CN110516856A - The method of estimation Marine GIS temperature based on convolutional neural networks - Google Patents
The method of estimation Marine GIS temperature based on convolutional neural networks Download PDFInfo
- Publication number
- CN110516856A CN110516856A CN201910738699.4A CN201910738699A CN110516856A CN 110516856 A CN110516856 A CN 110516856A CN 201910738699 A CN201910738699 A CN 201910738699A CN 110516856 A CN110516856 A CN 110516856A
- Authority
- CN
- China
- Prior art keywords
- data
- temperature
- ocean
- estimation
- neural networks
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Development Economics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Geophysics And Detection Of Objects (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to ocean science technical field, in particular to the method for a kind of estimation Marine GIS temperature based on convolutional neural networks.In the method for estimation Marine GIS temperature provided by the invention, by obtaining the observation data about sea surface feature, and then extract the highest data point of the surface layer degree of association in observation data, and convolutional calculation is carried out with this, finally obtain the Temperature estimate value of ocean subsurface, the acquisition efficiency of sea surface Temperature estimate value is improved, and then improves the accuracy of Marine GIS Temperature estimate value.
Description
Technical field
The present invention relates to ocean science technical field, in particular to a kind of estimation ocean based on convolutional neural networks time table
The method of layer temperature.
Background technique
In global climate system, ocean plays unrivaled effect in terms of storage water and heat.Accurately detection and
The underground heat structure of description global ocean is the important research direction of ocean dynamics.
In the underground heat arrangement works for detecting and describing global ocean, research and accurately estimation Marine GIS temperature
Degree, for understanding that the mechanism and process of entire ocean and entire Earth climate system are vital, while also can be right
The building of mid-deep strata oceanographic data collection and ocean increase warm analysis and provide data support.
Summary of the invention
The method for the estimation Marine GIS temperature based on convolutional neural networks that the purpose of the present invention is to provide a kind of, institute
The method of stating includes the following steps:
S1: the observation data about ocean surface superficial feature are obtained；
S2: the highest data point of the surface layer degree of association and characterization centre data point in the observation data are obtained；
S3: carrying out convolution budget to the data point, obtains two numbers that are connected of the temperature of characterization Marine GIS
The estimated value of central point between strong point.
Further, step S1 includes:
Obtain the Satellite Observations comprising extra large apparent height, sea-surface temperature and sea surface salinity.
Further, the observation data are the moonscope numbers being observed by remote sensing satellite to sea surface
According to.
In the method for the estimation Marine GIS temperature provided by the invention based on convolutional neural networks, closed by obtaining
The observation data of Yu Haiyang's superficial feature, and then extract the highest data point of the surface layer degree of association and centre data in observation data
Point, and convolutional calculation is carried out with this, it finally obtains the Temperature estimate value of ocean subsurface, improves sea surface Temperature estimate value
Acquisition efficiency, and then improve the accuracy of Marine GIS Temperature estimate value.
Detailed description of the invention
Fig. 1 is method and step flow chart provided in an embodiment of the present invention；
Fig. 2 is CNN network structure provided in an embodiment of the present invention；
Fig. 3 a and Fig. 3 b is that the inshore point provided in an embodiment of the present invention randomly selected respectively and off-lying sea bank point are Bu Tong deep
Spend the comparison diagram of lower measured value and predicted value；
Fig. 4 a and Fig. 4 b are EI Nino regional analysis figures provided in an embodiment of the present invention
Fig. 5 a and Fig. 5 b are EI Nino regional analysis figures provided in an embodiment of the present invention
Fig. 6 is model prediction ability tendency chart provided in an embodiment of the present invention.
Specific embodiment
By above description content it is found that in global climate system, ocean is played in terms of storage water and heat without comparable
Quasi- effect.Accurately detecting and describe the underground heat structure of global ocean is the important research direction of ocean dynamics.
In the underground heat arrangement works for detecting and describing global ocean, research and accurately estimation Marine GIS temperature
Degree, for understanding that the mechanism and process of entire ocean and entire Earth climate system are vital, while also can be right
The building of mid-deep strata oceanographic data collection and ocean increase warm analysis and provide data support.
However, the estimation for being used for ocean interior dynamic environment information of sea surface remote sensing observations data more frequently is ground
Study carefully.
According to studies have shown that THE EQUATORIAL PACIFIC thermocline will considerable time in keep disorder ixi state,
This causes the increasing of Eastern Equatorial Pacific is warm to weaken greater than equator exterior domain and Hai Wen gradient, this exactly leads to extreme EI Nino
One of the multiple main mechanism of event Intergovernmental Panel on Climate Change (Intergovernmental Panel on
Climate Change, IPCC) it reports, global mean sea level temperature (Sea Surface Temperature, SST) will be every
Increase about 0.2 DEG C within 10 years.It is thus determined that sea surface temperature SST and subsurface temperature (Subsurface Temperature
Anomalies, STA) to research EI Nino phenomenon and its how to respond greenhouse climate and warm be in climate science it is most important
One of problem.
Satellite remote sensing provides many useful ocean surface observations on different room and time scales, but this is only limited
The surface layer of Yu Haiyang.Since many important ocean processes and feature are located at sea or less and quite deep place, Er Qiexian
Some data cannot completely and accurately describe the internal structure and changing rule of ocean, therefore for constructing complete three
Dimension temperature-salinity structure is necessary [5].With the continuous development of satellite remote sensing technology, especially satellite remote sensing sea surface temperature
Spend (Sea Surface Temperature Anomalies SSTA) and salinity (Sea Surface Salinity
Anomalies, SSSA) data increasingly increases, and it provides a large amount of wide coverages, precision and spatial resolution is higher, the time connects
The continuous stronger extra large surface real time information of property.The data prediction Marine GIS information how to be obtained using satellite remote sensing, establishes one
Complete secondary surface layer three dimensional analysis forecasting system is covered, this is urgent need to solve the problem in international ocean research field.
In the prior art using artificial neural network (Artificial Neural Network, ANN) by using me
Sea surface temperature (SST), sea level height (SSH), wind-stress, net radiation flux and the Net heat flux data estimation ocean of primary sea-anchor system
Internal temperature structure.On average, 50% estimated value is in ± 0.5 DEG C of error range, and 90% error is in ± 1.0 DEG C of models
In enclosing.For the temperature field of 200 meters of depth in the North Atlantic Ocean, separately there is the prior art to pass through multiple linear regression combination oceanographic observation
System data set Argo and remotely-sensed data estimate ocean three-dimensional temperature field, and use Objective Analysis Method, are combining two kinds of data
When type, the mean square deviation of the mapping error of the large scale and low frequency temperature field of 200 meters of depth is reduced.
There are also propose to estimate mesoscale three-dimensional ocean heat structure in a kind of near real-time Altimetry Data in the prior art
New empirical method.This method uses the bilayer model of one group of new empirical parameter with layering.Scientific research personnel's use comes from
The data of the sea surface temperature (SST) of the monthly abnormal data set of Argo grid, height (SSH) and salinity (SSS) utilize self-organizing mind
Abnormal through network estimation North Atlantic Ocean subsurface temperature, this method has superperformance, related coefficient in 30 to 700 meters of depth
Greater than 0.8.
There are also scientific research personnel by combining numerical value and artificial neural network technology to predict ocean temperature, wherein daily
The error statistics of prediction are correlation coefficient r=0.37, root-mean-square error Root Mean Square Error RMSE=0.47 DEG C
With mean absolute error Mean Absolute Error MAE=0.38 DEG C, the error statistics of every weekly forecasting are r=0.27,
RMSE=0.78 DEG C and MAE=0.64 DEG C, the error statistics monthly predicted are r=0.11, RMSE=0.58 DEG C and MAE=0.46
℃[12].A kind of new satellite-based Geographical Weighted Regression (Geographically developed by related fields scientific research personnel
Weighted regression, GWR) model, it is used for inverting Indian Ocean subsurface temperature structure, final experimental result to be RMSE model
Enclosing is about 0.10 to 0.18, and R2 range is about 0.50 to 0.80.
Scientific research personnel estimates the abnormal (Subsurface of the subsurface temperature of the Indian Ocean by a series of satellite remote sensings
Temperature Anomalies, STA), propose a kind of support vector machines (Support Vector Machines, SVM)
Method, the performance of support vector regression (Support Vector Regression, SVR) is unstable on 400 meters of top,
With R2Reduction and mean square error (Mean Squared Error, MSE) increase, estimated accuracy is with depth 500m depth
Increase and be gradually reduced.
The study found that the mixed layer that the large-scale SST annual period of THE EQUATORIAL PACIFIC east is largely changed every year
It is deep-controlled, and the competitive effect that layer depth depends primarily on solar radiation and wind-force forces.Spring the torrid zone and temperate zone
Pacific temperature is extremely unstable.This unstability for the predictability obstacle of EI Nino event presence to Guan Chong
It wants.
Under the influence of global warming, great change may occur for the average climate of the Pacific region.Previous researcher
The influence of the data prediction model climate established as unit of year and seasonal variations.These data prediction models need to usually subtract
Climate mean state is removed to eliminate the influence to training pattern, but STA predictive ability is insufficient as the result is shown.Originally research and propose foundation by
Month data model eliminates influence of the weather to model training.STA precision of prediction significantly improves as the result is shown.
Previous researcher establishes STA prediction model using only single-point feature, this does not account for the flowing of ocean interior seawater
Or influence its precision of prediction of STA as the result is shown that ocean current transmits seawater heat is very poor.
Based on above-mentioned analysis, it is seen that for being typically based on dynamics from the existing method of sea parameter search underground heat structure
Model or statistical model.Existing dynamic method seldom pays close attention to the application and advanced machine learning model of global scale, and
And a small amount of surface parameter is only used to derive underground dynamic field.Therefore, estimation method itself and its global dimensional accuracy are still aobvious
Very big room for improvement is shown.Existing statistical method is not enough sea surface data parameter application, lacks deep learning
The application of model, mode input characteristic parameter is single, and learning ability is poor, and establishing model climate based on annual data influences greatly.
Therefore model prediction accuracy never significantly improves.
Below in conjunction with the drawings and specific embodiments to the estimation ocean proposed by the present invention based on convolutional neural networks time table
The method of layer temperature is described in further detail.According to following explanation and claims, advantages and features of the invention will more
It is clear.It should be noted that attached drawing is all made of very simplified form and using non-accurate ratio, only to convenient, apparent
The purpose of the ground aid illustration embodiment of the present invention.
In the detailed description of the present embodiment using such as " ... under ", " ... below ", " below ", " above "
Etc. spatial terminologies, it is therefore an objective to be easy the positional relationship of a description component and another component shown in the drawings, but these are only
It is that embodiment is not intended to limit the present invention.In addition to orientation shown in figure, spatial relation term will include using or operating
In device a variety of different orientation.Device can be positioned in other ways, such as be rotated by 90 ° or in other orientation, and
It is explained accordingly by spatial relation description as used herein symbol.
The present embodiment proposes a kind of method of estimation Marine GIS temperature based on convolutional neural networks.This method packet
Include following steps:
S1: the observation data about ocean surface superficial feature are obtained；
S2: the highest data point of the surface layer degree of association and centre data point in the observation data are obtained；
S3: convolution budget is carried out to the data point, obtains the Temperature estimate value of Marine GIS.
Further, step S1 includes:
Obtain the Satellite Observations comprising extra large apparent height, sea-surface temperature and sea surface salinity.
Further, the observation data are the moonscope numbers being observed by remote sensing satellite to sea surface
According to.
Specifically, the present embodiment is used was used for model training and is commented using 2004 to 12 years 2015 data sets
Estimate, among these using 2005 to 2014 annual data collection for establishing model, 2004 and 2015 annual data collection are commented for model
Estimate.12 groups of CNN models month by month are established by 12 groups of different data sets are divided into month to 2014 annual data collection within 2005.Every group
Data sample is 140,000, wherein 78400 (56%) data are used for training data model, 19600 (14%) data are used for
Model is verified, 42000 (30%) data are used for test model.
Evaluation method in the present embodiment is standardized to sea surface feature first, it ensure that different special
The dimension and magnitude of sign have same numberical range.The data feature values of single sample are subtracted into all numbers of training sample when processing
According to the average value of same feature, then divided by its variance.All data are gathered near 0 for each feature in this way,
Variance is 1.Specific publicity calculates as follows:
X is training set or test set single sample characteristic value, μ are the average value of training sample data, σ is number of training
According to standard deviation, X be standardization after characteristic value.
Scheme disclosed in the present embodiment is according to the coordinate value of longitude and latitude by the temperature number on the central point of Argo time surface layer
According to the corresponding position for being associated with CMEMS data, eliminating the unmatched influence of CMEMS and Argo resolution ratio in this way, (CMEMS is differentiated
Rate is 0.25 ° × 0.25 °, and Argo resolution ratio is 1 ° × 1 °, is unified for 0.25 ° × 0.25 °).
It is put centered on the training characteristics value that this research is chosen and the data point of surrounding (amounts to 625 data points, characteristic value number
1875) amount is.
In addition, convolutional neural networks (Convolutional neural networks, CNN) have more complicated network
Structure has more powerful feature learning and feature representation ability compared with conventional machines learning method.Each neuron regards one as
A filter (filter), window (receptive field) sliding, filter (filter) count local data
It calculates.The output result of convolutional layer is done Nonlinear Mapping by Relu excitation layer.Pond layer is clipped among continuous convolutional layer, this is used for
The amount of compressed data and parameter reduces over-fitting.Usually full articulamentum is in convolutional neural networks tail portion, this is with traditional nerve net
The connection type of network neuron is the same.
CNN advantage is shared convolution kernel, high-efficient to processing high dimensional data, can extract some advanced spies automatically
Sign, reduces the time of Feature Engineering, this just can be improved the precision of prediction subsurface temperature.Therefore it is proposed that utilization
Surface layer degree of association the maximum data feature has practical feasibility meaning using convolutional neural networks CNN building model.
As Fig. 2 shows the CNN network structure of setting.The structure includes 5 layers of convolution algorithm, is most followed by 4 layers of full articulamentum.
Regard the highest data point of the surface layer degree of association and centre data point as a width two dimensional image.The image carries out local view
Wild convolution algorithm, 3 features (SSTA, SSHA, SSSA) of each data point are that a convolution unit carries out operation.It uses
RELU activation primitive and Adam majorized function.
1st layer of convolutional layer step-length stride is (3,1), i.e. mobile 3 steps of horizontal direction, mobile 1 step of vertical direction.Filtering
Device filter size is 3 × 2, i.e., sums after the feature corresponding element of 2 data points being multiplied with filter corresponding element, is calculated
Other backward regions in 1 piece of region are mobile specified stride (3,1), until the two-dimensional matrix 57 × 33 all covers.
After 5 layers of convolution algorithm, data dimension is laid into one-dimensional data, full articulamentum is input to and carries out 4 layers of neural network fortune again
It calculates, finally exports 57 layers of predicted value.
CNN model, which is established month by month, by extra large table multi-source remote sensing observation data (SSTA, SSHA, SSSA) estimates the Pacific Ocean time table
The process of layer temperature anomaly (STA) (by taking 100m depth as an example).
Firstly, building training dataset.Point and its highest data of the surface layer degree of association centered on the training characteristics value of selection
Point (amounts to 625 data points, 1875) characteristic value quantity is.Using Argo actual measurement STA as training label and test badge, institute
There is the equal standardization of data set.
Second, training CNN model constructs optimal CNN model.The model selects Relu as activation primitive, and Adam is excellent
Change function.We are by analyzing each MSE, R2The optimum combination of convolutional layer, pond layer, full articulamentum is determined with convergence rate.
We use training dataset (SSTAx, SSHAx, SSSAx) as CNN training input data, use the STA of Argo as
Training label.
Finally, we use CMEMS data set (SSTA, SSHA, SSSA) as the input parameter of CNN model, prediction time
The STA on surface layer.We assess CNN model in the precision of prediction of each layer position in secondary surface layer (57 layers) using the STA of Argo actual measurement.
The present embodiment gives the analysis to above-mentioned estimated temperature, as Fig. 3 a and Fig. 3 b show that this experiment randomly selects
Under point a (21.50 ° of N, 122.50 ° of E) and point b (14.50 ° of N, 160.50 ° of E) (in October, 2015) different depth measured value and
The comparison diagram of predicted value (MSE of a, b point is respectively 0.1980/0.0980).As the result is shown it is proposed that using surface layer be associated with
Spend the feature of the maximum data point predict the method for central point not only to the prediction effect that there is regular data point at off-lying sea bank, together
Sample also has good prediction effect to anomaly number strong point at inshore.This shows the method to inshore anomaly number strong point
There is generalization ability.
Fig. 4 a, Fig. 4 b, the content that Fig. 5 a, Fig. 5 b are shown is for EI Nino regional analysis.It is desirable that it can in future
Anomalous year information is further studied, by the data set for collecting more anomalous years.
The low temperature seawater of tropical Pacific shows obviously under different depth, and especially when 100m is to 300m depth, low temperature is existing
As obvious, this is related with the 3, La Nina Phenomenon on tropical Pacific in April upper layer.Pacific Ocean east torrid areas sea area is different deep
Degree STA is substantially less than other sea areas.As depth increases, ocean temperature generally gradually tends towards stability, and STA change intensity is more next
Smaller, special heterogeneity is gradually unobvious, this and ocean interior are related with surface layer power difference.
The present embodiment uses MSE and R2As model-evaluation index.We only choose marine climate and change apparent month
(Isosorbide-5-Nitrae, 7,10) is as analysis data.By Tables 1 and 2 and Fig. 4 a, Fig. 4 b, Fig. 5 a, Fig. 5 b can be obtained, 2004 and 2015 1,
4,7, October, (MSE peak is respectively 0.8408/1.1034/ to MSE highest when depth layer position is located at 30m to 200m
1.8640/2.8231/0.5961/0.6750/1.1856/0.4788), this may with the dynamic process of Pacific Ocean upper layer complexity and
Mixed layer is related with the disturbance of thermocline.
Table 1
1 2015 years different depths of table, the corresponding mean square deviation (MSE) of different month CNN models and the coefficient of determination (R2) big
Small comparison.
Table 2
Table 2 indicate be 2004 under different depth, different months the corresponding mean square deviation of CNN model (MSE) with certainly
Determine coefficient (R2)。
As shown in fig. 6, MSE tends towards stability after being shown in 300m depth, R2It gradually decreasing, model prediction ability reduces,
This may be since mid-deep strata seawater stratification state is relatively stable, the physical phenomenon of the deeper ocean interior of depth is more difficult to surface layer spy
It levies and predicts and satellite remote sensing is more embarrassed to detect to cause.The performance of different year month model prediction accuracy can be obtained from table not
It is same:
(1) under the same time, by taking 100m depth in 2015 as an example, the MSE that is averaged January is that 0.2821 (maximum/minimum value is
0.8408/0.0062), average R2For 0.966 (maximum/minimum value is 0.993/0.891).MSE (the R in January2) small (height) in
4,7, October, model prediction accuracy promoted, this and winter ocean temperature integrally tend towards stability state, STA change intensity
It is small related.10 month of July, Pacific Ocean feeling the current was violent since the variation of summer and autumn seawater internal temperature is obvious, and MSE increases
And R2Reduce, the MSE that is averaged July is 0.6207 (maximum/minimum value is 1.8640/0.0095), average R2For 0.942 (maximum/
Minimum value is 0.993/0.842), the MSE that is averaged October is 0.8027 (maximum/minimum value is 2.8232/0.0034), average R2
For 0.956 (maximum/minimum value is 0.983/0.939).The predictive ability of 10 month of July CNN model indicated above reduces, this
Mainly since the variation of summer and autumn seawater internal temperature is violent and the influence of ocean current.
(2) in the case of different year, different months different depth MSE in 2004 are totally low, and (MSE average value is within 2004
0.5117) 0.2859,2015 year MSE average value is the R in 20152It is overall to be higher than (R in 20042Average value is 0.969,2015
Year R2Average value is 0.958) 2015.1,4,7, October, model was relatively reliable to prediction result in 2004, and 2015 have
Declined, this is mainly fluctuated by climate change in recent years, and Pacific Ocean temperature anomaly phenomenon is obvious, and uncertain factor aggravation leads to model
Precision of prediction is remarkably decreased.
In conclusion in the method for estimation Marine GIS temperature provided by the invention, by obtaining about ocean table
The observation data of layer feature, and then the highest data point of the surface layer degree of association and centre data point in observation data are extracted, and with
This carries out convolutional calculation, finally obtains the Temperature estimate value of ocean subsurface, improves the acquisition of sea surface Temperature estimate value
Efficiency, and then improve the accuracy of Marine GIS Temperature estimate value.
Foregoing description is only the description to present pre-ferred embodiments, not to any restriction of the scope of the invention, this hair
Any change, the modification that the those of ordinary skill in bright field does according to the disclosure above content, belong to the protection of claims
Range.
Claims (3)
1. a kind of method of the estimation Marine GIS temperature based on convolutional neural networks, which is characterized in that the method includes
Following steps:
S1: the observation data about ocean surface superficial feature are obtained；
S2: the highest data point of the surface layer degree of association and characterization centre data point in the observation data are obtained；
S3: carrying out convolution budget to the data point, obtains two data points that are connected of the temperature of characterization Marine GIS
Between central point estimated value.
2. the method for the estimation Marine GIS temperature based on convolutional neural networks as described in claim 1, which is characterized in that
Step S1 includes:
Obtain the Satellite Observations comprising extra large apparent height, sea-surface temperature and sea surface salinity.
3. the method for the estimation Marine GIS temperature based on convolutional neural networks as claimed in claim 2, which is characterized in that
The observation data are the Satellite Observations being observed by remote sensing satellite to sea surface.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910738699.4A CN110516856B (en) | 2019-08-12 | 2019-08-12 | Method for estimating ocean subsurface temperature based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910738699.4A CN110516856B (en) | 2019-08-12 | 2019-08-12 | Method for estimating ocean subsurface temperature based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110516856A true CN110516856A (en) | 2019-11-29 |
| CN110516856B CN110516856B (en) | 2023-02-03 |
Family
ID=68625798
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910738699.4A Active CN110516856B (en) | 2019-08-12 | 2019-08-12 | Method for estimating ocean subsurface temperature based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110516856B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111144666A (en) * | 2020-01-02 | 2020-05-12 | 吉林大学 | Ocean thermocline prediction method based on deep space-time residual error network |
| CN113051824A (en) * | 2021-03-26 | 2021-06-29 | 福州大学 | Ocean internal temperature field reconstruction method based on profile observation data |
| CN113807562A (en) * | 2021-06-25 | 2021-12-17 | 中国海洋大学 | Prediction method of ocean surface temperature |
| CN113935249A (en) * | 2021-11-23 | 2022-01-14 | 中国海洋大学 | Upper-layer ocean thermal structure inversion method based on compression and excitation network |
| CN114139590A (en) * | 2021-05-27 | 2022-03-04 | 中国海洋大学 | Method for estimating ocean temperature |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090187369A1 (en) * | 2008-01-23 | 2009-07-23 | University Of Delaware | Estimation of subsurface thermal structure using sea surface height and sea surface temperature |
| CN109543768A (en) * | 2018-11-30 | 2019-03-29 | 福州大学 | Ocean interior thermohaline information intelligent extracting method based on multi-source satellite remote sensing |
| CN109543356A (en) * | 2019-01-07 | 2019-03-29 | 福州大学 | Consider the ocean interior temperature-salinity structure remote sensing inversion method of Space atmosphere |
-
2019
- 2019-08-12 CN CN201910738699.4A patent/CN110516856B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090187369A1 (en) * | 2008-01-23 | 2009-07-23 | University Of Delaware | Estimation of subsurface thermal structure using sea surface height and sea surface temperature |
| CN109543768A (en) * | 2018-11-30 | 2019-03-29 | 福州大学 | Ocean interior thermohaline information intelligent extracting method based on multi-source satellite remote sensing |
| CN109543356A (en) * | 2019-01-07 | 2019-03-29 | 福州大学 | Consider the ocean interior temperature-salinity structure remote sensing inversion method of Space atmosphere |
Non-Patent Citations (1)
| Title |
|---|
| 孙春健等: "卫星遥感重构海洋次表层研究进展", 《海洋信息》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111144666A (en) * | 2020-01-02 | 2020-05-12 | 吉林大学 | Ocean thermocline prediction method based on deep space-time residual error network |
| CN113051824A (en) * | 2021-03-26 | 2021-06-29 | 福州大学 | Ocean internal temperature field reconstruction method based on profile observation data |
| CN113051824B (en) * | 2021-03-26 | 2022-06-14 | 福州大学 | Ocean internal temperature field reconstruction method based on profile observation data |
| CN114139590A (en) * | 2021-05-27 | 2022-03-04 | 中国海洋大学 | Method for estimating ocean temperature |
| CN113807562A (en) * | 2021-06-25 | 2021-12-17 | 中国海洋大学 | Prediction method of ocean surface temperature |
| CN113935249A (en) * | 2021-11-23 | 2022-01-14 | 中国海洋大学 | Upper-layer ocean thermal structure inversion method based on compression and excitation network |
| CN113935249B (en) * | 2021-11-23 | 2022-12-27 | 中国海洋大学 | Upper-layer ocean thermal structure inversion method based on compression and excitation network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110516856B (en) | 2023-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516856A (en) | The method of estimation Marine GIS temperature based on convolutional neural networks | |
| Lionello | The climate of the Mediterranean region: From the past to the future | |
| Han et al. | A convolutional neural network using surface data to predict subsurface temperatures in the Pacific Ocean | |
| Mulligan | WaterWorld: a self-parameterising, physically based model for application in data-poor but problem-rich environments globally | |
| McNicol et al. | Effects of seasonality, transport pathway, and spatial structure on greenhouse gas fluxes in a restored wetland | |
| Wu et al. | A localized contour tree method for deriving geometric and topological properties of complex surface depressions based on high-resolution topographical data | |
| Yu et al. | Investigating impacts of urban morphology on spatio-temporal variations of solar radiation with airborne LIDAR data and a solar flux model: a case study of downtown Houston | |
| CN102135531A (en) | Method for forecasting blue-green algae water bloom in large-scale shallow lake within 72 hours | |
| Tian et al. | Studies on the relationships between land surface temperature and environmental factors in an inland river catchment based on geographically weighted regression and MODIS data | |
| Liu et al. | Topographic mapping of offshore sandbank tidal flats using the waterline detection method: A case study on the Dongsha Sandbank of Jiangsu Radial Tidal Sand Ridges, China | |
| Zhao et al. | Variability of surface heat fluxes and its driving forces at different time scales over a large ephemeral lake in China | |
| CN113821925B (en) | Wetland dynamic boundary determination method based on three elements of aquatic soil | |
| Gangrade et al. | Sensitivity of probable maximum flood in a changing environment | |
| Jiang et al. | Estimation of land production and its response to cultivated land conversion in North China Plain | |
| Wahyudi et al. | Combining Landsat and landscape metrics to analyse large-scale urban land cover change: A case study in the Jakarta Metropolitan Area | |
| Stachelek et al. | Application of inverse path distance weighting for high-density spatial mapping of coastal water quality patterns | |
| Oyedotun | Quantitative assessment of the drainage morphometric characteristics of Chaohu Lake Basin from SRTM DEM Data: a GIS-based approach | |
| Guo et al. | How the variations of terrain factors affect the optimal interpolation methods for multiple types of climatic elements? | |
| Hamylton et al. | Geomorphological modelling of tropical marine landscapes: Optical remote sensing, patches and spatial statistics | |
| Chaudhuri et al. | Response of the Gulf Stream transport to characteristic high and low phases of the North Atlantic Oscillation | |
| Huang et al. | Spatio-temporal variation of landscape heterogeneity under influence of human activities in Xiamen City of China in recent decade | |
| Krug et al. | Delineation of ocean surface provinces over a complex marine domain (off SW Iberia): An objective abiotic-based approach | |
| Liu et al. | Analysis of coastline changes and the socio-economic driving mechanisms in Shenzhen, China | |
| Zhang et al. | An improved water budget for the El Yunque National Forest, Puerto Rico, as determined by the water supply stress index model | |
| Dyakonov et al. | Reproduction of interannual variability of the Caspian Sea level in a high-resolution hydrodynamic 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 |