CN111859304B - Satellite aerosol missing prediction method and system based on space-time autocorrelation - Google Patents
Satellite aerosol missing prediction method and system based on space-time autocorrelation Download PDFInfo
- Publication number
- CN111859304B CN111859304B CN202010738199.3A CN202010738199A CN111859304B CN 111859304 B CN111859304 B CN 111859304B CN 202010738199 A CN202010738199 A CN 202010738199A CN 111859304 B CN111859304 B CN 111859304B
- Authority
- CN
- China
- Prior art keywords
- aod
- grid
- target
- satellite
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 239000000443 aerosol Substances 0.000 title claims abstract description 22
- 238000012952 Resampling Methods 0.000 claims abstract description 25
- 238000005070 sampling Methods 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 11
- 238000012217 deletion Methods 0.000 claims description 17
- 230000037430 deletion Effects 0.000 claims description 17
- 230000007812 deficiency Effects 0.000 claims description 16
- 230000005540 biological transmission Effects 0.000 claims description 14
- 238000001556 precipitation Methods 0.000 claims description 9
- 238000004088 simulation Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 230000002123 temporal effect Effects 0.000 claims description 3
- 230000002354 daily effect Effects 0.000 description 25
- 238000012545 processing Methods 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 230000008821 health effect Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Algebra (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a satellite aerosol missing prediction method and system based on space-time autocorrelation, wherein the method comprises the following steps: creating grids covering the target area, and obtaining projection coordinates of a central point of each grid; resampling the AOD of the target area and variables mediating satellite AOD space-time autocorrelation into grids respectively, matching the grids with data, and forming a single data set from daily data; establishing a prediction data set based on satellite AOD space-time autocorrelation; repeatedly sampling the predicted data set for a plurality of times, and constructing a missing prediction model based on a generalized addition model for each sampling; and respectively predicting the AOD missing data based on the established multiple missing prediction models, and taking the average value of multiple predictions of each grid as a final AOD predicted value. The method predicts the loss of the satellite AOD while considering that the satellite AOD has time and space autocorrelation. The index of the obtained prediction accuracy reflects that the method has high missing prediction accuracy by comparing the prediction value with the remote sensing monitoring value.
Description
Technical Field
The invention relates to the technical field of satellite remote sensing aerosol missing prediction, in particular to a satellite aerosol missing prediction method and system based on space-time autocorrelation.
Background
PM2.5 refers to particles with aerodynamic particle size smaller than 2.5 μm in air, and a large number of researches at home and abroad prove that PM2.5 can cause damage to human health both in long-term and short-term exposure. However, due to the fact that the PM2.5 ground monitoring station often has defects in time and space, particularly the lack of continuous ground monitoring of PM2.5 before 2013 in China, estimation of pollution concentration and evaluation of related health effects are seriously affected. Estimating the ground PM2.5 based on satellite remote sensing Aerosol Optical Depth (AOD) provides a new idea to solve this problem. The satellite monitoring can carry out periodic and large-range dynamic scanning on the atmosphere, thereby compensating for the time and space deficiency of the ground monitoring station. Compared with the traditional ground monitoring, the method has the characteristics of wide coverage area, long accumulation time, high spatial resolution, lower cost and the like, and is widely applied to the prediction of space-time distribution of ground atmospheric particulate matters.
Despite rapid development, satellite AOD-based PM2.5 estimation remains a challenge. Satellite remote sensing AOD is easily affected by cloud coverage and bright ground surface, and in a period with more cloud layers and areas covered by ice and snow, a large number of satellite AODs are in non-random missing conditions, so that PM2.5 estimation fails in the periods or areas. At present, when the ground PM2.5 concentration is predicted based on satellite AOD data, the AOD deficiency is mostly not considered or other sources of AOD data are adopted for rough filling, so that the PM2.5 estimated value is largely deficient or the accuracy is reduced, and the evaluation of PM2.5 exposure and the estimation of health effect are influenced. In view of this, there is an urgent need to develop an AOD missing prediction technique that completely fills in the AOD missing values (100% filling), improves the time and space coverage, and ensures good accuracy of the satellite AOD data.
Disclosure of Invention
The invention aims to provide a satellite aerosol missing prediction method based on space-time autocorrelation, which predicts the missing of satellite AOD and has high accuracy.
In order to solve the above problems, the present invention provides a satellite aerosol missing prediction method based on space-time autocorrelation, which includes:
creating grids covering the target area, and obtaining projection coordinates of a central point of each grid;
resampling the AOD of the target area and variables mediating satellite AOD space-time autocorrelation into the grids respectively, matching the grids with data, and forming a single data set from daily data;
based on the satellite AOD space autocorrelation, respectively adding grid coordinates, square terms and product terms thereof in the independent data sets, simultaneously, based on the satellite AOD time autocorrelation, merging the independent data sets of the target deficiency date and the front and rear days thereof, adding corresponding day sequence variables, and establishing a prediction data set;
repeatedly sampling the prediction data set for a plurality of times, and constructing a deletion prediction model based on a generalized addition model for each sampling;
and respectively predicting the AOD missing data based on the established multiple missing prediction models, and taking the average value of multiple predictions of each grid as a final AOD predicted value.
As a further improvement of the present invention, the creating a grid covering the target area specifically includes:
based on the projection coordinates of the AOD data pixel points and the target spatial resolution, establishing a Thiessen polygon;
a mesh is created that covers the target area based on the tesen polygon.
As a further improvement of the present invention, the creating a Thiessen polygon based on the projection coordinates of the AOD data pixel points and the target spatial resolution includes:
when the target spatial resolution is consistent with the AOD product resolution, the projection coordinates of the AOD data pixel points can be directly utilized to establish a Thiessen polygon; when the target spatial resolution is higher or lower than the AOD product resolution, selecting dot information consistent with the target resolution to establish a Thiessen polygon.
As a further improvement of the present invention, before the creating the grid covering the target area, the method further includes:
judging whether the satellite AOD and the cloud parameter products are derived from the same satellite, and respectively matching and fusing the AOD and the cloud parameters of different sources when the satellite AOD and the cloud parameters are derived from different satellites to obtain a daily AOD average value and a daily cloud coverage average value;
and judging the time resolution of the meteorological data, and when the time resolution of the meteorological data is higher than that of each day (namely, the hour average or the minute average), reducing the time scale and calculating the average value of the daily meteorological variables.
As a further improvement of the present invention, the predicting AOD missing data based on the established multiple models, respectively, and after taking the average value of multiple predictions of each grid as the final AOD predicted value, further includes:
and (3) for the grid with the satellite AOD not missing, after the satellite AOD monitoring value is matched with the final AOD predicted value, a simple linear correlation model is established, and model prediction accuracy is evaluated according to a determination coefficient (R2) and a Root Mean Square Error (RMSE).
As a further refinement of the invention, the resampling of the AOD of the target area and the variables mediating satellite AOD space-time autocorrelation into the grid, respectively, and matching the grid with the data, forming daily data into separate data sets, comprises:
when the target spatial resolution is consistent with the satellite AOD product resolution, matching the satellite AOD into the grid according to the grid center point coordinates; when the target spatial resolution is lower than the AOD product resolution, resampling the AOD into a target grid by adopting a spatial interpolation method; when the target spatial resolution is higher than the AOD product resolution, performing spatial superposition on the AOD grids and the target grids, and further calculating an AOD average value in each target grid;
resampling the atmospheric transmission model simulation AOD daily average value into a target grid by adopting a spatial interpolation method; the daytime cloud parameter mean value is matched with a target grid by adopting a space superposition method; resampling the meteorological parameters into a target grid by adopting a spatial interpolation method; the elevation data grids are spatially overlapped with the target grids, and the elevation mean value in each target grid is calculated;
and matching the resampled satellite AOD with the prediction variable according to the grid center point position, so as to obtain a single data set of daily AOD missing prediction.
As a further improvement of the present invention, the adding grid coordinates and their square terms and product terms in the dataset based on satellite AOD spatial autocorrelation includes:
and selecting grid center point projection coordinates (X and Y) and square terms and product terms of the coordinates to represent prediction indexes of satellite AOD spatial distribution trend and spatial autocorrelation, and respectively generating projection coordinates X, projection coordinates Y, X square terms, square terms of Y and product term variables of X and Y in a data set.
As a further improvement of the present invention, the combining the data sets of the target deficiency date and the two days before and after the target deficiency date based on the satellite AOD time autocorrelation, and adding the corresponding day sequence variable, includes:
selecting a data set of the current day of the target filling date, and assigning the day sequence variable of the data set to be 3; the target fills up the data sets of the two days before and the day before, and the day sequence variable is respectively assigned to 1 and 2; the target fills the dataset one day and two days after the date, assign its day sequence variables to 4 and 5, respectively.
As a further improvement of the present invention, the deletion prediction model is constructed as follows:
in the formula, AOD gt Is AOD data for grid g on day t; beta 1 -β 8 Is the slope of the prediction index; CTM (CTM) gt AOD data on day t for grid g simulated by the atmospheric transmission model; cloud gt Cloud parameter data for grid g on day t; tem (Tem) gt 、RH gt 、BLH gt 、Al gt And TCW gt The average temperature, the relative humidity, the boundary layer height, the ground reflection and the total precipitation of the grid g at the t-th balance are respectively; elev g Is the elevation of grid g; d (D) t Is a dumb variable of the day; x is X g And Y g Geographic coordinate items of grid g; epsilon gt Is the error term for grid g on day t.
In order to solve the above problems, the present invention further provides a satellite aerosol missing prediction system based on space-time autocorrelation, which includes:
the grid creation module is used for creating grids covering the target area and obtaining projection coordinates of center points of each grid;
the data matching module is used for resampling the AOD of the target area and the variable mediating satellite AOD space-time autocorrelation into the grids respectively, matching the grids with the data and forming the daily data into an independent data set;
the prediction data set establishing module is used for respectively adding grid coordinates, square items and product items thereof into the independent data sets based on the satellite AOD space autocorrelation, merging the independent data sets of the target deficiency date and the front and rear days based on the satellite AOD time autocorrelation, and adding corresponding day sequence variables to establish a prediction data set;
the missing prediction model construction module is used for repeatedly sampling the prediction data set for a plurality of times, and each sampling is used for constructing a missing prediction model based on a generalized addition model;
and the missing prediction module predicts the AOD missing data based on the established multiple missing prediction models respectively, and takes the average value of multiple predictions of each grid as a final AOD predicted value.
The invention has the beneficial effects that:
the satellite aerosol deletion prediction method and the system based on the space-time autocorrelation predict the deletion of the satellite AOD while considering that the satellite AOD has the time and space autocorrelation. The index of the obtained prediction accuracy reflects that the method is high in missing prediction accuracy by comparing the prediction value with the remote sensing monitoring value, and has great significance for environmental protection.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention, as well as the preferred embodiments thereof, together with the following detailed description of the invention, given by way of illustration only, together with the accompanying drawings.
Drawings
FIG. 1 is a flow chart of a satellite aerosol loss prediction method based on spatio-temporal autocorrelation in a preferred embodiment of the present invention;
fig. 2 is a schematic diagram of a satellite aerosol loss prediction system based on space-time autocorrelation in a preferred embodiment of the present invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings and specific examples, which are not intended to be limiting, so that those skilled in the art will better understand the invention and practice it.
Example 1
As shown in fig. 1, a satellite aerosol deletion prediction method based on space-time autocorrelation in a preferred embodiment of the present invention comprises the following steps:
s10, creating grids covering the target area, and obtaining projection coordinates of a central point of each grid.
The creating a grid covering the target area specifically includes:
based on the projection coordinates of the AOD data pixel points and the target spatial resolution, establishing a Thiessen polygon;
a mesh is created that covers the target area based on the tesen polygon.
The construction of the Thiessen polygon based on the projection coordinates of the AOD data pixel points and the target spatial resolution specifically comprises the following steps:
when the target spatial resolution is consistent with the AOD product resolution, the projection coordinates of the AOD data pixel points can be directly utilized to establish a Thiessen polygon; when the target spatial resolution is higher or lower than the AOD product resolution, selecting dot information consistent with the target resolution to establish a Thiessen polygon.
Preferably, in consideration of boundary effects of mesh generation and data processing, the target area is enlarged by a certain buffer area when constructing the mesh for reducing errors in data processing and data calculation in the target area. Preferably, the buffer size is set to 100 km.
S20, resampling the AOD of the target area and the variable mediating satellite AOD space-time autocorrelation into the grids respectively, matching the grids with the data, and forming the daily data into a single data set. The method specifically comprises the following steps:
and selecting variables with space-time properties related to satellite AOD time and space distribution, resampling the daily AOD and predicted variable data of corresponding dates into a grid by adopting methods of spatial interpolation, spatial superposition, spatial averaging and the like, and matching to form a daily single data set.
Factors affecting or mediating the satellite AOD temporal and spatial distribution are extracted based on published evidence and the best performance of the model fitted using these factors is guaranteed. Selection of model prediction variables reference model decision coefficients (R2) and Root Mean Square Error (RMSE). The larger R2 and the smaller the RMSE, the better the model fit. The AOD prediction variables used in the embodiment of the invention comprise: the atmospheric chemical transmission model simulates AOD, MODIS cloud parameters, temperature, relative humidity, boundary layer height, albedo, precipitation, and altitude.
When the target spatial resolution is consistent with the satellite AOD product resolution, matching the satellite AOD into the grid according to the grid center point coordinates; when the target spatial resolution is lower than the AOD product resolution, resampling the AOD into a target grid by adopting a spatial interpolation method; when the target spatial resolution is higher than the AOD product resolution, performing spatial superposition on the AOD grids and the target grids, and further calculating an AOD average value in each target grid;
resampling the atmospheric transmission model simulation AOD daily average value into a target grid by adopting a spatial interpolation method; the daytime cloud parameter mean value is matched with a target grid by adopting a space superposition method; resampling the meteorological parameters into a target grid by adopting a spatial interpolation method; the elevation data grid is spatially superimposed with the target grids, and an elevation mean value in each target grid is calculated.
And matching the resampled satellite AOD with the prediction variable according to the grid center point position, so as to obtain a single data set of daily AOD missing prediction.
S30, based on satellite AOD space autocorrelation, grid coordinates, square items and product items thereof are respectively added in the independent data sets, meanwhile, based on satellite AOD time autocorrelation, the independent data sets of the target deficiency date and the front and rear days are combined, corresponding day sequence variables are added, and a prediction data set is established.
Based on satellite AOD space autocorrelation, grid coordinates and square terms and product terms thereof are added to the independent data set respectively, and the method comprises the following steps: and selecting grid center point projection coordinates (X and Y) and square terms and product terms of the coordinates to represent prediction indexes of satellite AOD spatial distribution trend and spatial autocorrelation, and respectively generating projection coordinates X, projection coordinates Y, X square terms, square terms of Y and product term variables of X and Y in a data set.
Based on satellite AOD time autocorrelation, combining data sets of target deficiency date and two days before and after the target deficiency date, and adding corresponding day sequence variables, wherein the method comprises the following steps:
selecting a data set of the current day of the target filling date, and assigning the day sequence variable of the data set to be 3; the target fills up the data sets of the two days before and the day before, and the day sequence variable is respectively assigned to 1 and 2; the target fills the dataset one day and two days after the date, assign its day sequence variables to 4 and 5, respectively.
And combining the five-day data sets for establishing a missing prediction model. The final output dataset is a CSV (Comma-separator values) formatted file.
S40, repeatedly sampling the prediction data set for a plurality of times, and constructing a deletion prediction model based on a generalized addition model by each sampling.
In order to solve the missing value variation due to the random error in the missing padding, in the present embodiment, the prediction data set formed in S30 is randomly sampled. A missing prediction model based on a generalized addition model is established using the sampled data set, and prediction variables related to satellite AOD time and space autocorrelation are added to the model. The sampling and modeling process was repeated ten times to build ten models in total. The specific flow is as follows:
s41: and randomly sampling the predicted data.
For the target date satellite AOD missing filling, firstly, a data set is generated in S4, and a predicted data set corresponding to the target date (comprising the front and rear days of the data set) is selected. And deleting the data record corresponding to the missing value of the satellite AOD in the data set. And randomly sampling the data records of the data set, and extracting 85% of data record samples each time to form a model fitting data set.
In this example, the process was repeated ten times to form ten different model fitting datasets in total. In other embodiments of the invention, the number of samples is not limited to ten, the greater the number the better.
S42: and constructing a deletion prediction model based on the generalized addition model.
And establishing a missing prediction model based on a generalized addition model, wherein model prediction variables comprise atmospheric and chemical transmission model simulation AOD, MODIS cloud parameters, temperature, relative humidity, boundary layer height, albedo, precipitation and altitude, and a projection coordinate X, a projection coordinate Y, X square term, a Y square term and a smoothing term of a product term of X and Y.
The constructed deletion prediction model is as follows:
in the formula, AOD gt Is AOD data for grid g on day t; beta 1 -β 8 Is the slope of the prediction index; CTM (CTM) gt AOD data on day t for grid g simulated by the atmospheric transmission model; cloud gt Cloud parameter data for grid g on day t; tem (Tem) gt 、RH gt 、BLH gt 、Al gt And TCW gt The average temperature, the relative humidity, the boundary layer height, the ground reflection and the total precipitation of the grid g at the t-th balance are respectively; elev g Is the elevation of grid g; d (D) t Is a dumb variable of the day; x is X g And Y g Geographic coordinate items of grid g; epsilon gt Is the error term for grid g on day t.
S50, respectively predicting the AOD missing data based on the established multiple missing prediction models, and taking the average value of multiple predictions of each grid as a final AOD predicted value.
In one embodiment, the following steps are further included after step S50:
and filling the grid with the loss of the satellite AOD by adopting a final AOD predicted value.
And (3) for the grid without loss of the satellite AOD, after the satellite AOD monitoring value is matched with the final AOD prediction mean value, a simple linear correlation model is established, and model prediction accuracy is evaluated according to a determination coefficient (R2) and a Root Mean Square Error (RMSE).
In one embodiment, the method further comprises the following steps before step S10:
judging whether the satellite AOD and the cloud parameter products are derived from the same satellite, and respectively matching and fusing the AOD and the cloud parameters of different sources when the satellite AOD and the cloud parameters are derived from different satellites to obtain a daily AOD average value and a daily cloud coverage average value;
and judging the time resolution of the meteorological data, and when the time resolution of the meteorological data is higher than that of each day (namely, the hour average or the minute average), reducing the time scale and calculating the average value of the daily meteorological variables.
Example two
A satellite aerosol loss prediction system based on space-time autocorrelation, comprising:
and the grid creation module is used for creating grids covering the target area and obtaining the projection coordinates of the center point of each grid.
The creating a grid covering the target area specifically includes:
based on the projection coordinates of the AOD data pixel points and the target spatial resolution, establishing a Thiessen polygon;
a mesh is created that covers the target area based on the tesen polygon.
The construction of the Thiessen polygon based on the projection coordinates of the AOD data pixel points and the target spatial resolution specifically comprises the following steps:
when the target spatial resolution is consistent with the AOD product resolution, the projection coordinates of the AOD data pixel points can be directly utilized to establish a Thiessen polygon; when the target spatial resolution is higher or lower than the AOD product resolution, selecting dot information consistent with the target resolution to establish a Thiessen polygon.
Preferably, in consideration of boundary effects of mesh generation and data processing, the target area is enlarged by a certain buffer area when constructing the mesh for reducing errors in data processing and data calculation in the target area. Preferably, the buffer size is set to 100 km.
And the data matching module is used for resampling the AOD of the target area and the variable mediating satellite AOD space-time autocorrelation into the grids respectively, matching the grids with the data and forming the daily data into an independent data set. The method specifically comprises the following steps:
and selecting variables with space-time properties related to satellite AOD time and space distribution, resampling the daily AOD and predicted variable data of corresponding dates into a grid by adopting methods of spatial interpolation, spatial superposition, spatial averaging and the like, and matching to form a daily single data set.
Factors affecting or mediating the satellite AOD temporal and spatial distribution are extracted based on published evidence and the best performance of the model fitted using these factors is guaranteed. Selection of model prediction variables reference model decision coefficients (R2) and Root Mean Square Error (RMSE). The larger R2 and the smaller the RMSE, the better the model fit. The AOD prediction variables used in the embodiment of the invention comprise: the atmospheric chemical transmission model simulates AOD, MODIS cloud parameters, temperature, relative humidity, boundary layer height, albedo, precipitation, and altitude.
When the target spatial resolution is consistent with the satellite AOD product resolution, matching the satellite AOD into the grid according to the grid center point coordinates; when the target spatial resolution is lower than the AOD product resolution, resampling the AOD into a target grid by adopting a spatial interpolation method; when the target spatial resolution is higher than the AOD product resolution, performing spatial superposition on the AOD grids and the target grids, and further calculating an AOD average value in each target grid;
resampling the atmospheric transmission model simulation AOD daily average value into a target grid by adopting a spatial interpolation method; the daytime cloud parameter mean value is matched with a target grid by adopting a space superposition method; resampling the meteorological parameters into a target grid by adopting a spatial interpolation method; the elevation data grid is spatially superimposed with the target grids, and an elevation mean value in each target grid is calculated.
And matching the resampled satellite AOD with the prediction variable according to the grid center point position, so as to obtain a single data set of daily AOD missing prediction.
The prediction data set establishing module is used for respectively adding grid coordinates, square terms and product terms thereof into the independent data sets based on satellite AOD space autocorrelation, merging data sets of a target deficiency date and two days before and after the target deficiency date based on satellite AOD time autocorrelation, and adding corresponding day sequence variables to establish a prediction data set.
Based on satellite AOD space autocorrelation, grid coordinates and square terms and product terms thereof are added to the independent data set respectively, and the method comprises the following steps: and selecting grid center point projection coordinates (X and Y) and square terms and product terms of the coordinates to represent prediction indexes of satellite AOD spatial distribution trend and spatial autocorrelation, and respectively generating projection coordinates X, projection coordinates Y, X square terms, square terms of Y and product term variables of X and Y in a data set.
Based on satellite AOD time autocorrelation, combining data sets of target deficiency date and two days before and after the target deficiency date, and adding corresponding day sequence variables, wherein the method comprises the following steps:
selecting a data set of the current day of the target filling date, and assigning the day sequence variable of the data set to be 3; the target fills up the data sets of the two days before and the day before, and the day sequence variable is respectively assigned to 1 and 2; the target fills the dataset one day and two days after the date, assign its day sequence variables to 4 and 5, respectively.
And combining the five-day data sets for establishing a missing prediction model. The final output dataset is a CSV (Comma-separator values) formatted file.
And the missing prediction model construction module is used for repeatedly sampling the prediction data set for a plurality of times, and each sampling is used for constructing a missing prediction model based on the generalized addition model.
In order to solve the missing value variation due to random error in the missing padding, in this embodiment, the above-mentioned prediction data set is randomly sampled. A missing prediction model based on a generalized addition model is established using the sampled data set, and prediction variables related to satellite AOD time and space autocorrelation are added to the model. The sampling and modeling process was repeated ten times to build ten models in total. The method specifically comprises the following steps:
and the sampling sub-module is used for randomly sampling the predicted data.
For the target date satellite AOD missing filling, firstly, a data set is generated in S4, and a predicted data set corresponding to the target date (comprising the front and rear days of the data set) is selected. And deleting the data record corresponding to the missing value of the satellite AOD in the data set. And randomly sampling the data records of the data set, and extracting 85% of data record samples each time to form a model fitting data set.
In this example, the process was repeated ten times to form ten different model fitting datasets in total. In other embodiments of the invention, the number of samples is not limited to ten, the greater the number the better.
The deletion prediction model construction submodule is used for constructing a deletion prediction model based on a generalized addition model.
And establishing a missing prediction model based on a generalized addition model, wherein model prediction variables comprise atmospheric and chemical transmission model simulation AOD, MODIS cloud parameters, temperature, relative humidity, boundary layer height, albedo, precipitation and altitude, and a projection coordinate X, a projection coordinate Y, X square term, a Y square term and a smoothing term of a product term of X and Y.
The constructed deletion prediction model is as follows:
in the formula, AOD gt Is AOD data for grid g on day t; beta 1 -β 8 Is the slope of the prediction index; CTM (CTM) gt AOD data on day t for grid g simulated by the atmospheric transmission model; cloud gt Cloud parameter data for grid g on day t; tem (Tem) gt 、RH gt 、BLH gt 、Al gt And TCW gt Respectively the average temperature, the relative humidity, the boundary layer height and the ground of the grid g at the t-th balanceSurface reflection and total precipitation; elev g Is the elevation of grid g; d (D) t Is a dumb variable of the day; x is X g And Y g Geographic coordinate items of grid g; epsilon gt Is the error term for grid g on day t.
And the missing prediction module predicts the AOD missing data based on the established multiple missing prediction models respectively, and takes the average value of multiple predictions of each grid as a final AOD predicted value.
In one embodiment, the missing prediction module further comprises the following modules:
and the filling module is used for filling the grid with the loss of the satellite AOD by adopting the final AOD predicted value.
And the accuracy evaluation module is used for matching the satellite AOD monitoring value with the final AOD prediction mean value for the grid which is not lost in the satellite AOD, establishing a simple linear correlation model, and evaluating the model prediction accuracy according to the decision coefficient (R2) and the Root Mean Square Error (RMSE).
In one embodiment, the grid creation module further includes the following modules:
the first judging module is used for judging whether the satellite AOD and the cloud parameter products are derived from the same satellite, and when the satellite AOD and the cloud parameters are derived from different satellites, the AOD and the cloud parameters from different sources are respectively matched and fused to obtain a daily AOD average value and a daily cloud coverage average value;
and the second judging module is used for judging the time resolution of the meteorological data, reducing the time scale and calculating the average value of daily meteorological variables when the time resolution of the meteorological data is higher than that of each day (namely, the average value of the hours or the average value of the minutes).
The invention selects a medium resolution imaging spectrometer (MODIS) sixth edition (C6) satellite remote sensing fusion AOD product which is a fusion of deep blue and dark target algorithms and is carried on American national aviation and aerospace agency (NationalAeronauticsandSpace Administration, NASA for short) Aqua and Terra satellites. Firstly, carrying out matching on AOD products of Aquara or Terra, then adopting simple linear regression to establish correlation of AODs of Aquara or Terra every day, and further utilizing the correlation coefficient to respectively carry out preliminary filling on AOD deletion. Daily averages of Terra and AquaAOD in each grid cell were taken as AOD values indicated in the present invention.
The satellite aerosol deletion prediction method and the system based on the space-time autocorrelation predict the deletion of the satellite AOD while considering that the satellite AOD has the time and space autocorrelation. While predicting the absence, the present method and system predicts the AOD of the AOD presence grid. By comparing the predicted value with the remote sensing monitored value, parameters of prediction accuracy, namely a decision coefficient (R2) and a Root Mean Square Error (RMSE), are obtained to reflect the prediction accuracy. When R2 is larger, the satellite AOD loss filling quality is better; the smaller the RMSE, the better the satellite AOD loss padding quality is explained. According to the method, the model prediction variable and the model prediction structure can be adjusted to obtain higher R2 and lower RMSE so as to ensure the accuracy requirement of satellite AOD data fusion.
The above embodiments are merely preferred embodiments for fully explaining the present invention, and the scope of the present invention is not limited thereto. Equivalent substitutions and modifications will occur to those skilled in the art based on the present invention, and are intended to be within the scope of the present invention. The protection scope of the invention is subject to the claims.
Claims (7)
1. A satellite aerosol loss prediction method based on space-time autocorrelation, comprising:
creating grids covering the target area, and obtaining projection coordinates of a central point of each grid;
resampling the AOD of the target area and variables mediating satellite AOD spatiotemporal autocorrelation into the grids, respectively, and matching the grids with the data, forming daily data into separate data sets, comprising: when the target spatial resolution is consistent with the satellite AOD product resolution, matching the satellite AOD into the grid according to the grid center point coordinates; when the target spatial resolution is lower than the AOD product resolution, resampling the AOD into a target grid by adopting a spatial interpolation method; when the target spatial resolution is higher than the AOD product resolution, performing spatial superposition on the AOD grids and the target grids, and further calculating an AOD average value in each target grid; resampling the atmospheric transmission model simulation AOD daily average value into a target grid by adopting a spatial interpolation method; the daytime cloud parameter mean value is matched with a target grid by adopting a space superposition method; resampling the meteorological parameters into a target grid by adopting a spatial interpolation method; the elevation data grids are spatially overlapped with the target grids, and the elevation mean value in each target grid is calculated; matching the resampled satellite AOD with the prediction variable according to the grid center point position, so as to obtain a single data set of daily AOD missing prediction;
based on the satellite AOD space autocorrelation, respectively adding grid coordinates, square terms and product terms thereof in the independent data sets, simultaneously, based on the satellite AOD time autocorrelation, merging the independent data sets of the target deficiency date and the front and rear days thereof, and adding corresponding day sequence variables, and establishing a prediction data set, wherein the method comprises the following steps: selecting a data set of the current day of the target filling date, and assigning the day sequence variable of the data set to be 3; the target fills up the data sets of the two days before and the day before, and the day sequence variable is respectively assigned to 1 and 2; the target fills up the data set of the date and the data set of the two days after the date, and the variable of the day sequence is respectively assigned to 4 and 5;
repeatedly sampling the prediction data set for a plurality of times, and constructing a deletion prediction model based on a generalized addition model for each sampling; the constructed deletion prediction model is as follows:
in the method, in the process of the invention,is AOD data for grid g on day t; />Is the slope of the prediction index; />AOD data on day t for grid g, which is an atmospheric transmission model; />Cloud parameter data for grid g on day t;and->The average temperature, the relative humidity, the boundary layer height, the ground reflection and the total precipitation of the grid g at the t-th balance are respectively; />Is the error term for grid g on day t,
and respectively predicting the AOD missing data based on the established multiple missing prediction models, and taking the average value of multiple predictions of each grid as a final AOD predicted value.
2. The satellite aerosol loss prediction method based on space-time autocorrelation according to claim 1, wherein said creating a grid covering a target area specifically comprises:
based on the projection coordinates of the AOD data pixel points and the target spatial resolution, establishing a Thiessen polygon;
a mesh is created that covers the target area based on the tesen polygon.
3. The satellite aerosol loss prediction method based on space-time autocorrelation as set forth in claim 2, wherein the creating a Thiessen polygon based on projection coordinates of AOD data pel points and target spatial resolution includes:
when the target spatial resolution is consistent with the AOD product resolution, the projection coordinates of the AOD data pixel points can be directly utilized to establish a Thiessen polygon; when the target spatial resolution is higher or lower than the AOD product resolution, selecting dot information consistent with the target resolution to establish a Thiessen polygon.
4. The satellite aerosol loss prediction method based on spatio-temporal autocorrelation of claim 1, further comprising, prior to creating the grid covering the target region:
judging whether the satellite AOD and the cloud parameter products are derived from the same satellite, and respectively matching and fusing the AOD and the cloud parameters of different sources when the satellite AOD and the cloud parameters are derived from different satellites to obtain a daily AOD average value and a daily cloud coverage average value;
and judging the time resolution of the meteorological data, and when the time resolution of the meteorological data is higher than that of each day, namely, the time scale is reduced, and calculating the average value of daily meteorological variables.
5. The satellite aerosol loss prediction method based on space-time autocorrelation as set forth in claim 1, wherein said predicting AOD loss data based on the established plurality of models, respectively, further comprises, after taking each grid multiple prediction average as a final AOD predicted value:
filling the grid with the loss of the satellite AOD by adopting a final AOD predicted value;
and (3) for the grid with the satellite AOD not missing, after the satellite AOD monitoring value is matched with the final AOD predicted value, a simple linear correlation model is established, and the model prediction accuracy is evaluated according to the determination coefficient R2 and the Root Mean Square Error (RMSE).
6. The satellite aerosol loss prediction method based on space-time autocorrelation as set forth in claim 1, wherein said satellite AOD spatial autocorrelation based increasing grid coordinates and their square term and product term respectively in said dataset includes:
and selecting grid center point projection coordinates X and Y and square term and product term of the coordinates to represent prediction indexes of satellite AOD spatial distribution trend and spatial autocorrelation, and respectively generating product term variables of projection coordinates X, projection coordinates Y, X square term, Y square term and X and Y in a data set.
7. A satellite aerosol loss prediction system based on space-time autocorrelation, comprising:
the grid creation module is used for creating grids covering the target area and obtaining projection coordinates of center points of each grid;
the data matching module is used for resampling the AOD of the target area and the variable mediating satellite AOD space-time autocorrelation into the grids respectively, matching the grids with the data, and forming the daily data into a single data set, and comprises the following steps: when the target spatial resolution is consistent with the satellite AOD product resolution, matching the satellite AOD into the grid according to the grid center point coordinates; when the target spatial resolution is lower than the AOD product resolution, resampling the AOD into a target grid by adopting a spatial interpolation method; when the target spatial resolution is higher than the AOD product resolution, performing spatial superposition on the AOD grids and the target grids, and further calculating an AOD average value in each target grid; resampling the atmospheric transmission model simulation AOD daily average value into a target grid by adopting a spatial interpolation method; the daytime cloud parameter mean value is matched with a target grid by adopting a space superposition method; resampling the meteorological parameters into a target grid by adopting a spatial interpolation method; the elevation data grids are spatially overlapped with the target grids, and the elevation mean value in each target grid is calculated; matching the resampled satellite AOD with the prediction variable according to the grid center point position, so as to obtain a single data set of daily AOD missing prediction;
the prediction data set establishing module is configured to respectively increase grid coordinates and square terms and product terms thereof in the separate data sets based on satellite AOD spatial autocorrelation, and simultaneously, based on satellite AOD temporal autocorrelation, combine data sets of a target deficiency date and two days before and after the target deficiency date, and increase corresponding day sequence variables, and establish a prediction data set, including: selecting a data set of the current day of the target filling date, and assigning the day sequence variable of the data set to be 3; the target fills up the data sets of the two days before and the day before, and the day sequence variable is respectively assigned to 1 and 2; the target fills up the data set of the date and the data set of the two days after the date, and the variable of the day sequence is respectively assigned to 4 and 5;
the missing prediction model construction module is used for repeatedly sampling the prediction data set for a plurality of times, and each sampling is used for constructing a missing prediction model based on a generalized addition model; the constructed deletion prediction model is as follows:
in (1) the->Is AOD data for grid g on day t; />Is the slope of the prediction index; />AOD data on day t for grid g, which is an atmospheric transmission model; />Cloud parameter data for grid g on day t;and->The average temperature, the relative humidity, the boundary layer height, the ground reflection and the total precipitation of the grid g at the t-th balance are respectively; />Is the error term for grid g on day t,
and the missing prediction module predicts the AOD missing data based on the established multiple missing prediction models respectively, and takes the average value of multiple predictions of each grid as a final AOD predicted value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010738199.3A CN111859304B (en) | 2020-07-28 | 2020-07-28 | Satellite aerosol missing prediction method and system based on space-time autocorrelation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010738199.3A CN111859304B (en) | 2020-07-28 | 2020-07-28 | Satellite aerosol missing prediction method and system based on space-time autocorrelation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111859304A CN111859304A (en) | 2020-10-30 |
| CN111859304B true CN111859304B (en) | 2023-11-21 |
Family
ID=72947987
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010738199.3A Active CN111859304B (en) | 2020-07-28 | 2020-07-28 | Satellite aerosol missing prediction method and system based on space-time autocorrelation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111859304B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114943303A (en) * | 2022-06-16 | 2022-08-26 | 福州大学 | Time sequence AOD reconstruction method based on multi-sensor remote sensing |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106442236A (en) * | 2015-07-30 | 2017-02-22 | 中国科学院遥感与数字地球研究所 | Ground PM2.5 inversion method and system based on satellite remote sensing |
| CN106933776A (en) * | 2017-03-02 | 2017-07-07 | 宁波大学 | A kind of method that MODIS AOD products missing data is repaired |
| CN108241779A (en) * | 2017-12-29 | 2018-07-03 | 武汉大学 | Ground PM2.5 Density feature vectors space filter value modeling method based on remotely-sensed data |
| CN109784552A (en) * | 2018-12-29 | 2019-05-21 | 武汉大学 | A kind of construction method of the space variable coefficient PM2.5 concentration appraising model based on Re-ESF algorithm |
| KR20190103661A (en) * | 2018-02-28 | 2019-09-05 | 박무성 | Method of predicting fine dustreflecting both indoor and outdoordata |
| CN110738252A (en) * | 2019-10-14 | 2020-01-31 | 广州地理研究所 | Space autocorrelation machine learning satellite precipitation data downscaling method and system |
| CN111104639A (en) * | 2019-12-24 | 2020-05-05 | 福州大学 | Point-surface fused time sequence PM2.5 spatial distribution estimation method |
-
2020
- 2020-07-28 CN CN202010738199.3A patent/CN111859304B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106442236A (en) * | 2015-07-30 | 2017-02-22 | 中国科学院遥感与数字地球研究所 | Ground PM2.5 inversion method and system based on satellite remote sensing |
| CN106933776A (en) * | 2017-03-02 | 2017-07-07 | 宁波大学 | A kind of method that MODIS AOD products missing data is repaired |
| CN108241779A (en) * | 2017-12-29 | 2018-07-03 | 武汉大学 | Ground PM2.5 Density feature vectors space filter value modeling method based on remotely-sensed data |
| KR20190103661A (en) * | 2018-02-28 | 2019-09-05 | 박무성 | Method of predicting fine dustreflecting both indoor and outdoordata |
| CN109784552A (en) * | 2018-12-29 | 2019-05-21 | 武汉大学 | A kind of construction method of the space variable coefficient PM2.5 concentration appraising model based on Re-ESF algorithm |
| CN110738252A (en) * | 2019-10-14 | 2020-01-31 | 广州地理研究所 | Space autocorrelation machine learning satellite precipitation data downscaling method and system |
| CN111104639A (en) * | 2019-12-24 | 2020-05-05 | 福州大学 | Point-surface fused time sequence PM2.5 spatial distribution estimation method |
Non-Patent Citations (5)
| Title |
|---|
| Jing Yang 等.Filling the missing data gaps of daily MODIS AOD using spatiotemporal interpolation.《Science of the Total Environment》.2018,第633卷677-683. * |
| Qingyang Xiao 等.Full-coverage high-resolution daily PM2.5 estimation using MAIAC AOD in the Yangtze River Delta of China.《Remote Sensing of Environment》.2017,第199卷437-446. * |
| 景悦 ; 孙艳玲 ; 徐昊 ; 陈莉 ; 张辉 ; 高爽 ; 付宏臣 ; 毛健 ; .基于混合效应模型的京津冀地区PM_(2.5)日浓度估算.中国环境科学.2018,(第08期),2890-2897. * |
| 李如仁 ; 李广超 ; 陈伟 ; 霍音娇 ; .京津冀气溶胶数据普通克里金插值研究.沈阳建筑大学学报(自然科学版).2020,(第01期),179-185. * |
| 沈焕锋 ; 李同文 ; .大气PM_(2.5)遥感制图研究进展.测绘学报.2019,(第12期),1624-1635. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111859304A (en) | 2020-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dullaart et al. | Advancing global storm surge modelling using the new ERA5 climate reanalysis | |
| CN108761574B (en) | Rainfall estimation method based on multi-source information fusion | |
| Bright et al. | Improved satellite-derived PV power nowcasting using real-time power data from reference PV systems | |
| CN111033318A (en) | Local weather prediction | |
| Kyriakidis et al. | Geostatistical mapping of precipitation from rain gauge data using atmospheric and terrain characteristics | |
| CN113297527B (en) | PM based on multisource city big data 2.5 Overall domain space-time calculation inference method | |
| Kim et al. | Statistical downscaling for daily precipitation in Korea using combined PRISM, RCM, and quantile mapping: Part 1, methodology and evaluation in historical simulation | |
| CN111323352A (en) | Regional PM2.5 remote sensing inversion model fusing fine particulate matter concentration data | |
| CN111859304B (en) | Satellite aerosol missing prediction method and system based on space-time autocorrelation | |
| Zhang et al. | Modeling and simulation of high-frequency solar irradiance | |
| CN114048944A (en) | Estimation method for people to be evacuated and houses to be damaged under rainstorm induced geological disaster | |
| CN115691049A (en) | Convection birth early warning method based on deep learning | |
| CN112001121A (en) | Large-area gentle region predictive soil mapping method based on solar radiation | |
| Lu et al. | TropNet: a deep spatiotemporal neural network for tropospheric delay modeling and forecasting | |
| Sgoff et al. | Assimilation of crowd‐sourced surface observations over Germany in a regional weather prediction system | |
| Yang et al. | Evaluating WRF-Solar EPS cloud mask forecast using the NSRDB | |
| CN112285808B (en) | Method for reducing scale of APHRODITE precipitation data | |
| CN108510098B (en) | Power transmission line corridor snow depth estimation and early warning method and system based on satellite remote sensing | |
| CN112990609B (en) | Air quality prediction method based on space-time bandwidth self-adaptive geographical weighted regression | |
| CN112668811B (en) | Low-vortex and shear line automatic identification and tracking method and system based on deep learning | |
| CN114818857A (en) | Deep snow fusion method | |
| Airey et al. | Evaluating climate model simulations of precipitation: methods, problems and performance | |
| Walstad et al. | Data assimilation for coastal observing systems | |
| Cheng et al. | Improved hourly estimate of aerosol optical thickness over Asian land by fusing geostationary satellites Fengyun-4B and Himawari-9 | |
| CN112905561B (en) | Meteorological parameter fine scale conversion method based on depth bimodal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |