CN113189014A - Ozone concentration estimation method fusing satellite remote sensing and ground monitoring data - Google Patents
Ozone concentration estimation method fusing satellite remote sensing and ground monitoring data Download PDFInfo
- Publication number
- CN113189014A CN113189014A CN202110401303.4A CN202110401303A CN113189014A CN 113189014 A CN113189014 A CN 113189014A CN 202110401303 A CN202110401303 A CN 202110401303A CN 113189014 A CN113189014 A CN 113189014A
- Authority
- CN
- China
- Prior art keywords
- ozone concentration
- remote sensing
- concentration estimation
- basic model
- layer
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/95—Lidar systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N2021/1765—Method using an image detector and processing of image signal
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N2021/1793—Remote sensing
- G01N2021/1795—Atmospheric mapping of gases
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Radar, Positioning & Navigation (AREA)
- Electromagnetism (AREA)
- Remote Sensing (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses an ozone concentration estimation method fusing satellite remote sensing and ground monitoring data, and belongs to the technical field of environmental monitoring. The method comprises the following steps: firstly, collecting, preprocessing and fusing multi-source sample data to obtain input parameters; step two, establishing an ozone concentration estimation operation basic model based on a multilayer mapping back propagation neural network; and thirdly, searching the optimized input parameter combination of the obtained ozone concentration estimation operation basic model based on the influence factors, the previous tracing time and the space range, accurately estimating the ground ozone concentration according to the optimized input parameter combination to obtain the space continuous distribution condition of the ozone concentration, and realizing the ozone concentration estimation method fusing satellite remote sensing and ground monitoring data. The method has the advantages of high accuracy, strong reliability and simple operation, the used multi-source sample data is free and open source, the universality is enhanced, and the ozone concentration can be quickly estimated and the continuous distribution map of the ozone in the target area can be drawn.
Description
Technical Field
The invention belongs to the technical field of environmental monitoring, and relates to an ozone concentration estimation method fusing satellite remote sensing and ground monitoring data.
Background
With the acceleration of urbanization and industrialization, the problem of air pollution becomes more serious, and ozone becomes the primary pollutant which affects the quality of environmental air and directly affects the physical and mental health of human beings. Therefore, monitoring and revealing the rule of continuous distribution of ozone in time and space has great significance for preventing and treating ozone pollution and preventing the harm of ozone to health.
At present, common ozone monitoring methods include ground monitoring and remote sensing monitoring. The ground monitoring is based on a monitoring station to carry out all-weather continuous observation, and the accurate information of the ozone concentration in the surface space and the change of the ozone concentration along with the time can be directly obtained. However, the monitoring sites are high in construction cost, limited in quantity and uneven in distribution, and continuous and accurate ozone concentration monitoring in a large-scale space is difficult to achieve. The research of monitoring ozone by using satellite remote sensing image data starts in the 80 th 20 th century, and mainly comprises a mode scale factor method, a semi-empirical method based on a physical mechanism, a statistical model method and the like. In the past decades, the above ozone estimation method, although widely used, still has the following problems: 1) the model structure and the simulation process are very complex, and the calculation cost is high; 2) the requirement on basic data is high, and the pollutant discharge list is always high in uncertainty, so that the estimation accuracy is limited; 3) the estimation result is greatly influenced by parameter setting, and the parameters are not only very complicated in calculation process, but also have obvious difference among different regions.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide the ozone concentration estimation method fusing satellite remote sensing and ground monitoring data, the ozone concentration estimation method based on machine learning is high in accuracy, strong in reliability and simple in operation, and the problems of high uncertainty of the conclusion obtained by the existing ozone concentration estimation method and complex calculation are solved.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
the invention discloses an ozone concentration estimation method fusing satellite remote sensing and ground monitoring data, which comprises the following steps:
firstly, collecting, preprocessing and fusing multi-source sample data to obtain input parameters;
step two, establishing an ozone concentration estimation operation basic model based on a multilayer mapping back propagation neural network;
and thirdly, based on three dimensions of influence factors, forward tracing time and space range, in combination with the input parameters obtained in the first step, exploring the optimized input parameter combination of the ozone concentration estimation operation basic model obtained in the second step, estimating the ground ozone concentration according to the optimized input parameter combination to obtain a space continuous distribution result of the ozone concentration, and realizing the ozone concentration estimation method fusing satellite remote sensing and ground monitoring data.
Preferably, in the first step, the multi-source sample data includes: satellite remote sensing image data, air quality monitoring data and meteorological data.
Preferably, in the first step, the preprocessing of the multi-source sample data includes cloud removal processing and obtaining of a normalized vegetation index NDVI: cloud layer identification and classification based on satellite remote sensing image data, a projection influence range of a cloud layer on the ground and cloud layer coverage rate, and cloud removing processing is carried out on the remote sensing image according to different cloud layer characteristics; and performing orthorectification and spatial position registration on the satellite remote sensing image data, and extracting the wave band reflectivity of each wave band in the satellite remote sensing image data to obtain a normalized vegetation index NDVI.
Preferably, in the first step, the fusing of the multi-source sample data includes the following operations: acquiring a meteorological site nearest to an environment monitoring site through a proximity analysis algorithm, and using meteorological data monitored by the meteorological site nearest to the environment monitoring site as meteorological information of the environment monitoring site to realize fusion of multi-source sample data; and establishing an index for tracing analysis aiming at the multi-source sample data.
Preferably, in the second step, based on the multilayer mapping back propagation neural network, establishing an obtained ozone concentration estimation operation basic model, which comprises an input layer, a hidden layer and an output layer; the neurons in the layers are all connected, the neurons in the same layer are not connected, and the neurons in each layer can receive the signal of the neurons in the previous layer and generate a signal to be output to the next layer.
Further preferably, an input layer, an output layer and L hidden layers are included, wherein L ≧ 1.
Further preferably, the number of nodes of the hidden layer is obtained by: continuously training and comparing the multilayer mapping back propagation neural network by gradually enlarging the number of nodes in the hidden layer; and when the predicted result is consistent with the real result, obtaining the node number of the hidden layer.
Preferably, in step two, the input parameters of the basic model for ozone concentration estimation operation include: the method comprises the following steps of (1) obtaining wave band reflectivity, normalized vegetation index and meteorological data of different wave bands in satellite remote sensing image data; and the output data of the ozone concentration estimation operation basic model is the ozone concentration value of the monitoring site during remote sensing image imaging.
Preferably, in step three, the optimization of the input parameter combination of the basic model for ozone concentration estimation operation obtained in the influencing factor searching step two comprises the following operations:
firstly, analyzing the correlation between the input parameter obtained in the step one and the ozone concentration by adopting a statistical method; then, classifying and grouping according to the strength of the obtained correlation and the class characteristics of the input parameters obtained in the first step, and inputting each group of input parameter data obtained by grouping into the ozone concentration estimation operation basic model obtained in the second step for training and verification; and finally, determining the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model obtained in the step two on the influence factor dimension based on the decision coefficient, the average error and the root mean square error.
Preferably, in step three, the operation of the optimized input parameter combination of the ozone concentration estimation operation basic model obtained in the previous trace time searching step two includes: by cyclic exploration, with tsStep-by-step increase in hoursThe range of the previous trace period; every increase t of the forward trace periodsWhen the time is short, the input parameters corresponding to the time interval are added to the input parameters of the ozone concentration estimation operation basic model obtained in the step two; determining the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model obtained in the step two on the basis of the decision coefficient, the average error and the root-mean-square error in the forward tracing time dimension; wherein, 5 is more than or equal to tsNot less than 1; and D, when the current time period exceeds a preset threshold value and the estimation result of the ozone concentration estimation operation basic model obtained in the step two is kept unchanged or continuously worsened, the exploration is terminated.
Preferably, in step three, the operation of estimating and calculating the optimized input parameter combination of the basic model based on the ozone concentration obtained in the space range searching step two comprises: training and verifying an ozone concentration estimation operation basic model, and recording a decision coefficient, an average error and a root-mean-square error; gradually enlarging the training and verifying area according to the step length set in the research area, and determining the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model obtained in the step two on the spatial range dimension through determining the coefficient, the average error and the root mean square error; and when the ozone concentration estimation operation basic model training and verification area exceeds a preset threshold value and the ozone concentration estimation result is kept unchanged or continuously worsened, terminating the exploration.
Compared with the prior art, the invention has the following beneficial effects:
the invention discloses an ozone concentration estimation method fusing satellite remote sensing and ground monitoring data, which can obtain input parameters through collection, pretreatment and fusion of multi-source sample data, can establish an ozone concentration estimation operation basic model through a multilayer mapping back propagation neural network and by combining machine learning, and can explore the optimal combination of the input parameters through three dimensions based on influence factors, forward tracing time and a space range so as to realize accurate estimation of the ground ozone concentration. In addition, the optimization processes in three dimensions of influence factors, forward tracing time and space range are not separated from each other and run independently, but are comprehensively considered, and the optimal input parameter combination can be found more comprehensively and accurately. Therefore, the ozone concentration estimation method based on machine learning is high in accuracy, strong in reliability and simple to operate, can estimate the ozone concentration, can indirectly obtain the ozone concentration in the environment monitoring work, and has high popularization and application values.
Furthermore, the multi-source sample data is free and open, the universality of the model is enhanced, and therefore the ozone concentration can be quickly estimated and the continuous distribution map of the ozone concentration in the target area can be drawn.
Furthermore, through cloud removing treatment and obtaining of the normalized vegetation index NDVI, the influence of cloud on the estimation result can be removed, and the accuracy of the model on the estimation of the ground ozone concentration is improved.
Furthermore, by adopting the fusion of multi-source sample data and establishing an index of the traceability analysis, the relevant data (such as satellite remote sensing image data, meteorological data and air quality data) from different sources can be summarized and fused, so that the ozone estimation model can be conveniently used.
Furthermore, the establishment of the multi-layer mapping back propagation neural network can adapt to the condition that the distribution of the ground ozone concentration has very complicated nonlinear relations with a plurality of factors such as temperature, relative humidity, atmospheric pressure, wind speed, wind direction and the like, and the multi-layer mapping back propagation neural network has relatively great advantages in processing and solving the nonlinear mapping problem due to the unique structure of the multi-layer mapping back propagation neural network.
Furthermore, from the influence factor dimensionality, the optimized input parameter combination of the ozone concentration estimation operation basic model is determined, and the accuracy of ground ozone concentration estimation is improved.
Further, in the process of the ozone concentration estimation operation, the influence on the estimation value includes not only the satellite imaging time (denoted as T)0) The meteorological parameters at the same time also include T0The front tracing time (denoted as T) from front tracing to a certain time1) During the time period (i.e.: t is0-T1) The weather conditions of (a); therefore, the method of the invention well solves the problem and determines the optimal method based on the forward time exploration modeThe time is traced forward, and the accuracy of the estimation of the ground ozone concentration is improved. Wherein, the step length t is set by a circular searching modes，5≥tsNot less than 1; if tsLess than 1, not only is the sample data difficult to obtain, but also the prediction process becomes very complicated due to excessive data volume; if tsIf the time is more than 5, the error of the optimal forward tracing time period is increased, so that the accuracy of the model is greatly reduced.
Further, when there are few monitoring stations in the research area, but there are a certain number of ground monitoring stations distributed around the research area, the optimal spatial range for model training and verification will likely be larger than the research area; therefore, the method overcomes the influence of the space range on the estimation result through space range exploration, determines the optimal space range and improves the accuracy of the estimation of the ground ozone concentration.
Drawings
FIG. 1 is a schematic flow chart of an ozone concentration estimation method according to the present invention, which combines satellite remote sensing and ground monitoring data;
FIG. 2 is a schematic diagram of an ozone concentration estimation operation basic model established by a multi-layer mapping back propagation neural network according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the optimal spatial range of the ozone concentration estimation operation basic model training, verification and testing according to the embodiment of the present invention; wherein (a) the optimal spatial extent is equal to the region of interest; (b) the optimal spatial range is larger than the study area;
FIG. 4 is a graph of a correlation fit between an estimated and monitored value of ozone concentration in the ground in Beijing, according to an embodiment of the present invention;
FIG. 5 is a comparison between the ozone concentration on the ground estimated by the Beijing model and the monitoring value of the monitoring station;
FIG. 6 is a graph showing the spatial distribution of the concentration of ozone in Beijing at different times according to the embodiment of the present invention: (a) UTC 2:53, 10 months, 1 day 2018; (b) UTC 2:53, 3 months, 26 days 2019.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention is described in further detail below with reference to the accompanying drawings:
referring to fig. 1, it can be seen that the ozone concentration estimation method based on machine learning and integrating satellite remote sensing and ground monitoring data according to the present invention includes the following steps:
step one, collecting, preprocessing and fusing multi-source sample data to obtain input parameters.
Wherein, multisource sample data mainly includes: satellite remote sensing image data (Landsat 8, MODIS, etc.), air quality monitoring data (O)3) And meteorological data (wind speed, wind direction, humidity, air temperature, air pressure, etc.)
Specifically, in one embodiment of the present invention, the satellite remote sensing image data is collected from Google Earth Engine (GEE), the air quality monitoring data is collected from the chinese environment monitoring central office (CNEMC), and the meteorological data is collected from the National Oceanic and Atmospheric Administration (NOAA).
The pretreatment of the multi-source sample data comprises the following steps: based on an API (application programming interface) provided by Google Earth Engine (GEE) and other related open source programs, cloud layer identification and classification of satellite remote sensing image data in a research area, calculation of a projection influence range of a cloud layer on the ground and cloud layer coverage rate are achieved, and cloud removing processing is carried out on the remote sensing image by adopting corresponding algorithms according to different cloud layer characteristics; orthorectification and spatial position registration are carried out on satellite remote sensing image data in a research area, meanwhile, the wave band reflectivity of each wave band in the remote sensing image is extracted, and a normalized vegetation index NDVI is obtained through the following formula:
NDVI ═ (NIR-R)/(NIR + R); wherein, NIR is the reflection value of a near infrared band, and R is the reflection value of a red light band.
The fusion of multi-source sample data comprises the following steps: and an index of the sample data is established, so that the estimation result can be conveniently subjected to source tracing analysis. And acquiring a meteorological site closest to the environment monitoring site through a proximity analysis algorithm, and using the meteorological data monitored by the meteorological site as the meteorological information of the environment monitoring site to realize the fusion of multi-source sample data.
And step two, establishing an ozone concentration estimation operation basic model based on the multilayer mapping back propagation neural network.
The basic model for ozone concentration estimation operation constructed by the invention consists of an input layer, L hidden layers (L is more than or equal to 1) and an output layer. Wherein, the neurons between the layers are all connected, and the neurons in the same layer are not connected. The neurons of each layer can receive the signal of the neurons of the previous layer and generate a signal to output to the next layer. When a group of sample data is provided for the ozone concentration estimation operation basic model, the input signal is propagated backwards layer by layer from the input layer to the output layer through the hidden layer; if the output layer cannot obtain the expected output result, the connection weight of the network is corrected from the output layer to the middle layer by layer along the direction of reducing the error until the input layer is reached; the forward calculation process and the backward propagation process are repeated, and the weight and the threshold value of each layer are continuously adjusted, so that the predicted output of the backward propagation neural network continuously approaches to the expected output. The ozone concentration estimation operation basic model adopts a tansig function as a transfer function between different hidden layers, adopts a purelin function as a transfer function between the last hidden layer and an output layer, and adopts a trainlm function of a Levenberg-Marquardt (LM) algorithm to calculate in a network training process. For the determination of the number of nodes in the hidden layer of the neural network, the invention selects the most appropriate network structure by gradually expanding the number of nodes in the hidden layer and continuously training and comparing the network according to the Kolmogorov theorem. Wherein, the input parameters of the ozone concentration estimation operation basic model mainly comprise: the output data of the ozone concentration estimation operation basic model is the ozone concentration value of the ground monitoring station during remote sensing image imaging.
Specifically, in one embodiment of the present invention, the basic model for ozone concentration estimation operation constructed by the present invention is composed of an input layer, L hidden layers (L ═ 2), and an output layer.
The invention selects the Mean Error (ME), the Root Mean Square Error (RMSE) and the coefficient of determination (R)2) And comprehensively and objectively evaluating the ozone concentration estimation operation basic model.
And step three, searching for an optimized input parameter combination of an ozone concentration estimation operation basic model from three different dimensions of influence factors, forward tracing time and a space range, accurately estimating the ground ozone concentration according to the obtained optimized input parameter combination to obtain a space continuous distribution result of the ozone concentration, quickly estimating the ozone concentration and drawing a continuous distribution graph of the ozone concentration in a target area, and realizing the ozone concentration estimation method based on machine learning.
Finding an optimized combination of influencing factors, the operations comprising: firstly, analyzing the correlation between the input parameter obtained in the step one and the ozone concentration by adopting a statistical method; and then, classifying and grouping all possible input parameters according to the strength of the correlation and the class characteristics of the input parameters obtained in the step one, and inputting each group of input parameter data obtained by grouping into an ozone concentration estimation operation basic model for training and verification. By the pair R2And comprehensively evaluating the three parameters of ME and RMSE to determine the optimized combination of the input data of the ozone concentration estimation operation basic model on the influence factor dimension.
Finding an optimized combination of forward trace times, comprising the operations of: adopting a cyclic exploration mode, and taking 1-5 hours as a step length (t)s) The range of the forward tracing period is gradually increased. Every 1-5 hours of the forward time interval, the input parameters corresponding to the time interval are added to the input parameters of the ozone concentration estimation operation basic model, and R is calculated2And comprehensively evaluating the three parameters including ME and RMSE to determine the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model in the forward tracing time dimension. The termination of the exploration process can be controlled according to two principles: 1) the forward tracing time period is long enough, namely the forward tracing time period exceeds a preset threshold; 2) the estimation result of the ozone concentration estimation operation basic model is kept unchanged or continuously worsened.
Specifically, in a certain embodiment of the present invention, the range of the forward trace period is increased stepwise in 3-hour steps. Finding the optimal combination of spatial ranges: firstly, training and verifying an ozone concentration estimation operation basic model in a research area, and recording R of the model2ME and RMSE. The training and validation area is then expanded in steps until one or more new monitoring sites are present in the study area, passing R2And comprehensively judging the quality of the corresponding estimation result by the three characteristic values of ME and RMSE, and determining the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model on the dimension of the space range. The termination of the exploration process can be controlled according to two principles: 1) ozone concentration estimation operationThe basic model training and verifying area is large enough and exceeds a preset threshold value; 2) the ozone concentration estimation results remain unchanged or continuously deteriorate.
The optimization processes in three dimensions of influence factors, forward tracing time and space range are not separated from each other and operated independently, but are combined together for comprehensive consideration to search the optimal combination of the input parameters of the ozone concentration estimation operation basic model, so that the optimal estimation result under the support of the existing sample data condition is obtained.
The ozone concentration estimation operation basic model established based on the back propagation neural network has high accuracy and strong reliability, can estimate the ozone concentration to a certain degree, and has strong application potential. Meanwhile, data used in the model estimation process are free and open, so that the universality of the model is enhanced. In addition, the back propagation neural network model selected by the invention is very suitable for adopting a processing method of distributed parallel computation, which can improve the operation efficiency to a great extent, thereby enabling the rapid estimation of the continuous distribution of the ozone concentration on the space to be possible.
In order to illustrate the technical solution of the present invention, the following description will be made by taking Beijing as a specific embodiment.
Step one, collecting, preprocessing and fusing multi-source sample data to obtain input parameters.
In the embodiment, Landsat 8 remote sensing image data of Beijing city from 5.1.2014 to 10.1.1.2019 are collected through Google Earth Engine (GEE, https:// Earth. Google. com /); collecting ground ozone concentration data of Beijing city at the same time through a China environmental monitoring central office (CNEMC, http:// www.cnemc.cn /); meteorological data was collected by the national oceanographic and atmospheric administration (NOAA, https:// gis. ncdc. NOAA. gov/maps/nci/cdo/hour) for the simultaneous Beijing City, including: wind speed, wind direction, humidity, air temperature, air pressure, etc.
Based on an API (application programming interface) provided by Google Earth Engine (GEE) and other related open source programs, cloud layer identification and classification of satellite remote sensing image data in a research area, calculation of a projection influence range of a cloud layer on the ground and cloud layer coverage rate are achieved, and cloud removing processing is carried out on the remote sensing image by adopting corresponding algorithms according to different cloud layer characteristics. Orthorectification and spatial position registration are carried out on satellite remote sensing image data in a research area, meanwhile, the wave band reflectivity of each wave band in the remote sensing image is extracted, and a normalized vegetation index NDVI is obtained through the following formula:
NDVI＝(band5–band4)/(band5+band4)
wherein the band5 is the reflectivity of Landsat 8 remote sensing image wave band5, and the band4 is the reflectivity of Landsat 8 remote sensing image wave band 4.
And an index of the sample data is established, so that the ozone concentration estimation result can be conveniently subjected to source tracing analysis. And extracting the wave band reflectivity and the NDVI data in a buffer area of 15 meters around the Chinese environment monitoring station, and endowing the average value to the corresponding Chinese environment monitoring station to realize the fusion of Landsat 8 remote sensing image data and air quality data. Secondly, acquiring a weather monitoring site closest to the Chinese environment monitoring site through a proximity analysis algorithm, and taking the monitored weather data as weather information of the Chinese environment monitoring site.
And step two, establishing a basic model of ozone concentration estimation operation based on the multilayer mapping back propagation neural network.
As shown in fig. 2, the basic model for ozone concentration estimation operation constructed in this embodiment is composed of an input layer, two hidden layers and an output layer. Wherein, the neurons between the layers are all connected, and the neurons in the same layer are not connected. The neurons of each layer can receive the signal of the neurons of the previous layer and generate a signal to output to the next layer. When a set of sample data is provided to the model, the input signal propagates from the input layer, layer-by-layer, back through the hidden layer, to the output layer. If the output layer does not obtain the expected output result, the connection weight of the network is corrected from the output layer to the input layer through the intermediate layers layer by layer along the direction of reducing the error. The forward calculation process and the backward propagation process are repeated, and the weight and the threshold value of each layer are continuously adjusted, so that the predicted output of the ozone concentration estimation operation basic model continuously approaches to the expected output.
In FIG. 2, X1，X2，…，XmThe input parameters of the ozone concentration estimation operation basic model mainly include the wave band reflectivity of different wave bands in Landsat 8 remote sensing image data, NDVI and related meteorological data, such as: wind speed, wind direction, humidity, air temperature, air pressure, etc. Y is the estimated value of the model, namely the ozone concentration value of the ground monitoring station when Landsat 8 remote sensing images are imaged.
Representing the weight on the connection from the jth neuron at level l-1 to the ith neuron at level l,
indicating the bias of the ith neuron in layer l,
represents the activation value of the ith neuron of the l layer. The ozone concentration estimation operation basic model adopts a tansig function as a transfer function between different hidden layers, adopts a purelin function as a transfer function between the last hidden layer and an output layer, and adopts a trainlm function of a Levenberg-Marquardt (LM) algorithm to calculate in a network training process.
In addition, for determining the number of nodes in the hidden layer of the neural network, the method selects the most appropriate network structure by gradually expanding the number of nodes in the hidden layer and continuously training and comparing the multi-layer mapping back propagation neural network according to the Kolmogorov theorem. Finally, the optimal number of nodes of the multi-layer mapping back propagation neural network is determined [15,15 ].
In this embodiment, a conventional leave-out method in machine learning model evaluation is adopted, and a training data set and a verification test data set are selected in a manner of multiple random sampling, wherein the training set accounts for 80% and the verification test set accounts for 20%. According to the determined neural network structure and transfer function, the maximum training time is set to be 500, the network training precision is set to be 0.001, and the learning rate is 0.1. Repeat 300 times in each experiment to obtain the averageThe final result of the model evaluation. Finally, Mean Error (ME), Root Mean Square Error (RMSE) and decision coefficient (R) are selected2) And comprehensively and objectively evaluating the ozone concentration estimation operation basic model.
And step three, searching for the optimized input parameter combination of the ozone concentration estimation operation basic model from three different dimensions of influence factors, forward tracing time and space range.
Finding the optimal combination of influencing factors: and analyzing the correlation between various data and the ozone concentration by adopting a statistical method. And classifying and grouping all possible input data according to the strength of the correlation and the class characteristics of the data. In this embodiment, the 17 alternative input influencing factors are classified into the following three groups according to the strength of the correlation with ozone and different data sources and characteristics, namely:
(1) the reflectivities of the waveband 1, the waveband 2 and the waveband 3 in Landsat 8OLI/TIRS have strong correlation with the concentration of ozone;
(2) the reflectivities of other bands, and NDVI calculated from band4 and band5 of Landsat 8 OLI;
(3) meteorological parameters, wind speed, wind direction, humidity, air temperature and air pressure.
Since the parameters in group (1) have a strong correlation with ozone concentration, the reflectance of band 1, band 2 and band 3 are used as the basis for ozone estimation, and thus these parameters are used throughout the process to explore the best combination of alternative input influencing factors. Inputting the three groups of different parameters into the model step by step for training, learning and verification according to ME, RMSE and R2Training results generated by different alternative input parameter combinations can be compared, so that the optimized input parameter combination of the input parameters and various influencing factors is realized.
Searching for the optimized input parameter combination of the previous tracing time: since the meteorological conditions change significantly with the passage of time, the meteorological parameters during satellite imaging and the previous meteorological conditions have significant influence on the estimation accuracy of the ozone concentration. The process of exploring the optimal forward time for a particular area of interest (e.g., a city) can be described by the following six steps:
the first step is as follows: by T0Representing the satellite imaging time, T1Representing the forward trace time, tsRepresenting a time step.
The second step is that: let T1＝T0And evaluating O of ozone concentration estimation operation basic model3Estimation of Performance, denoted P0。
The third step: since the meteorological parameters are collected by the ground monitoring station every 3 hours, t is sets＝3h，T1＝T0-n×tsWherein n is an integer greater than zero.
The fourth step: will [ T0,T1]Inputting all meteorological parameters in a time period into an ozone concentration estimation operation basic model, and estimating O of the ozone concentration estimation operation basic model3Estimate the performance, recorded as
The fifth step: if the current performance is
Superior to the previous P0Is provided with
n＝n+1。
And a sixth step: repeating the third step to the fifth step repeatedly until
Continuous ratio P0A difference of, [ T ]0,T1]The time period becomes sufficiently long, for example, greater than an empirically determined threshold.
T explored by the iterative procedure0,T1]The optimal forward trace time for the multi-layer mapping back propagation neural network training will be used.
Finding the optimal combination of spatial ranges: empirically, the optimal spatial extent of the multi-level mapping backpropagation neural network training does not have to be the same as the minimum bounding rectangle of the study region. In practice, it is related to the distribution of the monitoring stations and may be larger than the smallest bounding rectangle of the investigation region, see fig. 3. The process of exploring the optimal spatial extent for a particular area of interest (e.g., a city) can be described by the following six steps:
the first step is as follows: with S0And
the minimum bounding rectangle representing the study area and the spatial extent of the neural network training, respectively.
The second step is that: is provided with
And evaluating O of ozone concentration estimation operation basic model3Estimation of Performance, denoted P0。
The third step: expanding in steps in both longitude and latitude directions
Until one or more new monitoring stations are located
And (4) the following steps.
The fourth step: o of estimation operation basic model for estimating ozone concentration3Estimate the performance, recorded as
The fifth step: if the current performance is
Superior to the previous P0Is provided with
And a sixth step: repeating the third step to the fifth step repeatedly until the area is reached
Become large enough, e.g.
Greater than an empirically determined threshold.
Explored by the above iterative process
Will be the optimal spatial range for neural network training.
In the embodiment, optimization processes in three dimensions of influence factors, forward tracing time and space range are not separated from each other and run independently, but are combined together for comprehensive consideration to search the optimal combination of input parameters of the ozone concentration estimation operation basic model, so that the optimal estimation result under the support of the existing sample data condition is obtained.
FIG. 4 shows the correlation between the estimated and monitored values of surface ozone concentration in Beijing, where R is20.91, ME 1.2. mu.g/m3RMSE of 18.4. mu.g/m3. The slope of the fitted line approaches 1 and the correlation is very significant. FIG. 5 compares the estimated surface ozone concentration from the Beijing City model to the monitored values from the monitoring station, where the red line represents the actual observed ozone concentration value and the blue line represents the estimated ozone concentration value. In most cases, the estimated data and the monitored data are substantially in agreement with each other. Therefore, the ozone concentration estimation operation basic model established by the invention has the capability of accurately estimating the ground ozone concentration, and is expected to become a new important means for monitoring atmospheric pollution change and analyzing areas.
FIGS. 6(a) and 6(b) show the spatial distribution of ozone concentration in Beijing area at UTC 2:53 on 1/10/2018 and UTC 2:53 on 26/3/2019, respectively, with spatial resolution as high as 30 m. At both times, the ozone concentration on the ground shows a gradually decreasing trend from southeast to northwest, which is consistent with the high-northwest and low-southeast topography and industrialization and population density characteristics of Beijing.
It will be understood by those skilled in the art that the foregoing is illustrative of specific embodiments of this invention and is not to be construed as limiting the scope of the invention, which is defined by the appended claims.
Claims (10)
1. An ozone concentration estimation method fusing satellite remote sensing and ground monitoring data is characterized by comprising the following steps:
firstly, collecting, preprocessing and fusing multi-source sample data to obtain input parameters;
step two, establishing an ozone concentration estimation operation basic model based on a multilayer mapping back propagation neural network;
and thirdly, based on three dimensions of influence factors, forward tracing time and space range, in combination with the input parameters obtained in the first step, exploring the optimized input parameter combination of the ozone concentration estimation operation basic model obtained in the second step, estimating the ground ozone concentration according to the optimized input parameter combination to obtain a space continuous distribution result of the ozone concentration, and realizing the ozone concentration estimation method fusing satellite remote sensing and ground monitoring data.
2. The method for estimating the concentration of ozone by fusing satellite remote sensing and ground monitoring data according to claim 1, wherein in the first step, multi-source sample data comprises: satellite remote sensing image data, air quality monitoring data and meteorological data.
3. The method for estimating the ozone concentration by fusing the satellite remote sensing and the ground monitoring data according to claim 1, wherein in the first step, preprocessing of multi-source sample data comprises cloud removal processing and normalized vegetation index NDVI obtaining:
cloud layer identification and classification based on satellite remote sensing image data, a projection influence range of a cloud layer on the ground and cloud layer coverage rate, and cloud removing processing is carried out on the remote sensing image according to different cloud layer characteristics; and performing orthorectification and spatial position registration on the satellite remote sensing image data, and extracting the wave band reflectivity of each wave band in the satellite remote sensing image data to obtain a normalized vegetation index NDVI.
4. The method for estimating the concentration of ozone by fusing the satellite remote sensing data and the ground monitoring data according to claim 1, wherein in the first step, the fusing of the multi-source sample data comprises the following operations: acquiring a meteorological site nearest to an environment monitoring site through a proximity analysis algorithm, and using meteorological data monitored by the meteorological site nearest to the environment monitoring site as meteorological information of the environment monitoring site to realize fusion of multi-source sample data;
and establishing an index for tracing analysis aiming at the multi-source sample data.
5. The method for estimating the ozone concentration by fusing the satellite remote sensing data and the ground monitoring data according to claim 1, wherein in the second step, an obtained ozone concentration estimation operation basic model is established based on a multilayer mapping back propagation neural network, and the model comprises an input layer, a hidden layer and an output layer; the neurons in the layers are all connected, the neurons in the same layer are not connected, and the neurons in each layer can receive the signal of the neurons in the previous layer and generate a signal to be output to the next layer.
6. The method for estimating ozone concentration by fusing satellite remote sensing and ground monitoring data according to claim 5, characterized by comprising an input layer, an output layer and L hidden layers, wherein L is more than or equal to 1.
7. The method for estimating the concentration of ozone by fusing the satellite remote sensing data and the ground monitoring data according to claim 5, wherein the number of nodes of the hidden layer is obtained by the following operations: continuously training and comparing the multilayer mapping back propagation neural network by gradually enlarging the number of nodes in the hidden layer; and when the predicted result is consistent with the real result, obtaining the node number of the hidden layer.
8. The method for estimating ozone concentration by fusing satellite remote sensing and ground monitoring data according to claim 1, wherein in the third step, the optimized input parameter combination of the ozone concentration estimation operation basic model obtained in the second step based on influence factors is searched, and the operation comprises:
firstly, analyzing the correlation between the input parameter obtained in the step one and the ozone concentration by adopting a statistical method;
then, classifying and grouping according to the strength of the obtained correlation and the class characteristics of the input parameters obtained in the first step, and inputting each group of input parameter data obtained by grouping into the ozone concentration estimation operation basic model obtained in the second step for training and verification;
and finally, determining the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model obtained in the step two on the influence factor dimension based on the decision coefficient, the average error and the root mean square error.
9. The method for estimating ozone concentration by fusing satellite remote sensing and ground monitoring data according to claim 1, wherein in the third step, the optimized input parameter combination of the ozone concentration estimation operation basic model obtained in the second step based on the previous time search is operated as follows:
by cyclic exploration, with tsThe hour is the step length and gradually increases the range of the previous tracing period; every increase t of the forward trace periodsWhen the time is short, the input parameters corresponding to the time interval are added to the input parameters of the ozone concentration estimation operation basic model obtained in the step two; determining the optimized output of the input parameters of the ozone concentration estimation operation basic model obtained in the second step on the forward tracing time dimension based on the decision coefficient, the average error and the root mean square errorEntering parameter combination;
wherein, 5 is more than or equal to ts≥1；
And D, when the current time period exceeds a preset threshold value and the estimation result of the ozone concentration estimation operation basic model obtained in the step two is kept unchanged or continuously worsened, the exploration is terminated.
10. The method for estimating ozone concentration by fusing satellite remote sensing and ground monitoring data according to claim 1, wherein in step three, based on the optimized input parameter combination of the ozone concentration estimation operation basic model obtained in the space range exploration step two, the operation comprises:
training and verifying an ozone concentration estimation operation basic model, and recording a decision coefficient, an average error and a root-mean-square error; gradually enlarging the training and verifying area according to the step length set in the research area, and determining the optimized input parameter combination of the input parameters of the ozone concentration estimation operation basic model obtained in the step two on the spatial range dimension through determining the coefficient, the average error and the root mean square error;
and when the ozone concentration estimation operation basic model training and verification area exceeds a preset threshold value and the ozone concentration estimation result is kept unchanged or continuously worsened, terminating the exploration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110401303.4A CN113189014B (en) | 2021-04-14 | 2021-04-14 | Ozone concentration estimation method integrating satellite remote sensing and ground monitoring data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110401303.4A CN113189014B (en) | 2021-04-14 | 2021-04-14 | Ozone concentration estimation method integrating satellite remote sensing and ground monitoring data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113189014A true CN113189014A (en) | 2021-07-30 |
| CN113189014B CN113189014B (en) | 2023-05-02 |
Family
ID=76975794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110401303.4A Active CN113189014B (en) | 2021-04-14 | 2021-04-14 | Ozone concentration estimation method integrating satellite remote sensing and ground monitoring data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113189014B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113821987A (en) * | 2021-11-19 | 2021-12-21 | 浙江甲骨文超级码科技股份有限公司 | Complex terrain meteorological data prediction method, system and device |
| CN115825338A (en) * | 2022-11-03 | 2023-03-21 | 浙江时空智子大数据有限公司 | Ozone monitoring downscaling method, device, equipment and computer readable storage medium |
| CN115855761A (en) * | 2022-11-03 | 2023-03-28 | 生态环境部卫星环境应用中心 | Five-base cooperative sky-ground integrated atmospheric environment three-dimensional remote sensing monitoring system and method |
| CN116486278A (en) * | 2023-04-24 | 2023-07-25 | 河南大学 | Hour-level ozone estimation method based on space-time information mosaic |
| CN117216480A (en) * | 2023-09-18 | 2023-12-12 | 宁波大学 | Near-surface ozone remote sensing estimation method for deep coupling geographic space-time information |
| CN117592005A (en) * | 2024-01-19 | 2024-02-23 | 中国科学院空天信息创新研究院 | PM2.5 concentration satellite remote sensing estimation method, device, equipment and medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102011081197A1 (en) * | 2011-08-18 | 2013-02-21 | Siemens Aktiengesellschaft | Method for the computer-aided modeling of a technical system |
| CN106909722A (en) * | 2017-02-10 | 2017-06-30 | 广西壮族自治区气象减灾研究所 | A kind of large area accurate inversion method of temperature near the ground |
| CN110595968A (en) * | 2019-08-02 | 2019-12-20 | 中国科学院遥感与数字地球研究所 | PM2.5 concentration estimation method based on geostationary orbit satellite |
| CN110595960A (en) * | 2019-08-02 | 2019-12-20 | 中国科学院遥感与数字地球研究所 | PM2.5 concentration remote sensing estimation method based on machine learning |
| CN111310386A (en) * | 2020-02-13 | 2020-06-19 | 北京中科锐景科技有限公司 | Near-surface ozone concentration estimation method |
| AU2020101095A4 (en) * | 2020-06-23 | 2020-07-23 | Tarim University | A Satellite Remote Sensing Monitoring Method for the Content of Chlorophyll A in the Canopy of Jujube Trees |
| CN112052627A (en) * | 2020-08-21 | 2020-12-08 | 海南星瞰信息咨询中心(有限合伙) | Method, device, medium and equipment for estimating near-surface ozone space distribution |
| CN112163375A (en) * | 2020-09-28 | 2021-01-01 | 中国科学院空天信息创新研究院 | Long-time sequence near-surface ozone inversion method based on neural network |
-
2021
- 2021-04-14 CN CN202110401303.4A patent/CN113189014B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102011081197A1 (en) * | 2011-08-18 | 2013-02-21 | Siemens Aktiengesellschaft | Method for the computer-aided modeling of a technical system |
| CN106909722A (en) * | 2017-02-10 | 2017-06-30 | 广西壮族自治区气象减灾研究所 | A kind of large area accurate inversion method of temperature near the ground |
| CN110595968A (en) * | 2019-08-02 | 2019-12-20 | 中国科学院遥感与数字地球研究所 | PM2.5 concentration estimation method based on geostationary orbit satellite |
| CN110595960A (en) * | 2019-08-02 | 2019-12-20 | 中国科学院遥感与数字地球研究所 | PM2.5 concentration remote sensing estimation method based on machine learning |
| CN111310386A (en) * | 2020-02-13 | 2020-06-19 | 北京中科锐景科技有限公司 | Near-surface ozone concentration estimation method |
| AU2020101095A4 (en) * | 2020-06-23 | 2020-07-23 | Tarim University | A Satellite Remote Sensing Monitoring Method for the Content of Chlorophyll A in the Canopy of Jujube Trees |
| CN112052627A (en) * | 2020-08-21 | 2020-12-08 | 海南星瞰信息咨询中心(有限合伙) | Method, device, medium and equipment for estimating near-surface ozone space distribution |
| CN112163375A (en) * | 2020-09-28 | 2021-01-01 | 中国科学院空天信息创新研究院 | Long-time sequence near-surface ozone inversion method based on neural network |
Non-Patent Citations (2)
| Title |
|---|
| SARAH SAFIEDDINE等: "Tropospheric ozone variability during the East Asian summer monsoon as observed by satellite (IASI), aircraft (MOZAIC) and ground stations", pages 10489 - 10500 * |
| 张猛 等: "基于机器学习和卫星遥感的 PM2.5/10空间连续分布反演方法研究", 《地球环境学报》, 31 August 2020 (2020-08-31), pages 449 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113821987A (en) * | 2021-11-19 | 2021-12-21 | 浙江甲骨文超级码科技股份有限公司 | Complex terrain meteorological data prediction method, system and device |
| CN115825338A (en) * | 2022-11-03 | 2023-03-21 | 浙江时空智子大数据有限公司 | Ozone monitoring downscaling method, device, equipment and computer readable storage medium |
| CN115855761A (en) * | 2022-11-03 | 2023-03-28 | 生态环境部卫星环境应用中心 | Five-base cooperative sky-ground integrated atmospheric environment three-dimensional remote sensing monitoring system and method |
| CN115855761B (en) * | 2022-11-03 | 2023-09-26 | 生态环境部卫星环境应用中心 | Five-base collaborative sky-ground integrated atmospheric environment three-dimensional remote sensing monitoring system and method |
| CN116486278A (en) * | 2023-04-24 | 2023-07-25 | 河南大学 | Hour-level ozone estimation method based on space-time information mosaic |
| CN116486278B (en) * | 2023-04-24 | 2023-11-21 | 河南大学 | Hour-level ozone estimation method based on space-time information mosaic |
| CN117216480A (en) * | 2023-09-18 | 2023-12-12 | 宁波大学 | Near-surface ozone remote sensing estimation method for deep coupling geographic space-time information |
| CN117592005A (en) * | 2024-01-19 | 2024-02-23 | 中国科学院空天信息创新研究院 | PM2.5 concentration satellite remote sensing estimation method, device, equipment and medium |
| CN117592005B (en) * | 2024-01-19 | 2024-04-26 | 中国科学院空天信息创新研究院 | PM2.5 concentration satellite remote sensing estimation method, device, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113189014B (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113189014B (en) | Ozone concentration estimation method integrating satellite remote sensing and ground monitoring data | |
| CN113919448B (en) | Method for analyzing influence factors of carbon dioxide concentration prediction at any time-space position | |
| CN111709981A (en) | Registration method of laser point cloud and analog image with characteristic line fusion | |
| CN111639747B (en) | GNSS-R sea surface wind speed inversion method and system based on BP neural network | |
| CN110232471B (en) | Rainfall sensor network node layout optimization method and device | |
| CN111639746B (en) | GNSS-R sea surface wind speed inversion method and system based on CNN neural network | |
| CN111651545A (en) | Urban marginal area extraction method based on multi-source data fusion | |
| CN110866494B (en) | Urban group extraction method and system based on optical remote sensing image | |
| CN110595968B (en) | PM2.5 concentration estimation method based on geostationary orbit satellite | |
| CN110110025B (en) | Regional population density simulation method based on feature vector space filtering value | |
| CN114460013B (en) | Coastal wetland vegetation overground biomass GAN model self-learning remote sensing inversion method | |
| CN112287294A (en) | Time-space bidirectional soil water content interpolation method based on deep learning | |
| CN114169232B (en) | Reconstruction method and device for full-time three-dimensional atmospheric pollutants, computer equipment and storage medium | |
| CN113836808A (en) | PM2.5 deep learning prediction method based on heavy pollution feature constraint | |
| CN115690632A (en) | Water environment monitoring method for inland river water body | |
| CN115438848A (en) | PM based on deep mixed graph neural network 2.5 Long-term concentration prediction method | |
| CN114254692B (en) | Multi-scale intelligent thunderstorm classifying and identifying method based on multi-source lightning data | |
| CN114972984A (en) | Accumulated snow time-space analysis and prediction method based on random forest | |
| CN114005027A (en) | Urban single tree detection system and method based on unmanned aerial vehicle image | |
| CN116466368B (en) | Dust extinction coefficient profile estimation method based on laser radar and satellite data | |
| Chen et al. | A novel predictor for exploring PM2. 5 spatiotemporal propagation by using convolutional recursive neural networks | |
| CN117456342A (en) | Traffic accident risk assessment method and system based on street view image | |
| CN110321528B (en) | Hyperspectral image soil heavy metal concentration assessment method based on semi-supervised geospatial regression analysis | |
| CN116778334A (en) | Quantitative large-scale space grassland mouse entrance density prediction method and system | |
| CN115840908B (en) | Method for constructing PM2.5 three-dimensional dynamic monitoring field based on microwave link of LSTM model |
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 |