring the validity of the Antarctic ozone data. Furthermore, this invention employs a nonlinear regression model, using new factors EESC and PFP as independent variables for the final prediction.

[0017] This invention combines artificial intelligence-based data filling, polar vortex morphology constraint elimination, and nonlinear retrospective models to construct a complete system from "data filling – data screening – data prediction." This system significantly improves the accuracy of ozone data analysis at Antarctic stations and the predictive power of future data. Specifically, the results show that the average recovery time of total ozone column at multiple stations is delayed by approximately 15 years compared to before screening, effectively improving the reliability of Antarctic ozone recovery predictions. Attached Figure Description

[0018] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a technical flowchart of an embodiment of the present invention; Figure 2 This is a schematic diagram showing the relationship between the polar vortex morphology in October 2013, 2019, and 2022 and the location distribution of Faraday Station, Halley Station, and Showa Station, according to an embodiment of the present invention. Figure 3 This is a comparison chart of the regression prediction of total ozone column at Antarctic stations before and after polar vortex morphology constraints in an embodiment of the present invention. Detailed Implementation

[0019] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0021] Example 1 To improve the completeness of ozone observation sequences at Antarctic stations, the effectiveness of observation data, and the accuracy of prediction models, this invention employs artificial intelligence methods to fuse multi-source satellite and reanalysis data to intelligently fill in missing data. In addition, polar vortex morphology constraints are introduced to screen effective station observation data, and based on an improved nonlinear regression model, more accurate predictions of Antarctic ozone recovery are obtained.

[0022] First, to address the issue of missing data from Antarctic stations, we introduced an artificial intelligence algorithm to intelligently fill in the gaps by fusing multi-source observational data. Second, we performed validity screening on the station observational data based on the polar vortex morphology at the highest level of stratospheric ozone concentration (530K isentropic surface). Building upon this, we constructed a nonlinear regression model using equivalent effective stratospheric chlorine (EESC) and polar stratospheric cloud production potential (PFP) as independent variables to predict the recovery of Antarctic ozone.

[0023] like Figure 1As shown, this embodiment provides a method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints, including: S1. Data preprocessing; S2. Artificial intelligence methods fill in ozone observation data at Antarctic stations; S3. Construct the climatological distribution of polar vortex morphology on the 530K isentropic surface; S4. Based on the distribution of the polar vortex boundary, remove the observation data when the station is located outside the polar vortex; S5. Use the processed Antarctic ozone station observation data to construct a regression model; S6. Regression prediction of Antarctic ozone variation trend.

[0024] In step S1, appropriate preprocessing operations are performed on the data. The specific preprocessing operations are as follows: S101. Standardize the units of the collected station data, reanalysis data, and satellite data. Specifically, convert the TCO3 (Total Column Ozone) units of ERA-5, JRA-55, MERRA-2, and NIWA-BS to DU; convert the PV (Potential Vorticity) units of ERA-5 data to PVU. .

[0025] In step S2, due to the harsh natural conditions in Antarctica, ozone data sequences from Antarctic stations often contain missing data. Therefore, we consider using artificial intelligence methods to integrate satellite data and reanalysis data (including ERA-5, JRA-55, MERRA-2, and NIWA-BS) released by multiple meteorological centers to fill in the missing parts of the ozone observation data from Antarctic stations. The specific steps are as follows: S201. Preprocess the reanalysis data and satellite data. Based on the latitude and longitude coordinates of the observation stations, extract the time series of TCO3 parameters at the corresponding grid points from the ERA-5, JRA-55, MERRA-2 and NIWA-BS data. In addition, extract the time series of geopotential vorticity (PV) parameters at the location from the ERA5 data.

[0026] S202. Perform linear interpolation preprocessing on the ozone observation sequence of the Antarctic station, calculate its correlation coefficient with each parameter sequence described in S201, retain the parameter sequences with significant correlation, and construct a decision tree model to screen out all feature input parameters.

[0027] S203. Standardize each of the selected parameters and use them as features to train the LSTM (Long Short-Term Memory) model to obtain the model's prediction results. Then, perform destandardization to generate a new TCO3 time series.

[0028] S204. The new TCO3 time series is matched one-to-one with the original observation data of the Antarctic ozone station according to the time axis. The missing parts in the original station data are replaced with the newly generated TCO3 time series, and finally combined to form the restored Antarctic ozone station observation data series.

[0029] In step S3, the climatological distribution of the polar vortex boundary on the 530K isentropic surface is calculated using ERA-5 reanalysis data. The specific steps are as follows: S301. If PV data on the isentropic surface cannot be obtained directly, the PV data of the isobaric surface needs to be interpolated onto the 530K isentropic surface. First, according to the PV calculation formula: ...(1) Where g is the acceleration due to gravity, f is the entrainment vorticity, ζ is the relative vorticity, θ is the potential temperature, and p is the air pressure. The PV values ​​at the isobaric grid points are calculated; the unit of PV is PHU. Then, the potential temperature of unsaturated air is calculated using the formula: ...(2) Calculate the potential temperature at the grid points of the isobaric surface. It is known that the potential temperature increases monotonically with altitude. For each latitude and longitude grid point (taking grid point (a, b) as an example), find the two adjacent isobaric layers j and j+1 such that the lower j layer's θ is less than or equal to 530K and the higher j+1 layer's θ is greater than or equal to 530K. Then, use the interpolation formula: ...(3) The PV value on the 530K isentropic surface is obtained by combining the PV calculation results of layers j and j+1.

[0030] S302 and PV increase rapidly with altitude. To keep the potential vorticity of different characteristic layers on the same order of magnitude, the PV of each altitude layer needs to be corrected using the following formula, which is expressed as follows: ... (4) in .

[0031] S303. After correcting the PV, first divide the PV contour lines according to a 0.25PVU spacing, then calculate the area defined by each PV contour line, find the latitude line that defines the same area, and the latitude of this latitude line is the equivalent latitude of this PV contour line. Finally, obtain the equivalent latitude of all PV contour lines.

[0032] S304. Based on the distribution relationship between PV value and equivalent latitude, find the location where the PV value changes the fastest with the equivalent latitude (i.e., the location with the maximum gradient). This location is the polar vortex boundary. The corresponding PV contour line can be used to represent the polar vortex boundary on a certain day. The PV value at the polar vortex boundary is denoted as eg.

[0033] S305. Apply the steps of S302-S304 above to the daily isentropic surface PV data for June, July and August of 1979-2022 in the ERA-5 reanalysis dataset, average the PV values ​​at the polar vortex boundary during this period, and use the average PV contour lines to represent the climate-averaged polar vortex morphology. The PV value at the climate-averaged polar vortex boundary is denoted as egm, which is -28.8PVU in this embodiment.

[0034] In step S4, referring to the polar vortex boundary calculated in S3, the observation results from the Faraday, Showa, and Halley ozone observation data where the stations were located outside the polar vortex for a period of time are removed. The specific steps are as follows: S401. First, the correction method in S301 is applied to the daily PV dataset of the ERA-5 isentropic surface to keep the PV at the same order of magnitude. Then, referring to the climatological results (egm, -28.8PVU) of the polar vortex boundary on the 530K isentropic surface calculated in S3, for each station, the PV time series of the 530K isentropic surface at the nearest latitude and longitude grid point is extracted. Since the PV values ​​in the Southern Hemisphere are all negative, this embodiment assigns a value of 1 to time points where the PV value is less than or equal to egm, indicating that the station is inside the polar vortex at that time. Taking Halley station as an example, its PV value is... Figure 2 The three time periods ((a) October 2013, (b) October 2019, and (c) October 2022) are all within the polar vortex, and therefore are assigned a value of 1; time points with PV values ​​greater than egm are assigned a value of 0, indicating that the station is outside the polar vortex at that time, for example... Figure 2 (b) shows the Faraday and Showa stations in October 2019. Ultimately, each station will obtain a unique 1 / 0 time series.

[0035] S402. Using the polar vortex 1 / 0 dataset generated in S401, data from each station is removed. Specifically, the station data for the corresponding time period is multiplied by 1 / 0; if multiplied by 0, it is removed; if multiplied by 1, the original data is retained. This operation is then applied to the station observation data generated after S2, removing observation data from time periods when the station is outside the polar vortex. The final result is the Antarctic station ozone observation data after these removals.

[0036] In step S5, the ozone observation data from Antarctic stations after being removed, generated in S4, is used as the dependent variable y, and the calculation result of EESC is used as the independent variable. The calculation results of PFP were used as independent variables. Construct a regression model. The specific steps are as follows: S501. The Antarctic ozone station observation data generated in S4, after being filtered out, are processed into monthly averages. Specifically, the daily TCO3 data for October of each year are selected and averaged to obtain the monthly average time series of TCO3. Data from 1960 to 2022 are selected from this series as the historical data of the dependent variable y.

[0037] S502. Calculate the EESC time series from 1960 to 2100 using the ODS emission inventory provided in the 2022 Scientific Assessment Report on Ozone Depletion. Select the EESC index from 1960 to 2022 in the calculation results as... Historical data. The formula for calculating EESC is: …… (5) in, These are the number of atoms, halogen release efficiency, and mixing ratio of a specific ODS substance.

[0038] S503. The PFP index for the period 1960–2022 was calculated using stratospheric temperature and geopotential vorticity data from the ERA5 dataset. Historical data. The formula for calculating PFP is: ... (6) S504. Based on the data generated from S501, S502, and S503, according to the regression equation: …(7) A regression model was constructed and fitted to obtain the final regression model used to predict Antarctic ozone recovery data.

[0039] In step S6, the future EESC and PFP calculation results are substituted into the regression equation established in S5 to calculate the future TCO3 prediction result. The specific steps are as follows: S601. Based on the ODS emission inventory provided by the 2022 Ozone Depletion Scientific Assessment, calculate the EESC index for 2023-2100, denoted as... Based on CMIP6 data and stratospheric temperature and geopotential vorticity data under the SSP2-4.5 path, the PFP index for 2023-2100 is calculated and denoted as . .

[0040] S602, put and Substituting into the regression equation generated by S5, the calculated result is denoted as... The historical ozone observation results of Antarctica in 1980 were used as the standard for ozone recovery to estimate the time for Antarctic ozone recovery.

[0041] The final drawing results are shown in the attached document. Figure 3 , Figure 3 Taking Faraday Station, Halley Station, and Showa Station as examples, the black solid line is the station ozone change line without removal treatment, and the blue solid line is the station ozone change line after screening according to polar vortex morphology; the black and blue solid curves represent the regression curves before and after polar vortex screening, respectively; the black and blue dashed lines represent the total ozone column level in 1980 before and after screening, respectively.

[0042] As can be observed from the figure, at the Faraday and Showa stations, the total ozone column recovered to the 1980 level after screening was delayed by about 30 years compared to before screening, while at the Halley station, the difference before and after screening was smaller, and the recovery time after screening was delayed by about 10 years compared to before screening.

[0043] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting long-term changes in Antarctic ozone based on polar vortex morphology constraints, characterized in that, Includes the following steps: Preprocessing of multi-source ozone data; the multi-source ozone data includes ozone observation data from Antarctic stations, reanalysis data, and satellite data; Using artificial intelligence methods, the missing parts of ozone station observation data were filled in based on preprocessed reanalysis data and satellite data to obtain a complete ozone data sequence; The process of obtaining a complete ozone data sequence includes: Based on the latitude and longitude of the Antarctic station, the time series of total ozone column parameters and geopotential vorticity parameters for the corresponding locations were extracted from the preprocessed reanalysis data and satellite data. Linear interpolation is performed on the ozone observation sequence of the ozone observation data to calculate the correlation between the ozone observation sequence and the time series of the total ozone column parameter and the time series of the geopotential vorticity parameter. The parameter sequences with significant correlation are retained, and a decision tree model is constructed to screen out all feature input parameters. After standardizing all the selected feature input parameters, they are input into the long short-term memory neural network model for training. The model output is then de-standardized to generate a complete time series of total ozone column. The complete ozone column total time series is matched with the ozone observation data to fill in the missing parts and obtain a complete ozone data series. Based on the preprocessed reanalysis data, the average climate state of the polar vortex boundary on the 530K isentropic surface was determined. Based on the climate average state, the complete ozone data sequence was screened, and the observation data when the station was located outside the polar vortex were removed to obtain effective ozone observation data. A nonlinear regression model was constructed based on the effective ozone observation data, and the long-term trend of Antarctic ozone variation was predicted based on the nonlinear regression model.

2. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 1, characterized in that, Preprocessing of ozone observation data, reanalysis data, and satellite data from Antarctic stations includes converting the units of total ozone column to Dobson units and the units of geopotential vorticity to geopotential vorticity units.

3. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 1, characterized in that, The process of determining the climatological mean state of the polar vortex boundary on the 530K isentropic surface based on preprocessed reanalysis data includes: Interpolate the geopotential vorticity data on the isobaric surface to the 530K isentropic surface; perform dimensional correction on the interpolated geopotential vorticity data; calculate the equivalent latitude of each contour line in the corrected geopotential vorticity field; determine the daily polar vortex boundary based on the gradient of geopotential vorticity with the equivalent latitude; calculate the average value of the daily polar vortex boundary geopotential vorticity value over historical periods as the climate average state.

4. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 1, characterized in that, The process of obtaining effective ozone observation data includes: comparing daily geopotential vorticity values ​​with the climate average to generate a binary sequence that identifies the station's position relative to the polar vortex; and based on the binary sequence, performing masking on the complete ozone data sequence to retain data where the station is located inside the polar vortex.

5. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 1, characterized in that, Using the effective ozone observation data as the dependent variable and the stratospheric equivalent chlorine and polar stratospheric cloud generation potential as independent variables, a nonlinear regression model is constructed. Substituting future data on the stratospheric equivalent chlorine and polar stratospheric cloud generation potential into the nonlinear regression model, the long-term trend of Antarctic ozone variation is predicted.

6. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 5, characterized in that, The stratospheric equivalent chlorine is a time series calculated based on the emission inventory of ozone-depleting substances.

7. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 5, characterized in that, The polar stratospheric cloud generation potential is a time series calculated based on stratospheric temperature and geopotential vorticity data, as shown in the following formula: ； in, The volume of polar stratospheric clouds. Let be the volume of the polar vortex.

8. The method for predicting long-term Antarctic ozone variation based on polar vortex morphology constraints according to claim 1, characterized in that, The nonlinear regression model is as follows: ； This is the monthly average time series of the total ozone column from the valid ozone observation data. For EESC time series, For PFP time series, to For regression coefficients, This is the error term.

Images