Method, system, device and medium for predicting sub-seasonal precipitation in high-altitude regions

Abstract

The application provides a high-altitude area sub-seasonal precipitation prediction method, system, device and medium, and aims to construct a high-altitude area sub-seasonal precipitation prediction optimization system based on soil moisture memory. First, the physical mechanism of soil moisture affecting precipitation is revealed in depth and a statistical prediction model is constructed by fusing multi-source observation data. Then, the reproduction ability of the main dynamic model to the mechanism is evaluated and the deviation source is diagnosed. Finally, by fusing the physical statistics and the dynamic model, an optimization framework with high prediction skill and strong interpretability is constructed to significantly improve the sub-seasonal precipitation prediction level of the high-altitude area in the rainy season.

Description

Technical Field

[0001] This invention relates to the field of meteorological forecasting technology, and in particular to a method, system, equipment and medium for predicting subseasonal precipitation in high-altitude areas. Background Technology

[0002] Soil moisture, as a core variable in land-atmosphere coupling, exhibits strong memory (lasting for weeks to months) in high-altitude semi-arid to semi-humid regions, exemplified by the Qinghai-Tibet Plateau, where soil moisture is limited. It serves as a key precursor factor for sub-seasonal climate prediction. Its core mechanism involves using its memory to regulate subsequent surface evaporation or energy balance, thereby altering the thermodynamic state of the atmospheric boundary layer and ultimately driving regional precipitation anomalies. This manifests in two ways: firstly, local influences, such as the coupling relationship between spring soil moisture and subsequent summer precipitation on the Qinghai-Tibet Plateau (positive coupling in the east, negative coupling in the west); secondly, non-local influences, where spring soil moisture on the Qinghai-Tibet Plateau can serve as a key precursor signal, with its above-average levels in winter and spring persisting into summer. This non-adiabatic heating alters atmospheric circulation, leading to increased summer precipitation in North and South my country, and decreased precipitation in the Yangtze River and Huai River basins.

[0003] However, current research still has the following key shortcomings: First, existing results mostly focus on the regulatory effects of factors such as sea surface temperature on high-altitude precipitation, lacking in-depth analysis of the physical mechanisms (especially freeze-thaw processes) by which soil moisture memory influences subseasonal precipitation; second, mainstream dynamic models (such as the International Subseasonal Prediction System) suffer from insufficient parameterization of permafrost hydrothermal processes, leading to excessively rapid decay of the land surface initialization signal, making it difficult to effectively maintain the soil moisture memory signal within the model. Therefore, an effective method for predicting subseasonal precipitation in high-altitude areas is urgently needed to address these issues. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed to provide a method, system, device and medium for predicting subseasonal precipitation in high-altitude areas that overcomes or at least partially solves the above problems.

[0005] To achieve the above and other related objectives, the present invention provides a method for predicting subseasonal precipitation in high-altitude areas, the method comprising:

[0006] By integrating site observations, satellite fusion products, and reanalysis data, multi-source observation data of the Qinghai-Tibet Plateau is constructed. The integrated empirical mode decomposition method is used to extract the sub-seasonal precipitation signal, and the dominant mode and key areas of the influence of previous soil moisture on precipitation are identified through singular value decomposition.

[0007] A statistical prediction model for subseasonal precipitation considering soil moisture memory was constructed based on the partial least squares path method. The contribution of previous soil moisture to subseasonal precipitation was quantified using the Copula function. The physical mechanism was revealed by combining the water vapor budget and surface energy balance equations.

[0008] Based on the aforementioned key physical processes, the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation are evaluated using the multi-source observation data, the sources of their bias are analyzed, and the effectiveness of soil moisture in key areas as a predictability source in each dynamic model is identified.

[0009] Error correction was performed by constructing a dynamic multi-mode ensemble with dynamic weights using a Bayesian model, and feature selection was performed based on Lasso regression. By integrating physical statistics and dynamic models, a physically interpretable optimization model for predicting subseasonal precipitation at high altitudes was constructed.

[0010] Optionally, the step of extracting sub-seasonal precipitation signals using integrated empirical mode decomposition and identifying the dominant modes and key areas of the influence of previous soil moisture on precipitation through singular value decomposition includes:

[0011] The multi-source observation data were cleaned and quality controlled to obtain daily precipitation data and previous soil moisture data;

[0012] The daily precipitation data were decomposed using an integrated empirical mode decomposition method to extract the next season's precipitation signal;

[0013] The dominant mode and key regions between the previous soil moisture data and the next season precipitation signal were identified by the singular value decomposition method, and the precipitation time series and soil moisture time series corresponding to the key regions were extracted.

[0014] Optionally, the subseasonal precipitation statistical prediction model based on the partial least squares path method, which considers soil moisture memory, and uses the Copula function to quantify the contribution of previous soil moisture to subseasonal precipitation, reveals its physical mechanism by combining the water vapor budget and surface energy balance equations, including:

[0015] Based on the dominant modes identified by singular value decomposition and the corresponding precipitation and soil moisture time series, a statistical prediction model for sub-seasonal precipitation considering soil moisture memory is constructed using the partial least squares path method.

[0016] Using the Copula function, the contribution of early soil moisture in key areas identified by singular value decomposition to the changes in sub-seasonal precipitation is quantified, and the nonlinear dependence between the two is revealed.

[0017] For key areas identified by singular value decomposition, we use water vapor budget and surface energy balance equations to explore the physical mechanism by which soil moisture affects subseasonal precipitation and clarify its role in water vapor transport and non-adiabatic heating processes.

[0018] Optionally, the process of evaluating the predictive skill of mainstream dynamic models for high-altitude subseasonal precipitation using the multi-source observation data based on the aforementioned key physical processes, analyzing the sources of bias, and identifying the effectiveness of soil moisture in key areas as a predictability source in various dynamic models includes:

[0019] The multi-source observation data is compared with the subseasonal precipitation prediction data of the mainstream dynamic model to calculate the corresponding evaluation index; wherein the evaluation index includes mean absolute deviation, root mean square error and time correlation coefficient.

[0020] Based on the aforementioned evaluation indicators, the predictive skills of each dynamic mode are quantitatively evaluated and ranked according to their merits to obtain the evaluation results.

[0021] Based on the assessment results, we will conduct an in-depth analysis of the physical sources of prediction bias in each dynamic model, and examine whether each dynamic model can reproduce the physical mechanism by which soil moisture affects sub-seasonal precipitation, so as to explore the effectiveness of soil moisture in key areas as a predictability source in each dynamic model.

[0022] Optionally, the method employs a Bayesian model to construct a dynamic multi-model ensemble with average weights for error correction, and uses Lasso regression to screen key features, integrating physical statistics and dynamic models to form a physically interpretable optimized model for predicting subseasonal precipitation at high altitudes, including:

[0023] Using the soil moisture time series identified by singular value decomposition and the subseasonal precipitation prediction data of the mainstream dynamic model as the initial set of independent variables, and the subseasonal precipitation signal extracted by integrated empirical mode decomposition as the dependent variable;

[0024] Based on the previous soil moisture conditions, the posterior probability weights of each dynamic mode are calculated using the Bayesian model averaging method. The multiple dynamic modes are then weighted and averaged according to the posterior probability weights to generate a weighted set prediction result that has been constrained by soil moisture and corrected for errors.

[0025] The weighted set prediction results and the soil moisture time series are used to construct a new set of independent variables. The new set of independent variables and the dependent variable are then input into a Lasso regression model. The Lasso regression model is used for feature selection and model fusion to obtain an optimized high-altitude subseasonal precipitation prediction model with physical interpretability.

[0026] Optionally, after the step of selecting key features based on Lasso regression, integrating physical statistics and dynamic models to form a physically interpretable optimization model for high-altitude subseasonal precipitation prediction, the method further includes:

[0027] The latest multi-source observation data of key areas are collected, and the high-altitude subseasonal precipitation prediction optimization model is used to predict the latest observation data, generating and outputting the corresponding subseasonal precipitation prediction results.

[0028] Secondly, the present invention also provides a subseasonal precipitation prediction system for high-altitude areas, the system comprising:

[0029] The integration module is used to integrate site observations, satellite fusion products and reanalysis data to construct multi-source observation data of the Tibetan Plateau, and to extract the sub-seasonal precipitation signal using integrated empirical mode decomposition method. It also identifies the dominant mode and key areas of the influence of previous soil moisture on precipitation through singular value decomposition.

[0030] The quantification module is used to construct a statistical prediction model for subseasonal precipitation that considers soil moisture memory based on the partial least squares path method, and to quantify the contribution of previous soil moisture to subseasonal precipitation using the Copula function, and to reveal its physical mechanism by combining the water vapor budget and surface energy balance equations.

[0031] The analysis module is used to evaluate the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation based on the aforementioned key physical processes and the multi-source observation data, analyze the sources of deviation, and identify the effectiveness of soil moisture in key areas as a source of predictability in each dynamic model.

[0032] The fusion module is used to construct a dynamic multi-mode set with dynamic weights using a Bayesian model for error correction, and to screen key features based on Lasso regression, and to fuse physical statistics and dynamic models to form a physically interpretable high-altitude subseasonal precipitation prediction optimization model.

[0033] Thirdly, the present invention provides an electronic device comprising: a memory and a processor; the memory for storing a computer program; and the processor for executing the computer program stored in the memory to cause the electronic device to perform the steps of the high-altitude subseason precipitation prediction method as described above.

[0034] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by an electronic device, implements the steps of the high-altitude subseason precipitation prediction method as described above.

[0035] Fifthly, the present invention provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer implements the steps of the high-altitude subseason precipitation prediction method as described above.

[0036] The above-described one or more technical solutions provided by this invention can have the following advantages or at least achieve the following technical effects:

[0037] This invention aims to construct an optimization system for predicting subseasonal precipitation in high-altitude areas based on soil moisture memory. First, by fusing multi-source observation data, the physical mechanism by which soil moisture affects precipitation is revealed and a statistical prediction model is constructed. Then, the reproducibility of mainstream dynamic models of this mechanism is evaluated and the sources of their bias are diagnosed. Finally, by fusing physical statistics and dynamic models, an optimization framework with both high predictive skill and strong interpretability is constructed to significantly improve the prediction level of subseasonal precipitation in high-altitude areas during the rainy season. Attached Figure Description

[0038] Figure 1 The diagram shows a flowchart of a method for predicting subseasonal precipitation in high-altitude areas according to an embodiment of the present invention.

[0039] Figure 2 This is a schematic diagram illustrating the technical route for predicting subseasonal precipitation in high-altitude areas according to an embodiment of the present invention.

[0040] Figure 3 This is a schematic diagram of the functional modules of a high-altitude subseasonal precipitation prediction system according to an embodiment of the present invention.

[0041] Figure 4 The diagram shown is a schematic representation of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0042] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0043] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0044] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0045] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0046] Unless otherwise stated, the term "multiple" means two or more.

[0047] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0048] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0049] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings.

[0050] Please see Figure 1 An embodiment of the present invention provides a method for predicting subseasonal precipitation in high-altitude areas, the method comprising the following steps S10-S40:

[0051] Step S10: Integrate site observations, satellite fusion products, and reanalysis data to construct multi-source observation data of the Qinghai-Tibet Plateau. Use integrated empirical mode decomposition to extract sub-seasonal precipitation signals and identify the dominant modes and key areas of precipitation influence by previous soil moisture through singular value decomposition.

[0052] Among them, multi-source observation data is used to represent a comprehensive dataset constructed by collaboratively analyzing, complementing, and integrating three types of data: national meteorological densified station observations, satellite fusion products, and reanalysis data such as ERA5, MERRA2, and JRA55.

[0053] Subseasonal precipitation signals are used to represent precipitation change trends or modes that are hidden in the random fluctuations of daily precipitation, with a time scale of 15-60 days, and are persistent and predictable.

[0054] Dominant modes are used to represent the modes that have the main influence on precipitation variation, identified by the singular value decomposition (SVD) method. These modes typically reflect the periodic characteristics or main influencing factors of precipitation variation.

[0055] Key regions are identified using singular value decomposition (SVD) to determine the geographical areas that have the most significant impact on precipitation during the preceding soil moisture influence process. These regions are likely where the correlation between soil moisture and precipitation is strongest, or where changes in soil moisture have the greatest impact on precipitation.

[0056] As an example, in studies of the Tibetan Plateau, certain specific soil moisture monitoring stations or areas have been identified as key regions, where changes in soil moisture have a significant impact on precipitation changes.

[0057] Soil moisture time series is used to represent a set of soil moisture data recorded continuously at fixed time intervals (such as daily or weekly) in a key area.

[0058] Precipitation time series is used to represent a set of precipitation data continuously recorded at fixed time intervals (such as hourly or daily) in a key area (i.e., the time coefficient obtained from SVD analysis).

[0059] In practical implementation, multi-source observation data of the Tibetan Plateau can be constructed by integrating observations from national meteorological densification stations (ground-based measured precipitation), satellite fusion products (such as IMERG precipitation and SMP soil moisture), and reanalysis data from ERA5, MERRA2, and JRA-55. Then, through data cleaning and quality control, daily precipitation data is generated, and the Empirical Integrated Mode Decomposition (EEMD) algorithm is applied to separate the sub-seasonal precipitation signal from the daily precipitation data. Subsequently, the singular value decomposition method is used to analyze the coupled dominant mode and its key regions between anterior soil moisture and sub-seasonal precipitation. Through mode reconstruction, the precipitation and soil moisture time series corresponding to the dominant mode are extracted for subsequent physical mechanism analysis to elucidate the regulatory role of anterior soil moisture on sub-seasonal precipitation in high-altitude areas, providing a foundation for the construction of subsequent prediction models.

[0060] Step S20: Construct a statistical prediction model for subseasonal precipitation that considers soil moisture memory based on the partial least squares path method, and use the Copula function to quantify the contribution of previous soil moisture to subseasonal precipitation. Combine the water vapor budget and surface energy balance equations to reveal its physical mechanism.

[0061] Among them, the sub-seasonal precipitation statistical prediction model is a mathematical model used to predict the precipitation trend in the next 15-60 days based on the statistical relationships existing in multi-source observation data (i.e., historical observation data).

[0062] In practical implementation, based on the dominant modes obtained from singular value analysis and the corresponding precipitation and soil moisture time series, a statistical prediction model for subseasonal precipitation considering soil moisture memory can be constructed using the partial least squares path method (PLS-PM model). Then, the contribution of previous soil moisture to subseasonal precipitation can be quantified using the Copula function, analyzing the contribution of soil moisture to high-altitude precipitation. Subsequently, by combining the water vapor budget and surface energy balance equations, the physical mechanism by which previous soil moisture affects subseasonal precipitation can be revealed, clarifying its impact on water vapor transport characteristics and non-adiabatic heating mechanisms. This reveals the regulatory mechanism and physical processes of previous soil moisture on high-altitude subseasonal precipitation.

[0063] Step S30: Based on the above key physical processes, use the multi-source observation data to evaluate the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation, analyze the sources of their biases, and identify the effectiveness of soil moisture in key areas as a predictability source in each dynamic model.

[0064] Among them, mainstream dynamical models are used to represent numerical weather / climate prediction systems developed by authoritative international meteorological agencies. These include dynamical models such as BoM (Australian Bureau of Meteorology), CMA (China Meteorological Administration), ECMWF (European Centre for Medium-Range Weather Forecasts), and NCEP (National Center for Environmental Prediction).

[0065] Forecasting techniques, used to represent the statistical consistency between dynamic model outputs and observed values ​​(i.e., multi-source observation data), need to be quantified and evaluated using objective indicators. Examples include Bias (mean absolute deviation), RMSE (root mean square error), and TCC (time correlation coefficient).

[0066] The source of deviation is used to indicate the root cause of the systematic difference between the prediction results and the observed values ​​of the dynamic model, and needs to be identified through error diagnosis.

[0067] Predictability sources are used to represent key factors affecting the ability to predict subseasonal precipitation at high altitudes, and need to be identified through analysis of physical mechanisms.

[0068] In practical implementation, after clarifying the key areas and physical processes by which soil moisture affects precipitation in the high-altitude subseason, we can further evaluate the predictions and sources of bias of multi-source dynamic models (such as BoM, CMA, ECMWF, etc.) for high-altitude subseason precipitation. We can compare the multi-source observation data with the prediction results of mainstream dynamic models such as BoM, CMA, and ECMWF, using objective indicators such as bias, RMSE, and TCC to quantitatively evaluate the prediction skills of each dynamic model. We can then rank the models based on the aforementioned bias, RMSE, and TCC indicators. Combining the ranking results with the prediction skill evaluation, we can analyze the sources of prediction bias in each dynamic model, focusing on whether the dynamic models can reproduce the identified soil moisture-precipitation physical processes (such as water vapor transport and non-adiabatic heating mechanisms), and determining whether soil moisture in key areas can serve as an effective predictability source in each dynamic model. Through the above analysis, we can comprehensively evaluate the prediction performance of each dynamic model, clarify the physical mechanisms of bias, and reveal the predictability contribution of soil moisture in the dynamic models.

[0069] Step S40: A dynamic multi-mode set with dynamic weights is constructed using a Bayesian model for error correction. Key features are then selected based on Lasso regression. By integrating physical statistics and dynamic models, a physically interpretable optimization model for predicting subseasonal precipitation at high altitudes is formed.

[0070] Among them, the high-altitude subseasonal precipitation prediction optimization model is used to represent a collaborative prediction framework that integrates physical statistical methods and multi-source dynamic models (such as BoM, CMA, ECMWF, NCEP, etc.). It aims to improve the accuracy and physical interpretability of subseasonal precipitation predictions for high-altitude areas (such as the Three-River-Source Region and the Qinghai-Tibet Plateau) in the next 15-60 days through dynamic weight adjustment, feature selection, and physical constraints.

[0071] In practical implementation, the Bayesian Model Averaging (BMA) method can be adopted, with previous soil moisture as a constraint, to assign dynamic weights to various mainstream dynamic models (such as BoM, CMA, ECMWF, NCEP, etc.), and generate optimized dynamic multi-model ensemble prediction results through error correction. Then, using the optimized dynamic multi-model ensemble prediction results and the soil moisture time series of key areas as independent variables, and the sub-seasonal precipitation signal as the dependent variable, Lasso regression is used to screen key influencing factors (including soil moisture), construct a statistical model based on physical mechanisms (such as water vapor budget and surface energy balance), and synergistically couple it with the dynamic model prediction results. Finally, a physically interpretable high-altitude sub-seasonal precipitation prediction optimization model is formed, realizing the dual optimization of statistical correction of dynamic models and physical mechanism constraints.

[0072] Further, please refer to Figure 2 , Figure 2This demonstrates a technical approach for predicting subseasonal precipitation in high-altitude areas; in one embodiment, step S10 may include sub-steps S101~S103:

[0073] Sub-step S101 involves cleaning and quality control of the multi-source observation data to obtain daily precipitation data and previous soil moisture data;

[0074] Sub-step S102: The daily precipitation data is decomposed using the integrated empirical mode decomposition method to extract the next season precipitation signal;

[0075] Sub-step S103 involves identifying the dominant mode and key regions between the previous soil moisture data and the next season precipitation signal using the singular value decomposition method, and extracting the precipitation time series and soil moisture time series corresponding to the key regions.

[0076] Daily precipitation data represents the daily precipitation record, indicating the total amount of liquid or solid precipitation accumulated within a given day. It is input from multi-source observation data and, after cleaning and quality control, is used to extract the next season's precipitation signal (e.g., through integrated empirical mode decomposition (EEMD)).

[0077] Pre-hospital soil moisture data represents soil moisture data for a specific period (e.g., 1-3 months prior) before the target prediction period (e.g., next season precipitation). This data characterizes the lagged impact of surface moisture conditions on subsequent precipitation. As a key predictor, it is correlated with precipitation signals through singular value decomposition (SVD) analysis to identify key regions and time series influencing precipitation. Finally, it is input into a Lasso regression model for feature selection and prediction optimization.

[0078] Subseasonal precipitation signals, used to represent fluctuations on a 15-60 day scale in precipitation sequences, are extracted using the EEMD method and used to analyze short-term climate mechanisms and optimize prediction models.

[0079] Please see Figure 2 This paper presents a technical roadmap for predicting subseasonal precipitation in high-altitude areas. The roadmap demonstrates a complete research process, from observational impact analysis (identification of dominant modes and key regions, and construction of statistical models) to dynamic model evaluation (assessment of prediction skills and sources of bias) and then to model optimization (optimization using Bayesian averaging and Lasso regression). The aim is to improve the prediction skills for subseasonal precipitation in the Sanjiangyuan region. Observational impact analysis primarily focuses on the regulatory role of soil moisture in subseasonal precipitation in the Sanjiangyuan region, specifically including three core stages: identification of dominant modes and key regions, construction of statistical models, and analysis of key physical processes by which soil moisture affects precipitation. Through observational impact analysis, the crucial role of soil moisture in subseasonal precipitation in the Sanjiangyuan region can be clarified, providing a theoretical foundation and data support for subsequent dynamic model evaluation and model optimization.

[0080] Among them, the dominant mode and key area identification: the integrated empirical mode decomposition (EEMD) method is used to decompose the soil moisture and precipitation observation data at multiple time scales, effectively separating the sub-seasonal precipitation signals of different frequencies; then, singular value decomposition (SVD) is used to extract the dominant mode of the soil moisture-precipitation coupling system, and identify the key areas affecting the sub-seasonal precipitation changes in the Three-River-Source region and the corresponding soil moisture and precipitation time series.

[0081] In practical implementation, after constructing multi-source observation data, data cleaning (such as removing outliers and filling missing values) and quality control (consistency verification) can be performed on the multi-source observation data to obtain daily precipitation data. Simultaneously, the previous soil moisture data (such as the average soil moisture of the previous 15-60 days) in the reanalysis data can be extracted. Then, the integrated empirical mode decomposition method (EEMD) is applied to decompose the daily precipitation data into intrinsic mode functions at different time scales, and the sub-seasonal precipitation signal of 15-60 days is screened and reconstructed. Then, singular value decomposition (SVD) is performed on the previous soil moisture data and the sub-seasonal precipitation signal to identify the dominant mode of spatial coupling between the two. Based on the SVD analysis results, the key areas of soil moisture influence on sub-seasonal precipitation are located. Finally, for the key areas, the corresponding precipitation time series and soil moisture time series are extracted for subsequent modeling analysis, thereby revealing the regulatory relationship between previous soil moisture and sub-seasonal precipitation at high altitudes.

[0082] Further, please refer to Figure 2 In one embodiment, step S20 may include sub-steps S201-S203:

[0083] Sub-step S201: Based on the dominant modes identified by singular value decomposition and the corresponding precipitation time series and soil moisture time series, a statistical prediction model for next-season precipitation considering soil moisture memory is constructed using the partial least squares path method.

[0084] Sub-step S202 uses the Copula function to quantify the contribution of early soil moisture in key areas identified by singular value decomposition to the changes in precipitation in the next season, and reveals the nonlinear dependence between the two.

[0085] Sub-step S203 involves using the water vapor budget and surface energy balance equations to explore the physical mechanism by which soil moisture affects subseasonal precipitation in key areas identified by singular value decomposition, and to clarify its role in water vapor transport and non-adiabatic heating processes.

[0086] Please see Figure 2The presentation showcases the observational impact analysis (statistical model construction and analysis of key physical processes by which soil moisture affects precipitation). Specifically, the statistical model construction involves using a partial least squares path model (PLS-PM) to construct a statistical prediction model of the impact of soil moisture on precipitation, based on identified key regions and dominant modes. This quantifies the causal relationship and impact pathways between the two, establishing physically meaningful statistical prediction equations. The key physical process analysis combines various analytical tools, including Copula functions, water vapor budget equations, and surface energy balance equations, to deeply reveal the key physical mechanisms by which soil moisture affects precipitation, validating the scientific basis for soil moisture as a source of predictability.

[0087] In practical implementation, based on the dominant modes of precipitation and soil moisture identified by SVD analysis and their corresponding time series, a statistical prediction model for subseasonal precipitation considering soil moisture memory can be constructed using the partial least squares path method (PLS-PM). Then, the contribution of anterior soil moisture in key areas to subseasonal precipitation changes is quantitatively analyzed using the Copula function, revealing the nonlinear dependence between the two. Subsequently, for key areas identified by SVD, the physical mechanisms by which soil moisture affects subseasonal precipitation are analyzed using the water vapor budget equation and the surface energy balance equation, clarifying its role in regulating water vapor transport paths and non-adiabatic heating effects. Finally, it elucidates how anterior soil moisture affects subseasonal precipitation in high-altitude areas through land-atmosphere interactions, forming a complete analytical framework that combines statistical prediction with physical mechanisms.

[0088] Furthermore, in one embodiment, after step S40, the method may further include step S50:

[0089] Step S50: Collect the latest multi-source observation data of key areas, and use the high-altitude sub-seasonal precipitation prediction optimization model to predict the latest observation data, generate and output the corresponding sub-seasonal precipitation prediction results.

[0090] Among them, the latest multi-source observation data is used to represent the diverse datasets collected in real time or near real time during the prediction phase and used as model inputs; it includes, but is not limited to: station precipitation data, soil moisture data, satellite products and reanalysis data (ERA5, MERRA2, etc.).

[0091] Sub-seasonal precipitation forecast results are used to represent the precipitation forecast information for the next 15-60 days output by the high-altitude sub-seasonal precipitation forecast optimization model; including but not limited to: precipitation probability or precipitation amount expressed in specific values ​​or levels (such as "light rain" or "heavy rain") (such as "probability of precipitation in the second week"), precipitation anomaly percentage based on ensemble forecast (such as "precipitation is 20% higher than normal, confidence level 75%), marking of high-risk areas, precipitation evolution trends by week (such as weeks 1-4) or by ten-day period (such as "precipitation intensity reaches its peak in the second week"), etc.

[0092] In practical implementation, after obtaining a physically interpretable high-altitude subseasonal precipitation prediction optimization model, the latest multi-source observation data (including station precipitation, satellite-fused soil moisture, reanalysis data, etc.) of key areas are collected in real time; then, the high-altitude subseasonal precipitation prediction optimization model is used to comprehensively predict the latest multi-source observation data, generate and output the subseasonal precipitation prediction results corresponding to the latest observation data, and realize dynamic prediction application based on physical mechanisms and multi-source information.

[0093] In this embodiment, by fusing multi-source data, the physical mechanism by which soil moisture affects precipitation is revealed and a statistical prediction model is constructed. Then, the reproducibility of the mainstream dynamic model of this mechanism is evaluated and the source of its bias is diagnosed. Finally, by fusing physical statistics and dynamic models, an optimization framework with both high prediction skill and strong interpretability is constructed to significantly improve the prediction level of subseasonal precipitation in high-altitude areas during the rainy season.

[0094] Based on the foregoing embodiments, a second embodiment of the present invention for predicting subseasonal precipitation in high-altitude areas is proposed. Please refer to [link to embodiment]. Figure 2 In this embodiment, step S30 may include the following sub-steps S301~S303:

[0095] Sub-step S301 involves comparing the multi-source observation data with the sub-seasonal precipitation prediction data from the mainstream dynamic model to calculate the corresponding evaluation indicators; wherein the evaluation indicators include mean absolute deviation, root mean square error, and time correlation coefficient.

[0096] Among them, the evaluation indicators are a series of mathematical tools used to quantitatively measure the degree of agreement between sub-seasonal precipitation forecast data from mainstream dynamic models and multi-source observation data; these include, but are not limited to, mean absolute deviation (Bias), root mean square error (RMSE), and time correlation coefficient (TCC).

[0097] Please see Figure 2The dynamic model evaluation presented (assessing prediction techniques and sources of bias) aims to systematically evaluate the predictive performance of multiple mainstream dynamic models for subseasonal precipitation in the Sanjiangyuan region and diagnose the main sources of prediction bias, providing a scientific basis for subsequent model optimization. Specifically, a multi-source dynamic model ensemble evaluation method is used, employing indicators such as mean absolute bias (Bias), root mean square error (RMSE), and time correlation coefficient (TCC) to analyze the sources of predictability bias and determine whether soil moisture is a predictable source.

[0098] In practice, evaluation indicators such as mean absolute deviation (Bias), root mean square error (RMSE), and time correlation coefficient (TCC) can be calculated based on multi-source observation data and subseasonal precipitation forecast data from mainstream international dynamic models (such as BoM, CMA, ECMWF, HMCR, NCEP, etc.).

[0099] Sub-step S302: Based on the evaluation indicators, quantitatively evaluate the prediction skills of each dynamic mode and rank them according to their merits to obtain the evaluation results.

[0100] The evaluation results are used to represent a comprehensive judgment on the predictive performance of each dynamic model by quantitative analysis and horizontal comparison of the predictive data of the systematic comparative observation data and the sub-seasonal precipitation prediction data of each dynamic model, based on objective indicators such as mean absolute deviation (Bias), root mean square error (RMSE), and time correlation coefficient (TCC).

[0101] In practical implementation, the evaluation indicators obtained from the aforementioned calculations can be used to quantitatively evaluate the prediction skills of each dynamic mode and rank them according to their merits, thus obtaining the ranked evaluation results.

[0102] Sub-step S303 involves combining the evaluation results to conduct an in-depth analysis of the physical sources of prediction bias in each dynamic model, and to examine whether each dynamic model can reproduce the physical mechanism by which soil moisture affects the next season's precipitation, in order to explore the effectiveness of soil moisture in key areas as a predictability source in each dynamic model.

[0103] Among them, assessing the effectiveness of predictability sources involves verifying whether the anterior soil moisture in key areas serves as an effective predictability source in various dynamic models, i.e., whether the dynamic models can correctly predict subsequent precipitation changes based on the initial conditions or memory of soil moisture.

[0104] In practical implementation, the physical sources of prediction bias in each dynamic model can be analyzed in depth by combining the evaluation results after the sorting (such as insufficient precipitation simulation capability and lack of soil moisture-precipitation coupling mechanism). The focus is on examining the reproducibility of each dynamic model with the key physical mechanisms (water vapor transport regulation and non-adiabatic heating effect) identified in the aforementioned embodiments (such as steps S10~S20, sub-steps S101~S103, and sub-steps S201~S203), and analyzing the effectiveness of the early soil moisture in key areas as a predictability source in each dynamic model. This clarifies the mechanism of prediction bias in dynamic models and identifies the direction for improving high-altitude subseason precipitation prediction.

[0105] In this embodiment, by comparing the multi-source observation data with the subseasonal precipitation prediction data of the mainstream dynamic models, corresponding evaluation indicators are calculated. These evaluation indicators include mean absolute deviation, root mean square error, and time correlation coefficient. Based on these indicators, the prediction skills of each dynamic model are quantitatively evaluated and ranked to obtain evaluation results. Combining these evaluation results, the physical sources of prediction bias in each dynamic model are analyzed in depth, and the ability of each dynamic model to reproduce the physical mechanism by which soil moisture affects subseasonal precipitation is examined. This aims to explore the effectiveness of soil moisture in key areas as a predictability source in each dynamic model, thereby clarifying the mechanism of prediction bias in dynamic models and identifying directions for improvement in high-altitude subseasonal precipitation prediction.

[0106] Based on the foregoing embodiments, a third embodiment of the present invention for predicting subseasonal precipitation in high-altitude areas is proposed. Please refer to [link to embodiment]. Figure 2 In this embodiment, step S40 may include sub-steps S401 to S403:

[0107] Sub-step S401 uses the soil moisture time series identified by singular value decomposition and the subseasonal precipitation prediction data of the mainstream dynamic model as the initial set of independent variables, and the subseasonal precipitation signal extracted by integrated empirical mode decomposition as the dependent variable.

[0108] The initial set of independent variables represents the set of input variables used for prediction in the construction of the model (i.e., the high-altitude subseasonal precipitation prediction optimization model), and includes two types of data: soil moisture time series identified by singular value decomposition and subseasonal precipitation prediction data from the mainstream dynamic model.

[0109] The dependent variable is used to represent the target variable that the model (i.e., the high-altitude subseasonal precipitation prediction optimization model) needs to predict, which is the subseasonal precipitation signal extracted by the integrated empirical mode decomposition method.

[0110] Please see Figure 2The showcased optimization model construction (dynamic multi-model ensemble optimization, physical mechanism fusion, and final modeling) aims to build an optimized high-altitude subseason precipitation prediction model that combines physical interpretability and prediction accuracy, thereby improving prediction skill and physical interpretability. Specifically, "dynamic multi-model ensemble optimization" employs the Bayesian model averaging (BMA) method to dynamically weight and fuse the prediction results of multiple mainstream dynamic models, generating a dynamic ensemble prediction result with smaller errors and higher prediction skill, effectively correcting the systematic biases of the dynamic models. "Physical mechanism fusion and final modeling" uses the "dynamic ensemble prediction result" and the "key soil moisture signals" identified by SVD as independent variables, inputting them into a Lasso regression model; Lasso is then used for feature selection and model fusion, ultimately outputting an optimized prediction model (regression equation).

[0111] In practical implementation, the initial set of independent variables can be the soil moisture time series of key areas identified by SVD and the subseasonal precipitation prediction data of the mainstream dynamic model, and the subseasonal precipitation signal extracted by EEMD can be used as the dependent variable; thus laying the foundation for building a high-altitude subseasonal precipitation prediction optimization model that has both physical interpretability and prediction accuracy.

[0112] Sub-step S402: Using the previous soil moisture conditions as a condition, the posterior probability weight of each dynamic mode is calculated using the Bayesian model averaging method. Based on the posterior probability weight, multiple dynamic modes are weighted and averaged to generate a weighted set prediction result that has been constrained by soil moisture and corrected for errors.

[0113] The weighted ensemble prediction result is used to represent the integrated prediction result generated by dynamically weighting the sub-seasonal precipitation predictions of multiple mainstream dynamic models (such as ECMWF, CMA, BoM, etc.) using the Bayesian Model Averaging (BMA) method. The weights are determined by the posterior probability of each dynamic model, and the calculation requires the previous soil moisture status as a constraint to reflect the influence of soil moisture on the model's predictive ability.

[0114] In practice, the soil moisture conditions can be used as a constraint. The Bayesian model averaging (BMA) method can be used to calculate the predictive performance of each dynamic model under the influence of soil moisture. Then, the posterior probability weights can be calculated and weighted averaged to generate a weighted set prediction result that is constrained by soil moisture and corrected for systematic errors.

[0115] Sub-step S403 involves constructing a new set of independent variables by combining the weighted set prediction results with the soil moisture time series, and inputting the new set of independent variables and the dependent variable into a Lasso regression model. The Lasso regression model is then used for feature selection and model fusion to obtain a physically interpretable optimized model for predicting subseasonal precipitation at high altitudes.

[0116] The new set of independent variables represents the extended set of variables constructed based on the weighted set prediction results and the soil moisture time series of key areas identified by SVD, which serves as the input to the Lasso regression model.

[0117] In practical implementation, the weighted ensemble prediction results generated by BMA can be combined with soil moisture time series to construct a new set of independent variables, which is then input into a Lasso regression model for feature selection and model fusion. Leveraging the sparsity property of Lasso, combinations of predictive factors (including weighted ensemble prediction results and key regional soil moisture) that significantly contribute to high-altitude subseasonal precipitation signals are selected. The contribution weights of each factor are then quantified based on regression coefficients, ultimately forming an optimized model that combines physical interpretability and predictive accuracy. This model achieves statistical correction of the dynamic model through dynamic weight integration of BMA, while explicitly coupling soil moisture variables to quantify their physical constraints on precipitation, achieving synergistic optimization of statistical methods and physical mechanisms. Thus, based on the synergistic effect of soil moisture and dynamic models, a high-altitude subseasonal precipitation prediction optimization model with both physical interpretability and predictive accuracy is constructed.

[0118] In this embodiment, key soil moisture signals identified by SVD, dynamic model predictions, and statistical learning methods are used. Bayesian model averaging (BMA) is employed to dynamically weight and fuse the prediction results of multiple mainstream dynamic models, generating a weighted set prediction result with smaller errors and higher prediction skill. Then, the "weighted set prediction result" and the "key soil moisture signals" identified by SVD are used as input, and a Lasso regression model is used for feature selection and model fusion, outputting a regression equation that is the final optimized prediction model. By using BMA to statistically correct the systematic errors of the dynamic models and using Lasso to statistically quantify and integrate the key physical mechanism of soil moisture into the prediction system, a physically interpretable subseasonal precipitation prediction solution suitable for complex high-altitude environments is formed.

[0119] Based on the same inventive concept, the fourth embodiment of this invention also provides a high-altitude subseasonal precipitation prediction system corresponding to the high-altitude subseasonal precipitation prediction method of the foregoing embodiments. Since the principle of the system in the fourth embodiment of this invention is similar to the high-altitude subseasonal precipitation prediction method of the foregoing embodiments, the implementation of the system can refer to the implementation of the method; repeated details will not be elaborated further. Please refer to... Figure 3 The present invention provides a high-altitude subseasonal precipitation prediction system, the system comprising:

[0120] The integration module 10 is used to integrate site observations, satellite fusion products and reanalysis data to construct multi-source observation data of the Tibetan Plateau, and to extract sub-seasonal precipitation signals using integrated empirical mode decomposition method, and to identify the dominant mode and key areas of the influence of previous soil moisture on precipitation through singular value decomposition.

[0121] Quantization module 20 is used to construct a statistical prediction model for subseasonal precipitation based on the partial least squares path method, taking into account the memory of soil moisture, and to use the Copula function to quantify the contribution of previous soil moisture to subseasonal precipitation, and to reveal its physical mechanism by combining the water vapor budget and surface energy balance equations.

[0122] Analysis module 30 is used to evaluate the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation based on the above-mentioned key physical processes and the multi-source observation data, analyze the sources of deviation, and identify the effectiveness of soil moisture in key areas as a source of predictability in each dynamic model.

[0123] The fusion module 40 is used to construct a dynamic multi-mode set with dynamic weights using a Bayesian model for error correction, and to perform feature selection based on Lasso regression. It integrates physical statistics and dynamic models to construct a physically interpretable high-altitude subseason precipitation prediction optimization model.

[0124] In addition, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for predicting subseasonal precipitation in high-altitude areas.

[0125] Figure 4 This is a schematic block diagram of the electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device includes at least one processor 401, a memory 402, at least one network interface 403, and a user interface 405. The various components in the electronic device are coupled together via a bus system 404. It is understood that the bus system 404 is used to implement communication between these components. In addition to a data bus, the bus system 404 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 4 The general will label all buses as bus systems.

[0126] The user interface 405 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0127] It is understood that memory 402 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable categories of memory.

[0128] In this embodiment of the invention, the memory 402 is used to store various types of data to support the operation of the electronic device 400. Examples of this data include: any executable program for operation on the electronic device 400, such as the operating system 4021 and application programs 4022; the operating system 4021 contains various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. The application program 4022 may contain various applications, such as media players, browsers, etc., for implementing various application services. The high-altitude subseason precipitation prediction method provided in this embodiment of the invention can be included in the application program 4022.

[0129] The methods disclosed in the above embodiments of the present invention can be applied to or implemented by processor 401. Processor 401 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 401 or by instructions in the form of software. The processor 401 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 401 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. General-purpose processor 401 may be a microprocessor or any conventional processor, etc. The steps of the high-altitude subseason precipitation prediction method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in a memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0130] In an exemplary embodiment, the electronic device 400 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to perform the aforementioned method.

[0131] In summary, this invention aims to construct an optimization system for predicting subseasonal precipitation in high-altitude areas based on soil moisture memory. First, by fusing multi-source observational data, it delves into the physical mechanisms by which soil moisture affects precipitation and constructs a statistical prediction model. Then, it evaluates the reproducibility of mainstream dynamic models of this mechanism and diagnoses the sources of their biases. Finally, by integrating physical statistics and dynamic models, it constructs an optimization framework that combines high predictive skill with strong interpretability, thereby significantly improving the level of subseasonal precipitation prediction in high-altitude areas during the rainy season.

[0132] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for predicting subseasonal precipitation in high-altitude areas, characterized in that, The method includes: By integrating site observations, satellite fusion products, and reanalysis data, multi-source observation data of the Qinghai-Tibet Plateau is constructed. Subseasonal precipitation signals are extracted using integrated empirical mode decomposition. Singular value decomposition is used to identify the dominant modes and key areas of precipitation influenced by previous soil moisture. A statistical prediction model for subseasonal precipitation considering soil moisture memory was constructed based on the partial least squares path method. The contribution of previous soil moisture to subseasonal precipitation was quantified using the Copula function. The physical mechanism was revealed by combining the water vapor budget and surface energy balance equations. Based on the aforementioned key physical processes, the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation are evaluated using the multi-source observation data, the sources of their bias are analyzed, and the effectiveness of soil moisture in key areas as a predictability source in each dynamic model is identified. Error correction was performed by constructing a dynamic multi-mode ensemble with dynamic weights using a Bayesian model, and key features were selected based on Lasso regression. By integrating physical statistics and dynamic models, a physically interpretable optimization model for high-altitude subseasonal precipitation prediction was formed.

2. The method according to claim 1, characterized in that, The method of extracting sub-seasonal precipitation signals using integrated empirical mode decomposition (EMD) and identifying the dominant modes and key areas of precipitation influence by previous soil moisture through singular value decomposition (SVD) includes: The multi-source observation data were cleaned and quality controlled to obtain daily precipitation data and previous soil moisture data; The daily precipitation data were decomposed using an integrated empirical mode decomposition method to extract the next season's precipitation signal; The dominant mode and key regions between the previous soil moisture data and the next season precipitation signal were identified by the singular value decomposition method, and the precipitation time series and soil moisture time series corresponding to the key regions were extracted.

3. The method according to claim 1 or 2, characterized in that, The proposed statistical prediction model for subseasonal precipitation, based on the partial least squares path method and considering soil moisture memory, utilizes the Copula function to quantify the contribution of previous soil moisture to subseasonal precipitation. It also reveals the physical mechanism by combining the water vapor budget and surface energy balance equations, including: Based on the dominant modes identified by singular value decomposition and the corresponding precipitation and soil moisture time series, a statistical prediction model for sub-seasonal precipitation considering soil moisture memory is constructed using the partial least squares path method. Using the Copula function, the contribution of early soil moisture in key areas identified by singular value decomposition to the changes in sub-seasonal precipitation is quantified, and the nonlinear dependence between the two is revealed. For key areas identified by singular value decomposition, we use water vapor budget and surface energy balance equations to explore the physical mechanism by which soil moisture affects subseasonal precipitation and clarify its role in water vapor transport and non-adiabatic heating processes.

4. The method according to claim 1, characterized in that, Based on the aforementioned key physical processes, the multi-source observation data is used to evaluate the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation, analyze the sources of bias, and identify the effectiveness of soil moisture in key areas as a predictability source in various dynamic models, including: The multi-source observation data is compared with the subseasonal precipitation prediction data of the mainstream dynamic model to calculate the corresponding evaluation index; wherein the evaluation index includes mean absolute deviation, root mean square error and time correlation coefficient. Based on the aforementioned evaluation indicators, the predictive skills of each dynamic mode are quantitatively evaluated and ranked according to their merits to obtain the evaluation results. Based on the assessment results, we will conduct an in-depth analysis of the physical sources of prediction bias in each dynamic model, and examine whether each dynamic model can reproduce the physical mechanism by which soil moisture affects sub-seasonal precipitation, so as to explore the effectiveness of soil moisture in key areas as a predictability source in each dynamic model.

5. The method according to claim 1, characterized in that, The method employs a Bayesian model to construct a dynamic multi-model ensemble with average weights for error correction, and uses Lasso regression to screen key features. It integrates physical statistics and dynamic models to form a physically interpretable optimized model for high-altitude subseasonal precipitation prediction, including: Using the soil moisture time series identified by singular value decomposition and the subseasonal precipitation prediction data of the mainstream dynamic model as the initial set of independent variables, and the subseasonal precipitation signal extracted by integrated empirical mode decomposition as the dependent variable; Based on the previous soil moisture conditions, the posterior probability weights of each dynamic mode are calculated using the Bayesian model averaging method. The multiple dynamic modes are then weighted and averaged according to the posterior probability weights to generate a weighted set prediction result that has been constrained by soil moisture and corrected for errors. The weighted set prediction results and the soil moisture time series are used to construct a new set of independent variables. The new set of independent variables and the dependent variable are then input into a Lasso regression model. The Lasso regression model is used for feature selection and model fusion to obtain an optimized high-altitude subseasonal precipitation prediction model with physical interpretability.

6. The method according to claim 1 or 5, characterized in that, Following the steps of selecting key features based on Lasso regression, integrating physical statistics and dynamic models to form a physically interpretable optimized model for high-altitude subseasonal precipitation prediction, the following steps are also included: The latest multi-source observation data of key areas are collected, and the high-altitude subseasonal precipitation prediction optimization model is used to predict the latest observation data, generating and outputting the corresponding subseasonal precipitation prediction results.

7. A system for predicting subseasonal precipitation in high-altitude areas, characterized in that, The system includes: The integration module is used to integrate site observations, satellite fusion products and reanalysis data to construct multi-source observation data of the Tibetan Plateau, and to extract the sub-seasonal precipitation signal using integrated empirical mode decomposition method. It also identifies the dominant mode and key areas of the influence of previous soil moisture on precipitation through singular value decomposition. The quantification module is used to construct a statistical prediction model for subseasonal precipitation that considers soil moisture memory based on the partial least squares path method, and to quantify the contribution of previous soil moisture to subseasonal precipitation using the Copula function, and to reveal its physical mechanism by combining the water vapor budget and surface energy balance equations. The analysis module is used to evaluate the prediction skills of mainstream dynamic models for high-altitude subseasonal precipitation based on the aforementioned key physical processes and the multi-source observation data, analyze the sources of deviation, and identify the effectiveness of soil moisture in key areas as a source of predictability in each dynamic model. The fusion module is used to construct a dynamic multi-mode set with dynamic weights using a Bayesian model for error correction, and to screen key features based on Lasso regression, and to fuse physical statistics and dynamic models to form a physically interpretable high-altitude subseasonal precipitation prediction optimization model.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the processor to perform the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed, performs the steps of the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes computer instructions for causing a computer to perform the steps of the method according to any one of claims 1 to 6.

Images