KR102576775B1 - Method for predicting air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model and computing device using the same - Google Patents

Abstract

In the method of predicting air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, (a) a computing device, (i-1) at least one past point in time at a specific location Obtain t _p past weather data corresponding to the past time point t _p , input the t _p past weather data into a weather model (WRF, Weather Research and Forecasting), and allow the weather _model to to output at least one t _p+k past weather forecast data corresponding to at least one t _p+k past point in time, and at least one t _p+k past weather forecast data that is a processed value of the t _p+k past weather forecast data. Input the processed data into an air quality model (CMAQ, Community Multiscale Air Quality) to cause the air quality model to output at least one t _p+k past air quality forecast data corresponding to the t _p+ k past time, (i-2) a process of calculating background error covariance of the air quality model with reference to the t _p+k past air quality prediction data; and (ii-1) t _p past air quality observation data corresponding to the t _p past time at the specific location - the t _p past air quality observation data includes air quality observation data obtained from at least some of the ground and satellites. Includes - obtaining and performing (ii-2) a process of calculating observation error covariance with reference to the t _p past air quality observation data; (b) The computing device (i) acquires past weather data t _i corresponding to the past time t i, which is a past time point adjacent to the current time t _c at the specific location and is a future time _point than the t _p+k past time point. and (ii) inputting the t _i past weather data into the weather model to cause the weather model to output at least one t _i+k weather prediction data, and (iii) the t _i+k weather prediction data. Inputting at least one t _i+k processed data, which is a processed value, into the air quality model to cause the air quality model to output at least one t _i+k air quality prediction data; (c) The computing device, (i) acquires current air quality observation data corresponding to t _c , and provides t _c air quality prediction data corresponding to t _c among the t _i+k air quality prediction data. After determining the model initial field before data assimilation, calculating the model initial field after data assimilation with reference to the model initial field before data assimilation, the current air quality observation data, the observation error covariance, and the background error covariance; and (ii) acquiring current weather data corresponding to t _c at the specific location, and inputting the current weather data into the weather model to cause the weather model to detect at least one t that is after a preset time from t _c . To output at least one t _{c+k future} weather forecast data corresponding to a c+k future point in time, and to obtain at least one t _c+k future processed data that is a processed value of the t _c+k future weather forecast data. performing a process; and (d) the computing device inputs the t _{c + k} future processed data and the initial model length after data assimilation into the air quality model to enable the air quality model to predict the model at the t _{c + k} future time point. A method including a step of outputting a sheet and a computing device using the same are provided.

Description

Method for predicting air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model and computing device using the same {METHOD FOR PREDICTING AIR POLLUTION BY APPLYING AIR QUALITY OBSERVATION DATA OBTAINED FROM AT LEAST SOME OF THE GROUND AND SATELLITES TO AN AIR QUALITY MODEL AND COMPUTING DEVICE USING THE SAME}

The present invention relates to a method for predicting air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, and a computing device using the same. More specifically, it relates to fine dust obtained from ground observation stations. It relates to a method for improving the prediction accuracy of air pollutants by correcting air quality prediction data using at least some of the observation data and AOD observation data obtained from satellites and applying it to an air quality model, and a computing device using the same.

Air pollution includes artificial factors such as smoke generated during the combustion process of factories and automobiles, yellow dust, waste incineration, and food cooking, as well as natural factors such as sandstorms, volcanic ash, and forest fires. This air pollution causes human and material damage. Because significant damage occurs, high accuracy is required for current status and forecast information on air pollutants.

Conventionally, information on the status of air pollutants is confirmed through observation data acquired from ground observatories and satellites at home and abroad, and various numerical analysis models (e.g., WRF/CMAQ, WRF-Chem, etc.) that combine meteorological models and air quality models are used. ) was used to predict air pollutants.

In addition, in an attempt to improve prediction performance in relation to predictions of air pollutants, there are a method of improving the physical and chemical parameterization of the numerical analysis model itself and a data assimilation method. Among these, the data assimilation method is better than the parameterization improvement method. It is widely used because it has the advantage of improving prediction performance in a short period of time.

Previously, in relation to technology that improved prediction performance using data assimilation techniques, the aerosol optical depth (AOD) obtained from the Geostationary Ocean Color Imager (GOCI) mounted on the Cheollian satellite was used. There was a case where the data was assimilated and input into an air quality model (CMAQ, Community Multiscale Air Quality) to evaluate and analyze the transport of particulate air pollutants between countries (e.g., China-Korea).

However, in the above case, there was a problem in that correction through data assimilation was not possible in times when observation data could not be obtained from the sensor mounted on the satellite due to occlusion by clouds.

Therefore, improvement measures to solve the above problems are required.

The purpose of the present invention is to solve the above-mentioned problems.

In addition, the present invention inputs meteorological data into a current air quality forecast model (WRF/CMAQ) and refers to the weather forecast data and air quality forecast data output from the current air quality forecast model, respectively, to provide a background for the current air quality forecast model. After calculating the error covariance and performing preprocessing on each of the AOD observation data obtained from a satellite and the fine dust observation data obtained from a ground observatory, the observation error covariance is calculated using each of the AOD observation data and fine dust observation data. for a different purpose.

In addition, the present invention determines the air quality forecast data, which predicts the air quality at the current time at a past time adjacent to the current time, as the initial field of the model before data assimilation, and then uses the air quality observation data acquired at the current time, and the model before data assimilation. Calculate the model initial field after data assimilation by referring to the initial field, observation error covariance, and background error covariance of the air quality model, and improve the prediction performance of air pollutants by inputting the model initial field into the air quality model and correcting it after data assimilation. It serves another purpose.

In addition, the present invention identifies observable observation data among fine dust observation data obtained from a ground observatory at a specific location and AOD observation data obtained from a satellite, and reflects this to calculate the model initial field after data assimilation. It is for a different purpose.

The characteristic configuration of the present invention for achieving the purpose of the present invention as described above and realizing the characteristic effects of the present invention described later is as follows.

According to one aspect of the present invention, in the method of predicting air pollution by applying air quality observation data obtained from at least some of the ground and artificial satellites to an air quality model, (a) a computing device, (i-1) specifies Obtain t _p past weather data corresponding to at least one past time point in the location, t _p past time, and input the t _p past weather data into a weather model (WRF, Weather Research and Forecasting) to create the weather model. To output at least one t _p+k past weather forecast data corresponding to at least one t _p+k past time point after a preset time from _{t p} , and a value obtained by processing the t _p+k past weather forecast data. At least one t _p+k past processed data is input into an air quality model (CMAQ, Community Multiscale Air Quality) to cause the air quality model to generate at least one t p+k past air quality corresponding to the t p _{+ k} past time point. A process of outputting quality prediction data and (i-2) calculating background error covariance of the air quality model with reference to the t _p+k past air quality prediction data; and (ii-1) t _p past air quality observation data corresponding to the t _p past time at the specific location - the t _p past air quality observation data includes air quality observation data obtained from at least some of the ground and satellites. Includes - obtaining and performing (ii-2) a process of calculating observation error covariance with reference to the t _p past air quality observation data; (b) The computing device (i) acquires past weather data t _i corresponding to the past time t i, which is a past time point adjacent to the current time t _c at the specific location and is a future time _point than the t _p+k past time point. and (ii) inputting the t _i past weather data into the weather model to cause the weather model to output at least one t _i+k weather prediction data, and (iii) the t _i+k weather prediction data. Inputting at least one t _i+k processed data, which is a processed value, into the air quality model to cause the air quality model to output at least one t _i+k air quality prediction data; (c) The computing device, (i) acquires current air quality observation data corresponding to t _c , and provides t _c air quality prediction data corresponding to t _c among the t _i+k air quality prediction data. After determining the model initial field before data assimilation, calculating the model initial field after data assimilation with reference to the model initial field before data assimilation, the current air quality observation data, the observation error covariance, and the background error covariance; and (ii) acquiring current weather data corresponding to t _c at the specific location, and inputting the current weather data into the weather model to cause the weather model to detect at least one t that is after a preset time from t _c . To output at least one t _{c+k future} weather forecast data corresponding to a c+k future point in time, and to obtain at least one t _c+k future processed data that is a processed value of the t _c+k future weather forecast data. performing a process; and (d) the computing device inputs the t _{c + k} future processed data and the initial model length after data assimilation into the air quality model to enable the air quality model to predict the model at the t _{c + k} future time point. A method including a step of outputting a sheet is disclosed.

As an example, in step (c), the computing device refers to an observation operator, which is an operator for converting the model initial field before the data assimilation into an observation space, the observation error covariance, and the background error covariance after the data assimilation. Determine the optimal weight for calculating the model initial field, and determine the observation increment, which is the difference between the current air quality observation data and the model background field interpolated to the specific location by applying the observation operator to the model initial field before data assimilation. A method is disclosed wherein the initial length of the model after the data assimilation is calculated by applying the optimal weight and adding it to the initial length of the model before the data assimilation.

As an example, if the computing device determines that, as the current air quality observation data, there is (i) ground air quality observation data and satellite air quality observation data, the ground air quality is determined in a planetary boundary layer (PBL). The model initial field is calculated after the data assimilation by referring to the fine dust observation data included in the observation data, and in layers above the planetary boundary layer, the above is referred to the AOD (Aerosol Optical Depth) observation data included in the satellite air quality observation data. After data assimilation, the model initial field is calculated, and (ii) if it is determined that only the ground air quality observation data exists, the data in the planetary boundary layer is referred to the fine dust observation data included in the ground air quality observation data. After assimilation, the model initial field is calculated, and (iii) if it is determined that only the satellite air quality observation data exists, the AOD observation data included in the satellite air quality observation data is referred to in all layers of the air quality model. A method is disclosed, characterized in that the model initial length is calculated after the data assimilation.

As an example, when the computing device confirms the ground air quality observation data and the satellite air quality observation data as the current air quality observation data, (i) in the planetary boundary layer, before the data assimilation for fine dust data The model initial field after the data assimilation for the aerosol component data is calculated by referring to the ratio of the model initial field after the data assimilation to the model initial field, and the fine dust is calculated with reference to the model initial field after the data assimilation for the aerosol component data. Converting the model initial field after the data assimilation for the data into the model initial field after the data assimilation for the AOD, (ii) in a layer above the planetary boundary layer, the current air quality observation data is combined with the satellite air quality observation data , a method is disclosed, characterized in that the model initial field after the data assimilation for the AOD is calculated by replacing the difference in the model initial field after the data assimilation for the AOD in the planetary boundary layer.

As an example, in step (a), the computing device, (i) if the t _p past air quality observation data corresponds to ground air quality observation data acquired from a first ground observation station, 1 Measurement error including at least some of the error and the second error caused by differences in the setting method using the observation device, and the resolution of the t _p past air quality observation data and the t _{p + k} past air quality prediction data Calculate the representative error caused by the difference in scale, calculate the observation error for the ground air quality observation data using the measurement error and the representative error, and refer to the observation error and the average of the observation error. Calculate the observation error covariance for the t _p past air quality observation data, and (ii) if the t _p past air quality observation data corresponds to satellite air quality observation data acquired from a satellite, the t _p past air quality For each size of AOD according to quality observation data, the absolute error is calculated using the difference between the ground AOD obtained from the second ground observatory and the satellite AOD obtained from the satellite, with reference to the absolute error and the average of the absolute errors. A method for calculating the observation error covariance for the t _p past air quality observation data is disclosed.

As an example, in step (a), the computing device uses the t _p+k past air quality prediction data and the t _p+k past time point adjacent to the t _p+k past time point - where n is a positive number. - A method is disclosed, characterized in that calculating the background error covariance for the air quality model by estimating the difference with at least one t _{p + kn} past air quality prediction data corresponding to as the background error of the air quality model. do.

As an example, in step (a), the computing device measures previously measured data at a plurality of observation points within a radius of influence centered on the specific location, with respect to the t _p past air quality observation data at the specific location. A method is disclosed that is characterized in that it compares the average value of air quality observation data and determines it as noise when it is greater than a preset threshold.

As an example, in step (b), the computing device acquires weather forecast data at the past time t _i simulated through ECMWF (European Center for Medium-Range Weather Forecasts) as the past weather data t _i . A method characterized by this is disclosed.

As an example, in step (c), a method is disclosed wherein the computing device acquires weather prediction data at t _c simulated through ECMWF as the current weather data.

As an example, in step (b), the computing device uses a Meteorology Chemistry Interface (MCIP) to process the t _i+k weather forecast data into a data format that can be input to the air quality model and emissions model (MEGAN/SMOKE). Processor) to cause the MCIP to output t _i+k weather preprocessing data by adjusting the domain, and input the t _i+k weather preprocessing data to the emissions model to generate t _i+k emissions from the emissions model. A method is disclosed that outputs data and inputs the t _i+k meteorological preprocessing data and the t _i+k emission data into the air quality model as the t i _+k processed data.

As an example, in step (c), the computing device inputs the t _{c + k} future weather prediction data into MCIP for processing into a data form that can be input to the air quality model and emission model to cause the MCIP to enter the domain By adjusting t _{c + k} future weather preprocessing data to be output, and inputting the t _{c + k} future weather preprocessing data into the emission model to cause the emission model to output t _{c + k} future emission data, A method is disclosed, characterized in that inputting t _c+k future weather preprocessing data and t _c+k future emissions data into the air quality model as the t _c+k future processed data.

According to another aspect of the present invention, a computing device for predicting air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, comprising: at least one memory storing instructions; and at least one processor configured to execute the instructions, wherein the processor: (I) (i-1) t _p past weather data corresponding to at least one past time t _p at a specific location; Obtain and input the t _p past weather data into a weather model (WRF, Weather Research and Forecasting) to cause the weather model to correspond to at least one t _{p + k} past time point after a preset time from t _p . At least one t _p+k past weather forecast data is output, and at least one t _p+k past processed data, which is a processed value of the t _p+k past weather forecast data, is output to an air quality model (CMAQ, Community Multiscale Air Quality ) to cause the air quality model to output at least one t _{p+ k past air quality forecast data corresponding to the t p+} k past time, and (i-2) the t _p+k past air quality A sub-process for calculating background error covariance of the air quality model with reference to predicted data; and (ii-1) t _p past air quality observation data corresponding to the t _p past time at the specific location - the t _p past air quality observation data includes air quality observation data obtained from at least some of the ground and satellites. Includes - a process of obtaining and performing (ii-2) a subprocess of calculating observation error covariance with reference to the t _p past air quality observation data; (II) (i) Obtain past weather data t _i corresponding to the past time t _i , which is a past time point adjacent to the current time t _c at the specific location and is a time future than the past time point t _p+k , (ii) Input the t _i past weather data into the weather model to cause the weather model to output at least one t _i+k weather forecast data, and (iii) at least one value that is a processed value of the t _i+k weather forecast data. A process of inputting t _i+k processed data to the air quality model to cause the air quality model to output at least one t _i+k air quality prediction data; (III) (i) Obtain the current air quality observation data corresponding to t _c , and use the t _c air quality prediction data corresponding to t _c among the t _i+k air quality prediction data as the initial model before data assimilation. After determining as, a sub-process of calculating the model initial length after data assimilation with reference to the model initial length before data assimilation, the current air quality observation data, the observation error covariance, and the background error covariance; and (ii) acquiring current weather data corresponding to t _c at the specific location, and inputting the current weather data into the weather model to cause the weather model to detect at least one t that is after a preset time from t _c . To output at least one t _{c+k future} weather forecast data corresponding to a c+k future point in time, and to obtain at least one t _c+k future processed data that is a processed value of the t _c+k future weather forecast data. a process that performs a subprocess; and (IV) inputting the t _{c + k} future processed data and the model initial field after data assimilation into the air quality model to cause the air quality model to output a model prediction field at the t _{c + k} future time point. A computing device that performs a process is disclosed.

As an example, in the process (III), the processor refers to an observation operator, which is an operator for converting the initial field of the model before the data assimilation into an observation space, the observation error covariance, and the background error covariance, and the model after the data assimilation. Determine the optimal weight for calculating the initial field, and calculate the observation increment, which is the difference between the current air quality observation data and the model background field interpolated to the specific location by applying the observation operator to the model initial field before data assimilation. A computing device is disclosed, which calculates the model initial length after data assimilation by applying optimal weights and adding them to the initial length of the model before data assimilation.

As an example, if the processor determines that, as the current air quality observation data, (i) there is ground air quality observation data and satellite air quality observation data, the ground air quality is observed in a planetary boundary layer (PBL) The model initial field is calculated after the data assimilation by referring to the fine dust observation data included in the data, and in layers above the planetary boundary layer, the above data is referred to the AOD (Aerosol Optical Depth) observation data included in the satellite air quality observation data. After assimilation, the model initial field is calculated, and (ii) if it is determined that only the ground air quality observation data exists, the data assimilation in the planetary boundary layer is performed by referring to the fine dust observation data included in the ground air quality observation data. After calculating the model initial field, (iii) if it is determined that only the satellite air quality observation data exists, the AOD observation data included in the satellite air quality observation data is referred to in all layers of the air quality model. A computing device is disclosed, which is characterized in that it calculates the model initial field after data assimilation.

As an example, when the processor confirms the ground air quality observation data and the satellite air quality observation data as the current air quality observation data, (i) in the planetary boundary layer, the model before the data assimilation for fine dust data The model initial field after the data assimilation for the aerosol component data is calculated by referring to the ratio of the model initial field after the data assimilation to the initial field, and the fine dust data is calculated with reference to the model initial field after the data assimilation for the aerosol component data. Converting the model initial field after the data assimilation for AOD to the model initial field after the data assimilation for AOD, (ii) in a layer above the planetary boundary layer, the current air quality observation data, the satellite air quality observation data, A computing device is disclosed, wherein the model initial field after the data assimilation for the AOD is calculated by substituting the difference in the model initial field after the data assimilation for the AOD in the planetary boundary layer.

As an example, in the process (I), the processor: (i) If the t _p past air quality observation data corresponds to ground air quality observation data acquired from a first ground observation station, the first air quality observation data embedded in the observation device Measurement error including at least some of the error and the second error caused by differences in the setting method using the observation device, the resolution of the t _p past air quality observation data and the t _{p + k} past air quality prediction data, and Calculate representative errors caused by differences in scale, calculate observation errors for the ground air quality observation data using the measurement errors and representative errors, and refer to the observation errors and the average of the observation errors. Calculate the observation error covariance for the t _p past air quality observation data, and (ii) if the t _p past air quality observation data corresponds to satellite air quality observation data acquired from a satellite, the t _p past air quality For each size of AOD according to observation data, an absolute error is calculated using the difference between the ground AOD obtained from the second ground observatory and the satellite AOD obtained from the satellite, and the absolute error and the average of the absolute errors are referred to. t _p A computing device that calculates the observation error covariance for past air quality observation data is disclosed.

As an example, in the process (I), the processor processes the t _p+k past air quality prediction data and the t _p+k past time point adjacent to the t _p+k past time point - where n is a positive number - A computing device is disclosed, which calculates the background error covariance by estimating the difference with at least one t _{p + kn} past air quality prediction data corresponding to as the background error of the air quality model.

As an example, in the process (I), the processor measures the air quality at a plurality of observation points within a radius of influence centered on the specific location, with respect to the t _p past air quality observation data at the specific location. A computing device is disclosed, which is characterized in that it compares the average value of quality observation data and determines it as noise when it is greater than a preset threshold.

As an example, a computing device is disclosed, wherein the processor acquires weather forecast data at the past time t _i simulated through ECMWF (European Center for Medium-Range Weather Forecasts) as the past weather data t _i . do.

As an example, in the process (III), a computing device is disclosed, wherein the processor obtains weather prediction data at t _c simulated through ECMWF as the current weather data.

As an example, in the process (II), the processor uses a Meteorology Chemistry Interface Processor (MCIP) to process the t _i+k weather prediction data into a data form that can be input to the air quality model and emissions model (MEGAN/SMOKE). ) to cause the MCIP to output t _i+k meteorological preprocessing data by adjusting the domain, and input the t _i+k meteorological preprocessing data into the emission model to generate t _i+k emission data. A computing device is disclosed, wherein the t _i+k weather preprocessing data and the t _i+k emission data are input to the air quality model as the t _i+k processed data.

As an example, in the process (III), the processor inputs the t _{c + k} future weather forecast data into MCIP for processing into a data format that can be input to the air quality model and emission model to cause the MCIP to create a domain. By adjusting, t _{c + k} future weather pre-processing data is output, and the t _{c + k} future weather pre-processing data is input into the emission model to cause the emission model to output t _{c + k future emission data, and the t c + k} future weather pre-processing data is input to the emission model. A computing device is disclosed, characterized in that inputting _c+k future weather preprocessing data and t _{c+k future emissions data into the air quality model as the t c+k} future processed data.

The present invention inputs meteorological data into a current air quality forecast model, calculates the background error covariance of the current air quality forecast model by referring to each of the weather forecast data and air quality forecast data output from the current air quality forecast model, and calculates the background error covariance of the current air quality forecast model. After performing preprocessing on each of the AOD observation data obtained from and the fine dust observation data obtained from a ground observatory, the observation error covariance is calculated using each of the AOD observation data and the fine dust observation data.

In addition, the present invention determines the air quality forecast data, which predicts the air quality at the current time at a past time adjacent to the current time, as the initial field of the model before data assimilation, and then uses the air quality observation data acquired at the current time, and the model before data assimilation. Calculate the model initial field after data assimilation by referring to the initial field, observation error covariance, and background error covariance of the air quality model, and improve the prediction performance of air pollutants by inputting the model initial field into the air quality model and correcting it after data assimilation. It has an effect.

In addition, the present invention has the effect of identifying observable observation data among fine dust observation data obtained from a ground observatory at a specific location and AOD observation data obtained from a satellite, and calculating the model initial field after data assimilation by reflecting this. there is.

The following drawings attached for use in explaining embodiments of the present invention are only some of the embodiments of the present invention, and to those skilled in the art (hereinafter “those skilled in the art”), the invention Other drawings can be obtained based on these drawings without further work being done.
Figure 1 schematically shows a computing device that predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, according to an embodiment of the present invention.
Figure 2 shows the structure of a computing device according to an embodiment of the present invention in more detail,
Figure 3 schematically shows a process for calculating the observation error covariance for air quality observation data and the background error covariance of the air quality model, according to an embodiment of the present invention.
Figure 4 schematically shows a process for determining the initial model length before data assimilation, according to an embodiment of the present invention.
Figure 5 shows the process of calculating the model initial length after data assimilation with reference to the model initial length before data assimilation and inputting this into the air quality model to determine the model prediction length at a future point, according to an embodiment of the present invention. It is schematically shown,
Figure 6 schematically shows an example of calculating the background error covariance of an air quality model in the form of a matrix, according to an embodiment of the present invention.
Figure 7 schematically shows a process for performing preprocessing on air quality observation data, according to an embodiment of the present invention.
Figure 8 schematically shows a process for calculating observation error covariance when air quality observation data is acquired from a satellite, according to an embodiment of the present invention;
Figure 9 schematically shows an example of data resulting from calculating the observation error covariance according to the observation point where air quality observation data was obtained, according to an embodiment of the present invention.
Figure 10 schematically shows a process for confirming the presence or absence of current air quality observation data and performing data assimilation by reflecting this, according to an embodiment of the present invention.
11A and 11B show the difference between the model initial length before data assimilation and the model initial length after data assimilation, and the difference between the model initial length before data assimilation and the model prediction length after data assimilation, respectively, according to an embodiment of the present invention. This is a schematic illustration of the data resulting from the comparison.

The detailed description of the present invention described below refers to the accompanying drawings, which show by way of example specific embodiments in which the present invention may be practiced to make clear the objectives, technical solutions and advantages of the present invention. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention.

Additionally, throughout the description and claims, the word “comprise” and variations thereof are not intended to exclude other technical features, attachments, components or steps. Other objects, advantages and features of the invention will appear to those skilled in the art, partly from this description and partly from practice of the invention. The examples and drawings below are provided by way of example and are not intended to limit the invention.

Moreover, the present invention encompasses all possible combinations of the embodiments shown herein. It should be understood that the various embodiments of the invention are different from one another but are not necessarily mutually exclusive. For example, specific shapes, structures and characteristics described herein with respect to one embodiment may be implemented in other embodiments without departing from the spirit and scope of the invention. Additionally, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the detailed description that follows is not intended to be taken in a limiting sense, and the scope of the invention is limited only by the appended claims, together with all equivalents to what those claims assert, if properly described. Similar reference numbers in the drawings refer to identical or similar functions across various aspects.

The title or summary of the present invention provided herein is provided for convenience only and does not limit or construe the scope or meaning of these embodiments.

For reference, although each component is described below in the singular, this does not exclude the possibility of plurality.

Hereinafter, in order to enable those skilled in the art to easily practice the present invention, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.

Figure 1 schematically illustrates a computing device that predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, according to an embodiment of the present invention.

First, referring to FIG. 1, a computing device 100 that predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model includes a memory 110 and a processor 120. can do.

In addition, the memory 110 of the computing device 100, which predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, can store instructions to be performed by the processor 120. Specifically, there is a code generated for the purpose of enabling the computing device 100 to predict air pollution in a specific manner by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model. As such, it may be stored in a computer-usable or computer-readable memory capable of being directed to a computer or other programmable data processing equipment. Instructions may perform processes to execute functions described in the specification of the present invention.

And, the processor 120 of the computing device 100, which predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model, is an MPU (Micro Processing Unit) or CPU (Central Processing Unit). ), cache memory, and data bus. Additionally, the computing device 100 may further include an operating system and a software component of an application that performs a specific purpose.

Additionally, the computing device 100 may be a user terminal (not shown) such as a smartphone, tablet, or PC, and predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model. It may be a server for that purpose. In the latter case, the server may be linked to a separate user terminal (not shown) such as a smartphone, tablet, or PC.

Figure 2 shows the structure of a computing device according to an embodiment of the present invention in more detail.

Referring to FIG. 2, the computing device 100 includes, in addition to the aforementioned memory 110 and processor 120, a data acquisition unit 130, an observation error covariance calculation unit 140, a background error covariance calculation unit 150, and a standby It may further include a quality data assimilation unit 160 and a presentation system unit 170.

Here, at least some of the data acquisition unit 130, observation error covariance calculation unit 140, background error covariance calculation unit 150, air quality data assimilation unit 160, and display system unit 170 are processor 120. Functions may be performed in conjunction with, for example, each function may be performed by receiving predetermined instructions from the processor 120.

Below, we will look at each component of the computing device 100 in more detail.

First, the data acquisition unit 130 acquires a simulated weather forecast field through a weather model of the European Center for Medium-Range Weather Forecasts (ECMWF) and a weather forecast field mounted on an environmental satellite (Cheollian Satellite 2B). AOD observation data as air quality observation data observed from GEMS (Geostationary Environment Monitoring Spectrometer), an environmental sensor, and ultrafine dust (PM2.5) as air quality observation data observed from multiple first ground observation stations installed in Korea and China. Observation data can be obtained. In addition, the observation error covariance calculation unit 140 may obtain AOD observation data on the ground from AERONET (Aerosol Robotic Network) installed on the ground to calculate the observation error covariance using the AOD observation data of GEMS, but is limited to this. It's not like that.

At this time, conventionally, AOD observation data was obtained from GOCI, but as the mission performance period of the existing Cheollian satellite equipped with GOCI was completed, in the present invention, AOD observation data was acquired by replacing it with GEMS, but according to the judgment of those skilled in the art, other than GEMS It may be possible to obtain AOD observation data from AOD observation sensors mounted on other satellites.

Next, the observation error covariance calculation unit 140 performs a preprocessing process on the air quality observation data obtained from the data acquisition unit 130, and then calculates the observation error covariance for the fine dust observation data obtained from the first ground observatory. and the observation error covariance for the AOD observation data obtained from the sensor mounted on the satellite can be calculated.

In addition, the background error covariance calculation unit 150 inputs the weather data obtained from the data acquisition unit 130 into a weather model (WRF, Weather Research and Forecasting), and uses the weather prediction data output from the weather model to model air quality. By acquiring processed data so that it can be input into (CMAQ) and inputting the processed data into the air quality model, the background error covariance for the air quality model can be calculated by referring to the air quality prediction data output from the air quality model. there is.

In addition, the air quality data assimilation unit 160 acquires the observation error covariance calculated from the observation error covariance calculation unit 140 and the preprocessed air quality observation data at the time of data assimilation, and the background error covariance calculation unit ( Obtain the background error covariance of the air quality model calculated from 150), input meteorological data from a past time point adjacent to the time of data assimilation into the WRF/CMAQ model, and air quality forecast data output from the WRF/CMAQ model. Air quality prediction data at the time of data assimilation is determined and acquired as the initial field of the model before data assimilation, and preprocessed air quality observation data, background error covariance, observation error covariance, and data assimilation at the time of data assimilation are acquired. The model initial length after data assimilation is calculated by referring to the previous model initial length, and the model initial length after data assimilation can be input into the air quality model to have the air quality model calculate the model prediction length after data assimilation.

Then, when the display system unit 170 obtains the model forecast for the future output from the air quality model after performing data assimilation, it visualizes the model forecast for the future so that the user can easily check it and displays it on a predetermined display. It can be displayed through .

The operation process for predicting air pollution using the computing device 100 will be described in detail with reference to FIGS. 3 to 10.

Figure 3 schematically shows a process for calculating the observation error covariance of air quality observation data and the background error covariance of an air quality model, according to an embodiment of the present invention.

Referring to FIG. 3, the computing device 100 acquires t _p past weather data corresponding to t _p past time, which is at least one past time in a specific location (S310), and adds t _p past weather data to the weather model. The input can be used to cause the weather model to output at least one t _{p +k} past weather forecast data corresponding to at least one t _p+k past time point after a preset time from t p (S320).

For example, in step S310, the data acquisition unit 130 of the computing device 100 generates weather forecast data corresponding to a past time point t _p at a specific location simulated through the European Center for Medium-Range Weather Forecasts (ECMWF). t _p past weather data can be obtained by performing preprocessing into a form that can be input into the weather model. At this time, p may be an integer greater than 0.

For reference, as t _p past weather data, at least part of the simulated weather forecast data from November 1, 2021 to March 31, 2022 was obtained and used as raw weather data to input into the weather model. Depending on this, weather observation data observed from the first ground observatory during the same period may be used as raw weather data to input into the weather model. In addition, the weather model in the present invention is WRF v3.8.1, but is not limited to this.

In addition, in step S320, assuming that t _p is 12 o'clock on December 1, 2021 as an example of the past time, the computing device 100 uses the acquired past weather data at 12 o'clock on December 1, 2021 as a weather model. By entering , the weather model can output past weather forecast data predicted in hourly increments from 12:00 on December 1, 2021 to 72 hours later as t _p+k past weather forecast data. According to the above example, k may be an integer from 0 to 72, but is not limited thereto.

For reference, the t _p+k past weather forecast data output from the weather model is gridded data, and although the weather data at 12 o'clock on December 1, 2021 is not forecast data, the form of the data is different, so the weather model is 2021 Data as of 12 o'clock on December 1, 2018 can also be output.

Afterwards, the computing device 100 acquires at least one t _p+k past processed data, which is a value processed from t _p+k past weather forecast data (S330, S340), and uses the t _p+k past processed data to model the air quality. can be input to cause the air quality model to output at least one t _p+k past air quality forecast data corresponding to the t _p+k past time point (S350).

For example, in steps S330 and S340, the computing device 100 uses a Meteorology Chemistry Interface Processor (MCIP) to process t _p+k past weather forecast data into a data format that can be input into an air quality model and an emissions model (MEGAN/SMOKE). ) to have MCIP output t _p+k past weather preprocessing data by adjusting the domain, and input t _p+k past weather preprocessing data into the emissions model to have the emissions model output t _p+k past emissions data. It is output, and t _p+k past meteorological preprocessing data and t _p+k past emission data can be obtained as t _p+k past processed data. At this time, the air quality model in the present invention uses CMAQ v5.0.2 to have a grid resolution of 27km for the Northeast Asia region, MCIP uses v4.4, and the emission models include MEGAN v2.0.4 and SMOKE v3. Although .5 was used for each, it is not limited to this.

For example, the computing device 100 inputs t _p+k past weather forecast data predicted in hourly increments from 12:00 on December 1, 2021 to 72 hours later into MCIP and allows MCIP to create (i) an air quality model To output t _p+k past weather preprocessed data from 12:00 on December 1, 2021 to 72 hours later, (ii) t _p+k past weather forecast data is input into the emissions model. Convert to possible grid resolution and domain to output t _p+k past meteorological preprocessing data from 12:00 on December 1, 2021 to 72 hours later, and then input this into the emissions model to make the emissions model as of December 2021. You can output t _p+k past emissions data from 12:00 a.m. to 72 hours later. At this time, the t _p+k past emission data output from the emission model may also be preprocessed to have a grid resolution of 27km so that it can be input into the air quality model in the same way as the t _p+k past meteorological preprocessing data.

Thereafter, in step S350, the computing device 100 uses t _p+k past weather preprocessing data output from MCIP and t _p+k past emissions data output from the emission model to the air quality model as t _p+k past processed data. By entering it, you can have the air quality model output t _p+k past air quality forecast data from 12:00 on December 1, 2021 to 72 hours later.

In other words, the air quality model calculates aerosol component forecast data from 12:00 on December 1, 2021 to 72 hours later using t _p+k past weather pretreatment data and t _p+k past emissions data, and refers to this The converted ultrafine dust (e.g., PM2.5) forecast data can be output as t _p+k past air quality forecast data. Alternatively, the air quality model may output AOD prediction data converted with reference to aerosol component prediction data as t _p+k past air quality prediction data, but is not limited to this.

Additionally, the computing device 100 may calculate the background error covariance of the air quality model by referring to t _p+k past air quality prediction data (S360).

As an example, the background error covariance calculation unit 150 of the computing device 100 includes t _p+k past air quality prediction data and at least one t _p+kn past air quality prediction data corresponding to a t _p+kn past time point. By estimating the difference as the background error of the air quality model, the background error covariance for the air quality model can be calculated, which can be expressed as Equation (1) below. At this time, n may be a positive number.

Equation (1)

는 모델의 배경 오차 공분산, 는 모델의 참값, 는 모델의 예측장, 는 모델의 배경 오차로서 나타낼 수 있다.In equation (1) above,

is the background error covariance of the model, is the true value of the model, is the prediction field of the model, can be expressed as the background error of the model.

However, since the air quality model cannot know the true value of the model itself, the air quality model is modeled using NMC (National Meteorological Center), which is a method of calculating the error according to the model operation time by varying the model's prediction time. Calculate the background error covariance of the air quality model by assuming the t _p+kn past air quality forecast data output from as the true value of the air quality model, and assuming the t _p+k past air quality forecast data as the forecast field of the air quality model. can do.

For example, assuming that the past time t _p is 12 o'clock on December 1, 2021, the background error covariance calculation unit 150 of the computing device 100 calculates 12 on December 2, 2021, 24 hours after the past time t _p . The city's air quality forecast data is used as the true value of the air quality model, and the air quality forecast data at 12:00 on December 3, 2021, 48 hours after t _p , is used as the forecast field of the air quality model, 2021 The difference between the air quality forecast data at 12 o'clock on December 3 and the air quality forecast data at 12 o'clock on December 2, 2021 can be estimated as the background error of the air quality model. For reference, in the present invention, the difference between the air quality forecast data 24 hours after the air quality observation data was acquired and the air quality forecast data after 48 hours was estimated as the background error of the air quality model, but is not limited to this. .

Additionally, the background error covariance calculation unit 150 of the computing device 100 may calculate the background error covariance by referring to the background error of the estimated air quality model and the corresponding transpose matrix, which is explained with reference to FIG. 6. I will do it.

Figure 6 schematically shows an example of calculating the background error covariance of an air quality model in the form of a matrix.

Referring to Figure 6, for example, for each of a plurality of historical air quality forecast data from November 1, 2021 to March 31, 2022, the forecast field after 48 hours and the forecast after 24 hours from the past time t _p . The difference is 4 each as the error of the air quality model. 6 It is expressed in the form of a matrix, and the corresponding transpose matrices are also calculated, respectively. 4 You can see that it is expressed in the form of a matrix.

Then, the background error covariance calculation unit 150 of the computing device 100 multiplies the background error of the air quality model by the corresponding transpose matrix to finally obtain 6. 6 You can check the calculation as a matrix, and by averaging it, you can finally calculate the background error covariance of the air quality model.

Referring again to FIG. 3, the computing device 100 acquires t _p past air quality observation data corresponding to t _p past time at a specific location (S370), and observes the observation error by referring to t _p past air quality observation data. Covariance can be calculated (S380).

For reference, in Figure 3, for convenience of explanation, the background error covariance is first calculated and then the observation error covariance is numbered. However, it is also possible to calculate the observation error covariance first and then the background error covariance.

In step S370, the data acquisition unit 130 of the computing device 100 acquires air quality observation data of the past t _p as air quality observation data measured from at least some of the first ground observation station and the artificial satellite at the past time t _p . Later, the observation error covariance calculation unit 140 may perform preprocessing on t _p past air quality observation data, which will be explained with reference to FIG. 7.

Figure 7 schematically shows a process for performing preprocessing on air quality observation data, according to an embodiment of the present invention.

Referring to FIG. 7, for t _p past air quality observation data at a specific location 710 obtained from the data acquisition unit 130 of the computing device 100, the observation error covariance calculation unit 140 determines the specific location ( The deviation of the average value of past air quality observation data previously measured at a plurality of observation points 711 to 714 within the radius of influence 720 centered on 710 is a preset threshold (e.g., ), t _p past air quality observation data at a specific location 710 can be preprocessed by determining it as noise and removing it.

At this time, it is clear to those skilled in the art that the observation error covariance calculation unit 140 of the computing device 100 does not consider the observation point 715 outside the radius of influence to determine noise in the preprocessing process of t _p past air quality observation data. You will understand.

In addition, the observation error covariance calculation unit 140 of the computing device 100 may adjust at least some of the grid and domain in advance so that the preprocessed t _p past air quality observation data can be input into the air quality model.

Next, in step S380, the observation error covariance calculation unit 140 of the computing device 100, if t _p past air quality observation data corresponds to ground air quality observation data obtained from the first ground observatory, sends the observation device to the observation device. Measurement error including at least part of the inherent first error and the second error caused by differences in the setting method using the observation device, and the resolution of t _p past air quality observation data and t _{p + k} past air quality prediction data Representative errors caused by differences in scale can be calculated, respectively, and can be expressed in the formula as follows.

Equation (2)

Equation (3)

Equation (2) above is obtained by applying ultrafine dust (PM2.5 ₎ observation data, which is the past air quality observation data t _p measured from the first ground observation station, and predetermined parameter values that are constants. It shows calculating the measurement error, and Equation (3) is the measurement error of the past air quality observation data t _p calculated in Equation (2). , , It shows how to calculate the representative error of t _p past air quality observation data by applying parameters such as t p.

In addition, the observation error covariance calculation unit 140 of the computing device 100 calculates the observation error for the ground air quality observation data using the measurement error and representative error, and refers to the observation error and the average of the observation errors t _p The observation error covariance for past air quality observation data can be calculated, and this can be expressed in the formula as follows.

Equation (4)

Equation (5)

Equation (4) above calculates the observation error for t _p past air quality observation data on the ground by referring to the measurement error and representative error of t _p past air quality observation data calculated in Equation (2) and Equation (3). It represents a decision.

In addition, under the assumption that t _p past air quality observation data is not influenced by air quality observation data at other observation points, the observation error covariance calculation unit 140 of the computing device 100 calculates t _p as in equation (5). The difference between the observation error of past air quality observation data and the average of the observation error is calculated in the form of a matrix, and the observation error covariance for t _p past air quality observation data can be calculated by multiplying the corresponding diagonal component and averaging. You can confirm that it is possible.

Next, the observation error covariance calculation unit 140 of the computing device 100 relates to a process of calculating the observation error covariance when t _p past air quality observation data corresponds to satellite air quality observation data acquired from a satellite. This will be explained with reference to FIG. 8 .

Figure 8 schematically shows a process for calculating observation error covariance when air quality observation data is acquired from a satellite, according to an embodiment of the present invention.

Referring to FIG. 8, the computing device 100 uses the difference between the ground AOD obtained from the second ground observation station and the satellite AOD obtained from the satellite for each size of AOD according to t _p past air quality observation data to calculate the absolute error. , and the observation error covariance for t _p past air quality observation data can be calculated by referring to the absolute error and the average of the absolute error. At this time, the second ground observatory for acquiring AOD observation data on the ground may be AERONET, but is not limited to this.

As an example, the observation error covariance calculation unit 140 of the computing device 100 compares the AOD observation data obtained from AERONET on the ground with the AOD observation data as t _p past air quality observation data obtained from the GEMS mounted on the satellite. Since the accuracy of is known to be high, this can be assumed as the true value for calculating the observation error covariance.

For reference, in the present invention, t _p past air quality observation data for calculating observation error covariance is the AOD of 500 nm wavelength measured from 30 AERONET observation points at 12KST from November 1, 2021 to March 31, 2022. Observation data and AOD observation data with a wavelength of 550 nm from GEMS measured near the AERONET observation point at 11:45 KST during the relevant period were obtained. Among these, when AOD observation data was not observed and noise was detected during pre-processing of AOD observation data. A total of 3,780 ground and satellite AOD observation data sets were acquired, excluding cases determined to be.

First, the observation error covariance calculation unit 140 of the computing device 100 uses equation (6) below because the wavelengths of the ground AOD observation data obtained from AERONET and the satellite AOD observation data obtained from GEMS are different from each other. By matching by wavelength, comparison between ground AOD observation data and satellite AOD observation data can be made possible.

Equation (6)

Equation (6) above is the Ångstrom index in the 440~870nm wavelength range provided by AERONET. This shows converting 500 nm ground AOD observation data into 550 nm ground AOD observation data.

Thereafter, the observation error covariance calculation unit 140 of the computing device 100 uses the ground AOD observation data of 550 nm as a true value for calculating the observation error covariance, and uses the AOD obtained from the satellite AOD observation data as shown in equation (7) below. The absolute error can be calculated for each section according to the size of .

Equation (7)

The definition of the section in equation (7) above means arranging the size of the AOD obtained from 3,780 satellite AOD observation data in descending order and then grouping the AOD into 90 units starting from the smallest size, with a total of 42 A graph of the results of calculating the absolute error for each section can be seen in (a) of FIG. 8.

Referring to the graph in (a), the x-axis represents the size of the AOD confirmed from the 550nm satellite AOD observation data, and the y-axis represents the absolute error, corresponding to the top 68% of the 90 absolute errors in each of the 42 sections. It can be seen that the linear graph 810 shown through linear regression using the values of It can be seen that the correlation is 0.93. Additionally, the linear error equation can be expressed as Equation (8) below.

Equation (8)

Through equation (8) above, it can be confirmed that the size of AOD (GEMSAOD ₅₅₀ ) obtained from satellite AOD observation data can be expressed as an absolute error by substituting.

Equation (9)

In addition, under the assumption that the satellite AOD observation data at a specific location is not influenced by the AOD observation data at other observation points, the observation error covariance calculation unit 140 of the computing device 100 calculates the absolute error as shown in equation (9). And the difference from the average of the absolute error is calculated in the form of a matrix, and the observation error covariance for the satellite AOD observation data as t _p past air quality observation data can be calculated by multiplying the corresponding diagonal component and then averaging. You can check it.

In this way, the observation error covariance calculation unit 140 of the computing device 100 calculates fine dust observation data obtained from the first ground observatory and satellites during the period from November 1, 2021 to March 31, 2022. The results of calculating the observation error covariance with reference to each AOD observation data can be confirmed with reference to FIG. 9.

Figure 9 schematically shows an example of data resulting from calculating observation error covariance according to the observation point where air quality observation data was obtained, according to an embodiment of the present invention.

Referring to Figure 9, (a) is the result data of calculating the observation error covariance for the AOD observation data acquired from a satellite, and (b) is the ultrafine dust (PM2.5) observation data obtained from the first ground observatory. This is the result data of calculating the observation error covariance for .

First, checking (a), for the AOD observation data, the observation error covariance within Korea is small, so it can be judged that the accuracy of the observation data is high, and some regions in China (areas marked in red) Since the value of the observation error covariance is large, the accuracy of the observation data can be judged to be relatively low compared to Korea.

Next, checking (b), the observation error covariance at each of the first plurality of ground observatories (points indicated by dots) has been calculated for the ultrafine dust (PM2.5) observation data, and the observation error covariance has been calculated for the ultrafine dust (PM2.5) observation data, and the observation error covariance at each of the first plurality of ground observatories (points indicated by dots) has been calculated, and the observation error covariance has been calculated for the ultrafine dust (PM2.5) observation data. Since the observation error covariance at the first ground observatory is small, the accuracy of the observation data can be judged to be high, and similar to the AOD observation data, some regions in China (points marked with red dots) have a small value of the observation error covariance. Because the value is large, it can be judged that the accuracy of the observation data is relatively low compared to Korea.

In this way, in a state where the computing device 100 calculates the background error covariance of the model and the observation error covariance of the observation data, the process of obtaining the model by determining the initial field before data assimilation required to perform data assimilation is shown in FIG. 4. I will explain with reference.

Figure 4 schematically shows a process for determining the initial model length before data assimilation, according to an embodiment of the present invention.

Referring to FIG. 4, the computing device 100 acquires past weather data t _i corresponding to the past time t _i , which is a past time point adjacent to the current time t _c at a specific location and is a future time point than the t _p+k past time point. (S410), t _i past weather data is input into the weather model to cause the weather model to generate at least one t i+ _k weather forecast data corresponding to at least one t _i+k time point after a preset time from t _i . It can be printed (S420).

As an example, the data acquisition unit 130 of the computing device 100 performs preprocessing of weather forecast data corresponding to a past time t _i at a specific location simulated _through ECMWF into a form that can be input into a weather model, thereby Meteorological data can be obtained. At this time, i may be an integer greater than or equal to 0.

For example, if the past time t _i is 12 o'clock on March 14, 2022, if the computing device 100 inputs the acquired past weather data at 12 o'clock on March 14, 2022 into the weather model, the weather model will be 3 in 2022. Past weather forecast data predicted in hourly increments from 12:00 on the 14th of the month to 72 hours later can be output as t _i+k weather forecast data. According to the above example, k may be an integer from 0 to 72.

Thereafter, the computing device 100 acquires at least one t _i+k processed data, which is a value obtained by processing t _i+k weather prediction data (S430, S440), and inputs the t _i+k processed data into the air quality model. The air quality model is made to output at least one t _i+k air quality prediction data (S450), and the t _c air quality prediction data corresponding to t _c among the t _i+k air quality prediction data is used in the model before data assimilation. It can be determined as the initial field (S460).

As an example, the computing device 100 inputs t i _+k weather forecast data to MCIP for processing into data that can be input into an air quality model and emissions model, and allows MCIP to adjust at least some of the grid and domain, thereby allowing t _{i +k} meteorological pre-processing data is output, t _i+k meteorological pre-processing data is input into the emissions model, and the emission model outputs t _i+k emissions data, and t _i+k meteorological pre-processing data and t _i+k Emission data can be obtained as t _i+k processed data.

For example, the computing device 100 inputs t _i+k weather forecast data predicted in 1-hour increments from 12:00 on March 14, 2022 to 72 hours later into MCIP and allows MCIP to (i) enter the air quality model. For input, t _i+k weather preprocessed data from 12:00 on March 14, 2022 to 72 hours later is output, and (ii) grid resolution that allows t _i+k weather forecast data to be input into the emissions model. and domain to output t _i+k meteorological preprocessing data from 12:00 on March 14, 2022 to 72 hours later, and then input this into the emissions model to make the emissions model as of 12:00 on March 14, 2022. It is possible to output t _i+k emissions data up to 72 hours after the start. At this time, the t _i+k emission data output from the emission model may also be preprocessed to have a grid resolution of 27km so that it can be input into the air quality model in the same way as the t _i+k meteorological preprocessing data.

Afterwards, the computing device 100 inputs t _i+k weather preprocessing data output from MCIP and t _i+k emission data output from the emission model as t _i+k processed data into the air quality model, and converts them into an air quality model. It is possible to output t _i+k air quality forecast data from 12:00 on March 14, 2022 to 72 hours later.

In other words, the air quality model calculates aerosol component forecast data from 12:00 on March 14, 2022 to 72 hours later using t _i+k meteorological preprocessing data and t _i+k emission data, and converts it with reference to this. Ultrafine dust (PM2.5) forecast data can be output as t _i+k air quality forecast data. Alternatively, the air quality model may output AOD prediction data converted with reference to aerosol component prediction data as t _i+k air quality prediction data, but is not limited to this.

In addition, the computing device 100 performs data assimilation on the air quality forecast data as of 12 o'clock on March 15, 2022 among the air quality forecast data from 12:00 on March 14, 2022 to 72 hours later. It can be obtained by determining the model initial field.

For example, if you want to perform data assimilation at 12 o'clock on March 15, 2022, the computing device 100 performs data assimilation at 12 o'clock on March 15, 2022 through the WRF/CMAQ model process performed at 12 o'clock on March 14, 2022. Those skilled in the art will clearly understand that the air quality prediction data predicting the air quality of can be interpreted as being obtained by determining the initial field of the model before data assimilation.

In this way, with the initial model length before data assimilation obtained, the computing device 100 performs data assimilation to increase the accuracy of air quality prediction, thereby determining the model prediction length after data assimilation. Referring to FIG. 5. I will explain with reference.

Figure 5 shows the process of calculating the model initial length after data assimilation with reference to the model initial length before data assimilation and inputting this into the air quality model to determine the model prediction length at a future point, according to an embodiment of the present invention. It is schematically shown.

Referring to FIG. 5, the computing device 100 acquires current air quality observation data corresponding to t _c (S510), and calculates the model initial field before data assimilation, current air quality observation data, observation error covariance, and background error covariance. With reference to this, the model initial length can be calculated after data assimilation (S520).

For example, in step S510, the data acquisition unit 130 of the computing device 100 uses ultrafine dust (PM2.5) observation data observed from the first ground observation station and AOD observation data observed from GEMS as current air quality observation data. At least some of the data can be acquired. Before acquiring the current air quality observation data, as previously described with reference to FIG. 7, a plurality of other observation points within a radius of influence of 68 km from the specific location where the current air quality observation data was acquired. A preset threshold (e.g. ), it is judged to be noise and preprocessing to remove it can be performed in advance.

And, in step S520, the air quality data assimilation unit 160 of the computing device 100 includes the current air quality observation data obtained in step S510, the background error covariance calculated in step S360, the observation error covariance calculated in step S380, and By referring to the model initial length before data assimilation obtained in step S460, the optimal interpolation method can be used to calculate the model initial length after data assimilation for application to the air quality model. Of course, it will not be limited to the optimal interpolation method.

As an example, the air quality data assimilation unit 160 of the computing device 100 refers to the observation operator, which is an operator for converting the initial field of the model before data assimilation into the observation space, the observation error covariance, and the background error covariance to model the model after data assimilation. The optimal weight for calculating the initial field is determined, and the optimal weight is applied to the observation increment, which is the difference between the current air quality observation data and the model background field interpolated to a specific location by applying an observation operator to the model initial field before data assimilation. By adding this to the initial length of the model before data assimilation, the initial length of the model after data assimilation can be calculated. This can be expressed in the formula as follows.

Equation (10)

Equation (11)

Equation (12)

Equation (13)

The above equation (10) shows calculating the model initial field after data assimilation with reference to the AOD observation data obtained from GEMS, and equation (11) shows the ultrafine dust (PM2.5) observation obtained from the first ground observatory. It shows calculating the model initial length after data assimilation by referring to the data. Equation (12) shows calculating the optimal weight for calculating the model initial length after data assimilation using AOD observation data, and Equation (13) ) shows calculating the optimal weight for calculating the model initial field after data assimilation of fine dust observation data.

는 자료동화 전 모델 초기장,

는 최적 가중치,

는 현재 대기질 관측데이터,

는 관측연산자,

는 배경 오차 공분산,

는 관측 오차 공분산을 나타낸 것일 수 있다.In the above formulas,

is the initial length of the model before data assimilation,

is the optimal weight,

is the current air quality observation data,

is the observation operator,

is the background error covariance,

may represent the observation error covariance.

At this time, in equations (12) and (13), (i) if the observation error covariance is smaller than the background error covariance, the optimal weight approaches 1, so that the model initial field after data assimilation can converge to the current air quality observation data, ( ii) If the observation error covariance is greater than the background error covariance, the optimal weight approaches 0, allowing the model initial length after data assimilation to converge to the model initial length before data assimilation.

In this way, when the air quality data assimilation unit 160 of the computing device 100 calculates the model initial field after data assimilation, fine dust, which is ground air quality observation data acquired from the first ground observatory, is used as current air quality observation data. You can see that it is divided into observation data and AOD observation data, which is satellite air quality observation data acquired from satellites.

Therefore, the air quality data assimilation unit 160 of the computing device 100 can calculate the model initial length after data assimilation depending on the presence or absence of current air quality observation data at a specific location obtained from the data acquisition unit 130. This will be explained with reference to FIG. 10.

Figure 10 schematically shows a process for confirming the existence of current air quality observation data and performing data assimilation by reflecting this, according to an embodiment of the present invention.

Referring to FIG. 10, the air quality data assimilation unit 160 of the computing device 100 can check the current air quality observation data at a specific location obtained from the data acquisition unit 130, where (a) is the current air quality. As quality observation data, it shows calculating the initial model length after data assimilation in the case where both ground air quality observation data and satellite air quality observation data exist, and (b) is the current air quality observation data, which is ground air quality observation data. It shows calculating the model initial field after data assimilation when only satellite air quality observation data exists, and (c) shows calculating the model initial field after data assimilation when only satellite air quality observation data exists. will be. For reference, in the case where neither ground air quality observation data nor satellite air quality observation data exist as current air quality observation data, the process of calculating the model initial field after data assimilation is not performed, and the air quality output from the air quality model is not performed. Quality prediction data can be used to forecast air quality, but it is not limited to this.

As an example, the air quality data assimilation unit 160 of the computing device 100 checks the current air quality observation data at a specific location obtained from the data acquisition unit 130 to obtain ground air quality observation data and satellite air quality observation data. If it is determined that all exist, as in (a), in the Planetary Boundary Layer (PBL), the ultrafine dust (PM2.5) observation data included in the ground air quality observation data is referred to for fine dust data. The model initial field can be calculated after assimilation, and in layers above the planetary boundary layer, the model initial field can be calculated after data assimilation for AOD by referring to the AOD observation data included in the satellite air quality observation data. At this time, since vertical mixing is very good in the planetary boundary layer, it is assumed that the composition ratio of aerosols is the same. If the model initial field is calculated after data assimilation only in the first layer of the air quality model (e.g., the ground surface layer), the initial field of the model is calculated. Because the effect disappears quickly, the model initial field is calculated after assimilating the data up to the planetary boundary layer using fine dust observation data.

Specifically, the air quality data assimilation unit 160 of the computing device 100 calculates the model initial field after data assimilation in the planetary boundary layer by substituting the fine dust observation data obtained from the first ground observation station into the equation (11). can do.

However, the air quality data assimilation unit 160 of the computing device 100 uses AOD observation data acquired from a satellite to calculate the model initial field after data assimilation in the upper layer than the planetary boundary layer, and the AOD observation data is located at a specific location. It is expressed as the sum of the AOD values for all vertical layers from the ground, and there is a problem in not being able to distinguish AOD values by vertical layer.

Therefore, the air quality data assimilation unit 160 of the computing device 100 calculates the model initial field after data assimilation of the fine dust data in the planetary boundary layer using the fine dust observation data, and then calculates the AOD observation data. Using , certain preprocessing can be performed to prevent the situation in which the model initial field is calculated after data assimilation for AOD in the planetary boundary layer.

For example, the air quality data assimilation unit 160 of the computing device 100 refers to the ratio of the model initial field after data assimilation to the model initial field before data assimilation for fine dust data in the planetary boundary layer to generate aerosol component data in the planetary boundary layer. Calculate the model initial field after data assimilation, and refer to the model initial field after data assimilation of aerosol component data in the planetary boundary layer. After data assimilation of fine dust data in the planetary boundary layer, the model initial field is calculated as the model initial field in the planetary boundary layer. After data assimilation for AOD, it can be converted to model initial length. This can be expressed in a formula as follows.

Equation (14)

Equation (15)

In equation (14) above, shows the model initial field after data assimilation of aerosol data in the planetary boundary layer, represents the initial field of the model before data assimilation of aerosol data in the planetary boundary layer. Since it was previously assumed that the composition ratio of aerosols is the same in the planetary boundary layer, the air quality data assimilation unit 160 of the computing device 100 By referring to the ratio of the initial length of the model after data assimilation to the initial length of the model before data assimilation for fine dust data, can be calculated.

In addition, the air quality data assimilation unit 160 of the computing device 100 uses equation (15) to calculate the AOD value for each vertical layer. By substituting the model initial field after data assimilation for AOD in the planetary boundary layer, It can be converted to . At this time, represents the moisture absorption growth coefficient.

In addition, the air quality data assimilation unit 160 of the computing device 100 equates the model initial field with the AOD observation data in the planetary boundary layer after assimilating the data for the AOD in the planetary boundary layer, and identifies the current atmosphere in the layer above the planetary boundary layer. By replacing the quality observation data with the difference between the AOD observation data and the model initial field after data assimilation for AOD in the planetary boundary layer, the model initial field after data assimilation for AOD in the layer above the planetary boundary layer can be calculated using the formula: It is expressed as follows.

Equation (16)

In equation (16) above, shows the model initial field after data assimilation for AOD in the upper layer than the planetary boundary layer, represents the initial field of the model before data assimilation for AOD in the upper layer than the planetary boundary layer, shows AOD observation data acquired from a satellite.

As another example, the air quality data assimilation unit 160 of the computing device 100 checks the current air quality observation data at a specific location obtained from the data acquisition unit 130 and determines that only ground air quality observation data exists. , (b), the model initial field can be calculated after data assimilation of the fine dust data in the planetary boundary layer by referring to the ultrafine dust (PM2.5) observation data included in the air quality observation data. However, in this case, since AOD observation data was not obtained from the satellite, the model initial field may not be calculated after data assimilation in the upper layer than the planetary boundary layer.

As another example, the air quality data assimilation unit 160 of the computing device 100 checks the current air quality observation data at a specific location obtained from the data acquisition unit 130 and determines that only satellite air quality observation data exists. Then, the model initial field can be calculated after data assimilation in all vertical layers of the air quality model by referring to the AOD observation data included in the satellite air quality observation data. In other words, since the AOD observation data is the sum of AOD for all vertical layers, the model initial field can be calculated after assimilating the data for AOD in all vertical layers.

However, since the present invention seeks to confirm ultrafine dust (PM2.5) data as air quality prediction data output through an air quality model, the air quality data assimilation unit 160 of the computing device 100 provides data on AOD. When the model initial field is calculated after assimilation, assuming that fine dust data and AOD are directly proportional, the model initial field after data assimilation for AOD can be converted to the model initial field after data assimilation for fine dust data. , when expressed as a formula, it is as follows.

Equation (17)

Equation (18)

Equation (17) is the current air quality observation data. When both fine dust observation data and AOD observation data exist, the model initial field is calculated after data assimilation for AOD in the layer above the planetary boundary layer. It shows the conversion to the model initial field after data assimilation of dust data, and equation (18) is the current air quality observation data. When only AOD observation data exists, the model after data assimilation of AOD in all vertical layers This shows the conversion of the initial field to the model initial field after data assimilation of fine dust data in all vertical layers.

Referring again to FIG. 5, in a state where the computing device 100 calculates the model initial field after data assimilation, the current weather data corresponding to t _c is acquired at a specific location (S530), and the current weather data is added to the weather model. input to cause the weather model to output at least one t _c+k future weather prediction data corresponding to at least one t _c+k future time after a preset time from t _c (S540), and t _c+k future At least one t _c+k future processed data, which is a processed value of weather forecast data, can be obtained (S550, S560).

For example, in step S530, the data acquisition unit 130 of the computing device 100 preprocesses the current weather forecast data corresponding to t _C at a specific location simulated through ECMWF into a form that can be input into a weather model. Current weather data can be obtained.

For example, if t _c is 12 o'clock on March 15, 2022, in step S540, if the computing device 100 inputs the acquired current weather data at 12 o'clock on March 15, 2022 into the weather model, the weather model is 2022 Future weather forecast data predicted in hourly increments from 12:00 on March 15, 2018 to 72 hours later can be output as t _c+k future weather forecast data. According to the above example, k may be an integer from 0 to 72.

Then, in step S550, the computing device 100 inputs the t _c+k future weather prediction data into the MCIP for processing into a data form that can be input to at least some of the air quality model and the emissions model and allows the MCIP to adjust the domain t _c+k future weather preprocessing data is output, and in step S560, t _c+k future weather preprocessing data is input into the emissions model to cause the emissions model to output t _c+k future emissions data, and t _c+k future weather preprocessing data is input into the emissions model. Meteorological preprocessing data and t _c+k future emissions data can be obtained as t _c+k future processed data.

For example, the computing device 100 inputs t _c+k future weather forecast data predicted in hourly increments from 12:00 on March 15, 2022 to 72 hours later into MCIP and allows MCIP to create (i) an air quality model To output t _c+k future weather preprocessed data from 12:00 on March 15, 2022 to 72 hours later, (ii) input t _c+k future weather forecast data into the emissions model. Convert to possible grid resolution and domain to output t _c+k future weather preprocessing data from 12:00 on March 15, 2022 to 72 hours later, and then input this into the emissions model to make the emissions model produce 72 hours from 12:00 on March 15, 2022. You can output t _c+k future emissions data from 12:00 on the 15th to 72 hours later. At this time, the t _c+k future emissions data output from the emissions model may also be preprocessed to have a grid resolution of 27 km so that it can be input into the air quality model in the same way as the t _c+k future weather preprocessing data.

For reference, in Figure 5, for convenience of explanation, the model initial length is first calculated after data assimilation, and then t _{c + k} future processing data is numbered. However, on the contrary, t _{c + k} future processing data is acquired first, and then data assimilation is performed. It may be possible to calculate the post-model initial field.

Thereafter, the computing device 100 inputs the t _c+k future processed data and the model initial field after data assimilation into the air quality model to make the air quality model at least one t c+k future at t c _{+ k} future time. Air quality prediction data can be output (S570), and at least some of the t _c+k future air quality prediction data can be determined as a model prediction field at t _c+k future time (S580).

As an example, the computing device 100 includes t _c+k processed data including t c+k future weather preprocessing data output from MCIP and t _{c+k future} emissions data output from an emissions model, and an air quality data assimilation unit. After assimilating the data calculated from (160), the model initial field is input into the air quality model, and the air quality model is calibrated starting at 12:00 on March 15, 2022, from 12:00 on March 15, 2022 to 72 hours later. It is possible to output t _c+k air quality forecast data, and from this, the model forecast field at t _c+k future time can be determined depending on how many hours after the air quality forecast data to be forecast, and t _{c+ k} The model prediction field at a future point in time can be provided to the user as spatiotemporal data through the display system unit 170. However, the t _c+k air quality forecast data at 12:00 on March 15, 2022 is the same as the model initial field after data assimilation, so it can serve as an analysis field rather than a forecast field.

11A and 11B show the difference between the model initial length before data assimilation and the model initial length after data assimilation, and the difference between the model initial length before data assimilation and the model prediction length after data assimilation, respectively, according to an embodiment of the present invention. This is a schematic illustration of the data resulting from the comparison.

First, referring to Figure 11a, for the period from November 1, 2021 to March 31, 2022, the model prediction field compared to the ultrafine dust (PM2.5) observation data observed at the first ground observatory in Korea. As a scatter plot, it can be seen that the slope of the model initial field (1110) before data assimilation for fine dust data was 0.36, and the slope of the model initial field (1120) after data assimilation was 0.71.

Therefore, it can be seen that the initial field of the model before data assimilation underpredicted compared to the fine dust observation data, and since the initial field of the model after data assimilation predicted relatively closely compared to the fine dust observation data, the prediction performance was judged to have improved significantly. can do.

Referring to Figure 11b, (a) NMB (Normalized Mean Bias), (b) RMSE (Root Mean Square Error), and (c) IOA (Index of You can see that the result of comparing the Agreement is shown.

At this time, NMB can determine the tendency of the model's prediction field to under-predict or overpredict compared to the observed data, RMSE can determine the accuracy of the model, and IOA can determine the degree of agreement between the model's prediction field and the observed data. You can judge.

In (a), the NMB of the initial field of the model before data assimilation is -11.6. And the NMB of the initial length of the model after data assimilation is -8.7. Since it was found, it can be judged that the prediction performance has improved. At this time, if NMB has a positive value, it means that the model prediction field overpredicts compared to the observed data, and if NMB has a negative value, it may mean that the model prediction field underpredicts compared to the observed data.

In (b), the RMSE of the initial field of the model before data assimilation is 9.82. And, after data assimilation, the RMSE of the initial field of the model was 7.88. Since it was found, it can be judged that the prediction performance has improved. At this time, the value of RMSE is 0 The closer it is, the more similar the model prediction field is to the observed data.

In (c), the IOA of the initial length of the model before data assimilation was 0.83, and the IOA of the initial length of the model after data assimilation was 0.88, so it can be judged that the prediction performance has improved. At this time, IOA has a value between 0 and 1, and the closer it is to 1, the higher the model prediction field matches the observed data.

Additionally, the embodiments according to the present invention described above may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and usable by those skilled in the computer software field. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. media), and hardware devices specifically configured to store and perform program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include not only machine language code such as that created by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform processing according to the invention and vice versa.

In the above, the present invention has been described with specific details such as specific components and limited embodiments and drawings, but this is only provided to facilitate a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , a person skilled in the art to which the present invention pertains can make various modifications and variations from this description.

Therefore, the spirit of the present invention should not be limited to the above-described embodiments, and the scope of the patent claims described below as well as all modifications equivalent to or equivalent to the scope of the claims fall within the scope of the spirit of the present invention. They will say they do it.

Claims (22)

In a method of predicting air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model,
(a) A computing device, (i-1) acquires t _{p past weather data corresponding to t p} past time, which is at least one past time in a specific location, and uses the _{t p past} weather data to create a weather model (WRF, Weather Research and Forecasting) to cause the weather model to output at least one t _p+k past weather forecast data corresponding to at least one t p _+k past time after a preset time from t _p , At least one t _p+k past processed data, which is a processed value of the t _p+k past weather forecast data, is input into an air quality model (CMAQ, Community Multiscale Air Quality) to enable the air quality model to process the t _p+k past weather forecast data. Output at least one t _p+k past air quality forecast data corresponding to the time point, and (i-2) calculate the background error covariance of the air quality model by referring to the t _p+k past air quality forecast data. process; and (ii-1) t _p past air quality observation data corresponding to the t _p past time at the specific location - the t _p past air quality observation data includes air quality observation data obtained from at least some of the ground and satellites. Includes - obtaining and performing (ii-2) a process of calculating observation error covariance with reference to the t _p past air quality observation data;
(b) The computing device (i) acquires past weather data t _i corresponding to the past time t i, which is a past time point adjacent to the current time t _c at the specific location and is a future time _point than the t _p+k past time point. and (ii) inputting the t _i past weather data into the weather model to cause the weather model to output at least one t _i+k weather prediction data, and (iii) the t _i+k weather prediction data. Inputting at least one t _i+k processed data, which is a processed value, into the air quality model to cause the air quality model to output at least one t _i+k air quality prediction data;
(c) The computing device, (i) acquires current air quality observation data corresponding to t _c , and provides t _c air quality prediction data corresponding to t _c among the t _i+k air quality prediction data. After determining the model initial field before data assimilation, calculating the model initial field after data assimilation with reference to the model initial field before data assimilation, the current air quality observation data, the observation error covariance, and the background error covariance; and (ii) acquiring current weather data corresponding to t _c at the specific location, and inputting the current weather data into the weather model to cause the weather model to detect at least one t that is after a preset time from t _c . To output at least one t _{c+k future} weather forecast data corresponding to a c+k future point in time, and to obtain at least one t _c+k future processed data that is a processed value of the t _c+k future weather forecast data. performing a process; and
(d) The computing device inputs the t _{c + k} future processed data and the initial model length after data assimilation into the air quality model to cause the air quality model to have a model prediction length at the t _{c + k} future point in time. A step of outputting;
How to include . According to paragraph 1,
In step (c) above,
The computing device determines an optimal weight for calculating the model initial field after data assimilation by referring to an observation operator, which is an operator for converting the model initial field before data assimilation into an observation space, the observation error covariance, and the background error covariance. Determine, apply the optimal weight to the observation increment, which is the difference between the current air quality observation data and the model background field interpolated to the specific location by applying the observation operator to the model initial field before data assimilation, and apply this to the data. A method characterized by calculating the initial length of the model after assimilation of the data by adding it to the initial length of the model before assimilation. According to paragraph 2,
If the computing device determines that, as the current air quality observation data, (i) there is ground air quality observation data and satellite air quality observation data, the planetary boundary layer (PBL) includes the ground air quality observation data. After the data assimilation, the model initial field is calculated with reference to the included fine dust observation data, and in layers above the planetary boundary layer, the data is assimilated with reference to the AOD (Aerosol Optical Depth) observation data included in the satellite air quality observation data. Calculate the model initial field, and (ii) if it is determined that only the ground air quality observation data exists, the model is modeled after the data assimilation in the planetary boundary layer with reference to the fine dust observation data included in the ground air quality observation data. Calculate the initial field, and (iii) if it is determined that only the satellite air quality observation data exists, the data assimilation in all layers of the air quality model by referring to the AOD observation data included in the satellite air quality observation data. A method characterized by calculating the post-model initial field. According to paragraph 3,
When the computing device confirms the ground air quality observation data and the satellite air quality observation data as the current air quality observation data, (i) in the planetary boundary layer, the initial field of the model before the data assimilation for fine dust data Calculate the model initial field after data assimilation for the aerosol component data by referring to the ratio of the model initial field after the data assimilation, and calculate the model initial field after the data assimilation for the aerosol component data for the fine dust data. Converting the model initial field after the data assimilation into the model initial field after the data assimilation for AOD, (ii) in a layer above the planetary boundary layer, the current air quality observation data, the satellite air quality observation data, and the planet A method characterized in that the model initial field after the data assimilation for the AOD is calculated by replacing the difference in the model initial field after the data assimilation for the AOD in the boundary layer. According to paragraph 1,
In step (a) above,
The computing device, (i) when the t _p past air quality observation data corresponds to ground air quality observation data acquired from a first ground observation station, the first error inherent in the observation device and the setting method using the observation device A measurement error including at least part of the second error caused by the difference between and a representative error caused by the difference in resolution and scale of the t _p past air quality observation data and the t _{p + k} past air quality prediction data. Calculate each, and calculate the observation error for the ground air quality observation data using the measurement error and the representative error, and calculate the observation error for the ground air quality observation data by referring to the observation error and the average of the observation error to the t _p past air quality observation data. Calculate the observation error covariance for the t _p past air quality observation data, and (ii) if the t p past air quality observation data corresponds to satellite air quality observation data acquired from a satellite, according to the size of AOD according to the t _p past air quality observation data, The absolute error is calculated using the difference between the ground AOD obtained from the second ground observatory and the satellite AOD obtained from the satellite, and the absolute error and the average of the absolute error are referred to for the t _p past air quality observation data. A method characterized by calculating the observation error covariance. According to paragraph 1,
In step (a) above,
The computing device, the t _p+k past air quality prediction data and at least one t _p+ corresponding to the past time point t _{p+ kn, which is a past time point adjacent to the past time point t p+k} - where n is a positive number - _kn A method characterized in that the background error covariance is calculated by estimating the difference with past air quality prediction data as the background error of the air quality model. According to paragraph 1,
In step (a) above,
The computing device compares the t _p past air quality observation data at the specific location with the average value of air quality observation data previously measured at a plurality of observation points within a radius of influence centered on the specific location. A method characterized in that it is judged as noise when it is above a set threshold. According to paragraph 1,
In step (b) above,
A method characterized in that the computing device acquires weather forecast data at the time in the past t _i simulated through ECMWF (European Center for Medium-Range Weather Forecasts) as the past weather data in t _i . According to paragraph 1,
In step (c) above,
A method characterized in that the computing device acquires weather prediction data at t _c simulated through ECMWF as the current weather data. According to paragraph 1,
In step (b) above,
The computing device inputs the t _i+k weather prediction data into a MCIP (Meteorology Chemistry Interface Processor) to process the t i+k weather forecast data into a data form that can be input to the air quality model and emissions model (MEGAN/SMOKE), thereby allowing the MCIP to By adjusting t _i+k meteorological pre-processing data to be output, and inputting the t _i+k meteorological pre-processing data into the emission model to cause the emission model to output t _i+k emission data, the t _i+ A method characterized in that inputting _k meteorological preprocessing data and the t _i+k emission data into the air quality model as the t _i+k processed data. According to paragraph 1,
In step (c) above,
The computing device inputs the t _c+k future weather prediction data into an MCIP for processing into a data format that can be input to the air quality model and emissions model, and allows the MCIP to adjust the domain to pre-process the t _c+k future weather. output data, and input the t _c+k future weather preprocessing data into the emissions model to cause the emissions model to output t _c+k future emissions data, and input the t _c+k future weather preprocessing data and t A method characterized in that inputting _c+k future emission data into the air quality model as the t _c+k future processed data. In a computing device that predicts air pollution by applying air quality observation data obtained from at least some of the ground and satellites to an air quality model,
at least one memory storing instructions; and
At least one processor configured to execute the instructions,
The processor (I) (i-1) acquires t _p past weather data corresponding to t _p past time, which is at least one past time in a specific location, and uses the t _p past weather data to create a weather model (WRF, Weather Research and Forecasting) to cause the weather model to output at least one t _p+k past weather forecast data corresponding to at least one t p _+k past time after a preset time from t _p , At least one t _p+k past processed data, which is a processed value of the t _p+k past weather forecast data, is input into an air quality model (CMAQ, Community Multiscale Air Quality) to enable the air quality model to process the t _p+k past weather forecast data. Output at least one t _p+k past air quality forecast data corresponding to the time point, and (i-2) calculate the background error covariance of the air quality model by referring to the t _p+k past air quality forecast data. subprocess; and (ii-1) t _p past air quality observation data corresponding to the t _p past time at the specific location - the t _p past air quality observation data includes air quality observation data obtained from at least some of the ground and satellites. Includes - a process of obtaining and performing (ii-2) a subprocess of calculating observation error covariance with reference to the t _p past air quality observation data; (II) (i) Obtain past weather data t _i corresponding to the past time t _i , which is a past time point adjacent to the current time t _c at the specific location and is a time future than the past time point t _p+k , (ii) Input the t _i past weather data into the weather model to cause the weather model to output at least one t _i+k weather forecast data, and (iii) at least one value that is a processed value of the t _i+k weather forecast data. A process of inputting t _i+k processed data to the air quality model to cause the air quality model to output at least one t _i+k air quality prediction data; (III) (i) Obtain the current air quality observation data corresponding to t _c , and use the t _c air quality prediction data corresponding to t _c among the t _i+k air quality prediction data as the initial model before data assimilation. After determining as, a sub-process of calculating the model initial length after data assimilation with reference to the model initial length before data assimilation, the current air quality observation data, the observation error covariance, and the background error covariance; and (ii) acquiring current weather data corresponding to t _c at the specific location, and inputting the current weather data into the weather model to cause the weather model to detect at least one t that is after a preset time from t _c . To output at least one t _{c+k future} weather forecast data corresponding to a c+k future point in time, and to obtain at least one t _c+k future processed data that is a processed value of the t _c+k future weather forecast data. a process that performs a subprocess; and (IV) inputting the t _{c + k} future processed data and the model initial field after data assimilation into the air quality model to cause the air quality model to output a model prediction field at the t _{c + k} future time point. A computing device that performs a process; According to clause 12,
In process (III) above,
The processor determines optimal weights for calculating the model initial field after data assimilation by referring to an observation operator, which is an operator for converting the model initial field before data assimilation into an observation space, the observation error covariance, and the background error covariance. After applying the optimal weight to the observation increment, which is the difference between the current air quality observation data and the model background field interpolated to the specific location by applying the observation operator to the model initial field before the data assimilation, the data is assimilated. A computing device characterized in that it calculates the model initial length after the data assimilation by adding it to the previous model initial length. According to clause 13,
If the processor determines that, as the current air quality observation data, (i) there is ground air quality observation data and satellite air quality observation data, it is included in the ground air quality observation data in the planetary boundary layer (PBL) The model initial field is calculated after the data assimilation by referring to the fine dust observation data, and in layers above the planetary boundary layer, the model is calculated after the data assimilation by referring to the AOD (Aerosol Optical Depth) observation data included in the satellite air quality observation data. Calculate the initial field, and (ii) if it is determined that only the ground air quality observation data exists, the initial model after assimilating the data in the planetary boundary layer with reference to the fine dust observation data included in the ground air quality observation data. (iii) If it is determined that only the satellite air quality observation data exists, after assimilating the data in all layers of the air quality model with reference to the AOD observation data included in the satellite air quality observation data, A computing device characterized by calculating model initial fields. According to clause 14,
When the processor confirms the ground air quality observation data and the satellite air quality observation data as the current air quality observation data, (i) in the planetary boundary layer, compared to the initial field of the model before the data assimilation for fine dust data Calculate the model initial length after data assimilation for the aerosol component data with reference to the ratio of the model initial length after data assimilation, and calculate the model initial length after data assimilation for the aerosol component data with reference to the ratio of the model initial length after data assimilation to the fine dust data. Converting the model initial field after data assimilation to the model initial field after data assimilation for AOD, (ii) in a layer above the planetary boundary layer, the current air quality observation data, the satellite air quality observation data, and the planetary boundary layer A computing device characterized in that the model initial length after the data assimilation for the AOD is calculated by replacing the difference in the model initial length after the data assimilation for the AOD in . According to clause 12,
In process (I) above,
The processor, (i) when the t _p past air quality observation data corresponds to ground air quality observation data acquired from a first ground observation station, the first error inherent in the observation device and the setting method using the observation device A measurement error including at least part of the second error caused by the difference, and a representative error caused by the difference in resolution and scale of the t _p past air quality observation data and the t _{p + k} past air quality prediction data. Calculate each, calculate the observation error for the ground air quality observation data using the measurement error and the representative error, and refer to the observation error and the average of the observation error for the t _p past air quality observation data. Calculate the observation error covariance, and (ii) when the t _p past air quality observation data corresponds to satellite air quality observation data acquired from a satellite, according to the size of AOD according to the t _p past air quality observation data, 2 The absolute error is calculated using the difference between the ground AOD obtained from the ground observatory and the satellite AOD obtained from the satellite, and the absolute error and the average of the absolute error are referred to for the t _p past air quality observation data. A computing device characterized by calculating observation error covariance. According to clause 12,
In process (I) above,
The processor, the t _{p + k} past air quality prediction data and at least one t _{p + kn} corresponding to the past time point t p _{+ kn, which is a past time point adjacent to the t p + k} past time point - where n is a positive number - A computing device characterized in that the background error covariance is calculated by estimating the difference from past air quality prediction data as the background error of the air quality model. According to clause 12,
In process (I) above,
The processor compares the t _p past air quality observation data at the specific location with the average value of air quality observation data previously measured at a plurality of observation points within the radius of influence centered on the specific location, and preset A computing device characterized in that it is judged as noise when it is above a threshold. According to clause 12,
In process (II) above,
A computing device, characterized in that the processor acquires weather forecast data at the time in the past t _i simulated through ECMWF (European Center for Medium-Range Weather Forecasts) as the past weather data in t _i . According to clause 12,
In process (III) above,
A computing device, characterized in that the processor acquires weather prediction data at t _c simulated through ECMWF as the current weather data. According to clause 12,
In process (II) above,
The processor inputs the t _i+k weather forecast data into a MCIP (Meteorology Chemistry Interface Processor) to process the t i+k weather forecast data into a data form that can be input to the air quality model and emissions model (MEGAN/SMOKE), and causes the MCIP to create a domain. By adjusting, t _i+k meteorological pre-processing data is output, and the t _i+k meteorological pre-processing data is input into the emission model to cause the emission model to output t _i+k emission data, and the t _i+k A computing device characterized in that inputting meteorological preprocessing data and the t _i+k emission data into the air quality model as the t _i+k processed data. According to clause 12,
In process (III) above,
The processor inputs the t _c+k future weather forecast data into an MCIP for processing into a data format that can be input to the air quality model and emission model, and adjusts the domain of the MCIP to obtain t _c+k future weather preprocessing data. output, and input the t _{c + k} future weather preprocessing data into the emission model to cause the emission model to output t _{c + k} future emission data, and the t _{c + k} future weather preprocessing data and t _c A computing device characterized in that inputting _+k future emissions data into the air quality model as the t _c+k future processed data.