CN115392401A - Atmospheric water-reducing capacity data fusion method based on artificial neural network - Google Patents
Atmospheric water-reducing capacity data fusion method based on artificial neural network Download PDFInfo
- Publication number
- CN115392401A CN115392401A CN202211306842.0A CN202211306842A CN115392401A CN 115392401 A CN115392401 A CN 115392401A CN 202211306842 A CN202211306842 A CN 202211306842A CN 115392401 A CN115392401 A CN 115392401A
- Authority
- CN
- China
- Prior art keywords
- pwv
- precision
- data
- neural network
- model
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/14—Rainfall or precipitation gauges
-
- 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
-
- 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)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Environmental & Geological Engineering (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Hydrology & Water Resources (AREA)
- Health & Medical Sciences (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an atmospheric degradable water content data fusion method based on an artificial neural network. The method comprises the following steps: PWV data from different sources are kept consistent in time; step two: PWV data from different sources is kept consistent in space; step three: performing quality control; step four: constructing a generalized regression neural network model; step five: acquiring a GRNN model and precision evaluation information thereof; step six: calibrating and optimizing the low-precision PWV by using the trained GRNN model, wherein the PWV output by the trained GRNN model is the calibrated and optimized PWV; step seven: and combining the PWV subjected to GRNN calibration and optimization with the high-precision PWV to obtain a fusion PWV. The method has the functions of correcting the deviation of the PWV system and improving the PWV precision, can realize the fusion of multi-source PWV data, and has the advantages of simple method, high precision and high data utilization rate.
Description
Technical Field
The invention relates to a method for calibrating and fusing geological data, in particular to a method for calibrating and fusing atmospheric degradable water volume data based on an artificial neural network. More specifically, the method utilizes the generalized regression neural network to perform spatial fusion on multi-source geoscience data, and belongs to a data processing method in the fields of geoscience and surveying and mapping science and technology.
Background
The water vapor plays an important role in the weather and climate process and energy transmission, and meanwhile, the earth observation technology is adversely affected, so that the accurate water vapor monitoring has important significance for improving the weather forecast level, improving the earth observation precision and the like. However, monitoring the moisture state and predicting its changes still has many deficiencies due to the high dynamics of moisture and the limited observation techniques available today. The amount of atmospheric Water-reducible (PWV) is the most commonly used physical quantity for measuring the total amount of Water Vapor in the atmosphere, and is defined as the height of the Water Vapor contained in a vertical column with a unit cross-sectional area when all the Water Vapor is condensed into liquid. The rapid development of modern earth observation technology in recent 20 years makes water vapor observation technology unprecedentedly abundant, such as GNSS and synthetic aperture radar technology, and especially GNSS technology not only has the same high precision as radio sounding, but also has the advantages of continuous operation, high time resolution, no weather influence and the like, thereby making up the defects of the traditional observation means to a great extent, and also making water vapor observation data massively increased. In addition, numerical weather forecast models of international agencies, such as ECMWF and NCEP, can also provide data on the amount of water available in the atmosphere. Unprecedented enrichment of earth observation data and model data provides a great opportunity for development of earth science, but also provides new challenges, including the problems of common multisource heterogeneity (point-like, planar and net-like distribution), unequal precision, inconsistent resolution, serious system deviation and the like among water vapor observation values obtained by different technologies, which seriously hinders the joint utilization of multisource data, and data fusion is an effective way for solving the problems. Li (Li Z, 2004, production of regional 1 km x 1 km water boiler scales through The integration of GPS and MODIS data. In: proceedings of The 17th International Technical Meeting of The Satellite Division of The Institute of Navigation (ION GNSS 2004), long Beach, california, USA, 21-24 Sep 2004, pp.2396-2403) proposes a method of calibrating a medium Resolution Imaging spectrometer (MODIS) PWV with a GNSS PWV based on a linear model. Lindenbergh et al, (Lindenbergh, R., van der Marel, H., keshin, M., & De Haan, S. (2009). Validating time series of a combined GPS and MERIS Integrated Water value product. In Proceedings 2nd MERIS/(A) ATSR User works, september 22-26, 2008 ESA/ESRIN Fractification (round), 2009 (NB: all correction applied recent products version)) proposes a method of fusing PWV and Central Resolution Imaging Spectrometer (MERIS) PWV based on Kerrin interpolation. Alshawaf et al, (Alshawaf F, fersch B, hinz S, kunstmann H, mayer M, meyer FJ (2015) Water vacuum mapping by fusing InSAR and GNSS removal data and adhesion criteria. Hydro Earth System Sci 19 (12): 4747-4764) propose a fixed rank kriging interpolation method to fuse GNSS PWV, interferometric synthetic aperture radar (InSAR) PWV and WRF model PWV. However, the above fusion methods are all based on interpolation, and most of the existing few methods related to multi-source PWV data fusion are directly interpolated, and real data fusion is not realized; secondly, the existing method is simple in assumption, the accuracy problem of data from different sources is not considered, and the accuracy is damaged due to equal-weight use; thirdly, the existing method is difficult to realize the comprehensive utilization of different source data;
therefore, it is necessary to develop an atmospheric degradable water content data fusion method for realizing high-precision fusion of multi-source PWV data.
Disclosure of Invention
The invention aims to provide an atmospheric degradable water yield data fusion method based on an artificial neural network, which is a multisource PWV data fusion method based on a generalized regression neural network, the method has the functions of correcting PWV system deviation and improving PWV precision, high-precision fusion of multisource PWV data can be realized, and then a PWV data product with high precision and high space-time resolution is generated, and the method is simple, high in precision and high in data utilization rate; the problems of deviation calibration, weighting fusion and the like among multi-source PWV data are solved.
In order to realize the purpose, the technical scheme of the invention is as follows: a method for fusing data of a biodegradable Water Container (PWV) based on an artificial neural network is characterized in that: the system deviation calibration and precision optimization of the high-precision PWV to the low-precision PWV are realized by constructing an artificial neural network model between high-precision atmospheric degradable water yield data and low-precision atmospheric degradable water yield data;
the specific fusion method comprises the following steps of,
the method comprises the following steps: adopting a one-dimensional interpolation method to interpolate PWV data of different sources (the PWV data of different sources means that a plurality of PWV data have different precision and/or system deviation and/or inconsistent space-time resolution) to the same time, so that the PWV data of different sources are consistent in time;
step two: interpolating PWV data of different sources to the same sampling point by using a three-dimensional spatial interpolation method, so that the PWV data of different sources are consistent in space;
step three: performing quality control, removing poor PWV samples, and generating a high-quality PWV sample data set, wherein each high-quality PWV sample comprises longitude and latitude, elevation, time, low-precision PWV and high-precision PWV;
step four: constructing a Generalized Regression Neural Network (GRNN) model, wherein an input layer of the GRNN model is a longitude and latitude, elevation, time and low-precision PWV, and an output layer of the GRNN model is a high-precision PWV;
step five: training and testing the model by using a 10-fold cross-validation method, determining optimal GRNN model parameters according to the minimum root-mean-square error of a test result, and obtaining a final GRNN model and precision evaluation information thereof; in the fourth step, the fourth step and the fifth step, a generalized regression neural network model is constructed for calibrating system deviation among different source PWV data and improving PWV precision, and a PWV data product with high precision and high space-time resolution is generated through a corresponding inspection and evaluation method, so that the precision is high, and the data utilization rate is high;
step six: calibrating and optimizing a plurality of low-precision PWVs by using the trained GRNN model, wherein the low-precision PWV output by the trained GRNN model is the calibrated and optimized low-precision PWV;
step seven: and when the low-precision PWV and the high-precision PWV which are calibrated and optimized by the GRNN have no system deviation and have equivalent precision, combining the low-precision PWV and the high-precision PWV which are calibrated and optimized by the GRNN to obtain a fusion PWV.
In the above technical solution, in the first step, the one-dimensional temporal interpolation method at least includes a linear interpolation, a nearest neighbor interpolation, and a spline interpolation method, so as to implement temporal registration of data from different sources.
In the above technical solution, in the second step, the three-dimensional spatial interpolation is to interpolate PWV data with high spatial resolution to sampling points of data with low resolution by using a spherical cap harmonic model or a kriging interpolation method, so as to realize spatial registration of different data.
In the technical scheme, the generalized regression neural network model is used for constructing a nonlinear regression relationship between high-low-precision atmospheric degradable water yield data, and is a core innovation point of the method;
in the fourth step and the sixth step, the constructed generalized regression neural network model is used for calibrating the system deviation between different source PWV data (namely calibrating the system deviation of the low-precision PWV, so that the low-precision PWV calibrated and optimized by GRNN has no system deviation with the high-precision PWV), improving the precision of the low-precision PWV, enabling the low-precision PWV calibrated and optimized by GRNN to have the same precision as the high-precision PWV, obtaining the optimal model by the checking and evaluating method in the fifth step, and obtaining the calibrated and optimized low-precision PWV by the sixth step.
In the above technical solution, in step five, the accuracy evaluation information includes a system deviation, a median error, a root mean square error, a correlation coefficient, and the like.
The invention has the following advantages:
compared with the prior art, the method does not need to apply smooth constraint or hypothesis (namely, prior information or hypothesis), can self-adaptively construct a nonlinear regression relationship between a high-precision PWV and a low-precision PWV by utilizing an artificial neural network, can calibrate the system deviation of the low-precision PWV in a space differentiation manner by utilizing a trained neural network, optimizes the precision of the low-precision PWV, and does not artificially introduce the systematic deviation; in addition, the method can also maintain the resolution of the original data and correct the local deviation;
as a novel geoscience data fusion method, the artificial neural network model architecture used by the method can be expanded to other geoscience data (data containing time, geography and attribute information), and is suitable for system deviation calibration, precision optimization, data fusion and the like of various geoscience data.
Drawings
FIG. 1 is a GRNN model architecture diagram of the PWV fusion design of the present invention;
FIG. 2 is a flow chart of the GRNN model and PWV fusion process according to the present invention;
FIG. 3 shows the daily deviation (Bias) of MODIS PWV and ERA5 PWV before and after optimization of an embodiment of the invention;
FIG. 4 is a graph of standard deviation per day (STD) of MODIS PWV and ERA5 PWV before and after optimization of an embodiment of the invention;
FIG. 5 is a daily mean square error (RMSE) of MODIS PWV and ERA5 PWV before and after optimization of an embodiment of the invention.
Detailed Description
The embodiments of the present invention will be described in detail with reference to the drawings, but the present invention is not limited to these embodiments. While the advantages of the invention will be clear and readily understood by the description.
The invention provides a multisource PWV data calibration and fusion method based on a Generalized Regression Neural Network (GRNN). GRNN is a radial basis function network model based on mathematical statistics, and the theoretical basis is nonlinear regression analysis. GRNN has strong nonlinear mapping capability and learning speed, has more advantages than radial basis functions, has excellent prediction effect when the sample data is less, and can process unstable data;
the method can realize unbiased fusion of different PWV data sets, does not need prior information or hypothesis, can avoid introducing system deviation caused by artificial constraint, and can also keep the resolution of original data and correct local deviation.
Example (b): the invention is explained in detail by taking the embodiment of trying to fuse the data of the water capacity reduction in the atmosphere in a certain area, and the invention also has a guiding function of fusing the data of the water capacity reduction in the atmosphere in other areas.
In this embodiment, an experiment for calibrating and optimizing an MODIS PWV (atmospheric water reduction amount measured by a medium resolution imaging spectrometer) and an ERA5 PWV (atmospheric water reduction amount measured by a fifth generation atmospheric re-analysis data of a european middle-scale weather forecast center) by using a GNSS PWV (atmospheric water reduction amount measured by a global navigation satellite system) is carried out in a certain area (30 ° N-50 ° N and 126 ° W-102 ° W) by using the method of the present invention, the experimental data are 36,976 GNSS-MODIS samples and 56,291 GNSS-ERA5 samples in 2018, and finally, the method of the present invention is used to realize fusion of three PWV data, wherein the specific implementation mode is as follows:
A. interpolating GNSS PWV (atmospheric water reduction measured by a global navigation satellite system) data with high time resolution to the sampling time of ERA5 PWV (fifth generation atmospheric re-analysis data atmospheric water reduction of European mesoscale weather forecast center) and MODIS PWV (atmospheric water reduction measured by a medium resolution imaging spectrometer) with low resolution, and realizing time registration of different data;
B. on the basis of time registration, a spherical cap harmonic model is used for interpolating denser MODIS (ERA 5) PWV to sparse GNSS (global navigation satellite system) sites, and spatial registration of different data is realized;
C. performing quality control by using a medium error triple method to remove bad samples;
D. constructing a sample pair with input of (longitude, latitude, elevation, time, MODIS or ERA5 PWV) and output of (GNSS PWV), and carrying out normalization processing on all variables;
E. determining an optimal smooth parameter by a ten-fold cross verification method, training and testing a GRNN model to obtain an optimal model and precision information thereof;
F. respectively calibrating and optimizing MODIS PWV and ERA5 PWV by using the trained GRNN model, and fusing the optimized MODIS PWV and ERA5 PWV with GNSS PWV to obtain fused PWV;
it should be noted that: the above process of this embodiment needs to be performed on MODIS PWV and ERA5 PWV, that is, the MODIS PWV and ERA5 PWV are calibrated and optimized by GNSS PWV, and two GRNN models are finally obtained;
wherein ERA5 is the fifth generation atmosphere reanalysis data issued by the European mesoscale weather forecast center;
ERA5 PWV is data of the amount of atmospheric water reducible from ERA 5;
the GNSS is a global navigation satellite system;
the GNSS PWV is the atmospheric water reducible quantity measured by a global navigation satellite system;
MODIS is a medium resolution imaging spectrometer;
MODIS PWV is the atmospheric water reducible quantity measured by the medium resolution imaging spectrometer;
the GNSS-MODIS sample is a data sample obtained after space-time registration of the atmospheric degradable water content measured by a global navigation satellite system and a medium-resolution imaging spectrometer;
the GNSS-ERA5 sample is a data sample obtained by space-time registration of a global navigation satellite system and the atmospheric degradable water volume from ERA 5.
This example uses Root Mean Square Error (RMSE), standard Deviation (STD), bias (Bias), and correlation coefficient (R) to evaluate the performance of the present invention in data fusion. Experiments for calibrating and optimizing MODIS PWV and ERA5 PWV by utilizing GNSS PWV are developed in a certain area, so that the performance of the method is evaluated, and specific accuracy information for calibrating and optimizing by utilizing the GNSS PWV is shown in the following table 1;
TABLE 1 precision information (in mm) of the optimized PWV versus the original MODIS and ERA5 PWV
Table 1 above shows that the deviation Bias of MODIS PWV (atmospheric water reduction measured by medium resolution imaging spectrometer) is-2.1 mm, and the deviation Bias of ERA5 PWV (atmospheric water reduction data from ERA 5) is very small, 0.1 mm. After the calibration and optimization of the GRNN model, both the deviations are reduced to 0mm, which indicates that the GRNN model effectively calibrates the deviations. STD (standard deviation) estimates the accuracy of the data unaffected by the systematic deviation, and the results show: after GRNN optimization, the standard deviation STD of MODIS PWV (atmospheric water reduction measured by medium resolution imaging spectrometer) was reduced from 2.8 mm to 2.1 mm, and the standard deviation STD of ERA5 PWV (atmospheric water reduction data from ERA 5) was reduced from 1.9 mm to 1.6 mm. The improvement of the results of the Bias and the standard deviation STD shows that the method provided by the invention not only can calibrate the Bias, but also can improve the precision. After GRNN calibration and optimization, the root mean square error RMS of MODIS PWV (atmospheric water loss from mid resolution imaging spectrometer measurement) was reduced from 3.5 mm to 2.2 mm, ERA5 PWV (atmospheric water loss from ERA5 data) was reduced from 1.9 mm to 1.6 mm, with an overall accuracy improvement of 37.1% and 15.8%, respectively. The result improvement is very significant given that the accuracy of GNSS or radiosonde PWV is about 1-2 mm (the highest accuracy of current technology measuring PWV). The correlation coefficient R of MODIS PWV (atmospheric water reduction measured by medium resolution imaging spectrometer) increased from 0.95 to 0.96, the correlation coefficient R of ERA5 PWV (atmospheric water reduction data from ERA 5) increased from 0.97 to 0.98;
FIGS. 3, 4 and 5 respectively show the daily deviation Bias, standard deviation STD and root mean square error RMSE of MODIS PWV and ERA5 PWV before and after optimization by the method of the present invention;
FIG. 3 (a) is a graph showing the deviation per day for the optimization of the anterior and posterior MODIS PWV; in the legend, MODIS data (see the solid circle portion in the (a) diagram in fig. 3) represents Original MODIS PWV data (i.e., original MODIS, which is MODIS PWV data before optimization); the optimized MODIS data (see the open triangle part in the (a) diagram in fig. 3) are represented as Modified MODIS, i.e., optimized MODIS PWV data; the abscissa of the graph (a) in fig. 3 represents the yearly cumulative date of 2018 in units of: day; the ordinate represents the systematic deviation in units of: millimeter;
FIG. 3 (b) is a graph showing the deviation of the ERA5 PWV before and after optimization per day; in the legend, the ERA5 data (see the solid circle portion in the (b) diagram in fig. 3) represents the Original ERA5 data (i.e., original ERA5, ERA5 PWV data before optimization); the optimized ERA5 PWV data (see the open triangle part in the (b) diagram in fig. 3) is represented as Modified ERA5, i.e. optimized ERA5 PWV data; the abscissa of the graph (b) in fig. 3 represents the yearly cumulative date of 2018 in units of: day(s); the ordinate represents the system deviation in units of: millimeter;
FIG. 4 (a) is a graph showing the standard deviation of the MODIS PWV before and after optimization per day; in the legend, MODIS data (see the solid circle portion in the (a) diagram in fig. 4) represents Original MODIS PWV data (i.e., original MODIS, which is MODIS PWV data before optimization); the optimized MODIS data (see the open triangle part in the (a) diagram in fig. 4) are represented as Modified MODIS, i.e., optimized MODIS PWV data; the abscissa of the graph (a) in fig. 4 represents the yearly time of 2018 in units of: day; the ordinate represents the standard deviation in units of: millimeter;
FIG. 4 (b) is a graph showing the standard deviation per day for optimizing pre-and post-ERA 5 PWV; in the legend, the ERA5 data (see the solid circle portion in the (b) diagram in fig. 4) represents the Original ERA5 PWV data (i.e., original ERA5, ERA5 PWV data before optimization); the optimized ERA5 data (see the open triangle part in the (b) diagram in fig. 4) are represented as Modified ERA5, i.e. optimized ERA5 PWV data; the abscissa of the graph (b) in fig. 4 represents the yearly cumulative date of 2018 in units of: day; the ordinate represents the standard deviation in units of: millimeter;
FIG. 5 (a) is a graph showing the root mean square error per day for both before and after optimization of MODIS PWV; in the legend, MODIS data (see the solid circle portion in the (a) diagram in fig. 5) represents Original MODIS PWV data (i.e., original MODIS, which is MODIS PWV data before optimization); the optimized MODIS data (see the hollow triangular part in the diagram (a) in fig. 5), namely Modified MODIS, are optimized MODIS PWV data; the abscissa of the graph (a) in fig. 5 represents the yearly cumulative date of 2018 in units of: day; the ordinate represents the root mean square error in units of: millimeter;
FIG. 5 (b) is a graph showing the root mean square error for each day of ERA5 PWV before and after optimization; in the legend, the ERA5 data (see the solid circle portion in the (b) diagram in fig. 5) represents the Original ERA5 PWV data (i.e., original ERA5, ERA5 PWV data before optimization); the optimized ERA5 data (see the hollow triangular part in the graph (b) in FIG. 5), namely Modified ERA5, is optimized ERA5 PWV data; the abscissa of the graph (b) in fig. 5 represents the yearly time of 2018 in units of: day; the ordinate represents the root mean square error in units of: millimeter;
in fig. 3, 4 and 5, blank portions are all missing from the original observed data. As can be seen from fig. 3, fig. 4 and fig. 5, after the optimization of the method of the present invention, bias, STD and RMSE of MODIS PWV and ERA5 PWV are significantly reduced most of the time, which further proves the effectiveness of the present invention in weakening system deviation and improving accuracy;
in fig. 1, PWV is the amount of atmospheric reducible water;
x 1 、x 2 、x 3 、x 4 and x 5 Respectively representing latitude, radius, elevation, time and low precision PWV (namely low precision atmospheric water reducible quantity);
in fig. 2, PWV is the amount of atmospheric reducible water;
MODIS PWV is the amount of atmospheric water reducible by the medium resolution imaging spectrometer;
the GNSS PWV is the atmospheric degradable water volume measured by a global navigation satellite system;
ERA5 PWV is the amount of atmospheric water reducible from ERA 5.
Other parts not described belong to the prior art.
Claims (5)
1. An atmospheric degradable water yield data fusion method based on an artificial neural network is characterized by comprising the following steps: by constructing a generalized regression neural network model, calibrating system deviation among different source PWV data and improving PWV precision, and realizing system deviation calibration and precision optimization of high-precision PWV to low-precision PWV;
the specific fusion method comprises the following steps of,
the method comprises the following steps: adopting a one-dimensional interpolation method to interpolate PWV data of different sources to the same time, so that the PWV data of different sources are kept consistent in time;
step two: interpolating PWV data of different sources to the same sampling point by using a three-dimensional spatial interpolation method, so that the PWV data of different sources are consistent in space;
step three: performing quality control, removing poor PWV samples, and generating a high-quality PWV sample data set, wherein each high-quality PWV sample comprises longitude and latitude, elevation, time, low-precision PWV and high-precision PWV;
step four: constructing a generalized regression neural network model, wherein the input layer of the GRNN model is longitude and latitude, elevation, time and low-precision PWV, and the output layer is high-precision PWV;
step five: training and testing the model by using a 10-time cross validation method, determining an optimal GRNN model parameter according to the minimum root mean square error of a test result, and obtaining a final GRNN model and precision evaluation information thereof;
step six: calibrating and optimizing the low-precision PWV by using the trained GRNN model, wherein the PWV output by the trained GRNN model is the calibrated and optimized PWV;
step seven: and when the low-precision PWV and the high-precision PWV which are calibrated and optimized by the GRNN have no system deviation and have equivalent precision, combining the low-precision PWV and the high-precision PWV which are calibrated and optimized by the GRNN to obtain a fusion PWV.
2. The atmospheric degradable water volume data fusion method based on the artificial neural network as claimed in claim 1, wherein: in step one, the one-dimensional interpolation method in time at least comprises linear interpolation, nearest neighbor interpolation and spline curve interpolation.
3. The atmospheric degradable water volume data fusion method based on the artificial neural network as claimed in claim 1 or 2, wherein: in the second step, the three-dimensional spatial interpolation is to interpolate the PWV data with high spatial resolution to the sampling points of the data with low resolution by using a spherical cap harmonic model or a kriging interpolation method, so as to realize the spatial registration of different data.
4. The artificial neural network-based atmospheric degradable water content data fusion method according to claim 3, characterized in that: establishing a nonlinear regression relation between high-low precision atmospheric degradable water yield data by utilizing a generalized regression neural network model;
in the fourth step and the sixth step, the constructed generalized regression neural network model is used for calibrating the system deviation of the low-precision PWV and improving the precision of the low-precision PWV, the optimal model is determined through the checking and evaluating method in the fifth step, and the calibrated and optimized low-precision PWV is obtained through the sixth step.
5. The artificial neural network-based atmospheric degradable water content data fusion method according to claim 4, characterized in that: in the fifth step, the precision evaluation information comprises system deviation, mean error, root mean square error and correlation coefficient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211306842.0A CN115392401A (en) | 2022-10-25 | 2022-10-25 | Atmospheric water-reducing capacity data fusion method based on artificial neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211306842.0A CN115392401A (en) | 2022-10-25 | 2022-10-25 | Atmospheric water-reducing capacity data fusion method based on artificial neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115392401A true CN115392401A (en) | 2022-11-25 |
Family
ID=84128486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211306842.0A Withdrawn CN115392401A (en) | 2022-10-25 | 2022-10-25 | Atmospheric water-reducing capacity data fusion method based on artificial neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115392401A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542132A (en) * | 2023-04-07 | 2023-08-04 | 武汉大学 | Water vapor data calibration and optimization method combining spherical crown harmonic model and neural network |
-
2022
- 2022-10-25 CN CN202211306842.0A patent/CN115392401A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542132A (en) * | 2023-04-07 | 2023-08-04 | 武汉大学 | Water vapor data calibration and optimization method combining spherical crown harmonic model and neural network |
| CN116542132B (en) * | 2023-04-07 | 2024-04-16 | 武汉大学 | Water vapor data calibration and optimization method combining spherical crown harmonic model and neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112905560A (en) | Air pollution prediction method based on multi-source time-space big data deep fusion | |
| CN111401602B (en) | Assimilation method for satellite and ground rainfall measurement values based on neural network | |
| CN113297527A (en) | PM based on multisource city big data2.5Overall domain space-time calculation inference method | |
| Zhang et al. | Precipitable water vapor fusion: An approach based on spherical cap harmonic analysis and Helmert variance component estimation | |
| CN104764868B (en) | A kind of soil organic matter Forecasting Methodology based on Geographical Weighted Regression | |
| CN110909447B (en) | High-precision short-term prediction method for ionization layer region | |
| CN111695440B (en) | GA-SVR lake level measurement and prediction method based on radar altimeter | |
| CN116297068B (en) | Aerosol optical thickness inversion method and system based on earth surface reflectivity optimization | |
| CN115392401A (en) | Atmospheric water-reducing capacity data fusion method based on artificial neural network | |
| Zhang et al. | A new integrated method of GNSS and MODIS measurements for tropospheric water vapor tomography | |
| CN113408111B (en) | Atmospheric precipitation inversion method and system, electronic equipment and storage medium | |
| CN113468799A (en) | Method and system for acquiring near-ground PM2.5 concentration in static meteorological satellite observation | |
| CN114861840A (en) | Multi-source precipitation data fusion method | |
| CN115859789A (en) | Method for improving inversion accuracy of polar atmosphere temperature profile | |
| CN109884666B (en) | Troposphere delay correction method based on data assimilation technology | |
| Ma et al. | Improving the accuracy and spatial resolution of precipitable water vapor dataset using a neural network-based downscaling method | |
| CN114372508A (en) | Moon-scale star-ground precipitation fusion method based on random forest and land surface environment variables | |
| CN117219183A (en) | High coverage near ground NO in cloudy rain areas 2 Concentration estimation method and system | |
| CN116299247B (en) | InSAR atmospheric correction method based on sparse convolutional neural network | |
| CN115292968A (en) | Multi-source atmospheric degradable water yield data fusion method based on spherical cap harmonic model | |
| CN116699671A (en) | Ionosphere amplitude flicker index calculation method based on random forest regression | |
| CN115616637A (en) | Urban complex environment navigation positioning method based on three-dimensional grid multipath modeling | |
| CN116188705A (en) | Reconstruction method for kilometer-level resolution stereo distribution of regional atmospheric pollutants | |
| Xie et al. | Spatial downscaling of TRMM precipitation using an optimal regression model with NDVI in inner Mongolia, China | |
| CN116542132B (en) | Water vapor data calibration and optimization method combining spherical crown harmonic model and neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20221125 |