Sub-season heat wave prediction method and storage medium

Abstract

The invention provides a sub-season heat wave prediction method, which comprises the following steps of: determining all optimal prediction factors associated with any air temperature main mode time coefficient under preset prediction time efficiency according to a pre-obtained air temperature main mode of a target area and an air temperature main mode time coefficient-prediction time efficiency-optimal prediction factor relation table; and acquiring data of all the optimal forecasting factors through real-time non-band-pass filtering, inputting a pre-trained sub-season air temperature forecasting model, and acquiring a predicted air temperature abnormal field in the target area under preset time efficiency. And predicting the sub-season heat wave event in the target area according to the predicted air temperature abnormal field. According to the method, the interaction between multiple scales and multiple variables is considered by using the strong data mining capability of the convolutional neural network, the influence of the interannual scale signal on the large-scale sub-seasonal variation rate is fully considered, and the potential correlation between the early-stage multi-scale signal and the future air temperature is mined.

Description

Technical Field

[0001] The invention relates to a sub-seasonal heat wave prediction method and a storage medium, belonging to the technical field of sub-seasonal climate prediction. Background Art

[0002] Extreme heat or heat wave events can cause considerable harm to human health, ecosystems and social services. The Sixth Assessment Report of the Intergovernmental Panel on Climate Change pointed out that climate change has intensified extreme heat waves around the world. Improving the prediction capabilities of extreme weather events and extending the prediction time beyond the conventional weather scale (>10 days) is essential for effectively reducing risks and making emergency preparations. In view of the difficulties of current operational numerical dynamical models in accurately predicting extreme heat events on the sub-seasonal scale, the World Meteorological Organization launched the Sub-Seasonal-Seasonal S2S Prediction Project to improve the predictability of such high-impact events.

[0003] The tropical intraseasonal oscillation (MJO) and the extratropical intraseasonal variability (ETS) are the main modes of atmospheric subseasonal variability. These modes affect China's temperature changes by causing abnormal circulation and thermal conditions, and are therefore considered to be the key source of the predictability of subseasonal forecasts of high temperature heat waves in China. Previous researchers used large-scale intraseasonal signals as predictors of statistical models to establish subseasonal forecast models, which effectively improved the subseasonal forecast level of extreme weather and climate events in China. In addition, disturbances associated with the MJO and the extratropical intraseasonal variability are affected by background conditions related to the El Niño-Southern Oscillation (ENSO) and other interannual patterns of sea surface temperature. Therefore, incorporating interannual anomaly signals into statistical models can improve subseasonal forecast capabilities.

[0004] Although the sources of subseasonal predictability of extreme weather events are well understood, effectively using this information to improve forecasting capabilities is another challenge. With the use of deep learning technology, some studies have used deep learning models such as convolutional neural networks (CNN) and long short-term memory (LSTM) neural networks to correct systematic biases in MJO, precipitation, and temperature dynamics forecasts, and use precursor signals to predict future changes in regional precipitation and temperature. However, these models do not consider the impact of interannual-scale signals on large-scale subseasonal variability, and ignore the potential predictability information that may be brought about by multi-scale interaction processes. This is a key but unclear scientific issue in the field of subseasonal forecasting. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a sub-seasonal heat wave prediction method, which considers the interaction between multi-scale and multi-variable variables by utilizing the powerful data mining capability of convolutional neural networks, takes into account the modulation effect of interannual variability on atmospheric sub-seasonal variability, and predicts future temperature changes by mining the potential correlation between previous multi-scale signals and future temperatures, thereby performing sub-seasonal prediction of heat wave events.

[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention provides a sub-seasonal heat wave prediction method, comprising:

[0008] According to the pre-obtained temperature main mode of the target area and the temperature main mode time coefficient-forecast time effectiveness-optimal forecast factor component relationship table, all optimal forecast factors associated with any temperature main mode time coefficient under the preset forecast time effectiveness are determined to form an optimal forecast factor set;

[0009] Acquire the data of each optimal predictor in the optimal predictor set, and respectively extract the subseasonal component and the low-frequency background state LFBS component of each optimal predictor by real-time non-bandpass filtering;

[0010] Inputting the sub-seasonal component and the low-frequency background state LFBS component of each optimal forecast factor into a pre-trained sub-seasonal temperature forecast model to obtain the predicted value of each main mode time coefficient of the temperature in the target area under the time effect;

[0011] Obtaining the predicted temperature anomaly field under the time effect according to the predicted value of each temperature main mode time coefficient and the temperature main mode reconstruction;

[0012] Predicting sub-seasonal heat wave events in the target area based on the predicted temperature anomaly field under the time validity and the preset threshold standard;

[0013] The process of obtaining the predicted value of any main mode time coefficient of air temperature includes:

[0014] The sub-seasonal temperature forecast model performs feature extraction on the optimal forecast factor component corresponding to the main mode time coefficient of the temperature to obtain the deep features of the optimal forecast factor component;

[0015] The deep features of all the optimal prediction factor components corresponding to the main mode time coefficient of the temperature are fused to obtain the joint features containing multi-scale multi-variable interaction information;

[0016] The predicted value of the main modal time coefficient of the temperature is obtained by combining the joint features and the current month information.

[0017] Furthermore, the sub-seasonal temperature forecast model also includes: a univariate forecast model; the univariate forecast model is used to provide the temperature modal time coefficient-forecast timeliness-optimal forecast factor relationship table and a pre-trained convolution block for extracting deep features of each optimal forecast factor component during the training process of the sub-seasonal temperature forecast model.

[0018] Furthermore, the training process of the sub-seasonal temperature forecast model includes:

[0019] Determine a target area and determine candidate forecast factors influencing daily maximum temperature in the target area;

[0020] Obtain the daily maximum temperature data of the preset historical period in the target area and the data of the candidate forecasting factors, construct a data set and perform preprocessing, and divide the processed data into a training set, a validation set and a test set according to the year;

[0021] A non-bandpass filtering method is used to extract the sub-seasonal component and the low-frequency background state LFBS component of the candidate predictor;

[0022] The summer temperature data is obtained from the daily maximum temperature data in the training set, and the empirical orthogonal function EOF is used for analysis to obtain S main temperature modes and their corresponding time coefficients;

[0023] The annual data of each candidate forecast factor is obtained from the training set, and the subseasonal component and LFBS component of each candidate forecast factor are used as the input of the univariate forecast model respectively, and the time coefficient of S main temperature modes is used as the output of the univariate forecast model to train the univariate forecast model;

[0024] The sub-seasonal components or LFBS components of each candidate prediction factor in the validation set are input into the corresponding univariate prediction model, and the S main temperature modal time coefficients under different time effects are predicted respectively. By evaluating and comparing the prediction skills of the univariate prediction model, the a sub-seasonal components of the prediction factors and b LFBS components of the prediction factors with the best prediction performance for each main temperature modal time coefficient under each time effect are determined, and a relationship table of main temperature modal time coefficient-forecast time effect-optimal prediction factor component is formed; among which a and b are determined based on sensitivity tests;

[0025] Taking the data of the (a+b) optimal prediction factors corresponding to each main mode time coefficient of the temperature in the training set as input, taking each main mode time coefficient of the temperature as the output of the sub-seasonal temperature prediction model, combined with the pre-trained convolution blocks provided by the univariate prediction model corresponding to the (a+b) optimal prediction factors, training is performed;

[0026] Use the data in the validation set to evaluate the trained sub-seasonal temperature forecast model and optimize the model structure and parameters;

[0027] The independent forecast performance of the subseasonal temperature forecast model was tested using the data in the test set.

[0028] Furthermore, the univariate prediction model includes: 2 input layers, 1 convolutional block, 1 fully connected layer and 1 output layer;

[0029] For any candidate predictor:

[0030] The first input layer provides a corresponding characteristic map of the subseasonal component or LFBS component of the candidate predictor;

[0031] The convolution block is used to perform convolution and pooling operations on the feature map, so as to extract corresponding deep features;

[0032] The second input layer provides information for initializing the current month;

[0033] The fully connected layer integrates the deep features and the month information, and outputs the time coefficient of the specified temperature main mode under the future specified forecast time through the output layer;

[0034] The sub-seasonal temperature forecast model includes: (a+b+1) input layers, (a+b) convolution blocks, 2 two-dimensional convolution layers, 1 fully connected layer and 1 output layer;

[0035] Wherein, the (a+b) input layers are respectively arranged in (a+b) independent input channels; and are used to provide (a+b) optimal prediction factor corresponding input feature maps;

[0036] The (a+b) convolution blocks are respectively used to extract deep features corresponding to the (a+b) optimal prediction factors;

[0037] The two-dimensional convolution layer is used to perform further convolution operation on the result of splicing the deep features of the (a+b) optimal prediction factors in the channel dimension direction. The convolution result and the month information provided by the last input layer are integrated through the fully connected layer, and the time coefficient of the specified main mode of temperature under the future specified forecast time is output through the output layer.

[0038] Furthermore, the convolution block includes three two-dimensional convolutional layers and two maximum pooling layers arranged at intervals.

[0039] Furthermore, the extracting of the subseasonal component and the LFBS component of the candidate prediction factor by using a non-bandpass filtering method includes:

[0040] Subtract the daily climate state from the raw daily data of the candidate predictor to eliminate the climate annual cycle and obtain the daily anomaly; the daily climate state is the multi-year average of the same date from 1991 to 2020;

[0041] Subtract the average of the previous 45 days from the daily outliers to eliminate the signals over 90 days; get the outliers under 90 days;

[0042] The abnormal values ​​within 90 days are taken as the leading 5-day moving average to eliminate the high-frequency weather scale disturbance within 10 days, and the sub-seasonal component of the candidate forecast factor is obtained;

[0043] The LFBS component is then the 30-day average of the daily anomalies.

[0044] Furthermore, the filter size of the two-dimensional convolution layer is 3×3, the sliding step value is 1×1, and the activation functions of the two-dimensional convolution layer and the fully connected layer are rectified linear units with leakage parameters; the filter size of the maximum pooling layer is 2×2, and the sliding step value is 2×2;

[0045] Before the two-dimensional convolutional layer performs the convolution operation, it also includes: filling the boundary of the feature map input to the two-dimensional convolutional layer with a zero value so that the size of the feature map remains unchanged before and after the operation.

[0046] Furthermore, the candidate prediction factors include: 850 hPa temperature T850, 200 hPa zonal wind U200, 500 hPa zonal wind U500, 850 hPa zonal wind U850, 200 hPa meridional wind V200, 500 hPa meridional wind V500, 850 hPa meridional wind V850, 200 hPa geopotential height field H200, 500 hPa geopotential height field H500, 850 hPa geopotential height field H850, 700 hPa relative humidity RH700, precipitable water PW, sea level pressure SLP, surface soil moisture SSM and outward longwave radiation OLR.

[0047] Furthermore, during the training process of the sub-seasonal temperature forecast model, the Adam optimizer was used to optimize the training parameters, and the initial learning rate of the Adam optimizer was 1×10 −3 ; Mean square error is used as the loss function;

[0048] A dropout method and an early stopping method are used to prevent the model from overfitting; the dropout method includes: adding a dropout layer with a preset probability of discarding after each two-dimensional convolutional layer during the model training iteration;

[0049] The early stopping method includes: in response to no improvement in the skill of the validation set after T consecutive training iterations, stopping training to ensure effective use of the model weights corresponding to the minimum validation loss, where T is a preset value.

[0050] In a second aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the sub-seasonal heat wave prediction method as described in the first aspect is implemented.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] (1) The subseasonal temperature forecast model provided by the present invention utilizes the powerful data mining capability of convolutional neural networks to consider the interaction between multi-scale and multi-variable variables. The subseasonal component and LFBS component of the optimal forecast factor are extracted as model inputs by non-bandpass filtering. The LFBS component of the optimal forecast factor contains interannual scale signals, so that the model can not only learn the predictability information provided by large-scale subseasonal variability, but also consider the impact of interannual scale signals on large-scale subseasonal variability.

[0053] (2) During the training process of the sub-seasonal temperature forecast model provided by the present invention, the univariate forecast model is trained based on the sub-seasonal component or LFBS component of a single candidate forecast factor, the optimal forecast factor is determined according to the performance of many candidate forecast factors in the validation period, and the convolution block of the univariate forecast model corresponding to the optimal forecast factor is used to initialize the training sub-seasonal temperature forecast model to improve its sub-seasonal forecasting ability for extreme heat wave weather. The number of optimal forecast factors is determined based on sensitivity tests to ensure that the optimal forecast factor configuration has the most robust and skilled forecasting performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The result of applying the empirical orthogonal function EOF analysis to the Chinese summer temperature data provided in Example 1 of the present invention; in the figure, af is the eigenvectors of the first six main EOF modes of summer temperature anomaly in China from 1991 to 2010, g is the corresponding explained variance (black curve; left y-axis; unit: %) and error range (red line), and the blue curve in g is the PCC (right y-axis) between the temperature anomaly reconstructed using the first N (1-8) (x-axis) main EOF modes from 2016 to 2021 and the observed value;

[0055] Figure 2 The spatial distribution of the lead-lag correlation coefficient between PC1 of the summer temperature anomaly in China during 1979–2009 and the previous (af) PW and (gl) H500 subseasonal components and (mr) SLP and (sx) H850 LFBS components provided in Example 1 of the present invention;

[0056] Figure 3 The model architecture of the univariate forecast model (a) and the sub-seasonal temperature forecast model (b) provided in Example 1 of the present invention;

[0057] Figure 4 The TCC skill distribution of the (ae) CNN, (fj) CMA and (ko) ECMWF models for 10-30 days (2-6 pentads) ahead prediction during the test period provided in Example 2 of the present invention;

[0058] Figure 5 The PCC skills of the three models for the spatial mode of air temperature at different forecast times during the test period provided by Example 2 of the present invention;

[0059] Figure 6 The RMSE technique for predicting temperature anomalies provided in Example 2 of the present invention;

[0060] Figure 7 The EDI technique for predicting heat waves provided in Example 2 of the present invention;

[0061] Figure 8 The attribution graph corresponding to the accurate prediction of the PC1 high value period 20 days in advance during the test period provided in Example 2 of the present invention;

[0062] Fig. 9 During the test period provided in Example 2 of the present invention, the CNN model accurately predicted the large-scale sub-seasonal disturbance input field composite map corresponding to the initial reporting sample of the PC1 high value period (i.e., exceeding the 90th percentile threshold) 20 days in advance;

[0063] Fig.10 A distribution diagram of the lead-lag correlation coefficient between the LFBS field three months in advance and PC1 of the temperature anomaly in China during the test period provided in Example 2 of the present invention;

[0064] Fig.11 The model framework of the sensitivity experiment provided in Example 1 of the present invention;

[0065] Fig.12 The prediction skills of the CNN models based on different architectures provided in Example 1 of the present invention for temperature and heat waves during the test period. DETAILED DESCRIPTION

[0066] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0067] Example 1

[0068] The present invention provides a sub-seasonal heat wave prediction method, which predicts the spatial distribution of temperature anomalies in a target area in the next 10-30 days based on a pre-trained sub-seasonal temperature forecast model, and monitors the occurrence of heat wave events based on the predicted temperature anomalies.

[0069] In this embodiment, the application scenario of sub-seasonal heat wave prediction is China. This embodiment provides a process of building a sub-seasonal temperature forecast model based on CNN and a training process of the sub-seasonal temperature forecast model.

[0070] Firstly, empirical orthogonal function (EOF) analysis is applied to the summer temperature data in China to obtain the main temperature modes and their time coefficients in China. Candidate forecast factors are selected based on previous studies on the previous sub-seasonal signals related to the occurrence of heat waves in China.

[0071] Specifically, the EOF analysis was performed using the summer temperature data of China from 1991 to 2010. Figure 1 The eigenvectors (a–f) of the first six main EOF modes of summer temperature anomalies in China during 1991–2010 are shown. Figure 1 (g) shows the corresponding explained variance (black curve; left y-axis; unit: %) and error range (red line). The blue curve in (g) is the spatial correlation coefficient PCC (right y-axis) between the temperature anomaly reconstructed using the first N (1–8) (x-axis) main EOF modes and the observed value from 2016 to 2021, which is given by Figure 1 It can be seen that the predicted temperature anomaly field in China can be reconstructed by the first six main temperature modes and the corresponding time coefficients (PC1-PC6).

[0072] The large-scale fields that serve as candidate predictors include: 850 hPa temperature T850, 200 hPa zonal wind U200, 500 hPa zonal wind U500, 850 hPa zonal wind U850, 200 hPa meridional wind V200, 500 hPa meridional wind V500, 850 hPa meridional wind V850, 200 hPa geopotential height field H200, 500 hPa geopotential height field H500, 850 hPa geopotential height field H850, 700 hPa relative humidity RH700, precipitable water PW, sea level pressure SLP, surface soil moisture SSM and outward longwave radiation OLR.

[0073] In order to verify the influence of the selected candidate predictors on the temperature anomaly in China, the subseasonal component and low-frequency background state LFBS component of the candidate predictors were extracted by non-bandpass filtering method and experimental analysis was carried out. Figure 2 The correlation between the variability of the main mode time coefficient PC1 of China's temperature and the subseasonal component of the previous PW, the subseasonal component of H500, the LFBS component of SLP, and the LFBS component of H850 is shown.

[0074] similar Figure 2 Lead-lag correlation analysis was also conducted on other PCs, and it was found that there are three main paths of subseasonal disturbances that are crucial to regulating the variability of summer temperature in China: 1) southeastward propagation signals originating from northwestern Eurasia, 2) southeastward propagation signals originating from Northeast Asia, and 3) northward propagation signals originating from the tropical western Pacific and tropical Indian Ocean. The key early LFBS signals affecting temperature are also important over Eurasia and tropical oceans. Therefore, in this embodiment, the target area considered for constructing the subseasonal temperature forecast model is determined to be Eurasia and the tropics (2.5°S-75.0°N, 40°-167.5°E).

[0075] The daily maximum temperature data in China from 1979 to 2021 and the data of candidate forecasting factors in the target area were obtained, and a data set was constructed and preprocessed. The processed data were divided into training set, validation set and test set according to the year.

[0076] Specifically, SSM and OLR come from the National Centers for Environmental Prediction / National Center for Atmospheric Research reanalysis data and the National Oceanic and Atmospheric Administration polar-orbiting satellite, respectively, and the remaining variables come from the National Centers for Environmental Prediction / Department of Energy reanalysis data. SSM uses the T62 Gaussian grid format (192×94), and the horizontal resolution of the other large-scale variable fields is 2.5° ×2.5°. Before model training, the SSM field is interpolated to a horizontal resolution of 2.5° ×2.5°.

[0077] In order to carry out real-time applications, the subseasonal component (10–90 days) and the low-frequency background state LFBS component of the candidate predictor are extracted by non-bandpass filtering method. Specifically, for each variable, the daily climate state is first subtracted from the original daily data of the candidate predictor to eliminate the climate annual cycle and obtain the daily anomaly; the multi-year average of the same date in the daily climate state from 1991 to 2020; then, the average of the previous 45 days is subtracted from the daily anomaly to eliminate the signal of more than 90 days; the anomaly under 90 days is obtained; then, the anomaly under 90 days is subjected to the leading 5-day sliding average to eliminate the high-frequency weather scale disturbance under 10 days, and the subseasonal component of the candidate predictor is obtained; and the LFBS component is the 30-day average of the daily anomaly.

[0078] The historical data are divided into three stages: 1979-2009 is the training period, and relevant samples are used to train the model; 2010-2015 is the verification period, which is used to determine the optimal prediction factor components and model tuning; 2016-2021 is the testing period, which is used to detect the model's actual forecasting ability.

[0079] A sub-seasonal temperature forecast model is constructed based on CNN. The model is a multivariate forecast model, and its construction process involves the construction of a univariate forecast model. The univariate forecast model is used to determine the temperature main mode time coefficient-forecast timeliness-optimal forecast factor component relationship table, and provides a convolution block for extracting the deep features of the optimal forecast factor component. The convolution block is used to train the sub-seasonal temperature forecast model. Figure 3 The specific structures of the univariate forecast model and the sub-seasonal temperature forecast model considering multivariate information are shown.

[0080] like Figure 3 As shown in the figure, the univariate forecast model includes: 2 input layers, 1 convolution block containing 3 two-dimensional convolution (Conv2d) layers and 2 maximum pooling (MP) layers, 1 fully connected layer (FC) and 1 output layer. Among them, the first input layer (orange box) provides the input features corresponding to the forecast factors, and the second input layer (red box) provides the encoded initialization information of the current month to guide the model to consider the annual cycle impact of subseasonal variability.

[0081] The subseasonal temperature forecast model considering multivariate information includes: 8 input layers, 7 convolutional blocks including 3 Conv2d layers and 2 MP layers, 2 2D convolutional layers, 1 fully connected layer and 1 output layer; among them, 7 input layers are set in 7 independent input channels to consider the predictable information provided by 4 subseasonal forecast factors and 3 LFBS forecast factors. Another input layer (red box) provides the encoded initialization information of the current month.

[0082] First, the candidate predictors are used to train the univariate forecast model. Here, the model is trained using the full year data, not just the summer samples. The model trained by this method is better than the model trained using only the summer samples. The training of the univariate forecast model includes:

[0083] For any candidate predictor: obtain the full-year data of the candidate predictor from the training set, and use the non-bandpass filtering method to extract the sub-seasonal component and low-frequency background state LFBS component of the candidate predictor; use the sub-seasonal component or LFBS component of the candidate predictor as the input of the univariate prediction model, specify the main mode time coefficient of temperature as the output of the univariate prediction model, and train the univariate prediction model.

[0084] Specifically, for any candidate predictor: the first input layer provides the corresponding feature map of the sub-seasonal component or LFBS component of the candidate predictor; the convolution block is used to perform convolution and pooling operations on the input layer feature map to obtain the deep features provided by the candidate predictor component; the second input layer provides information for initializing the current month; the fully connected layer integrates the deep features and the month information, and outputs the time coefficient of the main mode of temperature under the future specified forecast time through the output layer.

[0085] Furthermore, the univariate forecasting model constructed based on the subseasonal component or LFBS component of each candidate forecasting factor on the validation set was evaluated for its forecasting skill for the six main temperature mode time coefficients 10 days, 15 days, 20 days, 25 days and 30 days in advance, and the four subseasonal forecasting factors and three LFBS forecasting factors with the best forecasting performance for each main temperature mode time coefficient under each forecast time were determined. The relationship table of main temperature mode time coefficient-forecast time effectiveness-optimal forecasting factor component was generated as shown below.

[0086]

[0087] Specifically, the dimension of the feature map (orange box) corresponding to the first input layer is N × 32 × 52. For the subseasonal forecast factor, it represents the large-scale subseasonal disturbance value every 5 days in the previous 30 days in the area of ​​2.5°S–75.0°N (32 grid points) and 40°–167.5°E (52 grid points) (N = 6). For the LFBS forecast factor, it represents the LFBS component every 30 days in the previous 90 days in the same area (N = 3). The output layer (black box) is used to predict the main mode time coefficient of temperature in the next 10, 15, 20, 25 or 30 days (i.e. 2–6 pentads).

[0088] It should be noted that the number of subseasonal forecast factors and LFBS forecast factors used to construct the subseasonal temperature forecast model considering multivariate information is determined by sensitivity tests. Such a forecast factor configuration has the most robust and skilled forecast performance.

[0089] The proficient prediction skills of the model constructed in this patent are attributed to the integration of information provided by multi-scale and multi-variable interactions. A series of sensitivity tests with different model structures are designed to confirm this:

[0090] (1) CNN-A Fig.11 a): Similar to the sub-seasonal temperature forecast model considering multivariate information in the present invention, but this model only uses the forecast information from four sub-seasonal forecast factors.

[0091] (2) CNN-L Fig.11 b): Compared with CNN-A ( Fig.11 a) Similar, but using only information from the 3 LFBS predictors.

[0092] (3) CNN-A-t0 Fig.11 c): Based on CNN-A ( Fig.11 a), but only using information from the most recent previous time step of the subseasonal predictors (N = 1). The input layer feature map associated with each large-scale predictor has size 1 × 32 × 52.

[0093] (4) CNN-A-ENS: No additional convolution operations are added, and prediction is made only by averaging the results of a single predictor.

[0094] (5) CNN-NP: Similar to the sub-seasonal temperature forecast model in the present invention, but this model does not use the pre-trained weights of the convolutional blocks obtained by training the optimal predictor model.

[0095] Fig.12 The results of the sensitivity test are shown: when the model adopts the same CNN structure as the sub-seasonal temperature forecast model in the present invention, but only uses the sub-seasonal forecast factor ( Fig.11 a) or LFBS predictor ( Fig.11 b) the prediction skill drops by 5%–15% and 25%–60% compared to the model that exploits information from both scales, respectively ( Fig.12 Introducing the temporal and spatial evolution characteristics of the predictors into the model can also improve the prediction ability. Compared with the model prediction skill using the subseasonal signal of the first 30 days (N = 6), the model prediction skill using only the first 5 days (i.e., N = 1; Fig.11 c) The model’s forecast skill for temperature and heat waves has decreased by 5%–30% ( Fig.12Purple and green histograms). The sensitivity test results also emphasize the importance of adding additional convolution operations to fuse the information of various predictors ( Figure 3 b) Without this feature fusion, the ensemble forecast obtained by averaging the outputs of the individual predictors shows lower skill ( Fig.11 Red and green bar graphs). In addition, not using pre-trained convolutional block weights to train the sub-seasonal temperature forecast model will also lead to a decrease in model prediction skills ( Fig.11 Orange and blue bar graphs), highlighting the advantages and necessity of pre-training to improve the generalization ability of the model.

[0096] The training of the sub-seasonal temperature forecast model includes: taking the data of the 7 optimal prediction factor components corresponding to each temperature main mode time coefficient in the training set as input, taking each temperature main mode time coefficient as the output of the sub-seasonal temperature forecast model, and combining the convolution blocks provided by the 7 optimal prediction factor components in the univariate forecast model to train the sub-seasonal temperature forecast model.

[0097] Specifically, the input features of the seven optimal predictor components are respectively extracted through corresponding convolution blocks to obtain the deep features of the seven optimal predictor components; the two-dimensional convolution layer is used to perform convolution processing on the channel connection results of the deep features of the seven optimal predictor components to obtain joint features, and the joint features and the month information encoded by the last input layer are integrated through the fully connected layer, and the main modal time coefficient of the specified temperature under the future specified forecast time is output through the output layer.

[0098] When training the model, the pre-trained convolutional blocks from the univariate prediction model corresponding to the best predictor component are used ( Figure 3 a) for training. After being processed by the convolutional blocks, the deep features extracted from multiple optimal predictors are concatenated in the channel dimension, followed by two convolution operations. This approach can utilize the joint prediction information from the optimal predictors and allow the model to fully consider the interaction between subseasonal and LFBS variability.

[0099] Use the data in the validation set to evaluate the trained sub-seasonal temperature forecast model and optimize the model structure and parameters;

[0100] The independent forecast performance of the subseasonal temperature forecast model was tested using the data in the test set.

[0101] The filter size of all two-dimensional convolutional layers in the sub-seasonal temperature forecast model and the univariate forecast model is 3×3, the sliding step value is 1×1, and the activation functions of the two-dimensional convolutional layer and the fully connected layer are rectified linear units with leakage parameters; the filter size of the maximum pooling layer is 2×2, and the sliding step value is 2×2. Before all two-dimensional convolutional layers perform convolution operations, it also includes: filling the boundaries of the feature maps of the input two-dimensional convolutional layers with zero values ​​to keep the size of the feature maps unchanged before and after the operation.

[0102] Preferably, during the training process of the sub-seasonal temperature forecast model, an Adam optimizer is used to optimize the training parameters, and the initial learning rate of the Adam optimizer is 1×10 −3 .

[0103] Preferably, the mean square error is used as the loss function.

[0104] Preferably, a dropout method and an early stopping method are used to prevent model overfitting; the dropout method includes: during the model training iteration, adding a dropout layer with a dropout probability of 50% after each two-dimensional convolutional layer; the early stopping method includes: in response to no improvement in the skill of the validation set for 15 consecutive training iterations, stopping training to ensure effective utilization of the model weights corresponding to the minimum validation loss.

[0105] Example 2

[0106] Based on Example 1, this embodiment provides a method for predicting sub-seasonal heat waves in China, including the following steps:

[0107] Step 1: According to the pre-obtained temperature main mode of the target area and the temperature main mode time coefficient-forecast time effectiveness-optimal prediction factor component relationship table, determine all the optimal prediction factors associated with any temperature main mode time coefficient under the preset forecast time effectiveness to form the optimal prediction factor set.

[0108] In this embodiment, the main components of the temperature mode in China are: PC1-PC6. From the relationship table of the main components of the temperature mode-forecast timeliness-optimal predictor factors, it can be obtained that when predicting 20 days in advance, the corresponding optimal predictor factors of PC1 include: the sub-seasonal components of PW, T850, U850 and OLR, and the LFBS components of T850, PW and H500; the corresponding optimal predictor factors of PC2 include: the sub-seasonal components of U200, U500, H500 and V850, and the LFBS components of U200, V200 and T850; similarly, when predicting PC3-PC6 20 days in advance, the corresponding optimal predictor factor components are obtained.

[0109] Step 2: Acquire the data of each optimal predictor in the optimal predictor set, and extract the sub-seasonal component and the low-frequency background state LFBS component of each optimal predictor respectively through real-time non-bandpass filtering.

[0110] Step 3: Input the sub-seasonal component and the low-frequency background state LFBS component of each optimal prediction factor into the pre-trained sub-seasonal temperature forecast model to obtain the predicted value of each main mode time coefficient of the temperature in the target area under the time effect.

[0111] Among them, the process of obtaining the predicted value of any main mode time coefficient of temperature includes: the sub-seasonal temperature forecast model extracts the features of the optimal prediction factor component corresponding to the main mode time coefficient of temperature to obtain the deep features of the optimal prediction factor component; the deep features of all the optimal prediction factor components corresponding to the main mode time coefficient of temperature are integrated to obtain the joint features containing multi-scale multi-variable interaction information; the predicted value of the main mode time coefficient of temperature is obtained by combining the joint features and the current month information.

[0112] Specifically, the values ​​of PC1-PC6 predicted 20 days in advance for the Chinese region include:

[0113] For the main mode time coefficient PC1 of air temperature:

[0114] like Figure 3 As shown in the figure, the optimal predictor factors including the sub-seasonal components of PW, T850, U850 and OLR, and the LFBS components of T850, PW and H500 are input into the sub-seasonal temperature forecast model, and the convolution blocks corresponding to the 7 input layers respectively extract the deep features contained in the input features provided by the optimal predictor components; the deep features of the 7 optimal predictor components corresponding to PC1 are stacked in the channel dimension and further convolved twice to obtain the joint features of the optimal predictor; the PC1 value is predicted by combining the joint features and the current month information.

[0115] The same method is used to predict the values ​​of PC2-PC6.

[0116] Step 4: Obtain the predicted temperature anomaly field under the time effect according to the predicted value of each temperature main mode time coefficient and the temperature main mode reconstruction;

[0117] In this embodiment, the predicted temperature anomaly field in China is obtained by reconstructing six temperature main mode time coefficients (PC1-PC6) and the pre-obtained EOF main mode.

[0118] Step 5: predicting the sub-seasonal heat wave event in the target area according to the predicted temperature anomaly field under the time validity and the preset threshold standard;

[0119] Specifically, the sub-seasonal heat wave events in China are predicted based on the predicted temperature anomaly field reconstructed from the main temperature mode time coefficients (PC1-PC6) in step 2 and the pre-obtained main temperature mode.

[0120] It should be noted that in observations or forecasts, when the grid temperature anomaly exceeds the 90th percentile for at least three consecutive days, it is determined to be a heat wave event.

[0121] Example 3

[0122] This example provides the performance of the method of the present invention in predicting sub-seasonal heat waves. The temporal correlation coefficient TCC and the spatial correlation coefficient PCC are used to evaluate the model's ability to predict the temporal evolution and spatial modal characteristics of temperature anomalies. The root mean square error RMSE is used to measure the magnitude of the forecast error, and the extreme dependence index EDI is used to evaluate the accuracy of predicting the occurrence of heat waves:

[0123]

[0124] where H is the hit rate, defined as the ratio of correctly predicted observed heatwaves to the total observed heatwaves, F is the false alarm rate, calculated as the ratio of the number of false alarms to the total number of observed non-events, and EDI ranges from −1 to 1, with 1 indicating a perfect prediction and 0 indicating no skill.

[0125] Figure 4 The CNN model used in the present invention and the TCC skill of the CMA and ECMWF models in predicting temperature anomalies are shown. This skill is used to measure the model's ability to predict the temporal evolution characteristics of temperature anomalies. The CNN model has a lower skill when it is 2-3 pentads in advance ( Figure 4 ab), not as good as the dynamic model ( Figure 4 fl, but outperforms the CMA and ECMWF models in forecasts beyond 20 days ahead ( Figure 4 Even up to 6 pentads ahead, the CNN model still has proficient (i.e., TCC at 99% confidence level) prediction skills in most parts of China ( Figure 4 e), while the dynamical model mainly performs well in western and southwestern China ( Figure 4 jo).

[0126] Figure 5 The CNN model used in the present invention and the PCC technique of predicting temperature anomalies by the CMA and ECMWF models are shown, and the technique is used to evaluate the predictive ability of each model on the spatial distribution of temperature anomalies in China. Figure 5a is the time average of the PCC of each forecast result. Similar to the TCC technique, the CNN model is difficult to match the dynamic model within 3 pentads. However, compared with the CMA and ECMWF models, they show superior prediction ability at 4-6 pentads. Figure 5 b shows the percentage of prediction frequencies of PCC that passed the 99% confidence level test, further confirming the superior prediction capability of the CNN model used in the present invention at 4-6 pentads in advance.

[0127] In addition, if Figure 6 As shown in Figure 2, when the forecast lead time exceeds 20 days, the RMSE value of the CNN model is smaller than that of the dynamic model, and the average RMSE of the Chinese region is always less than 1 standard deviation when the forecast lead time is 10-30 days. Figure 4 and Figure 6 It can be seen that regions with higher TCC generally show smaller prediction errors.

[0128] Figure 7 We further compared the EDI between CNN-based and dynamical predictions. 2-6 pentads in advance, both CNN and dynamical models can effectively predict heat waves in most parts of China (EDI > 0). In the shorter forecast period (3 pentads), the ECMWF model showed the highest skill among the three models ( Figure 7 kl), the CMA model also performs well in the 2-month-ahead forecast ( Figure 7 f). However, in long-term forecasts (more than 20 days), the CNN model has a higher forecast accuracy for temperature ( Figure 4-6 ), which has better forecasting ability for heat waves than the dynamic model ( Figure 7 3-5), especially in the densely populated eastern part of China.

[0129] Below, based on the Grad-CAM method, a physical explanation of the prediction using the CNN model is given.

[0130] Figure 8 The attribution map is shown when the CNN model accurately predicts the high value period (exceeding the 90th percentile threshold) of the time coefficient PC1 corresponding to the main mode of temperature 20 days in advance. Figure 8 ag shows the results of the last Conv2d layer of each convolutional block, highlighting the key features in the predictors located in Northwest Eurasia and the tropics that are important for the model to produce accurate predictions.

[0131] Figure 9-10The evolution characteristics of these predictors are shown, mainly highlighting the precursor signals from tropical and temperate regions. This shows that the CNN model is able to identify the origin and propagation of the previous subseasonal signals associated with the high PC1 period, and also shows the model's ability to capture the impact of interannual variability on temperature changes. Attribution map from the last Conv2d layer of the model ( Figure 8 h) shows the influence of all predictors on the sub-seasonal temperature forecast model ( Figure 3 The results show that large-scale signals from Northwest Eurasia, Northeast Asia, the tropical Indian Ocean, and local China provide a source of predictability for the 20-day forecast.

[0132] Example 4

[0133] According to a second aspect of the present invention, a computer storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the sub-seasonal heat wave prediction method as described in Example 1 is implemented.

[0134] The storage medium may include, for example, a storage component of a tablet computer, a hard disk of a computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer storage medium may be any combination of one or more computer storage media.

[0135] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A sub-seasonal heat wave prediction method, characterized in that: include: According to the pre-obtained temperature main mode of the target area and the temperature main mode time coefficient-forecast time effectiveness-optimal forecast factor component relationship table, all optimal forecast factors associated with any temperature main mode time coefficient under the preset forecast time effectiveness are determined to form an optimal forecast factor set; Acquire the data of each optimal predictor in the optimal predictor set, and respectively extract the subseasonal component and the low-frequency background state LFBS component of each optimal predictor by real-time non-bandpass filtering; Inputting the sub-seasonal component and the low-frequency background state LFBS component of each optimal forecast factor into a pre-trained sub-seasonal temperature forecast model to obtain the predicted value of each main mode time coefficient of the temperature in the target area under the time effect; Obtaining the predicted temperature anomaly field under the time effect according to the predicted value of each temperature main mode time coefficient and the temperature main mode reconstruction; Predicting sub-seasonal heat wave events in the target area based on the predicted temperature anomaly field under the time validity and the preset threshold standard; The process of obtaining the predicted value of any main mode time coefficient of air temperature includes: The sub-seasonal temperature forecast model performs feature extraction on the optimal forecast factor component corresponding to the main mode time coefficient of the temperature to obtain the deep features of the optimal forecast factor component; The deep features of all the optimal prediction factor components corresponding to the main mode time coefficient of the temperature are fused to obtain the joint features containing multi-scale multi-variable interaction information; The predicted value of the main modal time coefficient of the temperature is obtained by combining the joint features and the current month information.

2. The sub-seasonal heat wave prediction method according to claim 1, characterized in that: The sub-seasonal temperature forecast model also includes: a univariate forecast model; the univariate forecast model is used to provide the temperature modal time coefficient-forecast time effect-optimal forecast factor relationship table and a pre-trained convolution block for extracting deep features of each optimal forecast factor component during the training process of the sub-seasonal temperature forecast model.

3. The sub-seasonal heat wave prediction method according to claim 2, characterized in that: The training process of the sub-seasonal temperature forecast model includes: Determine a target area and determine candidate forecast factors influencing daily maximum temperature in the target area; Obtain the daily maximum temperature data of the preset historical period in the target area and the data of the candidate forecasting factors, construct a data set and perform preprocessing, and divide the processed data into a training set, a validation set and a test set according to the year; A non-bandpass filtering method is used to extract the sub-seasonal component and the low-frequency background state LFBS component of the candidate predictor; The summer temperature data is obtained from the daily maximum temperature data in the training set, and the empirical orthogonal function EOF is used for analysis to obtain S main temperature modes and their corresponding time coefficients; The annual data of each candidate forecast factor is obtained from the training set, and the subseasonal component and LFBS component of each candidate forecast factor are used as the input of the univariate forecast model respectively, and the time coefficient of S main temperature modes is used as the output of the univariate forecast model to train the univariate forecast model; The sub-seasonal components or LFBS components of each candidate prediction factor in the validation set are input into the corresponding univariate prediction model, and the S main temperature modal time coefficients under different time effects are predicted respectively. By evaluating and comparing the prediction skills of the univariate prediction model, the a sub-seasonal components of the prediction factors and b LFBS components of the prediction factors with the best prediction performance for each main temperature modal time coefficient under each time effect are determined, and a relationship table of main temperature modal time coefficient-forecast time effect-optimal prediction factor component is formed; among which a and b are determined based on sensitivity tests; Taking the data of the (a+b) optimal prediction factors corresponding to each main mode time coefficient of the temperature in the training set as input, taking each main mode time coefficient of the temperature as the output of the sub-seasonal temperature prediction model, combined with the pre-trained convolution blocks provided by the univariate prediction model corresponding to the (a+b) optimal prediction factors, training is performed; Use the data in the validation set to evaluate the trained sub-seasonal temperature forecast model and optimize the model structure and parameters; The independent forecast performance of the subseasonal temperature forecast model was tested using the data in the test set.

4. The sub-seasonal heat wave prediction method according to claim 3, characterized in that: The univariate forecasting model includes: 2 input layers, 1 convolutional block, 1 fully connected layer and 1 output layer; For any candidate predictor: The first input layer provides a corresponding characteristic map of the subseasonal component or LFBS component of the candidate predictor; The convolution block is used to perform convolution and pooling operations on the feature map, so as to extract corresponding deep features; The second input layer provides information for initializing the current month; The fully connected layer integrates the deep features and the month information, and outputs the time coefficient of the specified temperature main mode under the future specified forecast time through the output layer; The sub-seasonal temperature forecast model includes: (a+b+1) input layers, (a+b) convolution blocks, 2 two-dimensional convolution layers, 1 fully connected layer and 1 output layer; Wherein, the (a+b) input layers are respectively arranged in (a+b) independent input channels; and are used to provide (a+b) optimal prediction factor corresponding input feature maps; The (a+b) convolution blocks are respectively used to extract deep features corresponding to the (a+b) optimal prediction factors; The two-dimensional convolution layer is used to perform further convolution operation on the result of splicing the deep features of the (a+b) optimal prediction factors in the channel dimension direction. The convolution result and the month information provided by the last input layer are integrated through the fully connected layer, and the time coefficient of the specified main mode of temperature under the future specified forecast time is output through the output layer.

5. The sub-seasonal heat wave prediction method according to claim 4, characterized in that: The convolution block includes three two-dimensional convolutional layers and two maximum pooling layers arranged at intervals.

6. The sub-seasonal heat wave prediction method according to claim 5, characterized in that: The method of extracting the subseasonal component and the LFBS component of the candidate prediction factor by using a non-bandpass filtering method includes: Subtract the daily climate state from the raw daily data of the candidate predictor to eliminate the climate annual cycle and obtain the daily anomaly; the daily climate state is the multi-year average of the same date from 1991 to 2020; Subtract the average of the previous 45 days from the daily outliers to eliminate the signals over 90 days; get the outliers under 90 days; The abnormal values ​​within 90 days are taken as the leading 5-day moving average to eliminate the high-frequency weather scale disturbance within 10 days, and the sub-seasonal component of the candidate forecast factor is obtained; The LFBS component is then the 30-day average of the daily anomalies.

7. The sub-seasonal heat wave prediction method according to claim 5, characterized in that: The filter size of the two-dimensional convolution layer is 3×3, the sliding step value is 1×1, and the activation functions of the two-dimensional convolution layer and the fully connected layer are rectified linear units with leakage parameters; the filter size of the maximum pooling layer is 2×2, and the sliding step value is 2×2; Before the two-dimensional convolutional layer performs the convolution operation, it also includes: filling the boundary of the feature map input to the two-dimensional convolutional layer with a zero value so that the size of the feature map remains unchanged before and after the operation.

8. The sub-seasonal heat wave prediction method according to claim 5, characterized in that: The candidate prediction factors include: 850 hPa temperature T850, 200 hPa zonal wind U200, 500 hPa zonal wind U500, 850 hPa zonal wind U850, 200 hPa meridional wind V200, 500 hPa meridional wind V500, 850 hPa meridional wind V850, 200 hPa geopotential height field H200, 500 hPa geopotential height field H500, 850 hPa geopotential height field H850, 700 hPa relative humidity RH700, precipitable water PW, sea level pressure SLP, surface soil moisture SSM and outward longwave radiation OLR.

9. The sub-seasonal heat wave prediction method according to claim 5, characterized in that: During the training process of the sub-seasonal temperature forecast model, the Adam optimizer was used to optimize the training parameters. The initial learning rate of the Adam optimizer was 1×10 −3 ; The mean square error is used as the loss function; A dropout method and an early stopping method are used to prevent the model from overfitting; the dropout method includes: adding a dropout layer with a preset probability of discarding after each two-dimensional convolutional layer during the model training iteration; The early stopping method includes: in response to no improvement in the skill of the validation set after T consecutive training iterations, stopping training to ensure effective use of the model weights corresponding to the minimum validation loss, where T is a preset value.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the sub-seasonal heat wave prediction method according to any one of claims 1 to 9 is implemented.

Images