CN113435119A - Global ocean surface brightness temperature determination method and system - Google Patents
Global ocean surface brightness temperature determination method and system Download PDFInfo
- Publication number
- CN113435119A CN113435119A CN202110729585.0A CN202110729585A CN113435119A CN 113435119 A CN113435119 A CN 113435119A CN 202110729585 A CN202110729585 A CN 202110729585A CN 113435119 A CN113435119 A CN 113435119A
- Authority
- CN
- China
- Prior art keywords
- data set
- layer
- rttov
- layer data
- deep 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention relates to a global ocean surface brightness temperature determination method and a global ocean surface brightness temperature determination system. The method comprises the following steps: acquiring a single-layer data set and a barometer layer data set; determining a trained AI-RTTOV deep network by adopting a deep learning method according to the single-layer data set and the atmospheric pressure layer data set; and determining the global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation by utilizing the trained AI-RTTOV deep network according to the single-layer data set and the atmospheric pressure layer data set. The method can improve the efficiency of the atmospheric radiation transmission mode in the forward process of the satellite and reduce time consumption.
Description
Technical Field
The invention relates to the field of atmospheric radiation transmission, in particular to a global ocean surface brightness temperature determination method and system.
Background
RTTOV is one of the most common atmospheric radiation transmission models, and is mainly used for satellite data assimilation, quantitative information extraction and instrument performance stability monitoring. Through years of improvement and perfection, the method has developed into a fast radiation transmission mode for business operation, and satellite data assimilation based on the radiation mode obviously improves the accuracy of numerical weather forecast (NWP). Derber et al assimilate TOVS clear sky radiation data directly into the assimilation system of the national environmental forecast center of America, and improve the forecasting skill of the mode. The results of the simulation of the rainstorm process of Dingweiyu and the like show that the initial field water vapor and temperature distribution can be improved after the satellite data are assimilated, and the short-time rainfall forecast is positively influenced. Zhanghua and the like are based on a GRAPES three-dimensional variation assimilation system, assimilate and improve the data of a microwave detection device, and enable a forecasted typhoon path to be more accurate. The results of the researches on typhoons by utilizing RTTOV (real time to live) rapid radiation transmission modes of Tnweiyu and the like show that after the influence of clouds is considered by clear sky stripes, the simulation results corresponding to the low-layer channels are consistent with the observation results. The RTTOV transmittance coefficient was recalculated by Marangg et al and used for simulation of FY-2B infrared and water vapor channel light temperature. The RTTOV-SCATT is used instead by Sujie and the like to bring the information of the water condensate in the atmosphere into the calculation of assimilation, so that the phenomenon of overestimation of the water vapor content in the atmosphere caused by using the conventional method can be effectively solved. However, these widely used physical models also have drawbacks, such as 1.3 hours required for outputting daily data through the CRTM model. Therefore, at present, the amount of assimilable satellite data is very small, and with the increase of the amount of satellite observation data, how to improve the efficiency of application of radiation mode services such as RTTOV, and in particular, how to improve the assimilation efficiency of satellite data, is a problem that needs to be solved urgently.
Therefore, a new method or system is needed to improve the time-consuming problem of the physical algorithm-based radiation transmission mode in the forward direction of the satellite, so as to improve the forward performance and efficiency of the satellite.
Disclosure of Invention
The invention aims to provide a global ocean surface brightness temperature determination method and a global ocean surface brightness temperature determination system, which can improve the efficiency of an atmospheric radiation transmission mode in the satellite forward process and reduce time consumption.
In order to achieve the purpose, the invention provides the following scheme:
a global ocean surface brightness temperature determination method comprises the following steps:
acquiring a single-layer data set and a barometer layer data set; the single-layer dataset includes: satellite azimuth, satellite zenith angle, latitude, longitude, earth surface air pressure, 2 m specific humidity, 2 m temperature, 10 m U air volume and 10 m V air volume; the barometric pressure layer dataset comprising: temperature profile, ozone profile, specific humidity profile, and air pressure profile data;
determining a trained AI-RTTOV deep network by adopting a deep learning method according to the single-layer data set and the atmospheric pressure layer data set; the AI-RTTOV deep network comprises a local connection layer, a residual connection layer, a multi-layer perceptron layer and an output layer; the AI-RTTOV deep network firstly performs original feature mapping and residual error learning on an input single-layer data set and an air pressure layer data set, secondly inputs the learned features into a multi-layer sensor layer for deep feature extraction, fuses the features of different layers of networks, and outputs a global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation;
and determining the global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation by utilizing the trained AI-RTTOV deep network according to the single-layer data set and the atmospheric pressure layer data set.
Optionally, the determining a trained AI-RTTOV deep network according to the single-layer data set and the atmospheric pressure data set by using a deep learning method specifically includes:
the single-layer data set and the air pressure layer data set are respectively divided into a training data set, a verification data set and a test data set;
training and super-parameter tuning of an AI-RTTOV deep network are performed by using a training data set and a verification data set which are divided by the single-layer data set and the atmospheric pressure data set;
and testing the trained AI-RTTOV deep network by using the test set divided by the single-layer data set and the atmospheric pressure layer data set.
Optionally, the core size of the local connection layer is 3, and the step size is 1.
Optionally, the activation function of the output layer is a linear activation function.
Optionally, the activation functions of the local connection layer, the residual connection layer, and the multi-layer perceptron layer are relu activation functions.
A global ocean surface light temperature determination system comprising:
the data set acquisition module is used for acquiring a single-layer data set and an air pressure layer data set; the single-layer dataset includes: satellite azimuth, satellite zenith angle, latitude, longitude, earth surface air pressure, 2 m specific humidity, 2 m temperature, 10 m U air volume and 10 m V air volume; the barometric pressure layer dataset comprising: temperature profile, ozone profile, specific humidity profile, and air pressure profile data;
the deep network determining module is used for determining a trained AI-RTTOV deep network by adopting a deep learning method according to the single-layer data set and the air pressure layer data set; the AI-RTTOV deep network comprises a local connection layer, a residual connection layer, a multi-layer perceptron layer and an output layer; the AI-RTTOV deep network firstly performs original feature mapping and residual error learning on an input single-layer data set and an air pressure layer data set, secondly inputs the learned features into a multi-layer sensor layer for deep feature extraction, fuses the features of different layers of networks, and outputs a global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation;
and the brightness temperature value determining module is used for determining the global-range ocean brightness temperature value based on RTTOV simulation FY3-D MWHS183 +/-1 GHz by utilizing the trained AI-RTTOV depth network according to the single-layer data set and the atmospheric pressure layer data set.
Optionally, the deep network determining module specifically includes:
the data set dividing unit is used for dividing the single-layer data set and the air pressure layer data set into a training data set, a verification data set and a test data set respectively;
the deep network training unit is used for training an AI-RTTOV deep network and adjusting hyper-parameters by utilizing a training data set and a verification data set which are divided by the single-layer data set and the atmospheric pressure layer data set;
and the deep network testing unit is used for testing the trained AI-RTTOV deep network by utilizing the single-layer data set and the testing set divided by the air pressure layer data set.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a global ocean surface brightness temperature determination method and a system, wherein a single-layer data set and an air pressure layer data set are respectively used as input of an AI-RTTOV deep network, the input is firstly subjected to original feature mapping and residual error learning, then the learned features are input into a multilayer sensor layer for deep feature extraction, meanwhile, the AI-RTTOV is fused with different hierarchical network features to explore a complex nonlinear relation in an atmospheric radiation transmission process, and finally, the network output is a global ocean brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation. On the basis of the RTTOV calculation principle, forward simulation of the observed brightness temperature on the sea of the FY3-D MWHS183 +/-1 GHz channel is realized by constructing a deep nonlinear network AI-RTTOV to simulate a complex atmospheric radiation transmission process. The performance and the efficiency of the forward modeling of the satellite are improved, and the problems of low efficiency and serious time consumption of an atmospheric radiation transmission mode based on a physical algorithm in the forward modeling process of the satellite are solved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a method for determining global ocean surface brightness temperature according to the present invention;
FIG. 2 is a schematic diagram of an AI-RTTOV deep network structure provided by the present invention;
FIG. 3 is a schematic diagram of a residual connection layer structure provided in the present invention;
FIG. 4 is a graph of loss versus MAE during the AI-RTTOV deep network training phase;
FIG. 5 is a schematic diagram illustrating a correlation statistic comparison between an AI-RTTOV deep network and an RTTOV model;
FIG. 6 is a schematic diagram illustrating a statistical relationship between an AI-RTTOV deep network and an RTTOV model;
FIG. 7 is a schematic diagram of spatial distribution (ascending trajectory) of AI-RTTOV deep network and RTTOV model estimation results;
FIG. 8 is a schematic diagram of spatial distribution (falling trajectory) of the AI-RTTOV depth network and the RTTOV model estimation results;
FIG. 9 is a schematic diagram of the spatial distribution of the deviations between the AI-RTTOV depth network and the RTTOV model;
fig. 10 is a schematic structural diagram of a global ocean surface brightness temperature determination system provided by the present invention.
Detailed Description
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.
The invention aims to provide a global ocean surface brightness temperature determination method and a global ocean surface brightness temperature determination system, which can improve the efficiency of an atmospheric radiation transmission mode in the satellite forward process and reduce time consumption.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a schematic flow chart of a method for determining a global ocean surface brightness temperature according to the present invention, and as shown in fig. 1, the method for determining a global ocean surface brightness temperature according to the present invention includes:
s101, acquiring a single-layer data set and an air pressure layer data set; the single-layer dataset includes: satellite azimuth, satellite zenith angle, latitude, longitude, earth surface air pressure, 2 m specific humidity, 2 m temperature, 10 m U air volume and 10 m V air volume; the barometric pressure layer dataset comprising: temperature profile, ozone profile, specific humidity profile, and air pressure profile data;
s102, determining a trained AI-RTTOV deep network by adopting a deep learning method according to the single-layer data set and the atmospheric pressure layer data set; the AI-RTTOV deep network comprises a local connection layer, a residual connection layer, a multi-layer perceptron layer and an output layer; the AI-RTTOV deep network firstly performs original feature mapping and residual error learning on an input single-layer data set and an air pressure layer data set, secondly inputs the learned features into a multi-layer sensor layer for deep feature extraction, fuses the features of different layers of networks, and outputs a global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation;
s102 specifically comprises the following steps:
the single-layer data set and the air pressure layer data set are respectively divided into a training data set, a verification data set and a test data set;
training and super-parameter tuning of an AI-RTTOV deep network are performed by using a training data set and a verification data set which are divided by the single-layer data set and the atmospheric pressure data set;
and testing the trained AI-RTTOV deep network by using the test set divided by the single-layer data set and the atmospheric pressure layer data set.
As a specific example, the AI-RTTOV deep network has the ability to learn a time-periodic law. And dividing the single-layer data set and the atmospheric pressure layer data set from 1 month and 1 day in 2019 to 1 month and 30 days in 2019 into a training data set, a verification data set and test data according to a certain rule. The specific division rule is as follows: the data from 1/2019 to 7/2019/1/2019, from 10/2019/1/2019 to 17/2019/1/20/2019 to 27/2019/1/2019 are divided into training data sets, the data from 8/2019/1/9/2019, from 18/2019/1/18/2019 to 19/2019/1/19/2019 and from 29/2019/1/29/2019 are divided into verification data sets, and the data from 10/2019/1/2019, 20/2019/1/2019 and 30/2019/1/2019 are divided into test data sets. The training data set and the verification data set are used for training and super-parameter optimization of the AI-RTTOV deep network, and the performance evaluation of the AI-RTTOV deep network is performed on the test data set.
As shown in FIG. 2, the training of the AI-RTTOV deep network was done using a Tesla M608G GPU in the Linux operating system environment based on the Tensorflow Keras deep learning framework. The AI-RTTOV deep network iterates 100 times in total, taking 7 minutes. When the verification data set MAE reaches 0.6K, the continuous training of the AI-RTTOV deep network is stopped, so as to prevent the AI-RTTOV deep network from being over-fitted.
As shown in table 1, the core size of the local connection layer is 3, and the step size is 1.
TABLE 1
The activation function of the output layer is a linear activation function.
The activation functions of the local connection layer, the residual connection layer and the multi-layer perceptron layer are relu activation functions. And the batch size (batch size) is set to 4096.
As the number of network layers increases, the details of the original data gradually decrease in the forward transmission of the data in the deep network, and deep network learning becomes difficult, however, residual concatenation can effectively solve the problem. The structure of the residual connection layer used herein is shown in fig. 3, the parameter settings of the residual connection structure layer are shown in table 2, the residual connection module also mainly uses a local connection layer with a kernel size of 3 and a step size of 1, and each layer of activation function is also set to relu.
TABLE 2
Taking the average absolute error (MAE) and the correlation coefficient (R2) as AI-RTTOV performance evaluation indexes, wherein the calculation formulas of the average absolute error (MAE) and the correlation coefficient (R) are shown as formula (1) and formula (2):
in the formula (I), the compound is shown in the specification,
is AI-RTTOV estimation result, and y is RTTOV estimation result;
as a covariance of the AI-RTTOV and RTTOV estimates,
the variance of the result, Var [ y ], is estimated for AI-RTTOV]The resulting variance is estimated for RTTOV.
S103, determining a global marine brightness temperature value of 183 +/-1 GHz of the FY3-DMWHS based on RTTOV simulation by using the trained AI-RTTOV deep network according to the single-layer data set and the atmospheric pressure layer data set.
FIG. 4 shows a graph of MAE, validation data set loss, and training data set loss for the validation data set at each training iteration of the AI-RTTOV. Along with the increase of the training iteration number, the MAE and loss curve is continuously reduced, which shows that the AI-RTTOV learning ability is gradually improved. The MAE curve gradually approaches 0 in slope as the number of training iterations increases. This indicates that the model has converged to an optimal state, which can be a good indicator for AI-RTTOV stopping training. In the experiment, in order to prevent the AI-RTTOV from generating an overfitting phenomenon, a learning rate gradual linear attenuation training strategy with an initial learning rate of 0.001 is adopted. The stepwise linear attenuation is calculated as shown in equation (3):
in the formula (I), the compound is shown in the specification,
for the learning rate after attenuation, lr is the learning rate before attenuation, decadey is the learning rate attenuation coefficient, and is set to 0.0001 herein, and epoch is the current iteration number of the model.
Comparing the AI-RTTOV estimate with the RTTOV estimate, the AI-RTTOV performance can be evaluated, as described above, using data from month 10 of 2019, month 1 of 2019, month 20 of 2019, and month 30 of 2019, respectively. FIG. 5 is a scatter plot of the AI-RTTOV estimates and the RTTOV estimates showing that the AI-RTTOV estimates are centrally distributed around the 1:1 line (the 1:1 line shows the correlation between the AI-RTTOV estimate and the RTTOV estimate, and closer to the 1:1 line shows a tighter correlation). Meanwhile, the best fit straight line of the AI-RTTOV and the RTTOV is close to a 1:1 line, which shows that the AI-RTTOV estimation result is consistently related to the RTTOV estimation result.
In order to quantitatively describe the forward effect of AI-RTTOV on the observed light temperature of FY3-D MWHS183 +/-1 GHz, the quantitative evaluation indexes of AI-RTTOV and RTTOV are given in Table 3. The correlation coefficients (R2) of the AI-RTTOV estimate and the RTTOV estimate on the test data set are 0.988949, 0.988949, 0.990367, respectively, and the Mean Absolute Error (MAE) is 0.600686K, 0.624342K, 0.608329K, respectively.
TABLE 3
The AI-RTTOV and RTTOV estimate density histograms and both error density histograms are shown in FIG. 6. The AI-RTTOV and RTTOV estimated light temperature distribution range is centered between 230K and 275K, and the peak value is 240K. The third column in FIG. 6 shows the error density histogram of the AI-RTTOV and RTTOV estimation results, wherein in the error statistics, the error values are centrally distributed between + -2.5K, and the error range is smaller; the overall distribution form of the error is also close to the normal distribution, and the result similarity is higher.
Fig. 7 and 8 are graphs showing the test data of the global ocean surface AI-RTTOV and RTTOV of the ascending orbit and the descending orbit, respectively, and it can be seen that the AI-RTTOV estimation result and the RTTOV estimation result of either the ascending orbit or the descending orbit are relatively consistent in spatial distribution, and the light temperature distribution range is approximately 230K to 275K. FIG. 9 is a spatial distribution of error for both the global ocean surface for the ascending and descending tracks, respectively. The error space distribution of the two is relatively uniform, and the error distribution interval is approximately between-2.5K and 2.5K. This result further verifies the high consistency of the AI-RTTOV estimate with the RTTOV estimate.
Table 4 describes the time complexity of AI-RTTOV and RTTOV traffic operations. The CPU time required for processing FY3-D MWHS183 +/-1 GHz channel light temperature in one day under the same working environment is only 7 minutes for AI-RTTOV, and 1.3 hours for RTTOV. The processing efficiency of AI-RTTOV is much higher than that of RTTOV.
TABLE 4
Fig. 10 is a schematic structural diagram of a global ocean surface brightness temperature determination system provided by the present invention, and as shown in fig. 10, the global ocean surface brightness temperature determination system provided by the present invention includes:
a dataset acquisition module 1001 configured to acquire a single-layer dataset and a barolayer dataset; the single-layer dataset includes: satellite azimuth, satellite zenith angle, latitude, longitude, earth surface air pressure, 2 m specific humidity, 2 m temperature, 10 m U air volume and 10 m V air volume; the barometric pressure layer dataset comprising: temperature profile, ozone profile, specific humidity profile, and air pressure profile data;
a deep network determining module 1002, configured to determine a trained AI-RTTOV deep network by using a deep learning method according to the single-layer data set and the air pressure layer data set; the AI-RTTOV deep network comprises a local connection layer, a residual connection layer, a multi-layer perceptron layer and an output layer; the AI-RTTOV deep network firstly performs original feature mapping and residual error learning on an input single-layer data set and an air pressure layer data set, secondly inputs the learned features into a multi-layer sensor layer for deep feature extraction, fuses the features of different layers of networks, and outputs a global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation;
and a brightness temperature value determining module 1003, configured to determine, according to the single-layer data set and the atmospheric pressure layer data set, a global-range marine brightness temperature value based on RTTOV simulation, based on the trained AI-RTTOV deep network, of FY3-D MWHS183 ± 1 GHz.
The deep network determining module 1002 specifically includes:
the data set dividing unit is used for dividing the single-layer data set and the air pressure layer data set into a training data set, a verification data set and a test data set respectively;
the deep network training unit is used for training an AI-RTTOV deep network and adjusting hyper-parameters by utilizing a training data set and a verification data set which are divided by the single-layer data set and the atmospheric pressure layer data set;
and the deep network testing unit is used for testing the trained AI-RTTOV deep network by utilizing the single-layer data set and the testing set divided by the air pressure layer data set.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (7)
1. A global ocean surface brightness temperature determination method is characterized by comprising the following steps:
acquiring a single-layer data set and a barometer layer data set; the single-layer dataset includes: satellite azimuth, satellite zenith angle, latitude, longitude, earth surface air pressure, 2 m specific humidity, 2 m temperature, 10 m U air volume and 10 m V air volume; the barometric pressure layer dataset comprising: temperature profile, ozone profile, specific humidity profile, and air pressure profile data;
determining a trained AI-RTTOV deep network by adopting a deep learning method according to the single-layer data set and the atmospheric pressure layer data set; the AI-RTTOV deep network comprises a local connection layer, a residual connection layer, a multi-layer perceptron layer and an output layer; the AI-RTTOV deep network firstly performs original feature mapping and residual error learning on an input single-layer data set and an air pressure layer data set, secondly inputs the learned features into a multi-layer sensor layer for deep feature extraction, fuses the features of different layers of networks, and outputs a global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation;
and determining the global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation by utilizing the trained AI-RTTOV deep network according to the single-layer data set and the atmospheric pressure layer data set.
2. The method for determining the global ocean surface brightness temperature according to claim 1, wherein the determining the trained AI-RTTOV depth network according to the single-layer data set and the atmospheric pressure data set by using a deep learning method specifically comprises:
the single-layer data set and the air pressure layer data set are respectively divided into a training data set, a verification data set and a test data set;
training and super-parameter tuning of an AI-RTTOV deep network are performed by using a training data set and a verification data set which are divided by the single-layer data set and the atmospheric pressure data set;
and testing the trained AI-RTTOV deep network by using the test set divided by the single-layer data set and the atmospheric pressure layer data set.
3. The method according to claim 1, wherein the local connection layer has a kernel size of 3 and a step size of 1.
4. The method as claimed in claim 1, wherein the activation function of the output layer is a linear activation function.
5. The method according to claim 1, wherein the activation functions of the local connection layer, the residual connection layer and the multilayer sensor layer are relu activation functions.
6. A global ocean surface light temperature determination system, comprising:
the data set acquisition module is used for acquiring a single-layer data set and an air pressure layer data set; the single-layer dataset includes: satellite azimuth, satellite zenith angle, latitude, longitude, earth surface air pressure, 2 m specific humidity, 2 m temperature, 10 m U air volume and 10 m V air volume; the barometric pressure layer dataset comprising: temperature profile, ozone profile, specific humidity profile, and air pressure profile data;
the deep network determining module is used for determining a trained AI-RTTOV deep network by adopting a deep learning method according to the single-layer data set and the air pressure layer data set; the AI-RTTOV deep network comprises a local connection layer, a residual connection layer, a multi-layer perceptron layer and an output layer; the AI-RTTOV deep network firstly performs original feature mapping and residual error learning on an input single-layer data set and an air pressure layer data set, secondly inputs the learned features into a multi-layer sensor layer for deep feature extraction, fuses the features of different layers of networks, and outputs a global marine brightness temperature value of FY3-D MWHS183 +/-1 GHz based on RTTOV simulation;
and the brightness temperature value determining module is used for determining the global-range ocean brightness temperature value based on RTTOV simulation FY3-D MWHS183 +/-1 GHz by utilizing the trained AI-RTTOV depth network according to the single-layer data set and the atmospheric pressure layer data set.
7. The system according to claim 6, wherein the depth network determining module specifically comprises:
the data set dividing unit is used for dividing the single-layer data set and the air pressure layer data set into a training data set, a verification data set and a test data set respectively;
the deep network training unit is used for training an AI-RTTOV deep network and adjusting hyper-parameters by utilizing a training data set and a verification data set which are divided by the single-layer data set and the atmospheric pressure layer data set;
and the deep network testing unit is used for testing the trained AI-RTTOV deep network by utilizing the single-layer data set and the testing set divided by the air pressure layer data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110729585.0A CN113435119B (en) | 2021-06-29 | 2021-06-29 | Global ocean surface brightness temperature determination method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110729585.0A CN113435119B (en) | 2021-06-29 | 2021-06-29 | Global ocean surface brightness temperature determination method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113435119A true CN113435119A (en) | 2021-09-24 |
| CN113435119B CN113435119B (en) | 2022-08-05 |
Family
ID=77757758
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110729585.0A Active CN113435119B (en) | 2021-06-29 | 2021-06-29 | Global ocean surface brightness temperature determination method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113435119B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170140253A1 (en) * | 2015-11-12 | 2017-05-18 | Xerox Corporation | Multi-layer fusion in a convolutional neural network for image classification |
| CN109902748A (en) * | 2019-03-04 | 2019-06-18 | 中国计量大学 | A kind of image, semantic dividing method based on the full convolutional neural networks of fusion of multi-layer information |
| US20200309993A1 (en) * | 2019-03-25 | 2020-10-01 | Yandex Europe Ag | Method of and system for generating weather forecast |
| CN111737912A (en) * | 2020-06-15 | 2020-10-02 | 洛阳师范学院 | MWHTS simulated bright temperature calculation method based on deep neural network |
-
2021
- 2021-06-29 CN CN202110729585.0A patent/CN113435119B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170140253A1 (en) * | 2015-11-12 | 2017-05-18 | Xerox Corporation | Multi-layer fusion in a convolutional neural network for image classification |
| CN109902748A (en) * | 2019-03-04 | 2019-06-18 | 中国计量大学 | A kind of image, semantic dividing method based on the full convolutional neural networks of fusion of multi-layer information |
| US20200309993A1 (en) * | 2019-03-25 | 2020-10-01 | Yandex Europe Ag | Method of and system for generating weather forecast |
| CN111737912A (en) * | 2020-06-15 | 2020-10-02 | 洛阳师范学院 | MWHTS simulated bright temperature calculation method based on deep neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113435119B (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Comparison of support vector machine and copula-based nonlinear quantile regression for estimating the daily diffuse solar radiation: A case study in China | |
| Lv et al. | Wave characteristics analysis in Bohai Sea based on ECMWF wind field | |
| Mirzaei et al. | Wave climate simulation for southern region of the South China Sea | |
| CN109299812B (en) | Flood prediction method based on deep learning model and KNN real-time correction | |
| Castro et al. | Spatial estimation of daily precipitation in regions with complex relief and scarce data using terrain orientation | |
| CN114254561A (en) | Waterlogging prediction method, waterlogging prediction system and storage medium | |
| CN110213003B (en) | Wireless channel large-scale fading modeling method and device | |
| CN111260111A (en) | Runoff forecast improvement method based on meteorological big data | |
| CN111665218B (en) | Method for improving inversion accuracy of carbon dioxide differential absorption laser radar | |
| CN116449331B (en) | Dust particle number concentration estimation method based on W-band radar and meteorological satellite | |
| CN113281754A (en) | WRF-Hydro key parameter calibration method for quantitatively estimating rainfall by integrating rainfall station with radar | |
| Gámiz-Fortis et al. | Spatio-temporal variability in Ebro river basin (NE Spain): Global SST as potential source of predictability on decadal time scales | |
| Chen et al. | Postprocessing of ensemble weather forecasts using a stochastic weather generator | |
| Yang et al. | Effect of data assimilation using WRF‐3DVAR for heavy rain prediction on the northeastern edge of the Tibetan Plateau | |
| Moazami et al. | Spatiotemporal bias adjustment of IMERG satellite precipitation data across Canada | |
| Sun et al. | How do GPM and TRMM precipitation products perform in alpine regions? A case study in northwestern China’s Qilian Mountains | |
| CN113435119B (en) | Global ocean surface brightness temperature determination method and system | |
| CN111428421A (en) | Rainfall runoff simulation method for deep learning guided by physical mechanism | |
| Han et al. | A morphing-based technique for the verification of precipitation forecasts | |
| Chang et al. | Neural network with multi-trend simulating transfer function for forecasting typhoon wave | |
| Wongchuig et al. | Toward Discharge Estimation for Water Resources Management with a Semidistributed Model and Local Ensemble Kalman Filter Data Assimilation | |
| Mo et al. | Study on the growth of wind wave frequency spectra generated by cold waves in the northern East China Sea | |
| De Bruijn et al. | Rainfall prediction for a flooding event in Ireland caused by the remnants of Hurricane Charley | |
| CN111309778B (en) | Climate mode evaluation and deviation correction method based on information entropy | |
| Shen et al. | Development and preliminary validation of a land surface image assimilation system based on the common land 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 |