Methods and related devices for spatiotemporal reconstruction of water resources and drought conditions using coupled gravity satellite and multi-source high-resolution data

Abstract

This application provides a method and related apparatus for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite data with multi-source high-resolution data, relating to the field of hydrology. Specifically, by constructing a multi-scale hydrological relationship map including gravity grid nodes, watershed nodes, reservoir nodes, river segment nodes, and pixel nodes of a preset scale, macroscopic anomaly information can be progressively transmitted under the real hydrological connectivity. Under the constraint of mass conservation, the anomaly propagates along the map structure and is scaled down to the pixel level. Then, combined with the spatial influence range of each reservoir node, it is aggregated to generate a reservoir group hydrological drought index. In this way, the physical constraints of gravity satellite data on regional total water volume changes are preserved, while incorporating the spatial structure of the actual hydrological response of the reservoir group, enabling drought identification to be applied to specific reservoir areas and downstream water supply targets.

Description

Technical Field

[0001] This application relates to the field of hydrology, specifically to a method and related apparatus for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite data with multi-source high-resolution data. Background Technology

[0002] High-altitude regions (such as the Qinghai-Tibet Plateau) are often considered important water sources, and their reservoirs, lakes, snowmelt, active permafrost, and river networks together constitute a highly coupled high-altitude hydrological system. Influenced by multiple factors such as climate warming, changes in precipitation phases, premature snowmelt, permafrost degradation, and human-induced water regulation, the hydrological drought evolution of the complex river system and reservoir groups in this region exhibits strong nonlinear characteristics, spatial heterogeneity, and spatiotemporal lag.

[0003] Existing macrogravity satellites can provide information on regional total water storage anomalies, making them suitable for reflecting large-scale water volume changes. However, their spatial resolution is typically on the order of hundreds of kilometers, making it difficult to directly depict microscopic water deficit processes at the scale of reservoir shorelines, river sections, local catchment areas, and downstream water supply targets. Therefore, relying solely on GRACE or GRACE-FO products for drought identification can easily lead to problems such as insufficient spatial detail, unclear reservoir response, and blurred local drought boundaries.

[0004] While high-resolution radar and high-resolution optical remote sensing can provide detailed surface features such as water body boundaries, humidity, vegetation, snow cover, surface temperature, and topography, their observations mainly reflect surface conditions or near-surface responses and cannot directly represent anomalies in total water storage. Meanwhile, meteorological, hydrological, and reservoir management data generally suffer from problems such as sparse station coverage, inconsistent time scales, missing data for reservoir areas, and difficulty in separating the impacts of cross-basin regulation.

[0005] Against this backdrop, establishing the relationship between macro-scale water storage anomalies and smaller-scale reservoir group hydrological drought indices has become a critical issue that urgently needs to be addressed. Summary of the Invention

[0006] To overcome at least one deficiency in the prior art, this application provides a method and related apparatus for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite and multi-source high-resolution data. It can both retain the physical constraints of gravity satellite on the changes in total regional water volume and incorporate the spatial structure of the actual hydrological response of the reservoir group, so that drought identification can be applied to specific reservoir areas and downstream water supply objects.

[0007] In a first aspect, this application provides a method for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite data with multi-source high-resolution data, the method comprising: The system acquires water storage anomaly information and multi-source hydrological observation data for a target area. The target area is divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information is obtained from gravity satellite observation information. Based on water storage anomaly information and multi-source hydrological observation data, a multi-scale hydrological relationship map is constructed in the target area. The multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes, and pixel nodes of a preset scale in the target area. Based on the constraint of mass conservation, the water storage anomaly information is propagated along the multi-scale hydrological relationship map to obtain the water storage anomaly reconstruction field of the preset scale. Based on the reconstructed field of water storage anomalies and the reservoir nodes, the hydrological drought index of the reservoir group is obtained.

[0008] Secondly, this application provides a device for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite and multi-source high-resolution data, the device comprising: The water storage anomaly module is used to acquire water storage anomaly information in a target area, wherein the target area is divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information is obtained from the observation information of the gravity satellite. The hydrological relationship module is used to construct a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data. The multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes and pixel nodes of a preset scale within the target area. The water storage reconstruction module is used to propagate the water storage anomaly information along the multi-scale hydrological relationship map based on the constraint of mass conservation, so as to obtain the water storage anomaly reconstruction field of the preset scale. The drought analysis module is used to construct a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data.

[0009] Thirdly, this application provides a storage medium storing a computer program, which, when executed by a processor, implements the spatiotemporal reconstruction method for water resources and drought conditions using coupled gravity satellite and multi-source high-resolution data.

[0010] Fourthly, this application provides an electronic device, which includes a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the spatiotemporal reconstruction method for water resources and drought conditions using coupled gravity satellite and multi-source high-resolution data.

[0011] Compared with the prior art, this application has the following beneficial effects: The method and related apparatus for spatiotemporal reconstruction of water resources and drought conditions using coupled gravity satellite and multi-source high-resolution data provided in this application construct a multi-scale hydrological relationship map containing gravity grid nodes, watershed nodes, reservoir nodes, river section nodes, and pixel nodes of a preset scale. This allows macroscopic anomaly information to be transmitted step-by-step under the real hydrological connectivity. Under the constraint of mass conservation, the anomaly propagates along the map structure and is scaled down to the pixel level. Then, it is aggregated with the spatial influence range of each reservoir node to generate a reservoir group hydrological drought index. In this way, the physical constraints of gravity satellite on regional total water volume changes are preserved, while incorporating the spatial structure of the actual hydrological response of the reservoir group, enabling drought identification to be applied to specific reservoir areas and downstream water supply targets. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating the spatiotemporal reconstruction method for water resources and drought conditions using coupled gravity satellite and multi-source high-resolution data provided in this application embodiment; Figure 2 A schematic diagram of the inference process of a graph neural network provided in an embodiment of this application; Figure 3 A schematic diagram illustrating the training process of a graph neural network provided in an embodiment of this application; Figure 4 A schematic diagram of the structure of the spatiotemporal reconstruction device for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data provided in this application embodiment; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of this application (hereinafter referred to as "the embodiments") clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0015] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0016] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0017] In the description of this application, it should be noted that the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0018] Based on the above statements, as introduced in the background section, existing macrogravity satellites (GRACE / GRACE-FO) can provide information on regional total water storage anomalies, but their spatial resolution is limited to the hundred-kilometer level, making it difficult to describe small-scale water shortage processes in reservoirs, river sections, and local catchment areas. This results in insufficient spatial detail and unclear reservoir response when identifying drought conditions. While high-resolution radar and optical remote sensing can acquire detailed surface features such as water body boundaries, humidity, and vegetation, they only reflect near-surface conditions and cannot directly represent total water storage anomalies. Meteorological, hydrological, and reservoir scheduling data also suffer from problems such as sparse stations, inconsistent time scales, missing reservoir data, and difficulty in separating the impact of cross-basin regulation. Therefore, establishing the relationship between macro-level water storage anomalies and smaller-scale reservoir group hydrological drought indices has become an urgent problem to be solved.

[0019] It should be noted that the defects in the solutions in the prior art are the result of practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the embodiments of this application in the following text should be regarded as contributions to this application in the process of invention and creation, and should not be understood as technical content known to those skilled in the art.

[0020] Based on the discovery of the above-mentioned technical problems, this embodiment provides a method for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite data with multi-source high-resolution data. For example... Figure 1As shown, the method includes: S1, acquires information on water storage anomalies in the target area and multi-source hydrological observation data.

[0021] The target area was divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information was obtained from the observation information of gravity satellites.

[0022] S2. Based on water storage anomaly information and multi-source hydrological observation data, construct a multi-scale hydrological relationship map of the target area.

[0023] Among them, the multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes, and pixel nodes of a preset scale within the target area. S3, based on the constraint of mass conservation, the water storage anomaly information is propagated along the multi-scale hydrological relationship map to obtain the water storage anomaly reconstruction field of the preset scale. S4, based on the anomaly in water storage, reconstruct the field and reservoir nodes to obtain the hydrological drought index of the reservoir group.

[0024] This embodiment can be understood as follows: by constructing a multi-scale hydrological relationship map that includes gravity grid nodes, watershed nodes, reservoir nodes, river segment nodes, and pixel nodes of a preset scale, macroscopic anomaly information can be transmitted step by step under the real hydrological connectivity relationship. Under the constraint of mass conservation, the anomaly propagates along the map structure and is scaled down to the pixel level. Then, it is aggregated with the spatial influence range of each reservoir node to generate the reservoir group hydrological drought index.

[0025] In this way, the physical constraints of gravity satellites on changes in total regional water volume are preserved, while the spatial structure of the actual hydrological response of the reservoir group is incorporated, enabling drought identification to be applied to specific reservoir areas and downstream water supply targets.

[0026] It should be noted that the electronic device implementing this method for spatiotemporal reconstruction of water resources and drought conditions using coupled gravity satellite and multi-source high-resolution data can be, but is not limited to, a server, mobile terminal, tablet computer, laptop computer, desktop computer, etc. It should be explained that the server can be a single server or a group of servers. The server group can be centralized or distributed (e.g., the server can be a distributed system). In some embodiments, the server can be local or remote relative to the user terminal. In some embodiments, the server can be implemented on a cloud platform; by way of example only, the cloud platform can include private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, inter-cloud, multi-cloud, etc., or any combination thereof. In some embodiments, the server can be implemented on an electronic device having one or more components.

[0027] To make the solution provided in this embodiment clearer, a server is used as an example below, and in conjunction with... Figure 1 Each step of the method is described in detail. However, it should be understood that the operations in the flowchart may not be implemented in sequence, and steps without logical contextual relationships may be reversed in order or implemented simultaneously. Furthermore, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0028] See also Figure 1 Next, we will elaborate on step S1: S1, acquires information on water storage anomalies in the target area and multi-source hydrological observation data.

[0029] The target area was divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information was obtained from the observation information of gravity satellites.

[0030] Regarding this water storage anomaly information, it should be understood that the server can access raw observation data from the Gravity Recovery and Climate Experiment Follow-On (GRACE-FO) satellite and extract water storage anomaly information relative to the long-term mean of total land water storage based on its inversion results. This includes anomalies in the form of equivalent water thickness anomalies and the uncertainties corresponding to these anomalies. The anomalies reflect the increase or decrease in total land water storage (including the total volume of groundwater, soil water, snow cover, glacial meltwater, lakes, reservoirs, and rivers) relative to the long-term mean within a certain period.

[0031] Therefore, the information on water storage anomalies here is not obtained through direct measurement, but rather through inversion calculations based on satellite orbit measurement results. Its spatial coverage is a gravity grid on the order of hundreds of kilometers, with each grid representing an independent observation unit. The grid boundary is jointly determined by the spatial resolution of satellite observations and the data processing grid scheme.

[0032] In practical applications, the multi-source hydrological observation data synchronously accessed by the server includes high-resolution radar imagery, high-resolution optical imagery, digital elevation data, meteorological data, hydrological runoff data, reservoir water level and capacity data, and human activity data. Among them, high-resolution radar imagery is used to provide water body boundaries, surface humidity, freeze-thaw status, and reservoir bank change characteristics; high-resolution optical imagery is used to extract water body indices, vegetation indices, snow cover indices, and surface temperature; digital elevation data is used to derive slope, aspect, topographic humidity index, and runoff path; meteorological data covers precipitation, air temperature, evapotranspiration, wind speed, and radiation; hydrological runoff data includes measured or assimilated runoff and soil moisture; reservoir water level and capacity data includes water level, capacity, inflow, outflow, and scheduling records; and human activity data involves water consumption, irrigation intensity, land use change, engineering water storage and release behavior, and population or industrial water use intensity.

[0033] Given the diverse sources, formats, and spatiotemporal resolutions of this data, the server also needs to perform unified coordinate transformation, monthly-scale temporal resampling, resolution labeling, outlier identification and removal on all data, and generate cloud, snow, and water body masks and quality masks to ensure that data from different sources can be processed under the same spatiotemporal reference. The quality mask characterizes the reliability of each spatial pixel or observation data. It should be understood that due to cloud and snow cover, radar decoherence, terrain obscuring, sensor anomalies, or low signal-to-noise ratios, unreliable areas exist in remote sensing / observation information. This quality mask is used to dynamically reduce weights or mask invalid inputs in data quality control, graph neural network training, and quality conservation correction.

[0034] Specifically, gravity satellite data itself includes leakage correction coefficients and scale factors to correct signal attenuation and edge distortion caused by spatial filtering during satellite inversion. Synthetic aperture radar backscattering, interferometric coherence, and optical surface reflectivity in high-resolution remote sensing data are all indirect observation variables reflecting the physical state of the surface. They do not directly characterize water volume, but can constrain the distribution pattern and dynamic response of water on the surface. Snow water equivalent and permafrost active layer thickness in meteorological and hydrological data further reflect the special hydrological processes in the high-altitude and cold environment of the Qinghai-Tibet Plateau. The thickness of the permafrost active layer is not a fixed parameter, but a dynamic physical quantity that fluctuates with seasons and climate, directly affecting infiltration capacity and runoff generation mechanism.

[0035] Based on the above embodiments regarding the anomaly information on water storage and multi-source hydrological observation data, the following will continue to discuss... Figure 1 Step S2 will be explained below: S2. Based on water storage anomaly information and multi-source hydrological observation data, construct a multi-scale hydrological relationship map of the target area.

[0036] Among them, the multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river segment nodes, and pixel nodes of preset scale within the target area.

[0037] As an optional implementation, this embodiment provides the following optional implementation of step S2: S2-1: Based on the watershed vector boundary, reservoir area spatial range, river network topology and pixel size in the target area, construct watershed nodes, reservoir nodes, river segment nodes and pixel nodes.

[0038] S2-2, based on the relationship between hydrological physical connectivity and human regulation, establishes connection edges representing real hydrological processes between gravity grid nodes, watershed nodes, reservoir nodes, river segment nodes and pixel nodes.

[0039] The connection information of the connecting edge includes time delay information, edge weight, distance, quality identifier, and edge type. The time delay information represents the time delay characteristics of the hydrological process response, the edge weight represents the intensity of hydrological influence between nodes, the distance represents the physical path scale of the hydrological connection between nodes, the quality represents the reliability of the data on which the connecting edge depends, and the edge type represents the hydrological physical relationship between nodes.

[0040] S2-3, based on water storage anomaly information and multi-source hydrological observation data, configure attribute information for watershed nodes, reservoir nodes, river section nodes and pixel nodes to form a multi-scale hydrological relationship map.

[0041] Taking the Qinghai-Tibet Plateau as an example, it can be understood that constructing a multi-scale hydrological relationship map requires organizing the real hydrological connections in the complex water system of the Qinghai-Tibet Plateau into a graph data model that can be recognized and calculated by graph neural networks.

[0042] In practical applications, the server first determines natural confluence units based on the watershed vector boundary of the target area, using them as watershed nodes; it then extracts the geometric center and influence boundary based on the spatial range vector contour of the reservoir area, forming reservoir nodes; based on the river centerline, confluence relationships, and flow direction information in the river network topology, it divides river segment nodes with clear upstream and downstream relationships; finally, it uniformly rasterizes the entire study area into 30-meter resolution surface pixels, with each pixel serving as an independent pixel node. These four types of nodes, together with gravity grid nodes, constitute five layers. The gravity grid nodes correspond to the 100-kilometer-scale spatial units observed by the GRACE-FO gravity satellite, which are much larger than the preset-scale pixel grid.

[0043] Based on this, the server establishes connection edges between these five types of nodes according to the hydrological and physical connectivity and human regulation. Specifically, gravity grid nodes are connected to watershed nodes that they cover or overlap with, and the edge weight of this connection edge is jointly determined by the overlapping area of ​​the two, the GRACE-FO leakage correction coefficient, and grid uncertainty; watershed nodes are connected to reservoir nodes, and the edge weight is comprehensively determined by the proportion of the watershed's contribution to the reservoir's catchment area, the reservoir's controlled area, and the historical inflow; reservoir nodes are connected to river segment nodes, and the edge weight is jointly characterized by the river network connectivity direction, the actual length of the river segment, and the time delay characteristics reflected in the scheduling instructions; reservoir nodes and river segment nodes are further connected to 30-meter pixel nodes, and the edge weight is quantified based on spatial adjacency, topographic slope aspect, distance attenuation effect from the pixel to the shoreline or river channel, and similarity of surface types (e.g., bare soil to bare soil, water body to water body).

[0044] Therefore, the connection information for all connected edges includes time delay information, edge weight, distance, quality identifier, and edge type. Among them, the time lag information is characterized by the time delay characteristics of the hydrological process response in months. For example, the lag time for snowmelt replenishment to reservoir water level change may be in months, and the lag time for precipitation infiltration to groundwater replenishment may also be in months. The edge weight is characterized by a dimensionless value to represent the relative intensity of hydrological influence between nodes. The larger the value, the more direct and significant the influence. The distance is the actual hydrological distance along the water flow path, rather than the straight-line distance between two points. For example, the distance from the outlet of a sub-basin to the reservoir dam site needs to be measured along the main river channel. The quality label reflects the credibility level of the data on which the edge depends. For example, the edge of the inflow calculated based on measured hydrological station data is labeled as high, while the edge based solely on remote sensing water surface area is labeled as medium or low. The edge type clearly distinguishes different hydrological physical relationships such as "gravity-basin affiliation", "basin-reservoir confluence", "reservoir-river section scheduling", "reservoir-pixel water supply", and "river section-pixel confluence" to ensure that the physical source of each edge can be traced.

[0045] In addition, the server configures attribute information for each node based on water storage anomaly information and multi-source hydrological observation data. Among them, the attributes of gravity grid nodes include GRACE-FO equivalent water thickness anomaly value, leakage correction coefficient, scale factor and its uncertainty; the attributes of watershed nodes include watershed vector boundary, multi-year average precipitation, evapotranspiration, baseflow index, and permafrost distribution ratio; the attributes of reservoir nodes include reservoir area spatial range, water level-storage capacity curve, inflow and outflow records, and scheduling rules; the attributes of river segment nodes include river network topology number, river length, roughness parameter, and ice condition observation status; the attributes of 30-meter pixel nodes include spatial coordinates, elevation, slope aspect, D8 flow direction, Synthetic Aperture Radar (SAR) backscattering intensity, optical surface reflectivity, Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), snow cover index, surface temperature, and topographic humidity index.

[0046] Finally, the server records the node fields of the above multi-scale hydrological relationship graph as node number, node type, spatial range, scale identifier, feature vector and uncertainty; and the edge fields as start node number, end node number, edge type, hydrological distance, time lag month, edge weight and quality identifier, thus constructing the multi-scale hydrological relationship graph.

[0047] In this way, the spatial scale identifier and quality mask of each type of node are preserved, and the cross-scale attention aggregator can automatically identify and reduce the weight of missing or low-quality observation edges when propagating features, avoiding mistaking unreliable connections as strong supervision signals. Ultimately, when the mass conservation correction layer aggregates the output of 30-meter pixel nodes back to the gravity grid nodes, the loss calculation has physical consistency and data reliability.

[0048] Based on the multi-scale hydrological relationship map constructed in the above embodiments, the following will continue to... Figure 1 Step S3 will be explained below: S3, based on the constraint of mass conservation, propagates the water storage anomaly information along the multi-scale hydrological relationship map to obtain the water storage anomaly reconstruction field of the preset scale.

[0049] The preset scale can be, but is not limited to, 10 meters, 20 meters, 30 meters, 40 meters, 50 meters, etc. Taking 30 meters as an example, it should be noted that the water storage anomalies at the scale of hundreds of kilometers provided by GRACE or GRACE-FO gravity satellites reflect the trend of total regional water volume changes, but cannot express local deficiencies at the reservoir shoreline, river section, or 30-meter pixel scale. Similarly, 30-meter surface features acquired by high-resolution radar and optical remote sensing, such as water body boundaries, soil moisture, or vegetation status, only characterize surface or near-surface responses and cannot represent the total water storage in underground, snow, ice, and permafrost. In related technologies, simply splicing or linearly interpolating these two methods cannot guarantee the conservation of macroscopic total volume, nor can it reflect real hydrological processes such as watershed confluence and reservoir operation, resulting in fitted results that fail to reflect the actual situation.

[0050] To address the above problems, this embodiment also provides the following optional implementation methods for step S3: S3-1 uses the water storage anomaly value of each gravity grid node as the total constraint; through graph neural network, cross-scale feature propagation is performed in the multi-scale hydrological relationship map to obtain the prior water storage anomaly value of each pixel node.

[0051] As an optional implementation, the water storage anomaly information for each gravity grid node also includes the uncertainty of the water storage anomaly value. For each gravity grid node, the server can encode the water storage anomaly value and uncertainty of the gravity grid node to obtain a coarse-scale feature vector; based on the watershed affiliation, river network connectivity, and reservoir adjacency of the gravity grid node in the multi-scale hydrological relationship map, cross-scale attention aggregation is performed on the coarse-scale feature vector to transmit the target pixel node that has a connection relationship with the gravity grid node; the attribute information of the target pixel node is fused to decode and generate the prior water storage anomaly value of the target pixel node.

[0052] S3-2, the prior water storage anomaly values ​​of all pixel nodes in the same gravity grid are weighted by their respective areas and then regressed to obtain a water storage anomaly reconstruction field of a preset scale.

[0053] It should be understood that the water storage anomaly values ​​obtained from gravity satellite observations are not absolutely accurate measurement results. The inversion process is affected by factors such as orbital errors, atmospheric and ocean dealiasing model biases, leakage effects, and spatial smoothing, and is inherently uncertain.

[0054] like Figure 2 , Figure 3As shown, the graph neural network includes coarse-scale encoding, a cross-scale attention aggregator, a fine-scale decoder, and a mass conservation correction layer. In practical applications, the server first uses coarse-scale encoding to jointly encode the water storage anomalies and uncertainties of each gravity grid node, generating a physically meaningful coarse-scale feature vector. Therefore, when the uncertainty of a certain gravity grid... When the value is large, its corresponding feature vector will be reduced in weight during subsequent propagation to avoid over-imposing unreliable macroscopic signals into the fine-scale space.

[0055] During this process, the server performs cross-scale attention aggregation operations through a cross-scale attention aggregator based on the defined connections in the multi-scale hydrological relationship map. For any gravity grid node, it identifies the watershed nodes, sub-watershed nodes, or reservoir nodes directly connected to it in the map, and further extends the coarse-scale features downstream to target pixel nodes with real hydrological connections along the river network connectivity path, topographic confluence direction, and reservoir adjacency relationship.

[0056] Therefore, cross-scale attention aggregation must follow the logic of real-world hydrology on the plateau. If a gravity grid covers multiple sub-basins, its features are preferentially propagated to sub-basin nodes with large overlapping areas and low uncertainty. If a sub-basin contains a reservoir, the features are further directed to target pixel nodes with close physical connections along the river network connectivity, reservoir adjacency relationships, and topographic confluence paths. These connections are solidified in the graph structure as edges, and the propagation process is dynamically adjusted by edge weights.

[0057] As described in the above embodiments, the edge weights between gravity grid nodes and sub-basin nodes The overlapping area of ​​the two With gravity grid uncertainty and small perturbation terms The decision is made jointly, and the expression is:

[0058] Therefore, the greater the uncertainty, the lower the weight of the connection channel. Similarly, the expressions for calculating the edge weights between the reservoir node and the pixel node, and between adjacent pixel nodes, are as follows:

[0059]

[0060] In the formula, the shoreline distance is... Elevation difference D8 Flow Direction Marker and the difference in topographic humidity index All of these are embedded as physical constraints on the edge weights to ensure that the propagation path conforms to the actual hydrological conditions of the plateau.

[0061] Finally, the server fuses the attribute information of the target pixel node itself through a fine-scale decoder to decode the prior water storage anomaly value of the generated target pixel node. This attribute information includes, but is not limited to, the water surface / wet surface state characterized by the backscatter intensity of high-resolution synthetic aperture radar (SAR), water body indices extracted from optical images (such as NDWI), vegetation indices (such as NDVI), snow indexes (such as NDSI), surface temperature, topographic humidity index (TWI) derived from digital elevation model (DEM), slope, flow direction, and precipitation, evapotranspiration and soil moisture anomalies provided by meteorological data.

[0062] In this way, by combining these information together in the decoding, the output a priori water storage anomaly value is both constrained by macroscopic water volume and embedded with local surface process response characteristics.

[0063] It should be noted that this prior water storage anomaly is merely an intermediate result of the model inference and does not yet meet the physical conservation requirements. During the business inference phase, the server will separately invoke the mass conservation correction layer to perform area-weighted regression correction on all pixel nodes within the same gravity grid. Specifically, it first calculates the area-weighted mean of the prior values ​​for each pixel within the grid. and credibility-weighted average Then correct it using the following expression:

[0064]

[0065]

[0066] In the formula, This is the original observed water storage anomaly for this gravity grid. This represents the prior water storage anomaly value for a specific pixel output by the GNN. Represents a cell geographical area Represents gravity grid The set of all pixels, The corresponding confidence level is estimated. Therefore, in the correction layer, the overall bias is redistributed proportionally according to the relative confidence level of each pixel. High-confidence pixels receive more correction, while low-confidence pixels basically maintain the original prediction. In this way, the final output is a 30-meter-level water storage anomaly reconstruction field. It preserves local textures such as water body boundaries revealed by SAR images, vegetation cover reflected by optical images, and topographic confluence represented by DEM, while strictly satisfying the regional total water conservation limit defined by GRACE-FO observations.

[0067] During the training phase, the server utilizes supervised samples from historical months, including monthly values ​​of Grace-FO, SAR, optical data, DEM meteorological and hydrological data, and reservoir capacity, to optimize the graph neural network model using a composite loss function constructed from the following expression:

[0068]

[0069] In the formula, This indicates the aggregation error within the gravity grid. Watershed water balance error Error in reservoir capacity variation Runoff closure error Time or space smoothing error and uncertainty weighted error Perform a weighted summation. This is used to force the area-weighted aggregate value of all cell nodes within the same gravity grid to approach the observed value of that grid. (By...) The edge attention weights and time delay weights in the graph neural network model are optimized. The graph neural network also incorporates time delay memory units to learn the time lags of precipitation, snowmelt, permafrost degradation, water storage scheduling, water intake and downstream runoff in response to water storage anomalies, and passes the feature weights within different time delay windows to the high-resolution spatiotemporal drought index generation stage.

[0070] Based on the water storage anomaly reconstruction field obtained from the above embodiments, the following will continue to... Figure 1 Step S4 will be explained in detail below: S4, based on the anomaly in water storage, reconstruct the field and reservoir nodes to obtain the hydrological drought index of the reservoir group.

[0071] It should be noted that when generating hydrological drought indices for reservoir groups, related technologies often rely on data from a single source, which makes it difficult to reflect the true drought situation of plateau reservoirs. For example, they may only use remote sensing to invert reservoir water level changes or only use precipitation anomalies from meteorological stations.

[0072] To address the above problems, this embodiment also provides the following optional implementation methods for step S4: S4-1, for each reservoir node, each pixel node within the influence range of the reservoir node is taken as an associated node.

[0073] S4-2, based on the edges and their weights between the reservoir node and each associated node, spatial weighted aggregation is performed on the water storage anomaly anomaly, snow water equivalent anomaly, soil moisture anomaly, and frozen soil active layer water storage capacity anomaly of each associated node to obtain the aggregation result.

[0074] S4-3, the aggregation results are combined with the reservoir capacity anomaly, runoff anomaly and precipitation evapotranspiration deficit of the reservoir nodes to obtain the hydrological drought index of the reservoir group.

[0075] In this embodiment, the server can identify the actual spatial range affected by each reservoir node based on the connection relationship between reservoir nodes and pixel nodes defined in the hydrological relationship diagram, and regard all pixel nodes within that range as its associated nodes.

[0076] Based on this, the server performs spatial weighted aggregation of multiple anomaly variables according to the attributes of the edges between the reservoir node and its associated nodes. These multiple anomaly variables include the water storage anomaly anomaly obtained by standardizing the 30-meter-level water storage anomaly reconstruction field. Snow water equivalent anomaly Soil moisture anomaly and the water storage capacity anomaly of the active layer of permafrost Taking water storage anomalies as an example, the standardized operation expression is:

[0077] In this expression, the water storage anomaly of each pixel is converted into a dimensionless anomaly value based on the historical distribution of the same period; the other variables are also standardized in the same way according to their respective long-term series to ensure that all inputs are comparable and consistent in direction.

[0078] The aggregation process strictly depends on edge weights, which themselves reflect the strength of physical connections. For example, in areas dominated by spring snowmelt, the weight of snow accumulation or snowmelt equivalent anomaly automatically increases; in areas where permafrost degradation and runoff generation are significant in summer, the weight of permafrost active layer water storage capacity anomaly increases accordingly; and for pixels containing downstream water supply objects, the emphasis is placed on expressing reservoir capacity and runoff information. These weights are not preset constants, but are dynamically determined by the node type, edge type, and reservoir influence zone identifier in the hydrological map.

[0079] After completing the spatial weighted aggregation, the server compares the obtained results with the reservoir capacity anomaly of the reservoir node itself. Runoff Anomaly and precipitation evapotranspiration deficit The indicators are weighted and integrated to form the hydrological drought index of the reservoir group. The expression is:

[0080] In the formula, ~ The seven weights correspond to the relative importance of different variables in a specific spatial partition, and their values ​​are determined by the watershed runoff type, reservoir function positioning, and water supply needs.

[0081] When the hydrological drought index of the reservoir group is lower than the preset occurrence threshold and the duration exceeds the preset duration, the corresponding spatiotemporal unit is marked as the drought occurrence stage; when the hydrological drought index of the reservoir group continues to decrease and the spatial connectivity area expands, it is marked as the drought development stage; when the hydrological drought index of the reservoir group reaches a minimum value or the reservoir capacity deficit reaches a maximum value, it is marked as the drought peak stage; when the hydrological drought index of the reservoir group rises above the recovery threshold and continues to exceed the recovery period, it is marked as the drought dissipation stage.

[0082] It should also be noted that, in terms of drought attribution, traditional correlation analysis or single-factor sensitivity analysis is difficult to distinguish the independent contributions of climate change and human activities under different drought stages, different spatial units and different time lag windows. Although pure machine learning methods can fit nonlinear relationships, if the constraints of mass conservation, water balance and runoff generation process in high-altitude and cold watersheds are lacking, they are prone to results with insufficient physical meaning or repeated attributions.

[0083] To address the aforementioned issues, the spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data provided in this embodiment further includes: S5, obtain climate change factors and human activity factors, and construct their respective lag characteristics.

[0084] S6, based on the drought index benchmark value under the annual average state, fixes one set of factors and perturbs another set of factors through conditional permutation, and outputs the counterfactual drought index prediction value through an interpretable machine learning model.

[0085] S7 calculates the independent contribution rates of climate change and human activities to changes in the drought index based on the difference between the predicted value of the counterfactual drought index and the baseline value of the drought index.

[0086] The server can acquire two types of inputs: a set of climate change factors and a set of human activity factors. The set of climate change factors specifically includes precipitation, temperature, evapotranspiration, shortwave radiation, wind speed, snow cover or snow water equivalent, permafrost active layer thickness, atmospheric circulation index, and extreme climate event indicators; the set of human activity factors covers reservoir operation records, total regional water withdrawal, agricultural irrigation intensity, land use change type and rate, construction sequence of major water conservancy projects, population and industrial water use intensity, and inter-basin water transfer behavior, etc.

[0087] In practical applications, the server first constructs lag windows adapted to the hydrological response characteristics of the plateau for each type of factor. For example, a ten-day lag is set to address the rapid response of spring snowmelt runoff; a monthly lag is used for the soil moisture replenishment process; a seasonal lag is introduced to address the seasonal evolution of the thermal state of the active permafrost layer; and a separate snowmelt window is constructed for the critical period of concentrated snowmelt to capture its pulse-like impact on changes in inflow and reservoir capacity.

[0088] Based on this, the server feeds these time-delayed factor features into an interpretable machine learning model to train a model that can predict water storage anomalies or the hydrological drought index of reservoir groups. This model can use one of the following: gradient boosting tree, random forest, causal forest, generalized additive model, or neural network with interpretable attention mechanism, or a combination thereof, to ensure that while fitting complex nonlinear relationships, the traceability of the influence path of each factor is preserved.

[0089] The server uses the drought index benchmark value under the average annual condition as a reference and performs a conditional permutation operation. Specifically, it first fixes all values ​​of the climate change factor set and systematically perturbs only the human activity factor set to obtain a set of counterfactual drought index predictions; then, conversely, it fixes the human activity factor set and perturbs the climate change factor set to obtain another set of counterfactual predictions. The difference between each perturbation result and the benchmark value, after normalization, constitutes the independent contribution rate of the corresponding factor set to the change in the drought index.

[0090] To prevent double counting due to the high correlation between the two types of factors, the server further employs methods such as grouped SHAP, residualized regression, or causal graph constraints to introduce conditional dependency structures during the decomposition process. The final output includes not only the independent contribution rates of climate change and human activities, but also the interaction contribution rate resulting from their interaction, as well as the residuals that the model could not explain.

[0091] Furthermore, the server can conduct the above-mentioned attribution analysis for different evolution stages of the same drought event, such as the occurrence stage, development stage, peak stage, and dissipation stage, thereby identifying the initiation process dominated by precipitation deficit, the development process driven by warming and enhanced evapotranspiration, the local peak process caused by reservoir scheduling or agricultural water use, and the dissipation process driven by snowmelt replenishment or artificial water replenishment.

[0092] It should also be noted that in high-altitude cold basins, there is a significant hydrothermal coupling relationship between snow cover, snow water equivalent, active permafrost layer thickness, surface temperature, and runoff. The thickness of the active permafrost layer alters infiltration capacity, water storage capacity, and runoff generation zones; therefore, it cannot be simply treated as a fixed surface parameter. The hydrological drought index for reservoir groups needs to integrate multiple anomaly variables, including the water storage capacity anomaly of the active permafrost layer. Snow water equivalent anomaly and runoff anomaly If hydrological processes such as snow accumulation-permafrost-runoff are not taken into account, it will lead to deviations in the prediction of spring snowmelt runoff, summer permafrost degradation runoff, and reservoir inflow.

[0093] To address the aforementioned issues, in practical applications, the server can also integrate snow cover or snow water equivalent, permafrost active layer thickness, surface temperature, and runoff observation data for unified processing. In this process, the server first receives inputs such as snow cover, snow depth, snow water equivalent, and surface albedo, updating the current snowmelt estimate and the insulating effect of the snow layer on surface heat exchange. Simultaneously, the server also receives permafrost active layer thickness, surface temperature, freezing index, thawing index, or perennial permafrost thermal state inversion results to synchronously update the permafrost thermal state evolution process and the dynamic changes in active layer water storage capacity.

[0094] Based on this, the server dynamically adjusts runoff-related parameters according to the real-time updated thickness of the active permafrost layer. When the active layer is shallow, the permafrost's impermeability increases, and the server accordingly increases the generation ratio of surface runoff and interflow, while reducing infiltration capacity. When the active layer thickness increases and the water content rises, the server increases the estimated water storage capacity of the active layer and extends the runoff response lag time. When the active layer further thickens and forms effective infiltration channels, the server increases the weight of groundwater recharge and extends the baseflow response time. Thus, the thickness of the active permafrost layer is no longer used as a fixed parameter, but rather as a key dynamic constraint affecting runoff zoning, infiltration capacity, and water storage structure, participating in the entire calculation process.

[0095] The server employs ensemble Kalman filtering, particle filtering, four-dimensional variational assimilation, or a combination thereof, to perform assimilation operations. Specifically, observation errors in snow cover or snowmelt equivalent propagate to runoff state variables through the snowmelt process model; uncertainties in the thickness of the active permafrost layer are transmitted to the runoff generation simulation results via infiltration capacity and active layer water storage capacity parameters; and runoff observation errors, through an inflow ensemble correction mechanism, conversely affect the accuracy of reservoir capacity prediction.

[0096] This improves the simulation reliability of peak spring snowmelt runoff and summer permafrost degradation runoff processes, providing boundary conditions that are more in line with the physical laws of high-altitude watersheds for drought reconstruction and scheduling early warning.

[0097] In this embodiment, for high-altitude watersheds lacking hydrological station observations, the server can also utilize graph structure parameters, snow cover-permafrost state parameters, and reservoir scheduling priors from neighboring similar watersheds for migration. During the migration process, GRACE-FO macroscopic quality constraints are preserved, and the migration results are corrected using high-resolution remote sensing water body boundaries, snow cover, surface temperature, and vegetation moisture status.

[0098] The server can output spatiotemporal reconstruction maps of drought conditions in reservoir groups, HDSI time series, drought event lists, drought levels in different reservoir-affected areas, distribution maps of climate change and human activity contribution rates, and ensemble forecasts of inflows. For reservoirs or downstream water supply objects that reach the warning threshold, warning information can be generated, including drought level, expected duration, main driving factors, recoverable conditions, and recommended scheduling windows.

[0099] Based on the same inventive concept as the spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data provided in this embodiment, this embodiment also provides a device for spatiotemporal reconstruction of water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data. This device includes at least one software functional module that can be stored in a memory or embedded in an electronic device. The processor in the electronic device executes the executable module stored in the memory. For example, the software functional modules and computer programs included in this device. Please refer to... Figure 4 Functionally, the device may include: The water storage anomaly module 11 is used to obtain water storage anomaly information in the target area. The target area is divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information is obtained from the observation information of the gravity satellite. The hydrological relationship module 12 is used to construct a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data. The multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes and pixel nodes of preset scale within the target area. The storage reconstruction module 13 is used to propagate water storage anomaly information along a multi-scale hydrological relationship map based on the constraint of mass conservation, so as to obtain a water storage anomaly reconstruction field of a preset scale. The drought analysis module 14 is used to construct a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data.

[0100] Optionally, the water storage anomaly information for each gravity grid node includes anomaly values ​​for water storage. Based on the constraint of mass conservation, the water storage anomaly module 13 propagates water storage anomaly information along a multi-scale hydrological relationship map to obtain a water storage anomaly reconstruction field at a preset scale, including: The total amount is constrained by the water storage anomaly value of each gravity grid node; cross-scale feature propagation is performed in the multi-scale hydrological relationship map through graph neural network to obtain the prior water storage anomaly value of each pixel node. The prior water storage anomaly values ​​of all pixel nodes within the same gravity grid are weighted by their respective areas and then regressed to obtain a water storage anomaly reconstruction field of a preset scale.

[0101] Optionally, the water storage anomaly information for each gravity grid node also includes the uncertainty of the water storage anomaly value. The storage reconstruction module 13 obtains the prior water storage anomaly value for each pixel node by performing cross-scale feature propagation in the multi-scale hydrological relationship map through a graph neural network, including: For each gravity grid node, the water storage anomalies and uncertainties of the gravity grid node are encoded to obtain a coarse-scale feature vector; Based on the watershed affiliation, river network connectivity, and reservoir adjacency of gravity grid nodes in the multi-scale hydrological relationship map, cross-scale attention aggregation is performed on coarse-scale feature vectors to transmit target pixel nodes that have a connection relationship with gravity grid nodes. By fusing the attribute information of the target pixel node, the prior water storage anomaly value of the target pixel node is decoded and generated.

[0102] Optionally, the drought analysis module 14 obtains the hydrological drought index of the reservoir group by reconstructing the field and reservoir nodes based on water storage anomalies, including: For each reservoir node, each pixel node within the influence range of the reservoir node is regarded as an associated node; Based on the edges and their weights between the reservoir node and each associated node, spatial weighted aggregation is performed on the water storage anomaly ... of the active layer of permafrost for each associated node to obtain the aggregation result. The aggregation results are combined with the reservoir capacity anomaly, runoff anomaly and precipitation evapotranspiration deficit of the reservoir nodes to obtain the hydrological drought index of the reservoir group.

[0103] Optionally, the drought analysis module 14 is also used to acquire climate change factors and human activity factors, and to construct their respective lag characteristics; Based on the drought index baseline value under the annual average state, one set of factors is fixed and another set of factors is perturbed by conditional permutation, and the counterfactual drought index prediction value is output by an interpretable machine learning model. Based on the difference between the predicted value of the counterfactual drought index and the baseline value of the drought index, the independent contribution rates of climate change and human activities to the changes in the drought index are calculated.

[0104] Optionally, the drought analysis module 14 constructs a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data, including: Based on the watershed vector boundary, reservoir area spatial range, river network topology and pixel size in the target area, watershed nodes, reservoir nodes, river segment nodes and pixel nodes are constructed; Based on the relationship between hydrological physical connectivity and human regulation, connection edges representing real hydrological processes are established between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes and pixel nodes. Based on water storage anomaly information and multi-source hydrological observation data, attribute information is configured for watershed nodes, reservoir nodes, river section nodes, and pixel nodes to form a multi-scale hydrological relationship map.

[0105] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0106] It should also be understood that if the above embodiments are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0107] Therefore, this embodiment also provides a storage medium, which is a computer-readable storage medium. This storage medium stores a computer program, which, when executed by a processor, implements the spatiotemporal reconstruction method for water resources and drought conditions based on coupled gravity satellite and multi-source high-resolution data provided in this embodiment. The storage medium can be any medium capable of storing program code, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0108] Please refer to Figure 5 This embodiment also provides an electronic device, which may include a processor 22 and a memory 21. The memory 21 stores a computer program, and the processor reads and executes the computer program in the memory 21 corresponding to the above-described embodiments to implement the spatiotemporal reconstruction method for water resources and drought conditions based on coupled gravity satellite and multi-source high-resolution data provided in this embodiment.

[0109] See also Figure 5The electronic device also includes a communication unit 23. The memory 21, processor 22 and communication unit 23 are electrically connected to each other directly or indirectly through system bus 24 to realize data transmission or interaction.

[0110] The memory 21 can be an information recording device based on any electronic, magnetic, optical, or other physical principles, used to record execution instructions, data, etc. In some embodiments, the memory 21 can be, but is not limited to, volatile memory, non-volatile memory, memory drive, etc.

[0111] In some embodiments, the volatile memory may be random access memory (RAM); in some embodiments, the non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, etc.; in some embodiments, the storage drive may be a disk drive, solid-state drive, any type of storage disk (such as optical disc, DVD, etc.), or similar storage media, or a combination thereof.

[0112] The communication unit 23 is used to send and receive data over a network. In some embodiments, the network may include a wired network, a wireless network, a fiber optic network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, or a near field communication (NFC) network, or any combination thereof. In some embodiments, the network may include one or more network access points. For example, the network may include wired or wireless network access points, such as base stations and / or network switching nodes, through which one or more components of the service request processing system can connect to the network to exchange data and / or information.

[0113] The processor 22 may be an integrated circuit chip with signal processing capabilities, and may include one or more processing cores (e.g., a single-core processor or a multi-core processor). By way of example only, the processor described above may include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), an Application Specific Instruction-set Processor (ASIP), a Graphics Processing Unit (GPU), a Physics Processing Unit (PPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), a Programmable Logic Device (PLD), a controller, a microcontroller unit, a Reduced Instruction Set Computing (RISC) computer, or a microprocessor, or any combination thereof.

[0114] Understandable. Figure 5The structure shown is for illustrative purposes only. Electronic devices may also have more advanced features. Figure 5 Showing more or fewer components, or having with Figure 5 The different configurations shown. Figure 5 The components shown can be implemented using hardware, software, or a combination thereof.

[0115] It should be understood that the apparatus and methods disclosed in the above embodiments can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0116] The above descriptions are merely various embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for spatiotemporal reconstruction of water resources and drought conditions by coupling gravity satellite data with multi-source high-resolution data, characterized in that, The method includes: The system acquires water storage anomaly information and multi-source hydrological observation data for a target area. The target area is divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information is obtained from gravity satellite observation information. Based on water storage anomaly information and multi-source hydrological observation data, a multi-scale hydrological relationship map is constructed in the target area. The multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes, and pixel nodes of a preset scale in the target area. Based on the constraint of mass conservation, the water storage anomaly information is propagated along the multi-scale hydrological relationship map to obtain the water storage anomaly reconstruction field of the preset scale. Based on the reconstructed field of water storage anomalies and the reservoir nodes, the hydrological drought index of the reservoir group is obtained.

2. The spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data as described in claim 1, characterized in that, The water storage anomaly information for each gravity grid node includes anomaly values ​​for water storage. Based on the constraint of mass conservation, the water storage anomaly information is propagated along the multi-scale hydrological relationship map to obtain the water storage anomaly reconstruction field at the preset scale, including: The total amount is constrained by the water storage anomaly value of each gravity grid node; cross-scale feature propagation is performed on the multi-scale hydrological relationship map through graph neural network to obtain the prior water storage anomaly value of each pixel node. The prior water storage anomaly values ​​of all pixel nodes within the same gravity grid are weighted by their respective areas and then subjected to regression correction to obtain the water storage anomaly reconstruction field at the preset scale.

3. The spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data as described in claim 2, is characterized in that, The water storage anomaly information for each gravity grid node also includes the uncertainty of the water storage anomaly value. Through a graph neural network, cross-scale feature propagation is performed on the multi-scale hydrological relationship map to obtain the prior water storage anomaly value for each pixel node, including: For each gravity grid node, the water storage anomaly and uncertainty of the gravity grid node are encoded to obtain a coarse-scale feature vector; Based on the watershed affiliation, river network connectivity, and reservoir adjacency of the gravity grid nodes in the multi-scale hydrological relationship map, cross-scale attention aggregation is performed on the coarse-scale feature vectors to pass on target pixel nodes that have a connection relationship with the gravity grid nodes; By fusing the attribute information of the target pixel node, the prior water storage anomaly value of the target pixel node is decoded and generated.

4. The spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data as described in claim 1, characterized in that, Based on the reconstructed field of water storage anomalies and the reservoir nodes, the hydrological drought index of the reservoir group is obtained, including: For each reservoir node, each pixel node within the influence range of the reservoir node is regarded as an associated node; Based on the edges and their weights between the reservoir nodes and each associated node, spatial weighted aggregation is performed on the water storage anomaly ... of the active layer of permafrost for each associated node to obtain the aggregation result. The aggregation results are then weighted and fused with the reservoir capacity anomaly, runoff anomaly, and precipitation evapotranspiration deficit of the reservoir nodes to obtain the hydrological drought index of the reservoir group.

5. The spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data as described in claim 1, characterized in that, The method further includes: Obtain climate change factors and human activity factors, and construct their respective lag characteristics; Based on the drought index baseline value under the annual average state, one set of factors is fixed and another set of factors is perturbed by conditional permutation, and the counterfactual drought index prediction value is output by an interpretable machine learning model. Based on the difference between the predicted value of the counterfactual drought index and the baseline value of the drought index, the independent contribution rates of climate change and human activities to the changes in the drought index are calculated.

6. The spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data as described in claim 1, characterized in that, Based on anomaly information in water storage and multi-source hydrological observation data, a multi-scale hydrological relationship map of the target area is constructed, including: Based on the watershed vector boundary, reservoir area spatial range, river network topology and pixel size in the target area, the watershed node, the reservoir node, the river segment node and the pixel node are constructed; Based on the relationship between hydrological physical connectivity and human regulation, connection edges representing real hydrological processes are established between the gravity grid nodes, watershed nodes, reservoir nodes, river section nodes and pixel nodes. Based on the abnormal water storage information and multi-source hydrological observation data, attribute information is configured for the watershed nodes, reservoir nodes, river section nodes and pixel nodes to form the multi-scale hydrological relationship map.

7. The spatiotemporal reconstruction method for water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data according to claim 6, characterized in that, The connection information of the connecting edge includes time delay information, edge weight, distance, quality identifier, and edge type. The time delay information represents the time delay characteristics of the hydrological process response, the edge weight represents the intensity of hydrological influence between nodes, the distance represents the physical path scale of the hydrological connection between nodes, the quality identifier represents the reliability of the data on which the connecting edge depends, and the edge type represents the hydrological physical relationship between nodes.

8. A device for spatiotemporal reconstruction of water resources and drought conditions coupled with gravity satellite and multi-source high-resolution data, characterized in that, The device includes: The water storage anomaly module is used to acquire water storage anomaly information in a target area, wherein the target area is divided into multiple gravity grids according to the gravity satellite observation scale, and the water storage anomaly information is obtained from the observation information of the gravity satellite. The hydrological relationship module is used to construct a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data. The multi-scale hydrological relationship map describes the hydrological relationships between gravity grid nodes, watershed nodes, reservoir nodes, river section nodes and pixel nodes of a preset scale within the target area. The water storage reconstruction module is used to propagate the water storage anomaly information along the multi-scale hydrological relationship map based on the constraint of mass conservation, so as to obtain the water storage anomaly reconstruction field of the preset scale. The drought analysis module is used to construct a multi-scale hydrological relationship map of the target area based on water storage anomaly information and multi-source hydrological observation data.

9. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the spatiotemporal reconstruction method for water resources and drought conditions based on coupled gravity satellite and multi-source high-resolution data as described in any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the spatiotemporal reconstruction method for water resources and drought conditions based on any one of claims 1-7.

Images