NL2033915A - Monthly precipitation forecasting model with the temporal convolutional network - Google Patents
Monthly precipitation forecasting model with the temporal convolutional network Download PDFInfo
- Publication number
- NL2033915A NL2033915A NL2033915A NL2033915A NL2033915A NL 2033915 A NL2033915 A NL 2033915A NL 2033915 A NL2033915 A NL 2033915A NL 2033915 A NL2033915 A NL 2033915A NL 2033915 A NL2033915 A NL 2033915A
- Authority
- NL
- Netherlands
- Prior art keywords
- precipitation
- monthly precipitation
- monthly
- components
- convolutional network
- Prior art date
Links
- 238000001556 precipitation Methods 0.000 title claims abstract description 83
- 230000002123 temporal effect Effects 0.000 title claims abstract description 33
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000010187 selection method Methods 0.000 claims abstract description 7
- 238000000034 method Methods 0.000 claims description 16
- 238000001914 filtration Methods 0.000 claims description 11
- 230000001364 causal effect Effects 0.000 claims description 4
- 230000010355 oscillation Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 230000000306 recurrent effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000007637 random forest analysis Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 102000015097 RNA Splicing Factors Human genes 0.000 description 1
- 108010039259 RNA Splicing Factors Proteins 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- SESFRYSPDFLNCH-UHFFFAOYSA-N benzyl benzoate Chemical compound C=1C=CC=CC=1C(=O)OCC1=CC=CC=C1 SESFRYSPDFLNCH-UHFFFAOYSA-N 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000000116 mitigating effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- 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
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Environmental & Geological Engineering (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Medical Informatics (AREA)
- Hydrology & Water Resources (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Disclosed is a monthly precipitation forecasting model with the temporal convolutional 5 network, which comprises the following steps: collecting and sorting out monthly precipitation and candidate predictive factors, and dividing the monthly precipitation and candidate predictive factors into a training period and a prediction period respectively; decomposing the data of the training period and the prediction period to respectively obtain components with different decomposition scales; selecting important predictive factors for precipitation components based 10 on the importance of the components of candidate predictive factors to precipitation components; building a monthly precipitation component prediction model based on the monthly precipitation components and its important predictive factors; predicting the monthly precipitation components based on the monthly precipitation component prediction model, and adding the predicted values to obtain the final monthly precipitation predicted value. In this application, through discrete 15 wavelet decomposition and Boruta feature selection method, the useful information hidden in the data at different time scales is effectively extracted and the important factors affecting monthly precipitation are identified, thereby improving the prediction accuracy of the model.
Description
MONTHLY PRECIPITATION FORECASTING MODEL WITH THE TEMPORAL
CONVOLUTIONAL NETWORK
The application belongs to the technical fields of computer and meteorological prediction, and in particular to a monthly precipitation forecasting model with the temporal convolutional network.
Precipitation is one of the indicators that can directly describe climate change, and it is also an indispensable natural resource in human life. The uneven spatial and temporal distribution of precipitation often leads to the occurrence of natural disasters such as drought, flood and so on, which has a serious impact and damage on human life, economic development and ecological environment. Therefore, accurate precipitation prediction can provide useful information for water resources management, disaster prevention and mitigation, etc. With the development of artificial intelligence technology, the deep learning method is increasingly used in the field of monthly precipitation prediction. Although in the existing cyclic neural network architecture, such as long short-term memory network and gated recurrent neural network can alleviate the problem of gradient explosion, but the long short-term memory network and gated recurrent neural network have complex gating mechanism and high memory requirements. Compared with recurrent neural network, convolutional neural network, especially temporal convolutional network, has lower memory requirement, simpler structure and more stable training scheme.
Influenced by many factors, precipitation is nonlinear and non-stationary. The existing monthly precipitation prediction methods usually directly predict nonlinear and non-stationary series, which will lead to low prediction accuracy. Introducing the signal processing method such as discrete wavelet decomposition into the prediction model can effectively decompose the monthly precipitation series into different time scales, reduce the non-stationary characteristics of the time series data, and extract the effective information hidden in the data, thus improving the prediction accuracy. Using Boruta feature selection method can select important predictive factors of monthly precipitation from a large number of candidate predictive factors, and improve the prediction performance of the model.
The application provides a monthly precipitation forecasting model with the temporal convolutional network. The monthly precipitation prediction model is constructed by coupling temporal convolutional network, discrete wavelet decomposition and Boruta feature selection method, and the monthly precipitation predicted value can be obtained.
In order to achieve the above purpose, this application provides the following solutions:
The monthly precipitation forecasting model with the temporal convolutional network, includes the following steps: collecting and sorting out monthly precipitation and candidate predictive factors, and dividing the monthly precipitation and candidate predictive factors into a training period and a prediction period respectively; decomposing the data of the training period and the prediction period to respectively obtain components with different decomposition scales; selecting important predictive factors for precipitation components based on the importance of the components of candidate predictive factors to precipitation components; building a monthly precipitation component prediction model based on the monthly precipitation components and its important predictive factors; predicting the monthly precipitation components based on the monthly precipitation component prediction model, and adding the predicted values to obtain the final monthly precipitation predicted value.
Preferably, the candidate predictive factors include: historical monthly precipitation and climate index.
Preferably, the method of decomposing the data of the training period and the prediction period includes: the discrete wavelet decomposition.
Preferably, the discrete wavelet decomposition method includes: analysing signals with different frequencies with filters with different frequencies, including the high-pass filter and the low-pass filter.
Preferably, the high-pass filter comprises: filtering out the high-frequency part of the input signal and outputting the low-frequency part to obtain approximate components.
Preferably, the low-pass filter comprises: filtering out the low-frequency part and outputting the high-frequency part to obtain precise components.
Preferably, the method for obtaining the important predictive factors includes: adopting
Boruta feature selection method.
Preferably, the method for constructing the monthly precipitation component prediction model includes: constructing a temporal convolutional network.
Preferably, the temporal convolutional network includes causal convolution, dilated convolution and residual connection.
The application has the beneficial effects that:
The application discloses a monthly precipitation forecasting model with the temporal convolutional network. The application method firstly pre-processes the model input, decomposes the model input into precise components and approximate components with different time scales by using discrete wavelet transform, and effectively extracts useful information hidden in the original monthly precipitation data, thereby improving the accuracy of the prediction model. Boruta feature selection method is used to select important predictive factors from many candidate predictive factors, which effectively reduces the model input and improves the prediction performance. In addition, the temporal convolutional network technology is used to train and forecast the decomposed monthly precipitation, so that the implementer can effectively control the size of the field of view of the temporal convolutional network model, avoiding the gradient explosion or gradient disappearance in the conventional recurrent neural network, and making the model more robust.
In order to explain the technical scheme of this application more clearly, the following is a brief introduction of the drawings needed in the embodiments. Obviously, the drawings in the following description are only some of the embodiments of this application. For those of ordinary skill in this field, other drawings can be obtained according to these drawings without any creative labour.
FIG. 1 is a schematic diagram of a monthly precipitation forecasting model with the temporal convolutional network in the embodiment of the present application;
FIG. 2 is a schematic diagram of discrete wavelet decomposition in the embodiment of the present application.
The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the drawings in the embodiments of this application.
Obviously, the described embodiments are only part of the embodiments of this application, but not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative labour belong to the scope of protection in this application.
In order to make the above objects, features and advantages of this application more obvious and understandable, the application will be further explained in detail below with reference to the drawings and detailed description.
FIG. 1 is a schematic diagram of a monthly precipitation forecasting model with the temporal convolutional network in the embodiment of the present application;
Step 1, collecting and sorting out monthly precipitation and candidate predictive factors, and dividing the monthly precipitation and candidate predictive factors into a training period and a prediction period respectively; as shown in Table 1, the candidate predictive factors include: historical monthly precipitation and climate index;
Table 1
Category Climate index
Atlantic Multidecadal Oscillation (AMO); SSTs over Bay of Bengal (BB),
East China Sea (ECS), Kuroshio Current (KC) and South China Sea (SCS); Indian Ocean Dipole Mode Index (DMI); Bivariate ENSO
Timeseries (BEST); Nifio1+2 SST index (NINO1+2), Nifio3 SST index
Sea surface (NINOS), Nifio3.4 SST index (NINO3.4) and Nifio4 SST index (NINO4); temperature Co _ oo / (SST) Oceanic Nido Index (ONI); Pacific Decadal Oscillation (PDO); Tropical
Northern Atlantic Index (TNA); Trans-Nifio Index (TNI); Tropical
Southern Atlantic Index (TSA); Western Hemisphere Warm Pool (WHWP); SSTs at Western Pacific Warm Pool (WPWP) Atlantic
Meridional Mode (AMM); Pacific Meridional Mode (PMM)
Arctic Oscillation (AQ); North Atlantic Oscillation (NAO); Eastern
Atmospheric Atlantic/Western Russia (EAWR); North Pacific Pattern (NP); Pacific circulation North American Index (PNA); Quasi-Biennial Oscillation (QBO);
Southern Oscillation Index (SOI); Western Pacific Index (WP)
Other Solar Radiation Flux (SFX); Global Mean Temperature Anomaly (GlobalT); Bivariate ENSO index (BEST); Mixed ENSO index (MEI)
Step 2, decomposing the data of the training period and the prediction period to respectively obtain components with different decomposition scales, where the components include approximate components and precise components; the significance of discrete wavelet decomposition is that it can decompose signals on different scales, and different scales can be determined according to different goals. Analysing signals with different frequencies with filters with different frequencies, including the high-pass filter and the low-pass filter. The high-pass filter comprises: filtering out the high-frequency part of the input signal and outputting the low-frequency part to obtain approximate components.
The low-pass filter comprises: filtering out the low-frequency part and outputting the high- frequency part to obtain precise components.
Firstly, the original signal is divided into two parts: "low frequency approximation" and "high frequency detail"; Using the same operation, the last "low frequency approximation” part is subdivided into low frequency approximation and high frequency detail parts each time, and subdivides them successively (at most, each part is decomposed to only 1 point}. The high- frequency details separated each time are not decomposed. Therefore, the low-frequency approximate part decomposed each time is equivalent to "low-pass filtering" for this signal, and the high-frequency detailed part decomposed each time is equivalent to "high-pass filtering” for this signal. Therefore, each discrete wavelet decomposition is to use one low-pass filter and one high-pass filter to perform one low-pass filtering and one high-pass filtering for this signal. The key of wavelet decomposition lies in two (a group of) filters. Therefore, the signals in prediction period and training period are generated by two complementary filters, and the approximate details of the original signal can be obtained by calculating discrete wavelet coefficients, which is decomposition, and the inverse process. is reconstruction. Firstly, the precipitation signals in 5 different areas to be processed are sampling discretely to obtain x(z), then the wavelet decomposition and reconstruction of the signals can be realized by subband filtering, and the decomposition and reconstruction structure is shown in FIG. 2. In the figure, Fo(z) and F(z) are the filter coefficients corresponding to the low-pass filter and the high-pass filter, respectively,
Ho(z) and H4(z) are the filter coefficients corresponding to the mirror filters of the low-pass filter and the high-pass filter, respectively, satisfying Ho(z) = Fs {-z) and H(z) = F:(-z).
The signal decomposition process is as follows: on the one hand, the signal x(z) is "downsampled" (}2) after passing through the low-pass filter, and the average signal c(z) whose scale and resolution are halved, that is, the low frequency component; on the other hand, the signal x(z) is "downsampled" (| 2) after passing through the high-pass filter, and the detailed signal d(z) whose scale and resolution are halved, that is, the high frequency component.
Using discrete wavelet decomposition to decompose the monthly precipitation and candidate predictive factors during the training period, and obtaining high-frequency components, that is, detailed information (D+, Ds, ..., Dy) and low-frequency components that is, approximate information (Ay) with different time scales.
Step 3, selecting important predictive factors for precipitation components based on the importance of the components of candidate predictive factors to precipitation components; for each decomposition scale, using Boruta feature selection method to select the important predictive factors of monthly precipitation component from the candidate predictive factor components; taking the monthly precipitation components and selected important predictive factors as the model inputs of the temporal convolutional network.
Boruta method is a wrapper based on random forest classification method. Firstly, creating a copy of each month's precipitation candidate predictive factor, splicing the feature copy and the original feature to form a new feature matrix; scrambling randomly added attributes to eliminate their relevance to the response; running a random forest classifier on the extended feature matrix, and collecting the calculated Z-score; finding the maximum Z-Scors({MZSA) between shadow attributes, and then marking each attribute with a score higher than MZSA as important; performing a two-sided test equal fo MZSA for each attribute with undetermined importance; regarding the attributes whose importance is significantly lower than MZSA as "unimporiant” and permanently delete them from the feature set; regarding the atiribute whose importance is significantly higher than MZ3A as “important”, deleting all shadow attributes; repeating this process until all attributes are assigned importance, or the method has reached the previously set number of random forest runs.
Step 4, building a monthly precipitation component prediction model based on the monthly precipitation components and its important predictive factors;
Substituting the monthly precipitation component and its corresponding scale important predictive factor component into the temporal convolutional network model for training, and obtaining the prediction model of each monthly precipitation component. There are two specific designs of temporal convolutional network: causal convolution and one-dimensional full convolution network (FCN) architecture. By using causal convolution, the output of the model is only affected by current and past inputs. The temporal convolutional network adopts 1D-FCN architecture, so that the length of the network output sequence is consistent with that of the network input sequence. In addition, dilated convolution is also introduced into the temporal convolutional network to increase the receptive field of the model. The residual connection is used to combine the previous input with the convolution result to ensure that the addition operation gets the same tensor.
Step 5, predicting the monthly precipitation components based on the monthly precipitation component prediction model, and adding the predicted values to obtain the final monthly precipitation predicted value.
Inputting the components of monthly precipitation important predictive factors after treated by discrete wavelet decomposition and Boruta feature selection into the monthly precipitation prediction model based on temporal convolutional network with corresponding components, and obtaining the predicted values of each monthly precipitation component, and then accumulating the predicted values of each component to obtain the final monthly precipitation predicted value.
The above-mentioned embodiments only describe the preferred mode of this application, but do not limit the scope of this application. On the premise of not departing from the design spirit of this application, all kinds of modifications and improvements made by ordinary technicians in this field to the technical scheme of this application should fall within the scope of protection determined by the claims of this application.
Claims (9)
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211476547.XA CN115936196A (en) | 2022-11-23 | 2022-11-23 | Monthly rainfall model prediction method based on time sequence convolution network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| NL2033915A true NL2033915A (en) | 2024-05-30 |
| NL2033915B1 NL2033915B1 (en) | 2024-06-19 |
Family
ID=86650036
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| NL2033915A NL2033915B1 (en) | 2022-11-23 | 2023-01-06 | Monthly precipitation forecasting model with the temporal convolutional network |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN115936196A (en) |
| NL (1) | NL2033915B1 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449460B (en) * | 2023-04-23 | 2023-10-10 | 水利部交通运输部国家能源局南京水利科学研究院 | Regional month precipitation prediction method and system based on convolution UNet and transfer learning |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| ES2409481A2 (en) * | 2011-03-31 | 2013-06-26 | Universidad De Alcalá | Device and method of local frost and fog prediction based on neuronal computation with feature selection (Machine-translation by Google Translate, not legally binding) |
-
2022
- 2022-11-23 CN CN202211476547.XA patent/CN115936196A/en active Pending
-
2023
- 2023-01-06 NL NL2033915A patent/NL2033915B1/en active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| ES2409481A2 (en) * | 2011-03-31 | 2013-06-26 | Universidad De Alcalá | Device and method of local frost and fog prediction based on neuronal computation with feature selection (Machine-translation by Google Translate, not legally binding) |
Non-Patent Citations (1)
| Title |
|---|
| MING WEI: "Monthly rainfall forecasting by a hybrid neural network of discrete wavelet transformation and deep learning", WATER RESOURCES MANAGEMENT, vol. 36, no. 11, 22 July 2022 (2022-07-22), Dordrecht, pages 4003 - 4018, XP093158121, ISSN: 0920-4741, Retrieved from the Internet <URL:https://link.springer.com/content/pdf/10.1007/s11269-022-03218-w.pdf> [retrieved on 20240501], DOI: 10.1007/s11269-022-03218-w * |
Also Published As
| Publication number | Publication date |
|---|---|
| NL2033915B1 (en) | 2024-06-19 |
| CN115936196A (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhuo et al. | Physics-augmented deep learning to improve tropical cyclone intensity and size estimation from satellite imagery | |
| Li et al. | IncepTCN: A new deep temporal convolutional network combined with dictionary learning for strong cultural noise elimination of controlled-source electromagnetic data | |
| Li et al. | An analysis of extreme flood events during the past 400 years at Taihu Lake, China | |
| Li et al. | Long‐term streamflow forecasting using artificial neural network based on preprocessing technique | |
| NL2033915B1 (en) | Monthly precipitation forecasting model with the temporal convolutional network | |
| Chung et al. | Fuzzy inference system for modeling the environmental risk map of air pollutants in Taiwan | |
| CN111833224A (en) | Urban main and auxiliary center boundary identification method based on population grid data | |
| CN112001435A (en) | Method and system for constructing training sample set in regional landslide early warning and storage medium | |
| CN118036902A (en) | Knowledge graph-based ocean typical scene evaluation index system construction method and device, electronic equipment and storage medium | |
| Hao et al. | A prediction model of significant wave height in the south china sea based on attention mechanism | |
| CN118334347A (en) | SAR remote sensing image-based water body segmentation method and system | |
| CN114819260A (en) | Dynamic generation method of hydrologic time series prediction model | |
| Miao et al. | Monthly extended ocean predictions based on a convolutional neural network via the transfer learning method | |
| Wang et al. | A New Remote Sensing Change Detection Data Augmentation Method based on Mosaic Simulation and Haze Image Simulation | |
| Gerges et al. | Deep learning-based downscaling of temperatures for monitoring local climate change using global climate simulation data | |
| Jia et al. | Background noise suppression using trainable nonlinear reaction diffusion assisted by robust principal component analysis | |
| CN114943189B (en) | XGboost-based acoustic velocity profile inversion method and system | |
| Collins et al. | Markov models in geography | |
| Liu et al. | Application of convolutional neural network to GIS and physics | |
| Mu et al. | Prediction of North Atlantic Oscillation Index Associated with the Sea Level Pressure Using DWT‐LSTM and DWT‐ConvLSTM Networks | |
| CN112034512B (en) | Seismic data discontinuity detection method and system based on deep learning model | |
| Yan et al. | Association mining of coastline change and land use patterns to enhance conservation | |
| Qiu et al. | A short-term precipitation prediction model based on spatiotemporal convolution network and ensemble empirical mode decomposition | |
| Athanasiadis et al. | Linear contributions of different time scales to teleconnectivity | |
| Liu et al. | Prediction of tropical cyclone frequency with a wavelet neural network model incorporating natural orthogonal expansion and combined weights |