A Method and System for Predicting Terrestrial Precipitation in a Changing Environment

Abstract

The present invention provides a method and system for predicting terrestrial precipitation under changing environments, belonging to the technical field of artificial intelligence big data model prediction. It includes a data acquisition module, a training module, and a prediction module. Long-term multi-source data is obtained through ERA5 data and the WAM-2layers model. Combining the training and data processing of the long short-term memory network model, global and regional-level analyses are carried out. Since historical precipitation data, water vapor source data from different regions, and climate indices are combined during prediction, making full use of the correlation between the three types of data, the prediction of terrestrial precipitation has achieved high speed and high accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence big data model prediction, and particularly relates to a method and system for predicting terrestrial precipitation in a changing environment. Background Art

[0002] Terrestrial precipitation is crucial for the Earth's water cycle and is indispensable for maintaining ecosystems, agriculture, and human society. After continuous climate change, it has become increasingly important to understand the sources and dynamics of future precipitation, and changes in precipitation patterns are of great significance for research such as water resources, food security, and natural habitats. The change in the source of terrestrial precipitation is comprehensively affected by various factors, including the rising temperature caused by global warming, the increasing evaporation, and the long-term transformation of the atmospheric circulation pattern. The rising sea surface temperature enhances ocean evaporation, which in turn affects the contribution of the ocean to terrestrial precipitation. At the same time, deforestation and land use changes also inhibit terrestrial evapotranspiration, thus affecting the replenishment of terrestrial water sources to precipitation. Existing technologies for analyzing terrestrial precipitation often lack long-term and comprehensive methods required to capture broader patterns and trends. Most analyses focus on data limited to specific regions or times, ignoring long time periods and the influence of global factors. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for predicting terrestrial precipitation in a changing environment, to solve the problems of spatial, temporal, and data limitations in precipitation prediction in the background art, and to achieve faster and more accurate terrestrial precipitation prediction.

[0004] The present invention provides a method for predicting terrestrial precipitation in a changing environment, and the method includes:

[0005] Obtaining time series data of terrestrial precipitation amounts based on the reanalysis daily-scale terrestrial precipitation data of the ERA5 dataset;

[0006] Based on the WAM-2layers water vapor tracking model, obtaining time series of terrestrial water vapor contributions and time series of ocean water vapor contributions; performing cross-validation based on the time series data of terrestrial precipitation amounts and the time series of terrestrial water vapor contributions and ocean water vapor contributions;

[0007] If the cross-validation is consistent, based on the verified time series of terrestrial water vapor contributions and the WAM-2layers water vapor tracking model, calculating the time series of water vapor contributions for all Köppen climate regions respectively; based on the verified time series of ocean water vapor contributions and the WAM-2layers water vapor tracking model, calculating the time series of water vapor contributions for the four major oceans respectively;

[0008] Extract the principal components of the time series of water vapor contributions for all Köppen climate regions and the four major oceans respectively; conduct a correlation analysis between the principal components of the time series of water vapor contributions for all Köppen climate regions and the four major oceans and the major climate index data to obtain the key climate index data;

[0009] Build a model based on the long short-term memory network, and use the time series data of land precipitation, the time series of water vapor contributions for all Köppen climate regions, the time series of water vapor contributions for the four major oceans, and the key climate index data as the training set and the test set to train and test the model;

[0010] Predict future land precipitation based on the trained and tested model.

[0011] Further, the cross-validation based on the time series data of land precipitation and the time series of land water vapor contributions and ocean water vapor contributions includes: verifying whether the sum of the time series of land water vapor contributions and ocean water vapor contributions is consistent with the time series data of land precipitation.

[0012] Further, the obtaining of the time series data of land precipitation based on the reanalysis daily-scale land precipitation data of the ERA5 dataset includes:

[0013] Define a test statistic for the time series data of land precipitation;

[0014] Obtain the standardized test statistic based on the test statistic;

[0015] Obtain the qualitative change trend of land precipitation based on the standardized test statistic;

[0016] Conduct Sen's Slope test on the time series data of land precipitation to calculate the trend slope and obtain the quantitative change trend of land precipitation.

[0017] Further, the obtaining of the time series data of land precipitation based on the reanalysis daily-scale land precipitation data of the ERA5 dataset also includes: dividing the reanalysis daily-scale land precipitation data of the ERA5 dataset from 1941 to 2023 into four stages to calculate the time series data of land precipitation, and the four stages include: P1: 1941 - 1960, P2: 1961 - 1980, P3: 1981 - 2000, P4: 2001 - 2023.

[0018] Further, the obtaining of the time series of land water vapor contributions and the time series of ocean water vapor contributions based on the WAM-2layers water vapor tracking model includes: tracking and calculating water vapor based on the following formula:

[0019]

[0020] Among them, represents the moisture content at the bottom or top layer of a certain pixel, and are the horizontal and vertical wind speeds, and represent the moisture budget content in the horizontal and vertical directions, is the evaporation amount, is the precipitation amount, is the residual, is the vertical moisture transport amount between the upper and lower layers.

[0021] Furthermore, based on the verified terrestrial water vapor contribution time series and the WAM-2layers water vapor tracking model, calculate the water vapor contribution time series of all Köppen climate regions respectively; based on the verified marine water vapor contribution time series and the WAM-2layers water vapor tracking model, calculate the water vapor contribution time series of the four major oceans respectively, including:

[0022] Based on the Köppen climate region classification vector data, ocean boundary vector data, and the WAM-2layers water vapor tracking model, track and calculate the evaporation amounts of the land precipitation traced in all Köppen climate regions and the four major oceans respectively;

[0023] Based on the evaporation amounts of the land precipitation traced in all Köppen climate regions and the four major oceans and the total land precipitation, calculate the water vapor contribution ratios of the corresponding Köppen climate regions and oceans respectively;

[0024] Based on the water vapor contribution ratios of the corresponding Köppen climate regions and oceans, as well as the terrestrial water vapor contribution time series and the marine water vapor contribution time series, calculate the water vapor contribution time series of the corresponding Köppen climate regions and oceans respectively according to the four stages.

[0025] Furthermore, the main components of the water vapor contribution time series of all Köppen climate regions and the four major oceans are extracted respectively, including:

[0026] Based on the water vapor contribution time series of all Köppen climate regions and the four major oceans, extract the first three main components with a cumulative variance contribution rate greater than 60%.

[0027] Furthermore, the main climate index data includes ENSO index data, IDO index data, and NAO index data;

[0028] Perform a correlation analysis on the main components of the water vapor contribution time series of all Köppen climate regions and the four major oceans and the main climate index data, and the calculation formula for obtaining the key climate index data is:

[0029]

[0030] In the formula, and respectively represent the water vapor contribution data and the climate index data, and respectively represent the average values of the water vapor contribution data and the climate index data.

[0031] Furthermore, interpolation operations are performed on the key climate index data or the main climate index data;

[0032] Make the time scales of the key climate index data or the main climate index data the same as those of the land precipitation time series data, the water vapor contribution time series of all Köppen climate regions, and the water vapor contribution time series of the four major oceans.

[0033] Based on the above method for predicting land precipitation under a changing environment, the present invention also provides a system for predicting land precipitation under a changing environment, and the system includes:

[0034] A data acquisition module, which is used to obtain land precipitation time series data based on the reanalysis daily-scale land precipitation data of the ERA5 dataset; obtain the land water vapor contribution time series and the ocean water vapor contribution time series based on the WAM-2layers water vapor tracking model; perform cross-validation based on the land precipitation time series data, the land water vapor contribution time series, and the ocean water vapor contribution time series; if the cross-validation is consistent, calculate the water vapor contribution time series of all Köppen climate regions respectively based on the verified land water vapor contribution time series and the WAM-2layers water vapor tracking model; calculate the water vapor contribution time series of the four major oceans respectively based on the verified ocean water vapor contribution time series and the WAM-2layers water vapor tracking model; respectively extract the principal components of the water vapor contribution time series of all Köppen climate regions and the four major oceans; perform a correlation analysis on the principal components of the water vapor contribution time series of all Köppen climate regions and the four major oceans and the main climate index data to obtain the key climate index data;

[0035] A training module, which is used to establish a model based on a long short-term memory network, and use the land precipitation time series data, the water vapor contribution time series of all Köppen climate regions, the water vapor contribution time series of the four major oceans, and the key climate index data as the training set and the test set to train and test the model;

[0036] A prediction module, which is used to predict future land precipitation based on the trained and tested model.

[0037] One or more of the above technical solutions in the embodiments of the present application have at least one or more of the following technical effects:

[0038] The method and system for predicting terrestrial precipitation under changing environments provided by the embodiments of the present invention obtain long-term multi-source data through ERA5 data and the WAM-2layers model, combine with the long short-term memory network model, and through global and regional level analysis, combine historical precipitation data, water vapor source data in different regions and climate indices, providing a high-precision precipitation prediction method. By comprehensively using ERA5 reanalysis data and the WAM-2layers water vapor tracking model, it can more accurately track and analyze the sources and changes of terrestrial precipitation. The ERA5 dataset covers a wide range of meteorological elements, providing a detailed data basis for water vapor transport and precipitation processes. At the same time, the WAM-2layers model accurately calculates the proportion of water vapor contributions from different sources through a layered atmospheric structure, thereby achieving a high-precision dynamic assessment of precipitation sources. The method of multi-source data fusion overcomes the limitations of previous single data sources or simple models. Since historical precipitation data, water vapor source data in different regions and climate indices are combined during prediction, making full use of the correlation between the three types of data, the prediction of terrestrial precipitation is achieved quickly and with high precision, and can more comprehensively reveal the complex mechanism of precipitation sources. The above method and system require no manual intervention and achieve end-to-end prediction. Compared with traditional prediction methods that rely on artificial experience or simple statistical models, it can respond more quickly to precipitation changes and changes in climate indices, provide more timely and accurate information support for decision-makers, help improve the efficiency and effectiveness of water resource management, reduce the impact of human factors on prediction results, and ensure the objectivity and accuracy of predictions.

[0039] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the following specifically gives the specific embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flowchart of the prediction method in Embodiment 1 of the present invention;

[0041] Figure 2 It is an architecture diagram of the prediction system in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be completely conveyed to those skilled in the art.

[0043] Embodiment 1:

[0044] Reference appendix Figure 1 , this embodiment provides a method for predicting terrestrial precipitation under changing environments, and the method includes the following steps:

[0045] S1: Obtain the time series data of terrestrial precipitation based on the reanalysis daily-scale terrestrial precipitation data of the ERA5 dataset;

[0046] S2: Based on the WAM-2layers water vapor tracking model, obtain the time series of terrestrial water vapor contribution and the time series of ocean water vapor contribution; perform cross-validation based on the time series data of terrestrial precipitation, the time series of terrestrial water vapor contribution, and the time series of ocean water vapor contribution;

[0047] S3: If the cross-validation is consistent, calculate the time series of water vapor contribution for all Köppen climate regions respectively based on the time series of terrestrial water vapor contribution and the WAM-2layers water vapor tracking model; calculate the time series of water vapor contribution for the four major oceans respectively based on the time series of ocean water vapor contribution and the WAM-2layers water vapor tracking model;

[0048] S4: Extract the principal components of the time series of water vapor contribution for all Köppen climate regions and the four major oceans respectively; perform a correlation analysis on the principal components of the time series of water vapor contribution for all Köppen climate regions and the four major oceans and the main climate index data to obtain the key climate index data;

[0049] S5: Establish a model based on the long short-term memory network, and use the time series data of terrestrial precipitation, the time series of water vapor contribution for all Köppen climate regions, the time series of water vapor contribution for the four major oceans, and the key climate index data as the training set and the test set to train and test the model;

[0050] S6: Predict future terrestrial precipitation based on the trained and tested model.

[0051] For step S1, with the increase in global temperature, the evaporation rate increases, and the water vapor content in the atmosphere also increases accordingly, which leads to an increase in extreme precipitation events. Some regions may experience more frequent and severe rainstorms and floods, while other regions may suffer from long-term droughts. The uneven precipitation distribution not only affects agricultural production and water supply but may also cause environmental problems such as soil erosion, river course changes, and groundwater level decline. Terrestrial precipitation is a key component of the Earth's water cycle, and the center of the water cycle is the complex interaction between terrestrial and ocean water sources. To understand the trend of terrestrial precipitation and provide a basis for water vapor tracking, a long-term trend analysis of total terrestrial precipitation is first carried out. The present invention utilizes the ERA5 dataset, and its advantages lie in the improvements in model physics, dynamic core, data assimilation, and the increase in resolution, which play an important role in improving the quality of weather and climate data.

[0052] Based on the ERA5 reanalysis daily-scale terrestrial precipitation data, the terrestrial annual precipitation data from 1941 to 2023 was calculated to obtain the terrestrial precipitation time series data. , where represents the precipitation data of a certain year, and this time series data is a sample of independent and identically distributed random variables.

[0053] Specifically, the terrestrial precipitation time series data was obtained based on the ERA5 dataset's reanalysis daily-scale terrestrial precipitation data, including: dividing the reanalysis daily-scale terrestrial precipitation data of the ERA5 dataset from 1941 to 2023 into four stages to calculate the terrestrial precipitation time series data. The four stages are: P1: 1941 - 1960, P2: 1961 - 1980, P3: 1981 - 2000, P4: 2001 - 2023.

[0054] In this embodiment, after obtaining the terrestrial precipitation time series data based on the ERA5 dataset's reanalysis daily-scale terrestrial precipitation data, the long-term trend analysis of the total terrestrial precipitation can be carried out according to the obtained terrestrial precipitation time series data to analyze the change trend of terrestrial precipitation. The analysis methods include qualitative and quantitative analysis.

[0055] Among them, the specific steps of qualitative analysis are:

[0056] Define a test statistic for the terrestrial precipitation time series data. Specifically, use the Mann-Kendall trend test method to judge the trend significance and define the test statistic as follows:

[0057]

[0058]

[0059] Where , are the observed values corresponding to the , th time series respectively, and , is the sign function.

[0060] When the number of the water vapor contribution time series n is greater than 8, the statistic approximately follows a normal distribution. Without considering the existence of equal value data points in the sequence, the variance calculation formula is as follows:

[0061]

[0062] Based on the test statistic, a standardized test statistic is obtained. The standardized test statistic is calculated as follows:

[0063]

[0064] Based on the standardized test statistic the qualitative change trend of the terrestrial precipitation is obtained, where a positive value indicates an upward trend in terrestrial precipitation, and a negative value indicates a decreasing trend in terrestrial precipitation, and when the absolute value of is greater than or equal to 1.645, 1.96, 2.576, it indicates passing the significance tests with confidence levels of 90%, 95%, and 99% respectively.

[0065] The method for quantitative analysis is: perform Sen's Slope test on the terrestrial precipitation time series data to calculate the trend slope, and obtain the quantitative change trend of the terrestrial precipitation. The calculation formula is:

[0066]

[0067] where represents the change amount per unit time in the time series. When is a positive number, it indicates an upward trend in the time series. When is a negative number, the time series shows a downward trend, is the median function.

[0068] The above qualitative and quantitative trend analyses can qualitatively and quantitatively understand the overall change trend of terrestrial precipitation macroscopically, thereby providing data support for subsequent analyses of the causes and sources of precipitation changes.

[0069] For step S2, the change in the source of terrestrial precipitation is comprehensively affected by multiple factors, including the rising temperature, increasing evaporation, and long-term change of the atmospheric circulation pattern caused by global warming. The rising sea surface temperature enhances ocean evaporation, thereby affecting the contribution of the ocean to terrestrial precipitation. At the same time, deforestation and land use changes also have an inhibitory effect on terrestrial evapotranspiration, thus affecting the replenishment of terrestrial water sources to precipitation.

[0070] This embodiment adopts the WAM-2layers water vapor tracking model. WAM-2layers (Water Accounting Model-2layers) is an atmospheric water tracking model based on the Eulerian framework, mainly used to analyze the water vapor sources and their transport paths of regional precipitation. Its core features and applications are as follows:

[0071] Tracking mode: It can determine the location where precipitation initially evaporates (backward tracking) or the location where the evaporated moisture ultimately reaches (forward tracking), which is of great significance for understanding the water cycle.

[0072] Backtracking: Identify the water vapor evaporation source areas (i.e., "water sources") of precipitation in specific regions.

[0073] Forward tracking: Predict the ultimate fate of the evaporated moisture in the atmosphere (such as precipitation or recycling processes).

[0074] Stratified modeling: Vertically divide the atmosphere into two layers (the boundary layer and the free atmosphere), and combine humidity field and wind field data to quantify the contributions of water vapor transport at different heights.

[0075] The inherent capabilities of the WAM-2layers model enable it to more effectively identify the different water vapor contributions generated by different surface evaporation sources.

[0076] Based on the above WAM-2layers water vapor tracking model, time series of land water vapor contributions and ocean water vapor contributions are obtained, including: Tracking and calculating water vapor based on the following formula:

[0077] (6)

[0078] The above formula is based on the principle of atmospheric water balance, dividing the atmosphere into two levels, the bottom layer and the top layer, for water tracking and calculation, so as to better understand the transfer of water between different regions. The bottom layer mainly focuses on surface water evaporation and precipitation, while the top layer focuses on water vapor transport and redistribution in the atmosphere; among them, represents the water content in the bottom layer or top layer of a certain pixel, and are the horizontal and vertical wind speeds, and represent the water vapor budget contents in the horizontal and vertical directions, is the evaporation amount, is the precipitation amount, is the residual, is the vertical water transport amount between the upper and lower layers.

[0079] Based on the above WAM-2layers water vapor tracking model, the water vapor transport sources of land precipitation can be obtained, and the contributions of water vapor sources in different regions (such as the four major oceans: the Indian Ocean, the Arctic Ocean, the Atlantic Ocean, the Pacific Ocean; different Köppen climate zones: continental climate, polar climate, tropical climate, arid climate, temperate climate) to land precipitation can be obtained respectively. The water vapor contribution calculation formula is as follows:

[0080]

[0081] In the formula, Indicates the contribution ratio of specific source regions (the four major oceans, different Köppen climate zones) to terrestrial precipitation. Is the water vapor, i.e., the amount of water vapor, traced from specific regions (the four major oceans, different Köppen climate zones) to the land. Is the total terrestrial precipitation, obtained by statistically analyzing precipitation by region.

[0082] Based on the above principles, the contribution ratios of water vapor in terrestrial and oceanic regions can be statistically analyzed for four periods (P1: 1941 - 1960, P2: 1961 - 1980, P3: 1981 - 2000, P4: 2001 - 2023). The specific formula is as follows:

[0083]

[0084]

[0085]

[0086] In the formula, Is the total terrestrial precipitation, Is the water vapor contribution of the land to terrestrial precipitation, Is the water vapor contribution of the ocean to terrestrial precipitation, Is the land water vapor contribution ratio, Is the evaporation amount traced from the land to terrestrial precipitation, Is the ocean water vapor contribution ratio, Is the evaporation amount traced from the ocean to terrestrial precipitation.

[0087] The principle of cross - validation based on the terrestrial precipitation time - series data, the terrestrial water vapor contribution time - series, and the ocean water vapor contribution time - series is as follows: Theoretically, the precipitation in a specific region is equal to the water vapor transport. Specifically for terrestrial precipitation, the cross - validation method adopted in this embodiment is to verify whether the sum of the terrestrial water vapor contribution time - series and the ocean water vapor contribution time - series is consistent with the terrestrial precipitation time - series data. The above - mentioned terrestrial water vapor contribution time - series and ocean water vapor contribution time - series are And Long-time series time series data. Through the above verification, it can be proved that in the case of a long time scale, the water vapor contribution data obtained by tracking global water vapor using the WAM-2layers water vapor tracking model is generally consistent with the precipitation data. Therefore, there is an inherently consistent connection between the water vapor contribution data obtained by the above theoretical calculation and the reanalysis daily-scale land precipitation data of the ERA5 dataset. The reanalysis daily-scale land precipitation data of the ERA5 dataset during this period can be used for the prediction of land precipitation. If the above verification is inconsistent, it indicates that there are certain errors in the precipitation data during this period and it cannot be used for the training and testing of subsequent models. The above verification is particularly suitable for the verification of relatively early data and can effectively remove incorrect data and mask missing data.

[0088] For step S3, the Köppen climate classification therein is a climate classification system proposed by the German climatologist Wladimir Köppen. It divides the land into different climate types according to temperature and precipitation. This system divides the global land climate into five main climate zones (continental climate, polar climate, tropical climate, arid climate, temperate climate), and there are obvious differences in the contributions of different climate zones to land precipitation.

[0089] In this embodiment, based on the verified land water vapor contribution time series and the WAM-2layers water vapor tracking model, the water vapor contribution time series of all Köppen climate zones are calculated respectively; based on the verified ocean water vapor contribution time series and the WAM-2layers water vapor tracking model, the water vapor contribution time series of the four major oceans are calculated respectively, including:

[0090] Based on the Köppen climate zone classification vector data, the ocean boundary vector data and the WAM-2layers water vapor tracking model, the evaporation amounts of the tracked land precipitation in all Köppen climate zones and the four major oceans are tracked and calculated respectively;

[0091] Based on the evaporation amounts of the tracked land precipitation in all Köppen climate zones and the four major oceans and the total land precipitation, the water vapor contribution ratios of the corresponding Köppen climate zones and oceans are calculated respectively;

[0092] The method for calculating the water vapor contribution ratio of the corresponding Köppen climate zone is: obtain the Köppen climate zone classification vector data, extract the global evaporation amount data of the corresponding source area respectively, and based on the WAM-2layers model, calculate the evaporation amount of the th Köppen climate zone tracked to the land precipitation , calculate the water vapor contribution ratio, and the specific formula is as follows:

[0093] In the formula, represents the water vapor contribution of the th Köppen climate zone to the land precipitation.

[0094] The ocean not only has an important impact on the global climate, but is also a key component of the global water cycle and marine ecosystem. The ocean participates in the global water cycle through evaporation and precipitation, and has a significant impact on precipitation distribution and climate patterns.

[0095] To analyze the influence of different oceans on terrestrial precipitation, the boundary vector data of the Pacific Ocean, Indian Ocean, Atlantic Ocean, and Arctic Ocean are obtained respectively, and the global evaporation data of the corresponding source areas are extracted respectively. Based on the WAM-2layers model, calculate the evaporation of the th ocean traced to terrestrial precipitation

[0096] In the formula, represents the water vapor contribution of the th ocean area to terrestrial precipitation.

[0097] Using the same method as in step S2, based on the water vapor contribution ratios of the corresponding Köppen climate zones and oceans and the terrestrial water vapor contribution time series and ocean water vapor contribution time series, the water vapor contribution time series of the corresponding Köppen climate zones and oceans are calculated respectively according to the four stages.

[0098] For step S4, climate change is the main driving factor for the change in the structure of the water vapor contribution to terrestrial precipitation. Climate indices such as ENSO, IOD, and NAO play important roles in each climate zone, and there are many climate indices. Therefore, it is necessary to find the climate indices that have a greater impact on the change in the structure of the water vapor contribution to terrestrial precipitation. To find the climate indices related to the change in the structure of the water vapor contribution to terrestrial precipitation, first extract the principal components of the water vapor contribution time series of all Köppen climate zones and the four major oceans respectively. The specific method is to extract the first three principal components with a cumulative variance contribution rate greater than 60% based on the water vapor contribution time series of all Köppen climate zones and the four major oceans.

[0099] The specific formula is as follows:

[0100] For the water vapor contribution time series of the Köppen climate zone or ocean obtained in step S3 , this time series contains samples of the dataset, where each sample is a multi-dimensional vector, and calculate the mean vector :

[0101]

[0102] Perform centering processing, that is, subtract the corresponding mean from each feature to obtain the standardized data:

[0103]

[0104] Calculate the covariance matrix of the dataset :

[0105]

[0106] where is the number of samples, is a covariance matrix.

[0107] Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and the corresponding eigenvectors , sort the eigenvalues in descending order, and record the corresponding eigenvectors.

[0108] Based on the sorting of the eigenvalues, calculate the variance contribution rate of each eigenvalue, and select the first eigenvalues that can explain most of the variance. The formula for the cumulative variance contribution rate is:[[]]

[0109]

[0110] In this embodiment, it is set here to select the first three eigenvalues with a cumulative contribution rate greater than 60% as the principal components PC1, PC2, and PC3 for subsequent screening of climate index data. Since only three principal components are used for data operation, the amount of calculation can be greatly reduced while ensuring the accuracy of the operation.

[0111] The existing main climate index data includes ENSO index data, IDO index data, and NAO index data, etc. Different climate indices reflect the state of complex climate systems such as atmospheric circulation and ocean - atmosphere interaction from multiple dimensions, and these states will directly or indirectly affect the formation, distribution, and change of precipitation. By carefully analyzing the correlation between climate indices and water vapor contribution, key climate indices that have a significant indicative effect on precipitation in a specific region can be accurately screened. When predicting future precipitation subsequently, constructing a prediction model based on these screened climate indices can effectively improve the model's ability to capture the precipitation change trend and prediction accuracy. The formula for obtaining the key climate index data by performing a correlation analysis between the principal components of the water vapor contribution time series of all Köppen climate regions and the four major oceans and the main climate index data is:[[]]

[0112] (17)

[0113] In the formula,[[]] and respectively represent the water vapor contribution data and the climate index data,[[]] and They respectively represent the average values of water vapor contribution data and climate index data. The correlation relationship between the first three principal components of water vapor contribution in the Köppen climate regions and the climate index is obtained. The water vapor contribution data uses the three principal components obtained by the aforementioned steps, so as to obtain the correlation relationship between the first three principal components PC1, PC2, and PC3 of water vapor contribution in all Köppen climate regions and the four major oceans and the climate index.

[0114] For step S5, a model is established based on the long short-term memory network. Using the terrestrial precipitation time series data, the water vapor contribution time series of all Köppen climate regions, the water vapor contribution time series of the four major oceans, and the key climate index data as the training set and the test set, the model is trained and tested.

[0115] Among them, the terrestrial precipitation time series data is obtained from step S1, while the water vapor contribution time series of the Köppen climate regions and the water vapor contribution time series of the four major oceans are obtained from step S3. The key climate index data is selected through screening. For example, the ENSO index is publicly provided by the Physical Sciences Laboratory of the National Oceanic and Atmospheric Administration (NOAA / PSL) of the United States, the IOD index is obtained from the NOAA / OOPC website, and the NAO index comes from the website of the NOAA Climate Prediction Center (CPC). The long short-term memory network established model (LSTM) takes into account the characteristics of long-term dependencies and multi-step outputs. Using the above data as the input of the model, through calculating the correlation coefficient and performing weighted summation, interpolation operations are carried out on the key climate index data or the main climate index data; making the time scale of the key climate index data or the main climate index data the same as that of the terrestrial precipitation time series data, the water vapor contribution time series of all Köppen climate regions, and the water vapor contribution time series of the four major oceans, thus realizing the effective integration of the climate index data and the other two groups of time series data. As mentioned above, the time series data is divided into four periods: P1 (1941 - 1960), P2 (1961 - 1980), P3 (1981 - 2000), and P4 (2001 - 2023), and is further divided into the training set and the test set based on the principle of equal time division. Training and testing these data sets, the predicted output of the model aims to capture the future time series of terrestrial water vapor contribution based on the Köppen climate regions and ocean water vapor contribution based on the four major oceans.

[0116] Specifically, taking the daily-scale data as the input data as an example, it includes: the global terrestrial precipitation daily-scale time series , where represents the precipitation on the th day; the water vapor sources of the four major oceans , where represents the Vapor contribution vectors from the four major oceans (such as the Pacific Ocean, the Atlantic Ocean, the Indian Ocean, and the Arctic Ocean) in a day; vapor sources in different Köppen climate zones on land , where represents the th day's vapor contribution vector from different Köppen climate zones (such as the tropics, the temperate zone, the frigid zone, etc.). The vapor source is mainly affected by climate indices, namely ENSO, NAO, and IOD mentioned in the above steps. In this embodiment, monthly-scale climate index data is obtained through websites such as NOAA , where represents the climate index (such as ENSO, NAO, IOD, etc.) on the th day. Through interpolation, the time scale of the climate index data is unified to the daily scale to be consistent with other data. The above data is used as the input data of the LSTM model

[0117] Preprocess the data, perform standardization processing on the above input data. The reference formula is as follows

[0118] (18)

[0119] In the formula is the original data is the mean value is the standard deviation

[0120] Define the time step , and construct the standardized input data into time series samples. The reference formula is as follows

[0121]

[0122] In the formula is the input feature at the th time step

[0123] Secondly, construct the LSTM model. LSTM is divided into a cell structure and a network structure. The LSTM cell controls the information flow through a gating mechanism. Its mathematical description is as follows

[0124] Forget gate

[0125] (20)

[0126] Input gate

[0127] (21)

[0128] (22)

[0129] Cell state update

[0130] (23)

[0131] Output gate:

[0132] (24)

[0133] (25)

[0134] In the above formula, is the input at the current time step, is the hidden state at the previous time step, is the cell state at the previous time step, is the weight matrix, is the bias term, is the sigmoid function.

[0135] The network structure of the long short-term memory network LSTM model constructed in this embodiment includes:

[0136] Input layer: Accept the input feature with a time step of ;

[0137] LSTM layer: Contains multiple LSTM cells for extracting time series features;

[0138] Fully connected layer: Maps the output of the LSTM layer to the predicted value;

[0139] Output layer: Outputs the predicted precipitation .

[0140] After the LSTM model is constructed, it is trained. In order to predict future precipitation on a daily basis, this embodiment uses the daily-scale data from 1941 to 2023 to train and test the model, and uses the Adam optimization algorithm to update the model parameters during training to minimize the loss function. The specific formula of the loss function is as follows,

[0141] (26)

[0142] In the formula, is the actual precipitation, is the predicted precipitation, is the number of samples.

[0143] After training the model, step S6 is executed to predict future terrestrial precipitation based on the trained and tested model. The time series data is input into the trained LSTM model, and the model will output the predicted precipitation for future time steps Based on long - time - series historical data, the above - mentioned embodiment uses the LSTM model to summarize historical patterns, further refines the sources of terrestrial precipitation in the training and input data into the contribution sources of terrestrial and oceanic water vapor, and considers the correlation and influence between climate indices and water - vapor contribution sources. It can also reject the addition of future data with uncertainties, effectively improving the prediction accuracy of future precipitation.

[0144] Embodiment 2:

[0145] Refer to the appendix Figure 2 , based on the method for predicting terrestrial precipitation in a changing environment described in Embodiment 1, a system for predicting terrestrial precipitation in a changing environment is also provided. The system includes:

[0146] A data acquisition module, which is used to obtain time - series data of terrestrial precipitation based on the re - analyzed daily - scale terrestrial precipitation data of the ERA5 dataset; based on the WAM - 2layers water - vapor tracking model, obtain time - series of terrestrial water - vapor contribution and time - series of oceanic water - vapor contribution; perform cross - validation based on the time - series data of terrestrial precipitation, the time - series of terrestrial water - vapor contribution, and the time - series of oceanic water - vapor contribution; if the cross - validation is consistent, calculate the time - series of water - vapor contribution for all Köppen climate regions based on the verified time - series of terrestrial water - vapor contribution and the WAM - 2layers water - vapor tracking model respectively; calculate the time - series of water - vapor contribution for the four major oceans based on the verified time - series of oceanic water - vapor contribution and the WAM - 2layers water - vapor tracking model respectively; extract the principal components of the time - series of water - vapor contribution for all Köppen climate regions and the four major oceans respectively; perform correlation analysis on the principal components of the time - series of water - vapor contribution for all Köppen climate regions and the four major oceans and the main climate - index data to obtain key climate - index data;

[0147] A training module, which is used to establish a model based on a long - short - term memory network, and use the time - series data of terrestrial precipitation, the time - series of water - vapor contribution for all Köppen climate regions, the time - series of water - vapor contribution for the four major oceans, and the key climate - index data as the training set and the test set to train and test the model;

[0148] A prediction module, which is used to predict future terrestrial precipitation based on the trained and tested model.

[0149] The specific implementation method of this embodiment is the same as that of Embodiment 1, and will not be elaborated here. For details, refer to the description of Embodiment 1.

[0150] Using the land precipitation prediction method and system in a changing environment of the above embodiments, with the LSTM model, combined with historical precipitation data, water vapor source data from different regions, and climate indices, a high-precision precipitation prediction method is provided. The LSTM model can capture long-term dependencies in time series, effectively handle non-linear relationships and dynamic changes in time series, thereby significantly improving the accuracy and reliability of precipitation prediction. By comprehensively using ERA5 reanalysis data and the WAM-2layers water vapor tracking model, it is possible to more accurately track and analyze the sources and changes of land precipitation. The ERA5 dataset covers a wide range of meteorological elements, providing a detailed data basis for water vapor transport and precipitation processes. At the same time, the WAM-2layers model accurately calculates the proportion of water vapor contributions from different sources through a layered atmospheric structure, thereby achieving a high-precision dynamic assessment of precipitation sources. This method of multi-source data fusion overcomes the limitations of previous single data sources or simple models, fully considers the correlation and influence of climate indices and water vapor contribution sources, and can more comprehensively reveal the complex mechanism of precipitation sources. The above prediction method and system have a high degree of flexibility and adaptability, and can adapt to different data formats and time scales. Whether it is precipitation prediction on a daily, monthly, or annual scale, it can be achieved by adjusting the parameters and input data of the LSTM model. Compared with traditional prediction methods that rely on manual experience or simple statistical models, the present invention has a higher degree of automation, can respond more quickly to precipitation changes and changes in climate indices, provide more timely and accurate information support for decision-makers, help improve the efficiency and effectiveness of water resource management, reduce the influence of human factors on prediction results, and ensure the objectivity and accuracy of predictions.

[0151] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0152] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A method for predicting terrestrial precipitation under changing environments, characterized in that, The method includes: Obtaining time series data of land precipitation based on the reanalysis daily-scale land precipitation data of the ERA5 dataset; Based on the WAM-2layers water vapor tracking model, obtaining the time series of land water vapor contribution and the time series of ocean water vapor contribution; performing cross-validation based on the time series data of land precipitation, the time series of land water vapor contribution, and the time series of ocean water vapor contribution. If the cross-validation is inconsistent, the time series data of land precipitation in the corresponding period is not used for the training and testing of the subsequent model; If the cross-validation is consistent, based on the verified time series of land water vapor contribution and the WAM-2layers water vapor tracking model, calculating the time series of water vapor contribution for all Köppen climate regions respectively; based on the verified time series of ocean water vapor contribution and the WAM-2layers water vapor tracking model, calculating the time series of water vapor contribution for the four major oceans respectively; Extracting the principal components of the time series of water vapor contribution for all Köppen climate regions and the four major oceans respectively; performing a correlation analysis on the principal components of the time series of water vapor contribution for all Köppen climate regions and the four major oceans and the main climate index data, and screening out the key climate index data that has a significant indicative effect on land precipitation; Establishing a model based on a long short-term memory network, using the time series data of land precipitation, the time series of water vapor contribution for all Köppen climate regions, the time series of water vapor contribution for the four major oceans, and the key climate index data as the training set and the test set, and training and testing the model; Predicting future land precipitation based on the trained and tested model.

2. A method for predicting land precipitation in a changing environment according to claim 1, wherein: The cross-validation based on the time series data of land precipitation, the time series of land water vapor contribution, and the time series of ocean water vapor contribution includes: verifying whether the sum of the time series of land water vapor contribution and the time series of ocean water vapor contribution is consistent with the time series data of land precipitation.

3. A method for predicting land precipitation in a changing environment according to claim 1, wherein: The obtaining of the time series data of land precipitation based on the reanalysis daily-scale land precipitation data of the ERA5 dataset includes: Defining a test statistic for the time series data of land precipitation; Obtaining the standardized test statistic based on the test statistic; Obtaining the qualitative change trend of land precipitation based on the standardized test statistic; Performing Sen's Slope test on the time series data of land precipitation to calculate the trend slope and obtaining the quantitative change trend of land precipitation.

4. A method for predicting land precipitation in a changing environment according to claim 1, wherein: The reanalysis daily-scale terrestrial precipitation data based on the ERA5 dataset to obtain the terrestrial precipitation time series data further includes: dividing the reanalysis daily-scale terrestrial precipitation data of the ERA5 dataset from 1941 to 2023 into four stages to calculate the terrestrial precipitation time series data, and the four stages include: P1: 1941 - 1960, P2: 1961 - 1980, P3: 1981 - 2000, P4: 2001 - 2023.

5. A method for predicting terrestrial precipitation in a changing environment according to claim 4, characterized in that: Based on the WAM-2layers water vapor tracking model, obtaining the terrestrial water vapor contribution time series and the ocean water vapor contribution time series, including: tracking and calculating water vapor based on the following formula: Among them, represents the moisture content at the bottom or top of a certain pixel, and are the horizontal and vertical wind speeds, and represent the moisture budget content in the horizontal and vertical directions, is the evaporation rate, is the precipitation, is the residual, is the vertical moisture transport volume between the upper and lower layers.

6. A method for predicting terrestrial precipitation in a changing environment according to claim 5, characterized in that: Based on the verified terrestrial water vapor contribution time series and the WAM-2layers water vapor tracking model, calculating the water vapor contribution time series of all Köppen climate regions respectively; based on the verified ocean water vapor contribution time series and the WAM-2layers water vapor tracking model, calculating the water vapor contribution time series of the four major oceans respectively, including: Based on the Köppen climate region classification vector data, the ocean boundary vector data, and the WAM-2layers water vapor tracking model, tracking and calculating the evaporation amounts of the land precipitation traced to all Köppen climate regions and the four major oceans respectively; Calculating the water vapor contribution ratios of the corresponding Köppen climate regions and oceans respectively based on the evaporation amounts of the land precipitation traced to all Köppen climate regions and the four major oceans and the total terrestrial precipitation; Based on the water vapor contribution ratios of the corresponding Köppen climate regions and oceans, and the terrestrial water vapor contribution time series and the ocean water vapor contribution time series, calculating the water vapor contribution time series of the corresponding Köppen climate regions and oceans respectively according to the four stages.

7. A method for predicting terrestrial precipitation in a changing environment according to claim 1, characterized in that: Respectively extracting the principal components of the water vapor contribution time series of all Köppen climate regions and the four major oceans, including: Based on the water vapor contribution time series of all Köppen climate regions and the four major oceans, extracting the first three principal components with a cumulative variance contribution rate greater than 60%.

8. A method for predicting terrestrial precipitation in a changing environment according to claim 1, characterized in that: The main climate index data includes ENSO index data, IDO index data, and NAO index data; Performing a correlation analysis on the principal components of the water vapor contribution time series of all Köppen climate regions and the four major oceans and the main climate index data, and the calculation formula for obtaining the key climate index data is: In the formula, and respectively represent the water vapor contribution data and the climate index data, and respectively represent the principal component data of the water vapor contribution time series and the average value of the climate index data.

9. A method for predicting terrestrial precipitation in a changing environment according to claim 1, characterized in that: Performing an interpolation operation on the key climate index data or the main climate index data; Make the time scales of the key climate index data or the main climate index data the same as those of the terrestrial precipitation time series data, the time series of water vapor contributions in all Köppen climate regions, and the time series of water vapor contributions in the four major oceans.

10. A land precipitation prediction system under changing environments, characterized in that, The system includes: A data acquisition module, which is used to obtain the terrestrial precipitation time series data based on the reanalysis daily-scale terrestrial precipitation data of the ERA5 dataset; based on the WAM-2layers water vapor tracking model, obtain the time series of terrestrial water vapor contributions and the time series of ocean water vapor contributions; perform cross-validation based on the terrestrial precipitation time series data, the time series of terrestrial water vapor contributions, and the time series of ocean water vapor contributions; if the cross-validation is inconsistent, the terrestrial precipitation time series data for the corresponding period is not used for the subsequent training and testing of the model; if the cross-validation is consistent, based on the verified time series of terrestrial water vapor contributions and the WAM-2layers water vapor tracking model, calculate the time series of water vapor contributions in all Köppen climate regions respectively; based on the verified time series of ocean water vapor contributions and the WAM-2layers water vapor tracking model, calculate the time series of water vapor contributions in the four major oceans respectively; extract the principal components of the time series of water vapor contributions in all Köppen climate regions and the four major oceans respectively; perform a correlation analysis on the principal components of the time series of water vapor contributions in all Köppen climate regions and the four major oceans and the main climate index data, and screen out the key climate index data that has a significant indicative effect on terrestrial precipitation; A training module, which is used to establish a model based on a long short-term memory network, and use the terrestrial precipitation time series data, the time series of water vapor contributions in all Köppen climate regions, the time series of water vapor contributions in the four major oceans, and the key climate index data as the training set and the test set to train and test the model; A prediction module, which is used to predict future terrestrial precipitation based on the trained and tested model.

Images