Multi-scenario soil moisture data set generation method, system and computer equipment

Abstract

The present invention relates to a method, system, computer equipment and storage medium for generating a multi-scenario soil moisture data set. The method comprises: obtaining CMIP6 multi-scenario soil moisture and temperature sample data in a global range within a preset time, segmenting them according to a number of preset spatiotemporal dimensions, inputting them into a spatiotemporal feature extraction module, and obtaining a target spatiotemporal dimension; constructing a multi-scenario soil moisture and temperature sample database; constructing a soil moisture quantitative estimation model; training a soil moisture quantitative estimation model according to the multi-scenario soil moisture and temperature sample database to obtain a multi-scenario soil moisture quantitative estimation model; obtaining CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data in a global range within a target prediction time, and inputting them into a multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture data set, which has the advantage of improving the data accuracy of future soil moisture statistics.

Description

Technical Field

[0001] The present invention relates to the field of soil moisture, and in particular to a method, system, computer equipment and storage medium for generating a multi-scenario soil moisture data set. Background Art

[0002] The coupled climate models of the Sixth International Coupled Model Intercomparison Project (CMIP6) have become an important reference for the United Nations Intergovernmental Panel on Climate Change (IPCC) to prepare the Sixth Assessment Report (AR6) and the United Nations Climate Change Conference and other climate change negotiations. A series of scenarios describing different future socioeconomic development models and atmospheric greenhouse gas concentrations have been constructed around CMIP6, called shared socioeconomic pathway scenarios (SSPs). Many research institutions have developed future (2015-2100) multi-scenario surface soil moisture (0-10cm depth) data based on SSPs and Earth System Models (ESM). The massive open and shared CMIP6 soil moisture data are widely used to explore the future trends of key processes in the Earth's surface system, such as drought characteristics, heat wave patterns, vegetation evolution, and land-sea interactions. In order to enhance the trend representativeness of the data and reduce abnormal deviations, mainstream research usually simply calculates the ensemble average of multiple sets of future scenario data to characterize the global future spatial and temporal distribution of soil moisture, but it also faces the risk of being dominated by the data with the largest deviation. In addition, although ESM has been continuously optimized and iterated, there are still systematic deviations and uncertainties. However, there is a lack of systematic evaluation and verification research on soil moisture data for future CMIP6 scenarios. That is, the data set is widely used but its accuracy is still unknown, which makes the reliability of related research results questionable and the uncertainty difficult to quantify and estimate. Summary of the invention

[0003] Based on this, the purpose of the present invention is to provide a method, system, computer equipment and storage medium for generating a multi-scenario soil moisture dataset, which has the advantages of efficiently capturing the spatiotemporal dynamic coupled texture of soil moisture and improving the data accuracy of future soil moisture statistics.

[0004] A method for generating a multi-scenario soil moisture dataset comprises the following steps:

[0005] Obtain CMIP6 multi-scenario soil moisture and temperature sample data for the global range within the preset time;

[0006] The CMIP6 multi-scenario soil moisture and sample data are divided according to several preset spatiotemporal sizes, respectively, and input into a pre-built spatiotemporal feature extraction module, the regression loss of the spatiotemporal feature extraction module is calculated, and the preset spatiotemporal size corresponding to the minimum regression loss in the regression loss of the spatiotemporal feature extraction module is obtained as the target spatiotemporal size; wherein the spatiotemporal feature extraction module is a module constructed based on the spatiotemporal attention mechanism for extracting the spatiotemporal features of time series dynamic variables, and the time series dynamic variables include the CMIP6 soil moisture and temperature sample data;

[0007] Acquire soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data on a global scale within the preset time, unify the spatiotemporal dimensions of the CMIP6 soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data to the target spatiotemporal dimensions, and construct a multi-scenario soil moisture and temperature sample database;

[0008] Constructing a soil moisture quantitative estimation model including the spatiotemporal feature extraction module;

[0009] According to the multi-scenario soil moisture and temperature sample database, the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data and soil moisture error sample data are used as input, and the SMAP soil moisture data is used as output, the soil moisture quantitative estimation model is trained to obtain a multi-scenario soil moisture quantitative estimation model;

[0010] The CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data are obtained on a global scale within the target prediction time and input into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset.

[0011] A multi-scenario soil moisture dataset generation system, comprising:

[0012] A sample data acquisition unit is used to obtain CMIP6 multi-scenario soil moisture and temperature sample data on a global scale within a preset time;

[0013] A spatiotemporal size acquisition unit is used to divide the CMIP6 multi-scenario soil moisture and sample data according to a number of preset spatiotemporal sizes, input them into a pre-built spatiotemporal feature extraction module, calculate the regression loss of the spatiotemporal feature extraction module, and obtain the preset spatiotemporal size corresponding to the minimum regression loss in the regression loss of the spatiotemporal feature extraction module as the target spatiotemporal size; wherein the spatiotemporal feature extraction module is a module constructed based on the spatiotemporal attention mechanism for extracting spatiotemporal features of time series dynamic variables, and the time series dynamic variables include the CMIP6 soil moisture and temperature sample data;

[0014] A database construction unit is used to obtain soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data in the global range within the preset time, unify the spatiotemporal dimensions of the CMIP6 soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data to the target spatiotemporal dimensions, and construct a multi-scenario soil moisture and temperature sample database;

[0015] A model building unit, used to build a soil moisture quantitative estimation model including the spatiotemporal feature extraction module;

[0016] A model training unit is used to train the soil moisture quantitative estimation model according to the multi-scenario soil moisture and temperature sample database, using the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data and soil moisture error sample data as input and the SMAP soil moisture data as output, so as to obtain a multi-scenario soil moisture quantitative estimation model;

[0017] The product generation unit is used to obtain the CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data on a global scale within a target prediction time and input them into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture data set.

[0018] A computer device comprises: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the multi-scenario soil moisture dataset generation method as described above when executing the computer program.

[0019] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for generating a multi-scenario soil moisture dataset are implemented.

[0020] The multi-scenario soil moisture dataset generation method described in the present application comprises the following steps: obtaining CMIP6 multi-scenario soil moisture and temperature sample data on a global scale within a preset time; dividing the CMIP6 multi-scenario soil moisture and temperature sample data into CMIP6 multi-scenario soil moisture and temperature sample data corresponding to different spatiotemporal sizes according to a number of preset spatiotemporal sizes, and inputting the data into a pre-built spatiotemporal feature extraction module respectively; calculating the regression loss of the spatiotemporal feature extraction module, and taking the preset spatiotemporal size corresponding to the minimum regression loss as the target spatiotemporal size; obtaining soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data on a global scale within a preset time; and inputting the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data into the pre-built spatiotemporal feature extraction module respectively; calculating the regression loss of the spatiotemporal feature extraction module, and taking the preset spatiotemporal size corresponding to the minimum regression loss as the target spatiotemporal size; and obtaining soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data on a global scale within a preset time; and The spatial and temporal dimensions of the SMAP soil moisture data are unified to the target spatial and temporal dimensions, a multi-scenario soil moisture and temperature sample database is constructed, and a soil moisture quantitative estimation model including the spatial and temporal feature extraction module is constructed; according to the multi-scenario soil moisture and temperature sample database, the CMIP6 multi-scenario soil moisture and temperature sample data and soil physical property sample data, underlying surface property sample data, and soil moisture error sample data are used as input, and the SMAP soil moisture data is used as output, the soil moisture quantitative estimation model is trained to obtain a multi-scenario soil moisture quantitative estimation model; the CMIP6 soil moisture and temperature target data, soil physical property target data, underlying surface property target data, and soil moisture error target data within the target prediction time are obtained and input into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset.

[0021] The multi-scenario soil moisture dataset generation method described in the present application achieves efficient capture of the spatiotemporal dynamic coupled texture of soil moisture and improves the accuracy of multi-scenario soil moisture dataset generation.

[0022] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flowchart of a method for generating a multi-scenario soil moisture dataset in one embodiment of the present application;

[0024] Figure 2 A flowchart of the steps for obtaining CMIP6 multi-scenario soil moisture and temperature sample data in one embodiment of the present application;

[0025] Figure 3 A flowchart of the steps for constructing a multi-scenario soil moisture and temperature sample database in one embodiment of the present application;

[0026] Figure 4A flowchart of the steps of constructing a soil moisture quantitative estimation model in one embodiment of the present application;

[0027] Figure 5 A flowchart of the steps for obtaining the hyperparameters of the soil moisture quantitative estimation model in one embodiment of the present application;

[0028] Figure 6 A flowchart of the steps for training the soil moisture quantitative estimation model in one embodiment of the present application;

[0029] Figure 7 A flowchart of the steps for obtaining a multi-scenario soil moisture dataset in one embodiment of the present application;

[0030] Figure 8 This is a schematic diagram of the structure of a multi-scenario soil moisture dataset generation system in one embodiment of the present application;

[0031] Fig. 9 Schematic diagram of a computer device for generating a multi-scenario soil moisture dataset in an embodiment of the present application. DETAILED DESCRIPTION

[0032] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0033] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0034] It should be understood that, although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" / "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determination".

[0035] See also Figure 1 , Figure 1This is a flowchart of a method for generating a multi-scenario soil moisture dataset in one embodiment of the present application. A method for generating a multi-scenario soil moisture dataset is characterized by comprising the following steps:

[0036] S101, obtain the CMIP6 multi-scenario soil moisture and temperature sample data for the global range within the preset time;

[0037] S102, dividing the CMIP6 multi-scenario soil moisture and sample data according to a number of preset spatiotemporal sizes, inputting them into a pre-built spatiotemporal feature extraction module, calculating the regression loss of the spatiotemporal feature extraction module, and obtaining the preset spatiotemporal size corresponding to the minimum regression loss in the regression loss of the spatiotemporal feature extraction module as the target spatiotemporal size; wherein the spatiotemporal feature extraction module is a module constructed based on the spatiotemporal attention mechanism for extracting spatiotemporal features of time series dynamic variables, and the time series dynamic variables include the CMIP6 soil moisture and temperature sample data;

[0038] S103, obtaining soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data in the global range within the preset time, unifying the spatiotemporal dimensions of the CMIP6 soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data to the target spatiotemporal dimensions, and constructing a multi-scenario soil moisture and temperature sample database;

[0039] S104, constructing a soil moisture quantitative estimation model including the spatiotemporal feature extraction module;

[0040] S105, according to the multi-scenario soil moisture and temperature sample database, with the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data and soil moisture error sample data as input, and the SMAP soil moisture data as output, training the soil moisture quantitative estimation model to obtain a multi-scenario soil moisture quantitative estimation model;

[0041] S106, obtaining CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data on a global scale within a target prediction time, and inputting them into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset.

[0042] The multi-scenario soil moisture dataset generation method described in the present application comprises the following steps: obtaining CMIP6 multi-scenario soil moisture and temperature sample data on a global scale within a preset time; dividing the CMIP6 multi-scenario soil moisture and temperature sample data into CMIP6 multi-scenario soil moisture and temperature sample data corresponding to different spatiotemporal sizes according to a number of preset spatiotemporal sizes, and inputting the data into a pre-built spatiotemporal feature extraction module respectively; calculating the regression loss of the spatiotemporal feature extraction module, and taking the preset spatiotemporal size corresponding to the minimum regression loss as the target spatiotemporal size; obtaining soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data on a global scale within a preset time; and inputting the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data into the pre-built spatiotemporal feature extraction module respectively; calculating the regression loss of the spatiotemporal feature extraction module, and taking the preset spatiotemporal size corresponding to the minimum regression loss as the target spatiotemporal size; and obtaining soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data on a global scale within a preset time; and The spatial and temporal dimensions of the SMAP soil moisture data are unified to the target spatial and temporal dimensions, a multi-scenario soil moisture and temperature sample database is constructed, and a soil moisture quantitative estimation model including the spatial and temporal feature extraction module is constructed; according to the multi-scenario soil moisture and temperature sample database, the CMIP6 multi-scenario soil moisture and temperature sample data and soil physical property sample data, underlying surface property sample data, and soil moisture error sample data are used as input, and the SMAP soil moisture data is used as output, the soil moisture quantitative estimation model is trained to obtain a multi-scenario soil moisture quantitative estimation model; the CMIP6 soil moisture and temperature target data, soil physical property target data, underlying surface property target data, and soil moisture error target data within the target prediction time are obtained and input into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset.

[0043] The multi-scenario soil moisture dataset generation method described in the present application achieves efficient capture of the spatiotemporal dynamic coupled texture of soil moisture and improves the accuracy of multi-scenario soil moisture dataset generation.

[0044] For step S101, obtain CMIP6 multi-scenario soil moisture and temperature sample data on a global scale within a preset time;

[0045] Wherein, the CMIP6 is the sixth international coupled model comparison project, and the CMIP6 multi-scenario soil moisture and temperature sample data are sample data obtained from the CMIP6 future scenario soil moisture data. The scenario is an environment set for different soil environments. In one embodiment, the multiple scenarios include SSP1-2.6, SSP2-4.5, SSP3-7.0 and / or SSP5-8.5. The preset time is a time set in advance for obtaining sample data. In this embodiment, the preset time is set to 2015 to 2023.

[0046] See also Figure 2, Figure 2 This is a flowchart of the steps of obtaining CMIP6 multi-scenario soil moisture and temperature sample data for a preset time in an embodiment of the present application. In one embodiment, obtaining CMIP6 multi-scenario soil moisture and temperature sample data includes the following steps:

[0047] S201, obtain CMIP6 multi-scenario soil moisture and temperature data, SMAP soil moisture data and ERA5-Land soil moisture data for the global range within the preset time;

[0048] S202, sampling the CMIP6 multi-scenario soil moisture and temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data to a preset spatial resolution;

[0049] S203, inputting a set of CMIP6 single scenario soil moisture and temperature data in the CMIP6 multi-scenario soil moisture and temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data into an improved triple error matching model, and calculating soil moisture evaluation information of the CMIP6 single scenario soil moisture and temperature data;

[0050] S204: According to the soil moisture evaluation information, several groups of CMIP6 single scenario soil moisture and temperature data are selected from the CMIP6 multi-scenario soil moisture and temperature data to form the CMIP6 multi-scenario soil moisture and temperature sample data.

[0051] For steps S201 to S202, the preset spatial resolution is a pre-set index for characterizing the image resolution of ground target details. In one embodiment, the preset spatial resolution is set to 0.5°×0.5°, and in other embodiments, the preset spatial resolution can also be set to 0.1°×0.1°.

[0052] In this embodiment, by acquiring the CMIP6 multi-scenario soil moisture and temperature data, SMAP soil moisture data and ERA5-Land soil moisture data on a global scale within the preset time and sampling them to the preset spatial resolution, the uniformity of the data is ensured, which facilitates the calculation of soil moisture evaluation information.

[0053] Regarding step S203, the enhanced triple collocation model (ETC) considers that there is a linear relationship between three independent soil moisture data sets and the true soil moisture value.

[0054] The soil moisture evaluation information is evaluation information of the CMIP6 multi-scenario soil moisture and temperature data. In one embodiment, the soil moisture evaluation information includes error variance, correlation coefficient and signal-to-noise ratio.

[0055] In this embodiment, a group of CMIP6 single scenario soil moisture and air temperature data in the CMIP6 multi-scenario soil moisture and air temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data are input into the improved triple error matching model as three independent soil moisture data sets, and the error variance, correlation coefficient, and signal-to-noise ratio corresponding to the CMIP6 soil moisture and air temperature data are calculated to obtain soil moisture evaluation information corresponding to the soil moisture and air temperature data, so as to obtain accurate sample data for training the soil moisture quantitative estimation model.

[0056] In one embodiment, a set of CMIP6 single scenario soil moisture and temperature data in the CMIP6 multi-scenario soil moisture and temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data are input into an improved triple error collocation model to calculate soil moisture evaluation information of the CMIP6 single scenario soil moisture and temperature data, including the following steps:

[0057] A set of CMIP6 single scenario soil moisture and air temperature data in the CMIP6 multi-scenario soil moisture and air temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data are input into the improved triple error matching model:

[0058]

[0059] Among them, the SM true is the true value of soil moisture, SM1, SM2 and SM3 are a set of CMIP6 single scenario soil moisture and temperature data in the CMIP6 multi-scenario soil moisture and temperature data, the SMAP soil moisture data and the ERA5-Land soil moisture data, a1, a2, a3 and b1, b2, b3 are the intercept and slope of the corresponding linear equation, θ1, θ2 and θ3 are the errors of the corresponding soil moisture observation data set, and the mean is 0;

[0060] Arranging the above formulas yields:

[0061]

[0062] In the formula, and are the error variances of SM1, SM2 and SM3 respectively, <·> represents the mean calculation, SM′=SM- <sm>; <SM′1 2 >、 <SM′2 2 >、 <SM′3 2 > are the variances of SM1, SM2, and SM3, respectively,<SM′1SM′2> ,<SM′1SM′3> ,<SM′2SM′3> are the covariances between SM1, SM2, and SM3;

[0063]

[0064] Wherein, CC1, CC2 and CC3 are the correlation coefficients between SM1, SM2 and SM3 and the true value of soil moisture respectively;

[0065]

[0066] Where SNR1, SNR2 and SNR3 are the signal-to-noise ratios of SM1, SM2 and SM3 respectively, in dB;

[0067] The error variance, correlation coefficient and signal-to-noise ratio are used as soil moisture evaluation information of the CMIP6 single scenario soil moisture and air temperature data.

[0068] In step S204, according to the soil moisture evaluation information, several groups of CMIP6 single scenario soil moisture and temperature data are selected from the CMIP6 multi-scenario soil moisture and temperature data to form the CMIP6 multi-scenario soil moisture and temperature sample data;

[0069] In this embodiment, by obtaining the soil moisture evaluation information corresponding to the CMIP6 soil moisture and air temperature data, several groups of CMIP6 single scenario soil moisture and air temperature data with good soil moisture evaluation information are selected to form the CMIP6 soil moisture and air temperature sample data. In one embodiment, according to the soil moisture evaluation information, at least 10 groups of CMIP6 single scenario soil moisture and air temperature data with small error variance, high correlation coefficient and high signal-to-noise ratio are selected to form the CMIP6 soil moisture and air temperature sample data, so as to improve the accuracy of the multi-scenario soil moisture and air temperature sample database and ensure the accuracy of the generation of the multi-scenario soil moisture data set.

[0070] In step S102, the CMIP6 multi-scenario soil moisture and sample data are segmented according to a number of preset spatiotemporal sizes, respectively, and input into a pre-built spatiotemporal feature extraction module, the regression loss of the spatiotemporal feature extraction module is calculated, and the preset spatiotemporal size corresponding to the minimum regression loss among the regression losses of the spatiotemporal feature extraction module is obtained as the target spatiotemporal size;

[0071] Among them, the spatiotemporal feature extraction module is a module built based on the spatiotemporal attention mechanism for extracting the spatiotemporal features of time series dynamic variables. The time series dynamic variables include the CMIP6 soil moisture and temperature sample data. The time series dynamic variables refer to data that changes over time.

[0072] In one embodiment, the spatiotemporal attention mechanism includes a joint spatiotemporal attention mechanism, a split spatiotemporal attention mechanism, and a decomposition-based dot product spatiotemporal attention mechanism.

[0073] The joint spatiotemporal attention (JSTA) is an end-to-end spatiotemporal attention mechanism, which decomposes the input soil moisture sample sequence into a set of non-overlapping spatiotemporal image patches, and selectively focuses on the key changes in the spatial range and time series of soil moisture samples by calculating the attention between each pair of spatiotemporal image patches.

[0074] The temporal attention of the segmented spatiotemporal attention (DSTA) only compares image blocks at different times and the same spatial position, and the spatial attention only compares image blocks at the same time and different spatial positions, thereby avoiding tedious operations between all paired blocks and reducing computational costs.

[0075] The decomposition-based dot-product spatiotemporal attention (FDSTA) improves the serial calculation method of segmented spatiotemporal attention, and further improves the efficiency of spatiotemporal feature extraction through parallel calculation of spatial attention and temporal attention.

[0076] In this embodiment, the spatiotemporal feature extraction module is constructed by combining the joint spatiotemporal attention, the segmented spatiotemporal attention and the decomposition-based dot product spatiotemporal attention to extract spatiotemporal features of temporal dynamic variables.

[0077] In one embodiment, the spatiotemporal feature extraction module includes an embedding layer, n e The temporal dynamic variables are mapped to the E-dimensional space and encoded through the embedding layer, and then sequentially passed through the n e encoders, wherein each encoder includes a multi-head spatiotemporal attention sublayer (Multi-Head Attention) and a feedforward connection sublayer (Feed Forward), each sublayer is followed by residual connection and regularization calculation, and the spatiotemporal attention is calculated by the best method among the joint spatiotemporal attention, the split spatiotemporal attention and the decomposition-based dot product spatiotemporal attention; secondly, the encoder hidden layer state is mapped through a fully connected layer with k1 nodes; finally, the encoder hidden layer state is mapped through a fully connected layer with a Dropout rate of d r1 The Dropout layer improves the generalization ability of the model.

[0078] In this embodiment, the spatiotemporal feature extraction module is constructed based on the spatiotemporal attention mechanism to extract the spatiotemporal features in the spatiotemporal dynamic variables, which are used to generate a multi-scenario soil moisture dataset and improve the accuracy of the generation of the multi-scenario soil moisture dataset.

[0079] The regression loss is data used to represent model accuracy. In one embodiment, the model regression loss includes mean absolute error or mean square error.

[0080] The preset spatiotemporal size is a pre-set spatiotemporal size. The preset duration size includes a preset time size and a preset space size. In one embodiment, the preset time size includes 3 months, 6 months, 9 months or 12 months, and the preset space size includes 8 pixels, 16 pixels, 32 pixels or 64 pixels. The target spatiotemporal size is the optimal spatiotemporal size determined based on the regression loss of the model. In this embodiment, the spatiotemporal size corresponding to the minimum regression loss in the regression loss of the feature extraction module is selected as the target spatiotemporal size.

[0081] In this embodiment, the CMIP6 multi-scenario soil moisture and temperature sample data are divided into CMIP6 multi-scenario soil moisture and temperature sample data corresponding to different preset spatiotemporal sizes according to a plurality of preset spatiotemporal sizes. The CMIP6 multi-scenario soil moisture and temperature sample data corresponding to the different preset spatiotemporal sizes are respectively input into the spatiotemporal feature extraction module, the regression losses corresponding to the different spatiotemporal sizes are obtained, and the spatiotemporal size corresponding to the smallest regression loss is selected as the target spatiotemporal size.

[0082] In this embodiment, by calculating the regression loss of the CMIP6 multi-scenario soil moisture and temperature sample data corresponding to different preset spatiotemporal sizes, the spatiotemporal size corresponding to the minimum regression loss is obtained as the target spatiotemporal size, which is used to construct a multi-scenario soil moisture and temperature sample database, ensuring the highest model learning efficiency and output accuracy.

[0083] For step S103, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data are obtained in the global range within the preset time, and the spatiotemporal dimensions of the CMIP6 soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data are unified to the target spatiotemporal dimensions to construct a multi-scenario soil moisture and temperature sample database;

[0084] The soil physical property sample data includes soil property data, the underlying surface property sample data includes terrain data, geographic location data and climate data set, the soil moisture error sample data includes soil moisture data evaluation information, wherein the soil moisture data evaluation information includes error variance, correlation coefficient and signal-to-noise ratio; the soil property data includes reference bulk density, sand content, clay content and gravel content; the terrain data includes elevation, slope and slope direction; the geographic location data includes longitude and latitude. The SMAP soil moisture data is the soil moisture data obtained from the SMAP.

[0085] The multi-scenario soil moisture and temperature sample database is a database that integrates the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data and the SMAP soil moisture data for training a soil moisture quantitative estimation model.

[0086] In one embodiment, unifying the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data, and the SMAP soil moisture data to the target spatiotemporal dimensions comprises:

[0087] The spatiotemporal size of the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data, and the SMAP soil moisture data is set to the target spatiotemporal size;

[0088] The CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data and the SMAP soil moisture data are normalized using the maximum and minimum values.

[0089] Among them, the maximum and minimum value normalization processing is to unify the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data and the SMAP soil moisture data into the range of [0,1] according to the maximum and minimum values. For example, if the maximum and minimum values ​​of the temperature data are 40, then dividing all the temperature data by 40 is the normalization processing.

[0090] In this embodiment, by setting the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data, and the SMAP soil moisture data to the target spatiotemporal dimensions and performing normalization processing, the data uniformity and standardization of the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data, and the SMAP soil moisture data are ensured, thereby facilitating the construction of a multi-scenario soil moisture and temperature sample database.

[0091] See also Figure 3 , Figure 3 This is a flowchart of the steps of constructing a multi-scenario soil moisture and air temperature sample database in one embodiment of the present application. In one embodiment, the steps of constructing a multi-scenario soil moisture and air temperature sample database include:

[0092] S301, obtaining CMIP6 soil moisture and temperature sample data corresponding to a single scenario in the CMIP6 multi-scenario soil moisture and temperature sample data;

[0093] S302, extracting a SMAP soil moisture image block from the SMAP soil moisture data corresponding to the single scenario in the order of rows first and columns later using a sliding window of a preset specification;

[0094] S303, matching the SMAP soil moisture image block with the corresponding CMIP6 soil moisture and temperature sample data for 12 consecutive months and the soil physical property sample data, underlying surface property sample data, and soil moisture error sample data according to the temporal and spatial attributes to form an input-output sample pair as sample data;

[0095] S304, obtaining sample data of each single scenario in the CMIP6 multi-scenario soil moisture and temperature sample data to form the multi-scenario soil moisture and temperature sample database.

[0096] Regarding steps S401 to S404, the sliding window of preset specifications is a sliding window of specifications set in advance. In this embodiment, the preset specifications are set to 32*32.

[0097] In this embodiment, the SMAP soil moisture image blocks in the SMAP soil moisture data corresponding to the single scenario are extracted in the order of rows first and columns using the sliding window of the preset specifications, so as to facilitate the matching of the SMAP soil moisture data with the CMIP6 soil moisture and temperature sample data.

[0098] The SMAP soil moisture image block is matched with the corresponding CMIP6 soil moisture and air temperature sample data and the soil physical property sample data, the underlying surface property sample data, and the soil moisture error sample data for 12 consecutive months according to the spatiotemporal attributes to form an input-output sample pair, that is, the corresponding moment of the SMAP soil moisture image block is traced back for 12 consecutive months to obtain the corresponding CMIP6 soil moisture and air temperature sample data and the soil physical property sample data, the underlying surface property sample data, and the soil moisture error sample data, so as to obtain an input-output sample pair with the soil physical property sample data, the underlying surface property sample data, and the soil moisture error sample data of the CMIP6 soil moisture and air temperature sample data set as input and the SMAP soil moisture image block as output, and the input-output sample pair is used as the sample data.

[0099] In one embodiment, the sample data includes a training set, a test set, and a validation set, wherein the training set and the validation set are used to train the soil moisture quantitative estimation model, and the test set is used to evaluate the performance of the soil moisture quantitative estimation model. In this embodiment, the input-output sample pairs are divided into a training set, a validation set, and a test set according to a preset ratio to obtain the sample data, for example, the input-output sample pairs are divided into a training set, a validation set, and a test set according to a ratio of 7:1.5:1.5.

[0100] Finally, by obtaining the sample data of each single scenario in the CMIP6 multi-scenario soil moisture and temperature sample data, the multi-scenario soil moisture and temperature sample database is constructed.

[0101] In this embodiment, the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data, and the SMAP soil moisture data are unified to the target spatiotemporal dimensions to construct a multi-scenario soil moisture and temperature sample database, which is convenient for training the soil moisture quantitative estimation model.

[0102] For step S104, a soil moisture quantitative estimation model including the spatiotemporal feature extraction module is constructed;

[0103] Among them, the soil moisture quantitative estimation model is a model for estimating the soil moisture fusion estimation value based on CMIP6 multi-scenario soil moisture and temperature data, soil physical property data, underlying surface property data and soil moisture error data. In this embodiment, the soil moisture quantitative estimation model includes a spatiotemporal feature extraction module, a spatial feature extraction module and a feature fusion and relationship simulation module.

[0104] See also Figure 4 , Figure 4 This is a flowchart of the steps of constructing a soil moisture quantitative estimation model in one embodiment of the present application. In one embodiment, constructing a soil moisture quantitative estimation model including the spatiotemporal feature extraction module includes the following steps:

[0105] S401, constructing a spatial feature extraction module including several convolution submodules, wherein the spatial feature extraction module is used to extract spatial features of time series static variables, and the time series static variables include soil physical property sample data, underlying surface property sample data, soil moisture error sample data and soil physical property target data, underlying surface property target data, and soil moisture error target data;

[0106] S402, based on the fully connected layer, construct a feature fusion and relationship simulation module to obtain the soil moisture quantitative estimation model, wherein the feature fusion and relationship simulation module is used to fuse the spatiotemporal features and the spatial features and fit their nonlinear relationship with the SMAP soil moisture data.

[0107] For steps S401 to S402, the spatial feature extraction module is used to extract the spatial features of time series static variables, wherein the time series static variables are data that do not change with time. In one embodiment, the time series static variables include soil physical property sample data, underlying surface property sample data, and soil moisture error sample data.

[0108] In this embodiment, the spatial feature extraction module includes M convolution submodules, a two-dimensional global average pooling layer (GA Pooling 2D) and a Dropout layer, wherein the convolution submodule is used to extract spatial features, the two-dimensional global average pooling layer is used for context aggregation, and each convolution submodule is composed of m j It consists of a two-dimensional convolutional layer (Conv2D) with (j=1,2,…,M) convolution kernels of size r×s and a two-dimensional maximum pooling layer (MaxPooling2D) with a pooling window size of 2×2; the Dropout rate of the Dropout layer is d r2 .

[0109] The feature fusion and relationship simulation module is a module for fusing the spatiotemporal features extracted by the spatiotemporal feature extraction module and the spatial features extracted by the spatial feature extraction module to obtain target features, and fitting the nonlinear relationship between the target features and the SMAP soil moisture data. In this embodiment, the feature fusion and relationship simulation module includes two fully connected layers whose node data are k1 and k2 respectively.

[0110] In this embodiment, the soil moisture quantitative estimation model is obtained by further constructing the spatial feature extraction module and the feature fusion and relationship simulation module and combining the spatiotemporal feature extraction module.

[0111] For step S105, according to the multi-scenario soil moisture and temperature sample database, the CMIP6 multi-scenario soil moisture and temperature sample data and soil physical property sample data, underlying surface property sample data, and soil moisture error sample data are used as input, and the SMAP soil moisture data is used as output to train the soil moisture quantitative estimation model to obtain a multi-scenario soil moisture quantitative estimation model;

[0112] In this embodiment, the CMIP6 multi-scenario soil moisture and temperature sample data and soil physical property sample data, underlying surface property sample data, and soil moisture error sample data in the multi-scenario soil moisture and temperature sample database are used as input, and the SMAP soil moisture data is used as output to train the soil moisture quantitative estimation model to obtain a multi-scenario soil moisture quantitative estimation model.

[0113] See also Figure 5 , Figure 5 This is a flowchart of the steps of obtaining the hyperparameters of the soil moisture quantitative estimation model in one embodiment of the present application. In one embodiment, the following steps are also included before training the soil moisture quantitative estimation model:

[0114] S501, running hyperparameter iteration, inputting the preset initial hyperparameters into the maximization acquisition function, and obtaining a hyperparameter vector;

[0115] S502, inputting the hyperparameter vector into a probabilistic proxy model to obtain an objective function value;

[0116] S503, updating the probabilistic proxy model according to the objective function value and the hyperparameter vector;

[0117] S504, record the current number of iterations. If the current number of iterations exceeds the preset number of iterations, use the objective function value as the optimal hyperparameter. Otherwise, update the initial hyperparameter to the objective function value, and return to the step of inputting the initial hyperparameter into the maximization acquisition function to obtain the hyperparameter vector.

[0118] For steps S501 to S504, the hyperparameters are contents with multiple setting options in the soil moisture quantitative estimation model. In one embodiment, the hyperparameters include model structure hyperparameters and model training hyperparameters, wherein the model structure hyperparameters include model embedding dimension, number of encoders, number of attention heads, number of convolution sub-modules, convolution kernel size, number of convolution kernels, Dropout rate, number of fully connected layer nodes and / or activation function; the model training hyperparameters include initial learning rate and / or optimizer.

[0119] The preset initial hyperparameters are hyperparameters obtained by randomly initializing the soil moisture quantitative estimation model, and the probabilistic agent model is a mathematical model used to describe intelligent agents in the field of artificial intelligence.

[0120] The preset number of iterations is the number of cycles that need to be performed and is set in advance.

[0121] After running parameter iteration, the preset initial hyperparameters are input into the maximization acquisition function to obtain a hyperparameter vector, and the target hyperparameters are input into the probabilistic proxy model to obtain an objective function value, the probabilistic proxy model is updated according to the objective function value and the target hyperparameters, the current number of iterations is recorded, and it is determined whether the current number of iterations exceeds the preset number of iterations. If so, the objective function value is used as the optimal hyperparameter; otherwise, the objective function value is used as the initial hyperparameter, and the process returns to the step of inputting the initial hyperparameters into the maximization acquisition function to obtain a hyperparameter vector.

[0122] In this embodiment, by looping through steps S501 to S503, when the current number of iterations reaches the preset number of iterations, the objective function value is output as the optimal hyperparameter, thereby improving the accuracy of hyperparameter acquisition of the soil moisture quantitative estimation model, facilitating the construction of the optimal soil moisture quantitative estimation model, and improving the accuracy of generating multi-scenario soil moisture data sets.

[0123] See also Figure 6 , Figure 6 This is a flowchart of the steps of training the soil moisture quantitative estimation model in one embodiment of the present application. In one embodiment, the training of the soil moisture quantitative estimation model to obtain a multi-scenario soil moisture quantitative estimation model also includes the following steps:

[0124] S601, obtaining sample data corresponding to a single scenario in the multi-scenario soil moisture and temperature sample database;

[0125] S602, according to the sample data corresponding to the single scenario, respectively train the joint spatiotemporal attention mechanism, the segmentation spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism of the soil moisture quantitative estimation model to obtain the model regression losses corresponding to the joint spatiotemporal attention mechanism, the segmentation spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism.

[0126] S603, obtaining a spatiotemporal attention mechanism corresponding to the minimum model regression loss, and generating a soil moisture quantitative estimation model corresponding to the single scenario according to the spatiotemporal attention mechanism corresponding to the minimum model regression loss;

[0127] S604, repeating the above steps to obtain a soil moisture quantitative estimation model corresponding to each single scenario, and forming a multi-scenario soil moisture quantitative estimation model.

[0128] Regarding steps S601 to S604, the multi-scenario soil moisture quantitative estimation model includes soil moisture quantitative estimation models corresponding to different scenarios.

[0129] In this embodiment, a soil moisture quantitative estimation model is constructed based on the combination of the joint spatiotemporal attention mechanism, the segmented spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism, and the sample data corresponding to a single scenario in the multi-scenario soil moisture and temperature sample database are used to train the joint spatiotemporal attention mechanism, the segmented spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism in the soil moisture quantitative estimation model, respectively, and the model regression loss is obtained. The spatiotemporal attention method with the smallest model regression loss among the joint spatiotemporal attention mechanism, the segmented spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism is used as the optimal spatiotemporal attention to obtain the soil moisture quantitative estimation model corresponding to the single scenario.

[0130] For example, after the joint spatiotemporal attention, segmented spatiotemporal attention and decomposition-based dot product spatiotemporal attention are trained respectively on the sample data corresponding to scenario SSP1-2.6, the model regression loss corresponding to the joint spatiotemporal attention is the smallest, then the joint spatiotemporal attention is used as the optimal execution method in the soil moisture quantitative estimation model corresponding to the scenario SSP1-2.6, that is, the soil moisture quantitative estimation model corresponding to the scenario SSP1-2.6 uses the joint spatiotemporal attention as the main method to obtain the spatiotemporal characteristics of the time series dynamic variables during the execution process.

[0131] Repeat the above steps to obtain a soil moisture quantitative estimation model corresponding to each single scenario, and bind the soil moisture quantitative estimation model to the corresponding single scenario to form a multi-scenario soil moisture quantitative estimation model.

[0132] In this embodiment, the soil moisture quantitative estimation model is trained to obtain the multi-scenario soil moisture quantitative estimation model, and the optimal spatiotemporal attention method is selected for each single scenario to perform calculations, thereby further improving the accuracy of the multi-scenario soil moisture quantitative estimation model.

[0133] For step S106, the CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data, and soil moisture error target data within the target prediction time are obtained and input into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture data set.

[0134] The target prediction time is the time when soil moisture and temperature need to be preset. In one embodiment, the target prediction time is from 2015 to 2100. The multi-scenario soil moisture data set is a data set for representing soil moisture information within the target prediction time.

[0135] See also Figure 7 , Figure 7 This is a flowchart of the steps of obtaining a multi-scenario soil moisture dataset in one embodiment of the present application. In one embodiment, obtaining CMIP6 multi-scenario soil moisture and temperature target data and soil physical property target data, underlying surface property target data, and soil moisture error target data within the target prediction time and inputting them into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset includes the following steps:

[0136] S701, obtaining CMIP6 multi-scenario soil moisture and temperature prediction data, soil physical property prediction data, underlying surface property prediction data, and soil moisture error prediction data within the target prediction time;

[0137] S702, setting the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature prediction data, soil physical property prediction data, underlying surface property prediction data, and soil moisture error prediction data as target spatiotemporal dimensions and performing spatiotemporal matching to obtain the CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data, and soil moisture error target data;

[0138] S703, extracting, according to each grid pixel in the global range, the data sequence corresponding to the CMIP6 multi-scenario soil moisture and temperature target data and the auxiliary data image blocks corresponding to the soil physical property target data, the underlying surface property target data, and the soil moisture error target data;

[0139] S704, inputting the data sequence and the auxiliary data image block into the multi-scenario soil moisture quantitative estimation model respectively to obtain a soil moisture fusion estimation value;

[0140] S705, repeat the above steps to calculate the soil moisture fusion estimation value corresponding to each target spatiotemporal size to form a multi-scenario soil moisture dataset.

[0141] For steps S701 to S705, the grid pixel is the smallest unit of grid data. In this embodiment, the data sequence corresponding to the CMIP6 multi-scenario soil moisture and temperature target data and the auxiliary data image blocks corresponding to the soil physical property target data, the underlying surface property target data, and the soil moisture error target data are extracted based on the grid pixels in the global range to ensure the accuracy of the generation of the multi-scenario soil moisture data set.

[0142] The soil moisture fusion estimation value is an output result of soil moisture obtained according to the multi-scenario soil moisture quantitative estimation model.

[0143] In this embodiment, the CMIP6 multi-scenario soil moisture and temperature target data and soil physical property target data, underlying surface property target data, and soil moisture error target data are obtained by obtaining CMIP6 multi-scenario soil moisture and temperature prediction data and soil physical property prediction data, underlying surface property prediction data, and soil moisture error prediction data within the target prediction time for spatiotemporal matching; according to each grid pixel in the global range, the data sequence corresponding to the CMIP6 multi-scenario soil moisture and temperature target data and the auxiliary data image blocks corresponding to the soil physical property target data, the underlying surface property target data, and the soil moisture error target data are respectively extracted and input into the multi-scenario soil moisture quantitative estimation model to obtain a soil moisture fusion estimation value, and then according to the target spatiotemporal size, steps S701 to S704 are repeated to calculate the soil moisture fusion estimation value corresponding to each spatiotemporal size to obtain the multi-scenario soil moisture data set within the target prediction time.

[0144] The multi-scenario soil moisture dataset generation method described in the present application calculates soil moisture evaluation information by inputting CMIP6 soil moisture and temperature data, SMAP soil moisture data and ERA5-Land soil moisture data on a global scale within a preset time into an improved triple error matching model, and obtains data with small error variance, high correlation coefficient and high signal-to-noise ratio in the CMIP6 soil moisture and temperature data as CMIP6 multi-scenario soil moisture and temperature sample data.

[0145] The CMIP6 multi-scenario soil moisture and temperature sample data are divided into CMIP6 multi-scenario soil moisture and temperature sample data corresponding to different spatiotemporal sizes according to a number of preset spatiotemporal sizes, and then respectively input into the spatiotemporal feature extraction module, the regression loss of the spatiotemporal feature extraction module is calculated, and the preset spatiotemporal size corresponding to the minimum regression loss is used as the target spatiotemporal size.

[0146] The soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data and the SMAP soil moisture data are obtained in the global range within a preset time, the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data and the SMAP soil moisture data are unified to the target spatiotemporal dimensions, a corresponding relationship between the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data, the soil moisture error sample data and the SMAP soil moisture data is established, an input-output pair is formed with the CMIP6 multi-scenario soil moisture and temperature sample data, the soil physical property sample data, the underlying surface property sample data and the soil moisture error sample data as input and the SMAP soil moisture data as output, and a multi-scenario soil moisture and temperature sample database is constructed.

[0147] A soil moisture quantitative estimation model including the spatiotemporal feature extraction module, the spatial feature extraction module, and the feature fusion and relationship simulation module is constructed. According to the multi-scenario soil moisture and temperature sample database, the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, and soil moisture error sample data are used as input, and the SMAP soil moisture data is used as output. The soil moisture quantitative estimation model is trained to obtain a multi-scenario soil moisture quantitative estimation model.

[0148] The CMIP6 soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data within the target prediction time are obtained and input into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset.

[0149] The multi-scenario soil moisture dataset generation method described in the present application achieves efficient capture of the spatiotemporal dynamic coupled texture of soil moisture and improves the accuracy of multi-scenario soil moisture dataset generation.

[0150] See also Figure 8 , Figure 8 The structure diagram of a multi-scenario soil moisture dataset generation system in one embodiment of the present application is as follows. A multi-scenario soil moisture dataset generation system is characterized by comprising:

[0151] The sample data acquisition unit 11 is used to acquire the CMIP6 multi-scenario soil moisture and temperature sample data of the global scope within a preset time;

[0152] The spatiotemporal size acquisition unit 12 is used to divide the CMIP6 multi-scenario soil moisture and sample data according to a number of preset spatiotemporal sizes, input them into a pre-built spatiotemporal feature extraction module, calculate the regression loss of the spatiotemporal feature extraction module, and obtain the preset spatiotemporal size corresponding to the minimum regression loss in the regression loss of the spatiotemporal feature extraction module as the target spatiotemporal size; wherein the spatiotemporal feature extraction module is a module constructed based on the spatiotemporal attention mechanism for extracting the spatiotemporal features of time series dynamic variables, and the time series dynamic variables include the CMIP6 soil moisture and temperature sample data;

[0153] The database construction unit 13 is used to obtain soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data in the global range within the preset time, unify the spatiotemporal dimensions of the CMIP6 soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data to the target spatiotemporal dimensions, and construct a multi-scenario soil moisture and temperature sample database;

[0154] A model building unit 14, used to build a soil moisture quantitative estimation model including the spatiotemporal feature extraction module;

[0155] A model training unit 15 is used to train the soil moisture quantitative estimation model according to the multi-scenario soil moisture and temperature sample database, with the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data and soil moisture error sample data as input, and the SMAP soil moisture data as output, to obtain a multi-scenario soil moisture quantitative estimation model;

[0156] The product generation unit 16 is used to obtain the CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data in the global scope within the target prediction time and input them into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture data set.

[0157] It should be noted that the multi-scenario soil moisture dataset generation system provided in the above-mentioned embodiment only uses the division of the above-mentioned functional modules as an example when executing the multi-scenario soil moisture dataset generation method. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The multi-scenario soil moisture dataset generation system provided in the above-mentioned embodiment is used to execute the multi-scenario soil moisture dataset generation method described in the above-mentioned embodiment, and its operating method and principle are the same as the multi-scenario soil moisture dataset generation method recorded above, that is, the multi-scenario soil moisture dataset generation system provided in the above-mentioned embodiment and the multi-scenario soil moisture dataset generation method belong to the same concept, and its implementation process is detailed in the above-mentioned method embodiment, which will not be repeated here.

[0158] See also Fig. 9 , Fig. 9 Schematic diagram of a computer device for generating a multi-scenario soil moisture dataset in an embodiment of the present application. The computer device 21 includes: a control device 211, a memory 212, and a computer program 213 stored in the memory 212 and executable on the control device 211, for example: a multi-scenario soil moisture dataset generating program; the control device 211 executes the computer program 213 to implement the multi-scenario soil moisture dataset generating method described in the above embodiment.

[0159] Among them, the control device 211 includes a processor, and the processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts in the computer device 21, and executes various functions and processes data of the computer device 21 by running or executing instructions, programs, code sets or instruction sets stored in the memory 212, and calling the data in the memory 212. Optionally, the processor can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the touch display; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor, but implemented by a single chip.

[0160] Among them, the memory 212 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 212 includes a non-transitory computer-readable storage medium. The memory 212 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 212 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch instructions, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 212 may also be optionally at least one storage device located away from the aforementioned processor.

[0161] An embodiment of the present application also provides a computer-readable storage medium, which can store multiple instructions. The instructions are suitable for loading and executing the method steps of the above embodiment by a control device. The specific execution process can be found in the specific description of the above embodiment, which will not be repeated here.

[0162] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.< / sm>

Claims

1. A method for generating a multi-scenario soil moisture dataset, characterized in that: The following steps are involved: Obtain CMIP6 multi-scenario soil moisture and temperature sample data for the global range within the preset time; The CMIP6 multi-scenario soil moisture and temperature sample data are divided according to several preset spatiotemporal sizes, respectively, and input into a pre-constructed spatiotemporal feature extraction module, the regression loss of the spatiotemporal feature extraction module is calculated, and the preset spatiotemporal size corresponding to the minimum regression loss in the regression loss of the spatiotemporal feature extraction module is obtained as the target spatiotemporal size; wherein the spatiotemporal feature extraction module is a module constructed based on the spatiotemporal attention mechanism for extracting the spatiotemporal features of time series dynamic variables, the time series dynamic variables include the CMIP6 multi-scenario soil moisture and temperature sample data, and the spatiotemporal attention mechanism includes a joint spatiotemporal attention mechanism , a segmented spatiotemporal attention mechanism and a decomposition-based dot product spatiotemporal attention mechanism, wherein the joint spatiotemporal attention mechanism is an end-to-end spatiotemporal attention mechanism, which is used to decompose the input soil moisture sample sequence into non-overlapping spatiotemporal image blocks and calculate the attention between each pair of the spatiotemporal image blocks; the temporal attention of the segmented spatiotemporal attention mechanism compares the spatiotemporal image blocks at different times and the same spatial position, and the spatial attention compares the spatiotemporal image blocks at the same time and different spatial positions; the decomposition-based dot product spatiotemporal attention mechanism is to perform parallel calculations on the spatial attention and temporal attention in the segmented spatiotemporal attention mechanism; Acquire soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data on a global scale within the preset time, unify the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data to the target spatiotemporal dimensions, and construct a multi-scenario soil moisture and temperature sample database; Constructing a soil moisture quantitative estimation model including the spatiotemporal feature extraction module; According to the multi-scenario soil moisture and temperature sample database, the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data and soil moisture error sample data are used as input, and the SMAP soil moisture data is used as output, the soil moisture quantitative estimation model is trained to obtain the multi-scenario soil moisture quantitative estimation model, including: obtaining sample data corresponding to a single scenario in the multi-scenario soil moisture and temperature sample database; According to the sample data corresponding to the single scenario, the joint spatiotemporal attention mechanism, the segmentation spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism of the soil moisture quantitative estimation model are trained respectively to obtain the model regression losses corresponding to the joint spatiotemporal attention mechanism, the segmentation spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism respectively; Obtaining a spatiotemporal attention mechanism corresponding to a minimum model regression loss, and generating a soil moisture quantitative estimation model corresponding to the single scenario according to the spatiotemporal attention mechanism corresponding to the minimum model regression loss; Repeat the above steps to obtain the soil moisture quantitative estimation model corresponding to each single scenario, and form a multi-scenario soil moisture quantitative estimation model; The CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data are obtained on a global scale within the target prediction time and input into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture dataset.

2. The method for generating a multi-scenario soil moisture dataset according to claim 1, characterized in that: The steps of constructing a soil moisture quantitative estimation model including the spatiotemporal feature extraction module include: Constructing a spatial feature extraction module including several convolution submodules, wherein the spatial feature extraction module is used to extract spatial features of time series static variables, wherein the time series static variables include soil physical property sample data, underlying surface property sample data, soil moisture error sample data and soil physical property target data, underlying surface property target data, and soil moisture error target data; Based on the fully connected layer, a feature fusion and relationship simulation module is constructed to obtain the soil moisture quantitative estimation model, wherein the feature fusion and relationship simulation module is used to fuse the spatiotemporal features and the spatial features and fit their nonlinear relationship with the SMAP soil moisture data.

3. The method for generating a multi-scenario soil moisture dataset according to claim 1, characterized in that: Before training the soil moisture quantitative estimation model, the following steps are also included: Running hyperparameter iteration, inputting preset initial hyperparameters into a preset maximization acquisition function, and obtaining a hyperparameter vector, wherein the hyperparameters are contents with multiple setting options in the soil moisture quantitative estimation model, including: model structure hyperparameters and model training hyperparameters; Inputting the hyperparameter vector into the probabilistic proxy model to obtain the objective function value; updating the probabilistic proxy model according to the objective function value and the hyperparameter vector; Record the current number of iterations. If the current number of iterations exceeds the preset number of iterations, use the objective function value as the optimal hyperparameter. Otherwise, update the initial hyperparameter to the objective function value, return to the step of inputting the initial hyperparameter into the maximization acquisition function to obtain the hyperparameter vector.

4. The method for generating a multi-scenario soil moisture dataset according to claim 1, characterized in that: Obtaining CMIP6 multi-scenario soil moisture and temperature sample data for the global range within a preset time also includes the following steps: Obtain CMIP6 multi-scenario soil moisture and temperature data, SMAP soil moisture data, and ERA5-Land soil moisture data for the global range within the preset time; The CMIP6 multi-scenario soil moisture and temperature data, SMAP soil moisture data, and ERA5-Land soil moisture data are sampled to a preset spatial resolution; A set of CMIP6 single scenario soil moisture and temperature data in the CMIP6 multi-scenario soil moisture and temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data are input into the improved triple error matching model to calculate the soil moisture evaluation information of the CMIP6 single scenario soil moisture and temperature data. The improved triple error matching model is: Among them, the is the true value of soil moisture, , and They are a set of CMIP6 single scenario soil moisture and temperature data in the CMIP6 multi-scenario soil moisture and temperature data, the SMAP soil moisture data, and the ERA5-Land soil moisture data. , , and , , are the intercept and slope of the corresponding linear equation, , and are the errors of the corresponding soil moisture observation datasets, and , and The error variance, , and The correlation coefficient between the soil moisture value and , and The signal-to-noise ratio is used as the soil moisture evaluation information of the CMIP6 single scenario soil moisture and temperature data; According to the soil moisture evaluation information, several groups of CMIP6 single scenario soil moisture and temperature data in the CMIP6 multi-scenario soil moisture and temperature data are selected to form the CMIP6 multi-scenario soil moisture and temperature sample data.

5. The method for generating a multi-scenario soil moisture dataset according to claim 1, characterized in that: The steps to construct a multi-scenario soil moisture and temperature sample database include: Obtain the CMIP6 soil moisture and temperature sample data corresponding to a single scenario in the CMIP6 multi-scenario soil moisture and temperature sample data; Extracting the SMAP soil moisture image block in the SMAP soil moisture data corresponding to the single scenario in the order of first row and then column using a sliding window of preset specifications; The SMAP soil moisture image block is matched with the corresponding CMIP6 soil moisture and temperature sample data of 12 consecutive months and the soil physical property sample data, the underlying surface property sample data, and the soil moisture error sample data according to the temporal and spatial attributes to form an input-output sample pair as sample data; The sample data of each single scenario in the CMIP6 multi-scenario soil moisture and temperature sample data are obtained to form the multi-scenario soil moisture and temperature sample library.

6. The method for generating a multi-scenario soil moisture dataset according to claim 1, characterized in that: The steps of obtaining CMIP6 soil moisture and temperature target data, soil physical property target data, underlying surface property target data, and soil moisture error target data within the target prediction time and inputting them into the multi-scenario soil moisture quantitative estimation model include: Obtain CMIP6 multi-scenario soil moisture and temperature forecast data, soil physical property forecast data, underlying surface property forecast data, and soil moisture error forecast data within the target forecast time; The CMIP6 multi-scenario soil moisture and temperature prediction data are temporally and spatially matched with the soil physical property prediction data, the underlying surface property prediction data, and the soil moisture error prediction data to obtain the CMIP6 multi-scenario soil moisture and temperature target data, the soil physical property target data, the underlying surface property target data, and the soil moisture error target data; According to each grid pixel in the global range, the data sequence corresponding to the CMIP6 multi-scenario soil moisture and temperature target data and the auxiliary data image blocks corresponding to the soil physical property target data, the underlying surface property target data, and the soil moisture error target data are respectively extracted; Inputting the data sequence and the auxiliary data image block into the multi-scenario soil moisture quantitative estimation model to obtain a soil moisture fusion estimation value; Repeat the above steps to calculate the fused soil moisture estimation value corresponding to each target spatiotemporal size and form a multi-scenario soil moisture dataset.

7. A multi-scenario soil moisture dataset generation system, characterized in that: include: A sample data acquisition unit is used to obtain CMIP6 multi-scenario soil moisture and temperature sample data on a global scale within a preset time; The spatiotemporal size acquisition unit is used to divide the CMIP6 multi-scenario soil moisture and temperature sample data according to a number of preset spatiotemporal sizes, input them into a pre-built spatiotemporal feature extraction module, calculate the regression loss of the spatiotemporal feature extraction module, and obtain the preset spatiotemporal size corresponding to the minimum regression loss in the regression loss of the spatiotemporal feature extraction module as the target spatiotemporal size; wherein the spatiotemporal feature extraction module is a module constructed based on the spatiotemporal attention mechanism for extracting the spatiotemporal features of time series dynamic variables, the time series dynamic variables include the CMIP6 multi-scenario soil moisture and temperature sample data, and the spatiotemporal attention mechanism includes a joint spatiotemporal Attention mechanism, split spatiotemporal attention mechanism and decomposition-based dot product spatiotemporal attention mechanism, wherein the joint spatiotemporal attention mechanism is an end-to-end spatiotemporal attention mechanism, which is used to decompose the input soil moisture sample sequence into non-overlapping spatiotemporal image blocks and calculate the attention between each pair of spatiotemporal image blocks; the temporal attention of the split spatiotemporal attention mechanism compares the spatiotemporal image blocks at different times and the same spatial position, and the spatial attention compares the spatiotemporal image blocks at the same time and different spatial positions; the decomposition-based dot product spatiotemporal attention mechanism is to parallelly calculate the spatial attention and temporal attention in the split spatiotemporal attention mechanism; A database construction unit is used to obtain soil physical property sample data, underlying surface property sample data, soil moisture error sample data and SMAP soil moisture data in the global range within the preset time, unify the spatiotemporal dimensions of the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data, soil moisture error sample data and the SMAP soil moisture data to the target spatiotemporal dimensions, and construct a multi-scenario soil moisture and temperature sample database; A model building unit, used to build a soil moisture quantitative estimation model including the spatiotemporal feature extraction module; A model training unit is used to train the soil moisture quantitative estimation model according to the multi-scenario soil moisture and temperature sample database, with the CMIP6 multi-scenario soil moisture and temperature sample data, soil physical property sample data, underlying surface property sample data and soil moisture error sample data as input, and the SMAP soil moisture data as output, to obtain the multi-scenario soil moisture quantitative estimation model, including: obtaining sample data corresponding to a single scenario in the multi-scenario soil moisture and temperature sample database; According to the sample data corresponding to the single scenario, the joint spatiotemporal attention mechanism, the segmentation spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism of the soil moisture quantitative estimation model are trained respectively to obtain the model regression losses corresponding to the joint spatiotemporal attention mechanism, the segmentation spatiotemporal attention mechanism and the decomposition-based dot product spatiotemporal attention mechanism respectively; Obtaining a spatiotemporal attention mechanism corresponding to a minimum model regression loss, and generating a soil moisture quantitative estimation model corresponding to the single scenario according to the spatiotemporal attention mechanism corresponding to the minimum model regression loss; Repeat the above steps to obtain the soil moisture quantitative estimation model corresponding to each single scenario, and form a multi-scenario soil moisture quantitative estimation model; The product generation unit is used to obtain the CMIP6 multi-scenario soil moisture and temperature target data, soil physical property target data, underlying surface property target data and soil moisture error target data on a global scale within a target prediction time and input them into the multi-scenario soil moisture quantitative estimation model to obtain a multi-scenario soil moisture data set.

8. A computer device comprising: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for generating a multi-scenario soil moisture dataset according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for generating a multi-scenario soil moisture dataset are implemented.

Images