CN106407633A - Method and system for estimating ground PM2.5 based on space-time regression Kriging model - Google Patents

Abstract

The present invention provides a method and system for estimating ground PM2.5 based on a space-time regression kriging model, wherein the method includes: resampling the ground PM2.5 observation data in the area to be estimated into the created grid and performing matching; Among them, the matching process includes, in the created grid, averaging the ground PM2.5 observation data of all PM2.5 stations in the grid cell corresponding to the area to be estimated on the same day, and then averaging the data Assign the value to the corresponding grid unit; calculate the experimental variation function of the residual according to the ground PM2.5 observation data in the area to be estimated, and determine the space-time variation function model according to the experimental variation function of the residual; The variogram model was used for fitting; according to the fitting results of the spatiotemporal variogram model, the spatiotemporal regression Kriging model was used to estimate the ground PM2.5 concentration value in the area to be estimated. The estimation accuracy of PM2.5 can be improved through the invention.

Description

Space-time regression based kriging model Method and system for estimating ground PM2.5

technical field

The invention relates to the technical field of aerosol monitoring, and more specifically, to a method and system for estimating ground PM2.5 based on a time-space regression kriging model.

Background technique

With the rapid development of the economy and the sharp increase of man-made harmful gases such as industrial activities and motor vehicle exhaust, the air quality continues to deteriorate. PM2.5 refers to particulate matter with an aerodynamic particle size of less than 2.5 microns in the air. Compared with large particle size particles, PM2.5 particle size is small, rich in a large amount of toxic and harmful substances, and has a long residence time in the atmosphere and a long transportation distance. Therefore, it has a great impact on the human body and the quality of the atmospheric environment. A large number of epidemiological studies have proved that there is a certain relationship between PM2.5 and asthma, respiratory infections, lung cancer, and cardiovascular diseases. Therefore, the monitoring of PM2.5 has become one of the key points in the study of atmospheric environment.

Since the air monitoring data are often data with a certain time series, they have strong temporal and spatial variation characteristics, and the spatial interpolation method is the simplest and most effective method to obtain the distribution of pollutants in places with measurement stations.

At present, many scholars use spatial interpolation methods to estimate the distribution of pollutants. For example, the spatial interpolation method is applied to the research of nitrogen dioxide, ozone, PM10 and other pollutants, but it is rarely applied to the research of PM2.5, even if some researchers use the spatial interpolation method to study the distribution characteristics of PM2.5 , but these spatial interpolation methods do not consider the time distribution characteristics of PM2.5, and when the coverage of ground monitoring stations is insufficient, the estimation accuracy of PM2.5 is not high, so it is necessary to propose a spatial interpolation model with higher accuracy.

Contents of the invention

In view of the above problems, the object of the present invention is to provide a method and system for estimating ground PM2.5 based on a spatio-temporal regression Kriging model, so as to solve the problem that the estimation accuracy of ground PM2.5 is not high by existing spatial interpolation methods.

According to one aspect of the present invention, there is provided a method for estimating ground PM2.5 based on a space-time regression kriging model, comprising:

Create a grid according to the latitude and longitude range and resolution of the area to be estimated, resample the ground PM2.5 observation data of the area to be estimated into the created grid and perform matching; wherein, the matching process includes, in the grid, the Average the ground PM2.5 observation data monitored by all PM2.5 stations in the grid unit corresponding to the area to be estimated on the same day, and then assign the averaged data to the corresponding grid unit;

Calculate the experimental variogram of the residual according to the ground PM2.5 observation data in the area to be estimated, and then determine the spatiotemporal variogram model according to the experimental variogram of the residual;

The spatio-temporal variogram model was fitted by the least squares method;

According to the fitting results of the spatio-temporal variogram model, the spatio-temporal regression kriging model was used to estimate the ground PM2.5 concentration in the area to be estimated.

Among them, it also includes resampling the auxiliary variable data into the created grid and matching; wherein, the auxiliary variable data includes meteorological data, DEM and land use data.

Among them, in the process of resampling the auxiliary variable data into the created grid and performing matching, the resolution of the grid cell corresponding to the area to be estimated is compared with the resolution of the auxiliary variable data corresponding to the grid cell , the auxiliary variable data higher than the resolution of the corresponding grid unit is averaged and assigned to the corresponding grid unit, and the auxiliary variable data not higher than the resolution of the corresponding grid unit is assigned to the corresponding grid unit.

Among them, the auxiliary variable data corresponding to each PM2.5 station in the area to be estimated is extracted, and the multiple linear regression model between the PM2.5 observation data of each PM2.5 station and each auxiliary variable data is established. According to the calculation of the multiple linear regression model The results and PM2.5 observation data are used to obtain the ground PM2.5 residual value of the area to be estimated.

Among them, the interpolation results estimated by the space-time regression kriging model and the ground PM2.5 residual value are summed to obtain the ground PM2.5 concentration value in the area to be estimated.

In another aspect, the present invention provides a system for estimating ground PM2.5 based on a spatio-temporal regression kriging model, comprising:

The PM2.5 data matching unit is used to resample the ground PM2.5 observation data of the area to be estimated into the created grid and perform matching; wherein, the grid is created according to the latitude and longitude range and resolution of the area to be estimated, and the matching The process includes, in the grid, averaging the ground PM2.5 observation data monitored by all PM2.5 stations in the grid cell corresponding to the area to be estimated on the same day, and then assigning the averaged data to the corresponding network cell;

The spatio-temporal variogram model determination unit is used to calculate the experimental variogram of the residual error according to the ground PM2.5 observation data in the area to be estimated, and then determine the spatio-temporal variogram model according to the experimental variogram of the residual error;

A spatio-temporal variogram model fitting unit is used to fit the spatio-temporal variogram model by the method of least squares;

The PM2.5 concentration value estimation unit is used to estimate the ground PM2.5 concentration value of the area to be estimated by using the space-time regression kriging model according to the fitting result of the space-time variation function model.

Utilize the method and system for estimating ground PM2.5 based on the space-time regression Kriging model provided above according to the present invention, and verify the space-time correlation characteristics between the PM2.5 site data by using the spatial variation function and the time autocorrelation function, space-time If the correlation characteristic is high, it means that the data of the monitoring site is suitable for using the space-time regression kriging interpolation model. The multiple linear regression model among the auxiliary variables can reflect the impact factor of PM2.5; then, by calculating the spatio-temporal variogram of residuals after regression, an appropriate spatio-temporal variogram model is selected, and the time characteristic information is added to the model, so it can Better improve the estimation accuracy of PM2.5.

To the accomplishment of the above and related ends, one or more aspects of the invention comprise the features hereinafter described in detail and particularly pointed out in the claims. The following description and accompanying drawings detail certain exemplary aspects of the invention. These aspects are indicative, however, of but a few of the various ways in which the principles of the invention may be employed. Furthermore, the invention is intended to include all such aspects and their equivalents.

Description of drawings

By referring to the following description combined with the accompanying drawings and the contents of the claims, and with a more comprehensive understanding of the present invention, other objectives and results of the present invention will be more clear and easy to understand. In the attached picture:

Fig. 1 is the first flowchart of the method for estimating ground PM2.5 based on the space-time regression Kriging model according to an embodiment of the present invention;

Fig. 2 is the second schematic flow chart of the method for estimating ground PM2.5 based on the space-time regression Kriging model according to an embodiment of the present invention;

Fig. 3 is a logical structural block diagram of a system for estimating ground PM2.5 based on a space-time regression kriging model according to an embodiment of the present invention.

The same reference numerals indicate similar or corresponding features or functions throughout the drawings.

detailed description

Specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

Aiming at the problem that the aforementioned existing spatial interpolation method has low estimation accuracy of ground PM2.5, the present invention is to establish PM2.5 data by adding different auxiliary variable data (such as meteorological parameters, DEM, land use parameters, etc.) information The multiple linear regression model between the data and the auxiliary variable data, and then by calculating the spatio-temporal variation function of the residual after the regression, select the spatio-temporal variance function model, according to the spatio-temporal variance function model, the PM2. .5 Observational data uses the space-time regression kriging model to estimate the ground PM2.5 concentration, which can improve the estimation accuracy of PM2.5.

Before explaining the present invention, the space-time regression Kriging model will be described first.

Since the spatial interpolation method is the simplest and most effective method to obtain the distribution of pollutants in places with measurement stations, and the space-time regression kriging model (which is one of the spatial interpolation methods) has the characteristics of the best linear unbiased estimation It is used in the interpolation estimation of ozone and particulate matter at the local scale. The present invention proposes just on this basis.

In addition, it should be noted that PM2.5 in the present invention can also be written as PM _2.5 .

The method for estimating ground PM2.5 based on the space-time regression Kriging model provided by the present invention will be described in detail below.

FIG. 1 shows a first flow of a method for estimating ground PM2.5 based on a spatio-temporal regression kriging model according to an embodiment of the present invention.

As shown in Figure 1, the method for estimating ground PM2.5 based on the space-time regression kriging model provided by the present invention includes:

S110: Create a grid according to the latitude and longitude range and resolution of the area to be estimated, resample the ground PM2.5 observation data of the area to be estimated into the created grid and perform matching; wherein, the matching process includes, in the created grid In the grid, the ground PM2.5 observation data monitored by all PM2.5 stations in the grid unit corresponding to the area to be estimated is averaged on the same day, and then the averaged data is assigned to the corresponding grid unit.

S120: Calculate the experimental variation function of the residual according to the matched ground PM2.5 observation data of the area to be estimated, and determine the spatiotemporal variation function model according to the experimental variation function of the residual.

S130: Fitting the spatio-temporal variogram model by using the least square method.

S140: According to the fitting results of the spatio-temporal variogram model, use the spatio-temporal regression kriging model to estimate the ground PM2.5 concentration value in the area to be estimated.

Further, FIG. 2 shows the second process of the method for estimating ground PM2.5 based on the spatio-temporal regression kriging model according to an embodiment of the present invention, so as to further improve the estimation accuracy of ground PM2.5.

As shown in Figure 2, the method for estimating ground PM2.5 based on the space-time regression Kriging model provided by the present invention may include the following steps:

(1) S210: By using the spatial variation function and time correlation function to calculate the temporal-spatial correlation of PM2.5 observation data monitored by each PM2.5 station in the area to be estimated, investigate the distribution characteristics of PM2.5 stations.

Among them, if the calculation results show that there is a certain spatio-temporal correlation between the PM2.5 observation data monitored by the PM2.5 station, the spatial-temporal regression kriging interpolation model can be used to estimate the ground PM2.5 concentration in the area.

Specifically, the calculation method of the spatial variation feature is shown in Formula 1, and the calculation method of the time-related feature is shown in Formula 2 and Formula 3.

(Formula 1)

(Formula 2)

(Formula 3)

Among them, γ(h) in Formula 1 is the half of the variance of the difference between the values Z(s _i ) and Z(s _i +h) of the regionalized variables at s _i and s _i +h; h is the distance between two points , also known as the lag distance, N(h) is the logarithm of the samples used to calculate the variation function value between the distance h.

In formulas 2 and 3, n is the number of time series variables y _n , and y _nt is a variable with a time interval t distance from y _n , is the autocorrelation function of time distance t, is the mean value.

S220: Resample the collected ground PM2.5 data, meteorological data, DEM and land use data of the area to be estimated into the created grid and perform matching.

Specifically, the process of resampling the above data into the created grid for matching includes: determining the resolution of the grid corresponding to the area to be estimated. Among them, the matching method of PM2.5 data is: the ground PM2.5 observation data monitored by all PM2.5 stations in the grid unit corresponding to the area to be estimated on the same day is averaged and then assigned to the corresponding grid unit; The matching method of variable data (including meteorological data, DEM and land use data) is: compare the resolution of the grid cell corresponding to the area to be estimated with the resolution of the auxiliary variable data corresponding to the grid cell, which will be higher than The auxiliary variable data corresponding to the resolution of the grid unit is averaged and assigned to the corresponding grid unit, and the auxiliary variable data not higher than the resolution of the corresponding grid unit is assigned to the corresponding grid by using the interpolation method of inverse distance weighting unit.

It should be noted that during the matching process, the PM2.5 data comes from the environmental monitoring department, the meteorological data comes from the RAMS (Regional Atmospheric Modeling System, Regional Atmospheric Modeling System) model simulation, and the DEM comes from the SRTM (Shuttle Radar Topography Mission , space shuttle radar topographic mapping mission) 90m resolution elevation data, land use data from MODISLand Cover products; interpolation of meteorological data in RAMS into 3km resolution grid, DEM, land use data resampled to 3km grid network.

S230: Extract the auxiliary variable data corresponding to each PM2.5 station in the area to be estimated, establish a multiple linear regression model between the PM2.5 observation data of each PM2.5 station and each auxiliary variable data, and calculate according to the multiple linear regression model The results are combined with the PM2.5 observation data monitored by the PM2.5 station to obtain the ground PM2.5 residual value of the area to be estimated.

Specifically, the multiple linear regression model was used to establish the relationship between PM2.5 and auxiliary variables, and the least square method was used for fitting:

PM _2.5 ＝β+α ₁ *RH+α ₂ *PBL+α ₃ *Wind+α ₄ *DEM+α ₅ *LandUse (Formula 4)

In the formula, RH represents relative humidity, PBL represents boundary layer height, DEM represents elevation, LandUse represents land use type, β represents intercept, and α ₁ to α ₅ represent the slopes of auxiliary variables. The model obtained by formula 4 was initially used to calculate the PM2.5 trend, and the residual error between the observed PM2.5 value and the PM2.5 trend (that is, the PM2.5 observed data and the calculation result of the multiple linear regression model) was calculated.

S240: Calculate the experimental variation function of the residual.

Specifically, assuming that the observation value (ie PM2.5 observation data) at a spatial point s _i (i=1,...,n) in the area to be studied (ie, the area to be estimated) is z(s _i ), the point to be estimated (i.e. a point in the area to be estimated) the estimated value of the attribute at s ₀ is the weighted sum of the attribute values of the surrounding n known sample points, and the formula is expressed as:

(Formula 5)

In the formula, λ _i (i=1,…n) is the undetermined weight coefficient, is the estimated value of the spatial point s ₀ , and z(s _i ) is the attribute value of the sample point s _i . The spatio-temporal regression Kriging model is established under the condition of unbiased and minimum estimated variance, and Kriging equations can be established according to this condition.

(Formula 6)

In the formula, γ(s _i , s _j ) represents the variation function between the i-th point and the j-th point. It can be seen from formula 6 that the kriging weight λ can be calculated by calculating the variation function, and then the value of the point to be estimated ( That is, the estimated value at the point to be estimated) and the estimation error (the error between the estimated value at the point to be estimated and the actual observed value at the point to be estimated).

In the present invention, formula 7 is used to calculate the experimental variogram of the residual, and a suitable spatiotemporal variogram model is selected according to the distribution of the experimental variogram of the residual.

γ _st (h _s ,h _t )=0.5Var[Z(s+h _s ,t+h _t )-Z(s,t)] (Formula 7)

where γ _st (h _s ,h _t ) is the value Z( _s , _t ) and Z(s+h _s ,t+h _t ) half of the variance of the difference; h _s is the distance between two points in space, and h _t is the distance in time.

S250: According to the calculation result of step S240, construct and select a spatio-temporal variogram model that can describe the spatio-temporal characteristics of the PM2.5 observation data set, and use the least squares method to fit the spatio-temporal variogram model, and calculate the spatio-temporal variogram model parameters.

Among them, the parameters of the spatio-temporal variogram model include: the nugget value, sill value, and range of the space part; the nugget value, sill value, and range of the time part.

S260: Use the fitted spatio-temporal variogram model to perform spatio-temporal regression Kriging interpolation calculations, and sum the Kriging interpolation results and the trend in step S230 (the interpolation results estimated by using the spatio-temporal regression Kriging model and the ground PM2.5 residual value calculated in step S230 are summed and summarized), and the ground PM2.5 concentration value of the area to be estimated is estimated.

Furthermore, ten-fold cross-validation method can also be used to cross-validate the interpolation results of the space-time regression Kriging model. For the model that meets the accuracy requirements, it is applied to PM2.5 estimation, and the final result is output as a .tff format raster grid data.

It should be noted that, in the present invention, adding auxiliary variable data (i.e. meteorological data, DEM and land use data) is in order to further improve the estimation accuracy of PM2. In S220, the meteorological data, DEM and land use data of the area to be estimated may not be resampled into the created grid and matched, then there is no need to execute step S230, but directly enter step S240, so that The process shown in Figure 1.

Corresponding to the above method, the present invention provides a system for estimating ground PM2.5 based on a space-time regression kriging model. FIG. 3 shows the logical structure of the system for estimating ground PM2.5 based on the spatio-temporal regression Kriging model according to an embodiment of the present invention.

As shown in Figure 3, the system 300 for estimating ground PM2.5 based on the spatiotemporal regression Kriging model provided by the present invention includes a PM2.5 data matching unit 310, a spatiotemporal variogram model determination unit 320, and a spatiotemporal variogram model fitting unit 330 and PM2.5 concentration value estimation unit 340.

Wherein, the PM2.5 data matching unit 310 is used for resampling the ground PM2.5 observation data of the region to be estimated into the created grid and matching; The process of creating and matching includes, in the created grid, averaging the ground PM2.5 observation data of all PM2.5 stations in the grid cell corresponding to the area to be estimated on the same day, and then averaging the data Assigned to the corresponding grid unit.

The spatio-temporal variogram model determination unit 320 is used to calculate the experimental variogram of residuals based on the matched ground PM2.5 observation data in the region to be estimated, and determine the spatio-temporal variogram model according to the experimental variogram of residuals.

The spatio-temporal variogram model fitting unit 330 is used to fit the spatio-temporal variogram model using the least square method.

The PM2.5 concentration value estimating unit 340 is used for estimating the surface PM2.5 concentration value of the area to be estimated by using the space-time regression kriging model according to the fitting result of the space-time variation function model.

Further, in order to improve the estimation accuracy, the system 300 for estimating ground PM2.5 based on the space-time regression kriging model provided by the present invention may further include an auxiliary data matching unit 350 and a PM2.5 residual value obtaining unit 360 . Wherein, the auxiliary data matching unit 350 is used to resample the auxiliary variable data into the created grid and perform matching; wherein, the auxiliary variable data includes meteorological data, DEM and land use data; the PM2.5 residual value acquisition unit 360 uses Based on the extracted auxiliary variable data corresponding to each PM2.5 station in the area to be estimated, a multiple linear regression model between the PM2.5 observation data of each PM2.5 station and each auxiliary variable data was established. According to the multiple linear regression model The calculation results and PM2.5 observation data are used to obtain the ground PM2.5 residual value of the area to be estimated.

Specifically, when the auxiliary data matching unit 350 has auxiliary variable data matching, the PM2.5 residual value acquisition unit 360 is used to obtain the ground PM2.5 residual value of the area to be estimated, and then in the PM2.5 concentration value estimation unit 340 According to the fitting result of the spatio-temporal variogram model fitting unit 330, the interpolation result estimated by the spatio-temporal regression kriging model and the PM2.5 residual value acquisition unit 360 to obtain the ground PM2.5 residual value of the area to be estimated are carried out. Sum up and summarize to estimate the PM2.5 concentration value of the area to be estimated.

It can be seen from the above that the method and system for estimating ground PM2.5 based on the spatio-temporal regression kriging model provided by the present invention can establish a multiple linear regression model between PM2.5 data and auxiliary variables by adding different auxiliary variable data , which can reflect the impact factor of PM2.5; then, by calculating the spatio-temporal variogram of the residual after regression, an appropriate spatio-temporal variogram model is selected, which adds time characteristic information, so it can better improve the estimation of PM2.5 precision.

The method and system for estimating ground PM2.5 based on the spatio-temporal regression Kriging model according to the present invention are described above by way of example with reference to the accompanying drawings. However, those skilled in the art should understand that for the method and system for estimating ground PM2.5 based on the spatio-temporal regression Kriging model proposed in the present invention, various improvements can be made without departing from the content of the present invention. Therefore, the protection scope of the present invention should be determined by the contents of the appended claims.

Claims (10)

1. A method for estimating ground PM2.5 based on a space-time regression Kriging model, comprising: Create a grid according to the latitude and longitude range and resolution of the area to be estimated, resample the ground PM2.5 observation data of the area to be estimated into the created grid and perform matching; wherein, the matching process includes: In the grid, the ground PM2.5 observation data monitored by all PM2.5 stations in the grid unit corresponding to the area to be estimated is averaged on the same day, and then the averaged data is assigned to the corresponding grid unit; According to the ground PM2.5 observation data of the area to be estimated that is matched, calculate the experimental variogram of residual error, then determine the spatiotemporal variogram model according to the experimental variogram of described residual error; The spatio-temporal variogram model is fitted by the method of least squares; According to the fitting results of the spatio-temporal variogram model, the spatio-temporal regression kriging model is used to estimate the ground PM2.5 concentration value in the area to be estimated. 2. the method for estimating ground PM2.5 based on space-time regression Kriging model as claimed in claim 1, also comprises: The auxiliary variable data is resampled to the created grid and matched; wherein, the auxiliary variable data includes meteorological data, DEM and land use data. 3. The method for estimating ground PM2.5 based on the space-time regression kriging model as claimed in claim 2, wherein, in the process of resampling auxiliary variable data into the grid created and matching, Compare the resolution of the grid unit corresponding to the area to be estimated with the resolution of the auxiliary variable data corresponding to the grid unit, average the auxiliary variable data higher than the resolution of the corresponding grid unit and assign it to the corresponding Grid unit, the auxiliary variable data not higher than the resolution of the corresponding grid unit is assigned to the corresponding grid unit by using an interpolation method with inverse distance weighting. 4. the method for estimating ground PM2.5 based on space-time regression Kriging model as claimed in claim 3, also comprises, Extract the auxiliary variable data corresponding to each PM2.5 station in the area to be estimated, establish a multiple linear regression model between the PM2.5 observation data of each PM2.5 station and each auxiliary variable data, and calculate according to the multiple linear regression model The result and the PM2.5 observation data are used to obtain the ground PM2.5 residual value of the area to be estimated. 5. the method for estimating ground PM2.5 based on space-time regression kriging model as claimed in claim 4, wherein, will adopt the interpolation result obtained by space-time regression kriging model estimation and described ground PM2.5 residual value to carry out Sum and summarize to obtain the ground PM2.5 concentration value of the area to be estimated. 6. the method for estimating ground PM2.5 based on space-time regression Kriging model as claimed in claim 1, also comprises: Calculate the temporal-spatial correlation of PM2.5 observation data monitored by each PM2.5 station in the area to be estimated according to the spatial variation function and temporal correlation function; among them, When there is a temporal-spatial correlation between the PM2.5 observation data monitored by each PM2.5 station, the temporal-spatial regression Kriging model is used to estimate the surface PM2.5 in the area to be estimated. 7. The method for estimating ground PM2.5 based on the space-time regression kriging model as claimed in claim 1, wherein, the fitting result of the space-time regression kriging model is cross-validated using a ten-fold cross-validation method. 8. A system for estimating ground PM2.5 based on a space-time regression kriging model, comprising: The PM2.5 data matching unit is used to resample the ground PM2.5 observation data of the area to be estimated into the created grid and perform matching; wherein, the grid is created according to the latitude and longitude range and resolution of the area to be estimated , the matching process includes, in the grid, averaging the ground PM2.5 observation data monitored by all PM2.5 stations in the grid cell corresponding to the area to be estimated on the same day, and then averaging the The data is assigned to the corresponding grid unit; A spatio-temporal variogram model determining unit is used to calculate the experimental variogram of the residual according to the ground PM2.5 observation data in the area to be estimated, and then determine the spatio-temporal variogram according to the experimental variogram of the residual; A spatiotemporal variogram model fitting unit, configured to fit the spatiotemporal variogram model using the least squares method; The PM2.5 concentration value estimating unit is used for estimating the ground PM2.5 concentration value of the area to be estimated by using the space-time regression kriging model according to the fitting result of the space-time variation function model. 9. The system of estimating ground PM2.5 based on space-time regression Kriging model as claimed in claim 8, further comprising: The auxiliary data matching unit is configured to resample the auxiliary variable data into the created grid and perform matching; wherein, the auxiliary variable data includes meteorological data, DEM and land use data. 10. The system of estimating ground PM2.5 based on space-time regression Kriging model as claimed in claim 9, further comprising: The PM2.5 residual value acquisition unit is used to establish a multivariate relationship between the PM2.5 observation data of each PM2.5 station and each auxiliary variable data based on the extracted auxiliary variable data corresponding to each PM2.5 station in the area to be estimated. The linear regression model, according to the calculation result of the multiple linear regression model and the PM2.5 observation data, obtains the ground PM2.5 residual value of the area to be estimated.