CN115993668B - Polynomial correction and neural network-based PWV reconstruction method and system - Google Patents
Polynomial correction and neural network-based PWV reconstruction method and system Download PDFInfo
- Publication number
- CN115993668B CN115993668B CN202310285669.9A CN202310285669A CN115993668B CN 115993668 B CN115993668 B CN 115993668B CN 202310285669 A CN202310285669 A CN 202310285669A CN 115993668 B CN115993668 B CN 115993668B
- Authority
- CN
- China
- Prior art keywords
- data
- pwv
- gnss
- fitting
- neural network
- 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 63
- 238000012937 correction Methods 0.000 title claims abstract description 37
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 21
- 101100305901 Mus musculus Rnase2b gene Proteins 0.000 claims abstract description 28
- 238000013507 mapping Methods 0.000 claims abstract description 26
- 238000007689 inspection Methods 0.000 claims abstract description 20
- 238000003062 neural network model Methods 0.000 claims abstract description 18
- 238000012216 screening Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims description 29
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Chemical compound O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 8
- 238000011160 research Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 5
- 239000005442 atmospheric precipitation Substances 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 239000005436 troposphere Substances 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 2
- 241001429221 Passion fruit woodiness virus Species 0.000 description 143
- 230000008859 change Effects 0.000 description 13
- 238000010276 construction Methods 0.000 description 4
- 230000004927 fusion Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- 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
- Complex Calculations (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of data optimization, and discloses a PWV reconstruction method and system based on polynomial correction and a neural network, wherein the method comprises the following steps: step 1: initial data are obtained and preprocessed; step 2: performing space-time matching on the GNSS PWV and the EAR5 PWV; step 3: performing quality inspection on GNSS observation data in the initial data, and screening GNSS PWV and associated data corresponding to unqualified data; step 4: modifying EAR5 PWV by using GNSS PWV through a polynomial fitting method; step 5: unifying the resolution of the multi-source data; the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data are unified to be consistent with PWV data through an interpolation method; step 6: based on a BPNN neural network model, establishing a function mapping relation between the influence factors and the PWV; the impact factors include spatial location and weather parameters. The invention can realize continuous space-time distribution and PWV reconstruction with reliable precision in a certain range, and can obtain PWV data with high space-time resolution and high precision.
Description
Technical Field
The invention relates to the technical field of data optimization, in particular to a PWV reconstruction method and system based on polynomial correction and a neural network.
Background
Moisture, the major component of the atmosphere, is a very variable meteorological parameter and plays a very important role in many atmospheric activities. The method for accurately measuring the atmospheric water vapor content has very important significance in the fields of weather monitoring and forecasting, regional climate change, satellite signal processing, aerospace and the like. The conventional water vapor measurement technology mainly obtains water vapor data through a sounding balloon, a microwave radiometer and a satellite radar, and has the advantages but the limitations. Due to the continuous development of the global satellite navigation system, the GNSS water vapor inversion technology with the advantages of high precision, low operation cost and the like is increasingly researched and applied in recent years. However, the problems of lower spatial resolution and interruption of time distribution still exist to some extent due to the limitations of uneven station distribution and the like, and the PWV data (PWV, precipitable Water Vapor, atmospheric precipitation) obtained by inversion of GNSS observations.
ERA5 is a fifth generation analysis data set provided by the middle-term weather forecast center in europe (European Centre for Medium-Range Weather Forecasts), and analysis data is obtained by performing quality control on original data, and using an optimal combination numerical mode and observation data, and adopting a selected numerical mode to assimilate the observation data, wherein the data has higher usability and certain reliability. The ERA5 data set can provide spatial resolution of 0.25 degrees by 0.25 degrees, and time resolution of 1 hour, has high space-time distribution continuity, but reliable precision data is still difficult to obtain in areas lacking assimilation data.
The two data of GNSS PWV and ERA5 PWV are combined to make up for the limitation of space-time resolution of the former and the defect of precision of the latter, and some technical means are fused to construct a data fusion model and attempt of a data correction method at present. However, the existing PWV data fusion and correction methods based on multi-source data fusion such as GNSS PWV and ERA5 PWV still do not completely solve the problem of time distribution interruption and the problem of accuracy of PWV, and the upper limit of accuracy of multi-source PWV data correction is not high, so that a large lifting space still exists.
Disclosure of Invention
The invention aims to provide a polynomial correction and neural network-based PWV reconstruction method and system, which can realize continuous space-time distribution and reliable PWV reconstruction within a certain range and can obtain PWV data with high space-time resolution and high precision.
In order to achieve the above purpose, the present invention provides the following basic scheme:
scheme one
A PWV reconstruction method based on polynomial correction and neural network comprises the following steps:
step 1: initial data are obtained and preprocessed; the initial data comprises GNSS observation data, ERA5 data sets, meteorological parameter data, DEM data and NDVI data; processing and obtaining GNSS PWV from GNSS observation data, and processing and obtaining EAR5 PWV from ERA5 data set;
step 2: performing space-time matching on the GNSS PWV and the EAR5 PWV;
step 3: performing quality inspection on GNSS observation data in the initial data, and screening GNSS PWV and associated data corresponding to unqualified data;
step 4: modifying EAR5 PWV by using GNSS PWV through a polynomial fitting method;
step 5: unifying the resolution of the multi-source data; the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data are unified to be consistent with PWV data through an interpolation method;
step 6: based on a BPNN neural network model, establishing a function mapping relation between the influence factors and the PWV; the impact factors include spatial location and weather parameters.
Scheme II
A polynomial correction and neural network-based PWV reconstruction system, which is applied to the polynomial correction and neural network-based PWV reconstruction method in scheme one; the system comprises a data acquisition module, a preprocessing module, a matching module, a quality inspection module, a correction module, a data unification module and a mapping module;
the data acquisition module is used for acquiring initial data, wherein the initial data comprise GNSS observation data, ERA5 data sets, meteorological parameter data, DEM data and NDVI data; the preprocessing module is used for preprocessing initial data, processing the initial data from GNSS observation data to obtain GNSS PWV, and processing the ERA5 data to obtain EAR5 PWV;
the matching module is used for performing space-time matching on the GNSS PWV and the EAR5 PWV; the quality inspection module is used for performing quality inspection on GNSS observation data in the initial data and screening GNSS PWV and associated data corresponding to unqualified data; the correction module is used for correcting the EAR5 PWV by using the GNSS PWV through a polynomial fitting method; the data unifying module is used for unifying the resolution ratio of the multi-source data; the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data are unified to be consistent with PWV data through an interpolation method;
the mapping module is used for establishing a function mapping relation between the influence factors and the PWV based on the BPNN neural network model; the impact factors include spatial location and weather parameters.
Wherein PWV, precipitable Water Vapor, refers to the atmospheric precipitation;
DEM, digital Elevation Mode, refers to an elevation model, which is a discrete mathematical representation of the topography of the earth's surface; DEM represents a finite sequence of three-dimensional vectors over region D; DEM data, digital elevation model data;
NDVI, normalized Difference Vegetation Index, refers to normalized vegetation index.
The working principle and the advantages of the invention are as follows:
according to the scheme, by combining the characteristic that the space-time distribution and change of the PWV are closely related to various factors such as space position, meteorological parameters and ecological environment, multi-source data are collected, ERA5 PWV is corrected by using space-time information and GNSS PWV through a polynomial fitting method, and the obtained correction result and GNSS PWV further form a new data set. And then, establishing a function mapping relation between the influence factors and the PWV through a BPNN neural network model, namely, establishing a function mapping relation between space-time information, meteorological parameters and the like and the PWV, so that PWV reconstruction with continuous space-time distribution and reliable precision in a certain range is realized, PWV data with high space-time resolution and high precision can be obtained, and reliable data reference can be provided for researches on regional climate change, weather forecast and the like.
Particularly, the scheme breaks through the limitation of the conventional PWV processing means, and the method does not reconstruct or correct the data only around the PWV data, but expands and researches the real factors influencing the PWV change, such as the space position, the meteorological parameters, the ecological environment and the like, based on the actual distribution characteristic and the change characteristic of the PWV data; correspondingly, when acquiring data, the acquired multi-source data is different from the conventionally considered PWV data of different sources, but the acquired initial data particularly comprises meteorological parameter data, DEM data and NDVI data so as to acquire sufficient real factor related data such as meteorological, spatial position and the like; and the mapping relation between the real factors and the PWV is constructed through the neural network model so as to finish the PWV reconstruction based on the real influencing factors, and the data reconstruction based on the PWV data is more fit with the real data representation under the visual influence of factors such as space time, weather and the like, so that the corresponding actual PWV under different space time conditions can be restored and obtained, and the reliability is higher.
In addition, the method processes the PWV by sequentially adopting a polynomial correction method and a neural network model construction, wherein the polynomial fitting correction and the neural network model construction are basically construction of a variable function mapping relation, and different data processing methods are adopted, so that error accumulation possibly existing in the same data processing scheme can be effectively avoided while the PWV is fully corrected, further high space-time resolution PWV data with reliable precision is provided, more possibility of enriching and improving the existing method is provided, and PWV data processing dimension is expanded.
Drawings
Fig. 1 is a schematic flow chart of a method and a system for reconstructing PWV based on polynomial correction and neural network according to an embodiment of the present invention.
Detailed Description
The following is a further detailed description of the embodiments:
an example is substantially as shown in figure 1: a PWV reconstruction method based on polynomial correction and neural network comprises the following steps:
step 1: initial data are obtained and preprocessed; the initial data includes GNSS observations, ERA5 datasets, meteorological parameter data, DEM data, and NDVI data. The weather parameter data comprise temperature data, air pressure data, relative humidity data, wind speed data and rainfall data; in this embodiment, the weather parameter data may be obtained from different platforms, including a chinese weather data network, NOAA global ground station observation database, and the like. The NDVI data is specifically MODIS NDVI data; MODIS is vegetation index spatial distribution data set; NDVI data are normalized vegetation index data obtained from the MODIS dataset.
The collected multi-source data comprise carefully selected real influence data (weather-related temperature, air pressure, relative humidity, wind speed, rainfall and spatial position-related DEM and NDVI) which are highly related to the actual distribution characteristic and the change characteristic of the PWV, and the data reconstruction analysis dimension is expanded to the aspect of related data which influence the distribution and the change of the PWV. Particularly, in practical application, for the basic GNSS observation data and the determined PWV in the ERA5 dataset, the PWV data itself is susceptible to the acquisition conditions, site setting and numerical assimilation modes, and there is a certain error in practice, so that the upper limit of the accuracy of reconstruction is not high only based on the reconstruction performed by the PWV in the dataset itself. The scheme focuses on the problem, and by expanding the dimension of reconstruction analysis and then establishing the mapping relation between the reality influence data and the PWV, the upper limit of reconstruction accuracy can be further provided, and the truest PWV data in the air in actual practice can be obtained through reconstruction.
And processing and obtaining GNSS PWV from GNSS observation data, and processing and obtaining EAR5 PWV from ERA5 data set. Wherein when EAR5 PWV is obtained from ERA5 data centralized processing, the data spatial resolution is selected as
。
Specifically, when the GNSS PWV is obtained from the GNSS observation data in a processing way, the method comprises the following steps:
calculating GNSS observation data by using high-precision GNSS data processing software, and further obtaining total delay ZTD of the zenith troposphere;
the zenith dry delay component ZHD was calculated using the saastamonen model:
by the formula
Calculating zenith wet delay component ZWD; />
wherein ,
indicating the atmospheric pressure at the station in hPa; />
The latitude of the measuring station is expressed, and the unit is degree; h represents the station elevation in m.
According to the formula
And calculating GNSS PWV of the site position.
Wherein PWV represents the atmospheric precipitation amount of the site location;
for the water vapor conversion coefficient, the formula can be
Calculated, in the above formula, < >>
and />
Is refractive index of atmosphere>
，/>
The method comprises the steps of carrying out a first treatment on the surface of the Moisture gas constant->
The method comprises the steps of carrying out a first treatment on the surface of the Moisture Density->
Atmospheric weighted average temperature +.>
Can be automatically acquired by corresponding data processing software.
In step 1, the acquired NDVI data is also converted from the original HDF format to Geo-TIFF format and reconstructed daily NDVI data by time series harmonic analysis (i.e., harmonic Analysis of Time Series, HANTS).
Step 2: space-time matching is performed on the GNSS PWV and the EAR5 PWV.
When the GNSS PWV and the EAR5 PWV are subjected to space-time matching, ERA5 PWV data with the same time resolution is selected according to the time resolution of the GNSS PWV data, and a bilinear interpolation method is used for interpolating sparse GNSS sites to dense ERA5 PWV lattice points so as to obtain time-position-PWV observation pairs with the same spatial resolution.
Step 3: and performing quality inspection on the GNSS observation data in the initial data, and screening out GNSS PWV and associated data thereof corresponding to the unqualified data, wherein the associated data specifically refers to GNSS PWV interpolation. In practical application, the quality inspection rule is dynamically formulated according to the overall situation of the GNSS observation data, and in this embodiment, the quality inspection rule includes: the effective rate of the data is required to be more than 90%. mp1 and mp2 need to be less than 0.5m; the O/slps is required to be greater than 200. Wherein mp1 and mp2 respectively represent multipath effect combinations of L1 and L2 frequency point pseudo-ranges and carrier phase observed values of the GPS; the smaller mp1 and mp2 indicate the stronger multipath resistance. O/slips is the ratio of observed value to cycle slip, reflecting the condition of data cycle slip; the smaller the O/slips value, the more severe the occurrence of cycle slip. And judging the data which does not meet the rule of the quality inspection as unqualified data.
The inspection indexes for performing quality inspection include: data efficiency, cycle slip, and multipath effects.
Step 4: and correcting the EAR5 PWV by using the GNSS PWV through a polynomial fitting method.
The polynomial fitting method comprises the following steps: fitting is carried out by adopting a one-time fitting model which is easiest to process, and the fitting mode adopts fitting at single-time-period and single-point positions.
The fitting model is as follows:
；
in the formula ,
representing the PWV data obtained by fitting; />
、/>
、/>
、/>
、/>
Fitting coefficients;EAR5PWV is EAR5 PWV data corresponding to fitting time and fitting position;Latto fit the latitude of the location,Longto fit the longitude of the location,His the elevation of the fitting location.
The fitting criterion adopts a least square criterion; that is to say,
since the ERA5 dataset does not provide data accuracy, the matrix P is regarded as a unit matrix in the method, and the matrix V is defined as the difference between the fitting result PWV of a specified position and the GNSS PWV (interpolation result) of the position, namely:
wherein k is the number of observation pairs;
fitting PWV, namely, PWV data representing the ith single point position of the designated position obtained by fitting; />
Is the GNSS PWV data corresponding to the ith single point position. />
Step 5: unifying the resolution of the multi-source data; and unifying the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data to be consistent with PWV data through an interpolation method.
Step 6: based on a BPNN neural network model, establishing a function mapping relation between the influence factors and the PWV; the influencing factors include spatial location, weather parameters, time, etc. The DEM data and NDVI data both belong to spatial position related data.
The input layer of the BPNN neural network model is composed of 10 neurons (the 10 neurons are specifically longitude, latitude, elevation, time, NDVI and five meteorological parameters of temperature, air pressure, relative humidity, wind speed and rainfall), namely:
in the formula ,Longitudelongitude;Latitudeis latitude;DEMis an elevation;DOYtime is;{Met}is a meteorological parameter;NDVIis normalized vegetation index.
In order to divide the research area into grids, longitude, latitude, elevation, time, meteorological parameters and normalized vegetation index data corresponding to the ith grid point are obtained.
The output layer of the BPNN neural network model consists of 1 neuron, which together forms a PWV data set from a corrected ERA5 PWV observation pair and an interpolated GNSS PWV observation pair, i.e.:
in the formula ,MODIS-ERA5 PWVan observation pair for corrected ERA5 PWV;
after dividing the research area into grids, EAR5 PWV data which corresponds to the ith grid point and is corrected by GNSS PWV; i represents the ith data, j represents the jth data; there are n data points in total, each point has corresponding longitude and latitude, time, weather parameter, PWV and other data, i represents the ith and j represents the jth.
The training of the BPNN neural network model is to construct a function mapping relation between X and Y; and the function mapping relation between the influence factors and the PWV is established, so that the function mapping relation between X and Y is established. Wherein X is all
Is a collection of (1); y is all
Is a collection of (1).
The embodiment also provides a PWV reconstruction system based on polynomial correction and a neural network, which is applied to the PWV reconstruction method based on polynomial correction and the neural network as described in scheme one, and comprises a data acquisition module, a preprocessing module, a matching module, a quality inspection module, a correction module, a data unification module and a mapping module;
the data acquisition module is used for acquiring initial data, wherein the initial data comprise GNSS observation data, ERA5 data sets, meteorological parameter data, DEM data and NDVI data; the preprocessing module is used for preprocessing initial data, processing the initial data from GNSS observation data to obtain GNSS PWV, and processing the ERA5 data to obtain EAR5 PWV;
the matching module is used for performing space-time matching on the GNSS PWV and the EAR5 PWV; the quality inspection module is used for performing quality inspection on GNSS observation data in the initial data and screening GNSS PWV and associated data corresponding to unqualified data; the correction module is used for correcting the EAR5 PWV by using the GNSS PWV through a polynomial fitting method; the data unifying module is used for unifying the resolution ratio of the multi-source data; the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data are unified to be consistent with PWV data through an interpolation method;
the mapping module is used for establishing a function mapping relation between the influence factors and the PWV based on the BPNN neural network model; the impact factors include spatial location and weather parameters.
According to the PWV reconstruction method and system based on polynomial correction and neural network, the characteristics that the spatial-temporal distribution and change of PWV are closely related to various factors such as spatial position, meteorological parameters and ecological environment are combined, ERA5 PWV is corrected by using the spatial-temporal information and GNSS PWV through polynomial fitting, a data set is formed by correction results and GNSS PWV, and a function mapping relation of the spatial-temporal information, the meteorological parameters, vegetation indexes and the PWV is constructed through a BPNN neural network model, so that PWV reconstruction with continuous spatial-temporal distribution and reliable precision in a certain range is achieved, and the obtained high spatial-temporal resolution PWV data has important significance on researches such as regional climate change and weather forecast. In addition, the method adopts two means of polynomial fitting correction and neural network model to process PWV, so that error accumulation caused by the same data processing method can be effectively avoided, and higher data processing precision is achieved.
In particular, the method breaks through the limitation of the conventional PWV processing means, the existing processing schemes process the PWV data around, and data fusion is performed among PWVs of different sources, and the method is not established by considering deeper technology based on real factors (such as space-time information and weather factors) influencing PWV changes. The method is not only limited by a plurality of realistic factors influencing PWV change, but also has the advantages of huge data analysis amount, complex model and great processing difficulty; it is also limited by the conventional thinking that the PWV maintained in the database is itself data under the influence of real factors, so that it is not reconstructed in additional relation to real factors. In practice, however, the determined PWV in the GNSS observations, ERA5 dataset itself may be distorted by a range of acquisition conditions; in the area lacking assimilation data, ERA5 PWV is unreliable, GNSS PWV can not completely provide continuous reliable data for the area, correction effectiveness is limited, and the area which can not be accurately reconstructed through the original data exists.
The method and the device accurately find and fully consider the problems, firstly, the data reconstruction or correction is not carried out only around the PWV data, but the real factors influencing the PWV change such as the space position, the meteorological parameters and the ecological environment are expanded and researched based on the actual distribution characteristics and the change characteristics of the PWV data, the multi-source dimension of the multi-source data reconstruction is expanded, the PWV reconstruction is completed based on the real influencing factors, the data reconstruction is more fit with the real data representation, the real PWV data can be restored and obtained, and the method and the device are high in reliability and reality. Based on the real factors influencing the change of PWV, the reconstruction of the reconstruction area limited by the original data can be completed under the assistance of other influencing factors through effective mapping of the two factors, and the data reconstruction is complete and reliable.
Secondly, the scheme overcomes the difficulty of data processing and model construction in the development and research. Because the real influencing factors such as the space position and the meteorological parameters comprise a plurality of sub-factors, if the real influencing factors are directly combined with PWV data, the related modeling flow and model can be very complex, and compared with the existing processing scheme that the analysis data source is limited to a single PWV of different sources, the scheme increases the data processing capacity, but firstly selects the real influencing factors participating in PWV reconstruction and provides a reconstruction method with simple flow, optimizes the data processing difficulty and ensures the data reconstruction efficiency. The scheme selects the reality influence factors as weather-related temperature, air pressure, relative humidity, wind speed and rainfall; spatial location dependent DEM and NDVI; correspondingly, the input layer neurons of the BPNN model are simplified to 10, and the selection of the reality influence factors is more representative, so that the accurate reconstruction of PWV is facilitated, and meanwhile, the data processing capacity and the processing difficulty are greatly reduced. Training a BPNN model to naturally construct a real influence factor and PWV, wherein the real influence factor and PWV are the mapping relation of the corrected PWV, and the reconstruction of the PWV is completed; the whole flow is concise, the reconstruction efficiency is higher, and the accuracy is reliable.
The foregoing is merely an embodiment of the present invention, and a specific structure and characteristics of common knowledge in the art, which are well known in the scheme, are not described herein, so that a person of ordinary skill in the art knows all the prior art in the application day or before the priority date of the present invention, and can know all the prior art in the field, and have the capability of applying the conventional experimental means before the date, so that a person of ordinary skill in the art can complete and implement the present embodiment in combination with his own capability in the light of the present application, and some typical known structures or known methods should not be an obstacle for a person of ordinary skill in the art to implement the present application. It should be noted that modifications and improvements can be made by those skilled in the art without departing from the structure of the present invention, and these should also be considered as the scope of the present invention, which does not affect the effect of the implementation of the present invention and the utility of the patent.
Claims (8)
1. A polynomial correction and neural network based PWV reconstruction method, comprising the steps of:
step 1: initial data are obtained and preprocessed; the initial data comprises GNSS observation data, ERA5 data sets, meteorological parameter data, DEM data and NDVI data; processing and obtaining GNSS PWV from GNSS observation data, and processing and obtaining EAR5 PWV from ERA5 data set;
step 2: performing space-time matching on the GNSS PWV and the EAR5 PWV;
step 3: performing quality inspection on GNSS observation data in the initial data, and screening GNSS PWV and associated data corresponding to unqualified data;
step 4: modifying EAR5 PWV by using GNSS PWV through a polynomial fitting method;
step 5: unifying the resolution of the multi-source data; the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data are unified to be consistent with PWV data through an interpolation method;
step 6: based on a BPNN neural network model, establishing a function mapping relation between the influence factors and the PWV; the influence factors comprise spatial position and meteorological parameters;
in step 4, the polynomial fitting method is as follows: fitting is carried out by adopting a fitting model, and fitting at single-time intervals and single-point positions is adopted in a fitting mode;
the fitting criterion adopts a least square criterion; i.e.
The vision matrix P is a unit matrix, and the matrix V is defined as a difference value between the specified position fitting result PWV and the position interpolation result GNSS PWV, that is:
wherein k is the number of observation pairs;
fitting PWV, namely, PWV data representing the ith single point position of the designated position obtained by fitting; />
GNSS PWV data corresponding to the ith single point position;
the fitting model is as follows:
in the formula ,
representing the PWV data obtained by fitting; />
、/>
、/>
、/>
、/>
Fitting coefficients;EAR5PWV is EAR5 PWV data corresponding to fitting time and fitting position;Latto fit the latitude of the location,Longto fit the longitude of the location,Hthe elevation of the fitting position;
in step 6, the input layer of the BPNN neural network model is composed of 10 neurons, namely:
in the formula ,Longitudelongitude;Latitudeis latitude;DEMis an elevation;DOYtime is;{Met}is a meteorological parameter;NDVIis normalized vegetation index;
after dividing a research area into grids, longitude, latitude, elevation, time, meteorological parameters and normalized vegetation index data corresponding to an ith grid point;
the output layer of the BPNN neural network model consists of 1 neuron, namely:
in the formula ,MODIS-ERA5 PWVan observation pair for corrected ERA5 PWV;
after dividing the research area into grids, EAR5 PWV data which corresponds to the ith grid point and is corrected by GNSS PWV;
the function mapping relation between the influence factors and the PWV is established, and the function mapping relation between X and Y is established; wherein X is all
Is a collection of (1); y is all->
Is a collection of (1).
2. The method for reconstructing a PWV based on polynomial correction and neural network according to claim 1, wherein in step 1, when obtaining a GNSS PWV from GNSS observation data, the method comprises the steps of:
solving the GNSS observation data to obtain total zenith troposphere delay ZTD;
the zenith dry delay component ZHD was calculated using the saastamonen model:
by the formula
Calculating zenith wet delay component ZWD; wherein (1)>
Indicating the atmospheric pressure at the station in hPa; />
The latitude of the measuring station is expressed, and the unit is degree; h represents station elevation, and the unit is m;
according to the formula
Calculating GNSS PWV of the site position;
wherein PWV represents the atmospheric precipitation amount of the site location;
is the water vapor conversion coefficient;
wherein ,
, in the formula ,/>
and />
Is the atmospheric refractive constant; />
，
；/>
Is water vapor gas constant, < >>
；/>
Is water vapor density
；/>
The average temperature is weighted for the atmosphere.
3. The method for reconstructing PWV based on polynomial correction and neural network according to claim 2, wherein in step 1, the obtained NDVI data is further converted into Geo-TIFF format, and daily NDVI data is reconstructed by time-series harmonic analysis.
4. The method for reconstructing PWV based on polynomial correction and neural network according to claim 1, wherein the weather parameter data comprises temperature data, barometric pressure data, relative humidity data, wind speed data and rainfall data.
5. The polynomial correction and neural network based PWV reconstruction method according to claim 2, wherein in step 1, when EAR5 PWV is obtained from ERA5 dataset processing, the data spatial resolution is selected to be
。
6. The method according to claim 1, wherein in step 2, when performing space-time matching on the GNSS PWV and EAR5 PWV, the ERA5 PWV data with the same time resolution is selected according to the time resolution of the GNSS PWV data, and a bilinear interpolation method is used to interpolate sparse GNSS sites to dense ERA5 PWV lattice points to obtain time-position-PWV observation pairs with the same spatial resolution.
7. The method for reconstructing PWV based on polynomial correction and neural network according to claim 1, wherein in step 3, the inspection index for performing quality inspection comprises: data efficiency, cycle slip, and multipath effects.
8. A polynomial correction and neural network based PWV reconstruction system, characterized by being applied to a polynomial correction and neural network based PWV reconstruction method according to any one of claims 1-7; the system comprises a data acquisition module, a preprocessing module, a matching module, a quality inspection module, a correction module, a data unification module and a mapping module;
the data acquisition module is used for acquiring initial data, wherein the initial data comprise GNSS observation data, ERA5 data sets, meteorological parameter data, DEM data and NDVI data; the preprocessing module is used for preprocessing initial data, processing the initial data from GNSS observation data to obtain GNSS PWV, and processing the ERA5 data to obtain EAR5 PWV;
the matching module is used for performing space-time matching on the GNSS PWV and the EAR5 PWV; the quality inspection module is used for performing quality inspection on GNSS observation data in the initial data and screening GNSS PWV and associated data corresponding to unqualified data; the correction module is used for correcting the EAR5 PWV by using the GNSS PWV through a polynomial fitting method; the data unifying module is used for unifying the resolution ratio of the multi-source data; the spatial-temporal resolutions of the DEM data, the NDVI data and the meteorological parameter data are unified to be consistent with PWV data through an interpolation method;
the mapping module is used for establishing a function mapping relation between the influence factors and the PWV based on the BPNN neural network model; the impact factors include spatial location and weather parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310285669.9A CN115993668B (en) | 2023-03-22 | 2023-03-22 | Polynomial correction and neural network-based PWV reconstruction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310285669.9A CN115993668B (en) | 2023-03-22 | 2023-03-22 | Polynomial correction and neural network-based PWV reconstruction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115993668A CN115993668A (en) | 2023-04-21 |
| CN115993668B true CN115993668B (en) | 2023-05-30 |
Family
ID=85993776
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310285669.9A Active CN115993668B (en) | 2023-03-22 | 2023-03-22 | Polynomial correction and neural network-based PWV reconstruction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115993668B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104777488A (en) * | 2015-03-13 | 2015-07-15 | 中国科学院上海天文台 | Modeling method and device for zenith tropospheric delay as well as measuring method and device |
| CN115728842A (en) * | 2022-11-16 | 2023-03-03 | 南京信息工程大学 | Method, system and device for acquiring atmospheric degradable water and storage medium |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4611929A (en) * | 1983-03-21 | 1986-09-16 | The United States Of America As Represented By The Secretary Of The Navy | Satellite method for measuring sea surface temperature |
| CN106126826B (en) * | 2016-06-27 | 2019-02-26 | 中国科学院南京地理与湖泊研究所 | A kind of muddy water body in lake concentration of suspension evaluation method based on VIIRS sensor |
| CN106547840B (en) * | 2016-10-13 | 2019-09-06 | 国家卫星气象中心 | A kind of parsing and management method of whole world three-dimensional atmospheric data |
| EP3652636A4 (en) * | 2017-07-11 | 2021-04-07 | The Trustees of Princeton University | System and method for performing accurate hydrologic determination using disparate weather data sources |
| CN108120995A (en) * | 2017-12-07 | 2018-06-05 | 深圳市华信天线技术有限公司 | A kind of method and device for improving satellite navigation system data output frequencies |
| CN109117555A (en) * | 2018-08-14 | 2019-01-01 | 东南大学 | A kind of Bevis model refinement method of sub-region right mean temperature |
| CN109738926B (en) * | 2019-04-01 | 2019-06-28 | 湖南联智桥隧技术有限公司 | A kind of GNSS multipath effect correcting method based on BP neural network technology |
| CN111126466B (en) * | 2019-12-16 | 2020-09-22 | 西安科技大学 | Multi-source PWV data fusion method |
| CN111274707B (en) * | 2020-02-05 | 2022-11-25 | 东南大学 | Weighted average temperature calculation method based on reanalysis data and wireless sounding data |
| CN111273318B (en) * | 2020-02-25 | 2021-10-19 | 东南大学 | Regional troposphere wet delay calculation method based on parabola |
| CN112035448A (en) * | 2020-07-13 | 2020-12-04 | 国家海洋信息中心 | Neural network short-term rainfall forecasting method integrating water vapor and meteorological elements of foundation GNSS |
| CN112016052B (en) * | 2020-08-20 | 2021-07-09 | 广东省气象探测数据中心 | Near-surface daily maximum air temperature estimation method, system and terminal based on multi-source data |
| KR102437574B1 (en) * | 2021-10-06 | 2022-08-31 | 대한민국 | System for retrieving atmospheric water vapor |
| CN114910982B (en) * | 2022-07-05 | 2024-05-14 | 中国电建集团西北勘测设计研究院有限公司 | Rainfall early warning model construction method based on Beidou technology |
-
2023
- 2023-03-22 CN CN202310285669.9A patent/CN115993668B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104777488A (en) * | 2015-03-13 | 2015-07-15 | 中国科学院上海天文台 | Modeling method and device for zenith tropospheric delay as well as measuring method and device |
| CN115728842A (en) * | 2022-11-16 | 2023-03-03 | 南京信息工程大学 | Method, system and device for acquiring atmospheric degradable water and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115993668A (en) | 2023-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113297527B (en) | PM based on multisource city big data 2.5 Overall domain space-time calculation inference method | |
| CN112905560A (en) | Air pollution prediction method based on multi-source time-space big data deep fusion | |
| CN110427594A (en) | It is suitble to the meteorological element data-acquisition system of small-size laboratory | |
| CN106776481A (en) | A kind of NO emissions reduction bearing calibration for acting on satellite precipitation data | |
| CN112699959B (en) | Multi-source multi-scale precipitation data fusion method and device based on energy functional model | |
| Tonboe et al. | Product user manual for osi saf global sea ice concentration | |
| CN112785024A (en) | Runoff calculation and prediction method based on watershed hydrological model | |
| CN114120101A (en) | Soil moisture multi-scale comprehensive sensing method | |
| Zhang et al. | Spatial and temporal downscaling of TRMM precipitation with novel algorithms | |
| Zhang et al. | Comprehensive evaluation of mainstream gridded precipitation datasets in the cold season across the Tibetan Plateau | |
| CN114819737B (en) | Method, system and storage medium for estimating carbon reserves of highway road vegetation | |
| CN108154193A (en) | A kind of long-term sequence precipitation data NO emissions reduction method | |
| CN112100922A (en) | Wind resource prediction method based on WRF and CNN convolutional neural network | |
| Fan et al. | A comparative study of four merging approaches for regional precipitation estimation | |
| CN115808727A (en) | Hundred-meter-level gust objective forecasting method and device | |
| CN115203934A (en) | Mountain area water-reducing downscaling method based on Logistic regression | |
| CN113639893B (en) | Near-earth weighted average temperature information acquisition method based on multiple meteorological factors | |
| CN115993668B (en) | Polynomial correction and neural network-based PWV reconstruction method and system | |
| CN114169215B (en) | Surface temperature inversion method coupling remote sensing and regional meteorological model | |
| AU2021105536A4 (en) | A High Spatial-Temporal Resolution Method for Near-Surface Air Temperature Reconstruction | |
| CN115758856A (en) | Method for researching influence of landscape pattern and climate change on future water quality of drainage basin | |
| CN112270293B (en) | Daily paddy field CH 4 Remote sensing estimation method of flux | |
| CN115326721A (en) | Method for estimating crop water stress | |
| CN114324410A (en) | Multi-terrain microwave remote sensing soil humidity downscaling method | |
| Bi et al. | A deep learning-based model for tropospheric wet delay prediction based on multi-layer 1D convolution neural network |
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 |