Geological environment risk prediction and ecological restoration intelligent early warning method and device based on multi-source meteorological data, equipment and storage medium

Abstract

The application discloses a geological environment risk prediction and ecological restoration intelligent early warning method and device based on multi-source meteorological data, equipment and a storage medium, relates to the technical field of environmental monitoring, and comprises the following steps: collecting dynamic meteorological time series data and static spatial attribute data of a target area; obtaining long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time series data; recalibrating based on the static spatial attribute data to obtain calibrated static features; obtaining continuous values of drought indexes in a future preset period based on the long-term cross-seasonal trends, the short-term local trends and the calibrated static features; and performing ecological restoration intelligent early warning based on the continuous values of the drought indexes. The problems of insufficient prediction accuracy, lagging risk early warning for sudden changes and poor cross-region generalization ability caused by the difficulty in separating multi-scale meteorological features and the inability to realize adaptive association between static features and time series semantics are solved, so that the accuracy and balance of drought degree prediction are significantly improved.

Description

Technical Field

[0001] This application relates to the field of environmental monitoring technology, and in particular to a method, device, equipment and storage medium for geological environmental risk prediction and ecological restoration based on multi-source meteorological data. Background Technology

[0002] With the rapid development of multi-source meteorological observation technology and the increasingly refined needs for geological disaster prevention and control, effectively integrating dynamic meteorological time-series data with static geological attribute information to achieve accurate quantitative prediction of regional geological environmental risks has become a key technological requirement in the fields of disaster prevention, mitigation, and ecological restoration. Especially when facing geological environmental systems that combine the cumulative effects of long-term climate evolution with short-term extreme weather triggering mechanisms, there is an urgent need to construct intelligent prediction models capable of simultaneously capturing multi-scale time-series characteristics and achieving adaptive fusion of cross-modal information. This will improve the accuracy of trend judgments and the timeliness of early warnings for geological environmental risks such as drought.

[0003] While current mainstream methods for predicting geological environmental risks attempt to combine deep learning techniques such as convolutional neural networks and recurrent neural networks, they often employ a serial or simple stacking architecture to process multi-source heterogeneous data. On the one hand, these methods struggle to effectively separate the slow, cross-seasonal and cross-year climate trends from rapid, short-term weather fluctuations in meteorological time series, resulting in insufficient model perception of geological environmental evolution processes that possess both long-term cumulative effects and short-term triggering mechanisms. On the other hand, existing technologies, when dealing with the relationship between static geological attributes and dynamic meteorological time series, often use direct splicing or fixed gating methods, failing to achieve adaptive adjustment of the importance of static features as the meteorological context dynamically evolves. This causes static information to either become noise or become rigid, severely limiting the model's generalization ability and prediction accuracy across different risk scenarios.

[0004] Therefore, one of the technical problems to be solved by this invention is how to construct an intelligent prediction model that can decouple dynamic meteorological time series from multiple scales and achieve adaptive fusion of static geological features and temporal semantics, so as to accurately capture the risk evolution law with both long-term cumulative effects and short-term triggering mechanisms in complex geological environments, thereby improving the quantitative prediction accuracy of geological environmental risks such as regional future drought index. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, equipment and storage medium for geological environment risk prediction and ecological restoration based on multi-source meteorological data, aiming to solve the technical problem of how to improve the accuracy of quantitative prediction of geological environment risks such as regional future drought index.

[0006] To achieve the above objectives, this application proposes a method for intelligent early warning of geological environmental risk prediction and ecological restoration based on multi-source meteorological data. The method includes:

[0007] Collect dynamic time-series meteorological data and static spatial attribute data of the target area; Based on the aforementioned dynamic meteorological time series data, long-term cross-seasonal trends and short-term local trends are obtained; Recalibration is performed based on the static spatial attribute data to obtain calibration static features; Based on the long-term cross-seasonal trend, the short-term local trend, and the calibration static characteristics, the continuous value of the drought index within the future preset time period is obtained; Intelligent early warning for ecological restoration is based on the continuous values ​​of the drought index.

[0008] In one embodiment, obtaining the long-term cross-seasonal trend and the short-term local trend based on the dynamic meteorological time-series data includes: The dynamic meteorological time series data is divided according to a preset time step to obtain multiple dynamic meteorological time series features; Based on the dynamic meteorological time series characteristics, sequence correlation processing is performed to obtain long-term cross-seasonal trends; Convolution processing is performed based on the dynamic meteorological time series characteristics to obtain short-term local trends.

[0009] In one embodiment, the step of performing sequence correlation processing based on the dynamic meteorological time series characteristics to obtain a long-term cross-seasonal trend includes: Based on the dynamic meteorological time series characteristics, the query matrix, key matrix, and value matrix for the corresponding time step are obtained; Based on the query matrix and the key matrix at different time steps, a relevance matrix is ​​obtained; Based on the correlation matrix and the value matrix, a preliminary characterization is obtained; Based on the preliminary characterization, the long-term cross-seasonal trend is obtained.

[0010] In one embodiment, the convolutional processing based on the dynamic meteorological time-series features to obtain a short-term local trend includes: The dynamic meteorological time series features are convolved according to a preset convolution kernel to obtain the convolution result; The convolution result is nonlinearly activated to obtain a short-term local trend.

[0011] In one embodiment, the recalibration based on the static spatial attribute data to obtain calibrated static features includes: Based on the static spatial attribute data, the original static features are obtained; The original static features are projected from the static dimension to the hidden layer dimension to obtain the static feature structure information; Based on the static feature structure information, the control weights are obtained; The original static features are recalibrated based on the control weights to obtain calibrated static features.

[0012] In one embodiment, obtaining continuous drought index values ​​for a future preset period based on the long-term cross-seasonal trend, the short-term local trend, and the calibrated static characteristics includes: The long-term cross-seasonal trend, the short-term local trend, and the calibrated static features are concatenated to obtain the primary fusion feature; The primary fusion features are transformed to obtain the fusion result features; Based on the characteristics of the fusion results, continuous values ​​of the drought index for a future preset time period are obtained.

[0013] In one embodiment, transforming the primary fusion features to obtain the fusion result features includes: The primary fusion features are transformed to obtain the first predictor feature; The first predictor feature is transformed to obtain the second predictor feature; If the number of transformations has not reached the preset number, the second predictor feature is used as the first predictor feature, and the process returns to the step of transforming the first predictor feature to obtain the second predictor feature. When the number of transformations reaches a preset number, the second predictor feature is used as the fusion result feature.

[0014] Furthermore, to achieve the above objectives, this application also proposes an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data. The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data includes: The data acquisition module is used to collect dynamic meteorological time-series data and static spatial attribute data of the target area; The dynamic module is used to obtain long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time series data. The static module is used to perform recalibration based on the static spatial attribute data to obtain calibration static features; The prediction module is used to obtain continuous values ​​of the drought index for a future preset period based on the long-term cross-seasonal trend, the short-term local trend, and the calibration static characteristics. The early warning module is used to provide intelligent early warning for ecological restoration based on the continuous values ​​of the drought index.

[0015] Furthermore, to achieve the above objectives, this application also proposes an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data as described above.

[0016] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data as described above.

[0017] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data as described above.

[0018] This application collects dynamic meteorological time-series data and static spatial attribute data of the target area; based on the dynamic meteorological time-series data, it obtains long-term cross-seasonal trends and short-term local trends; based on the static spatial attribute data, it performs recalibration to obtain calibrated static features; based on the long-term cross-seasonal trends, the short-term local trends, and the calibrated static features, it obtains continuous drought index values ​​for a future preset period; based on the continuous drought index values, it performs intelligent early warning for ecological restoration. This solves the problems of insufficient prediction accuracy, delayed early warning of abrupt changes, and poor cross-regional generalization ability in existing technologies due to the difficulty in separating multi-scale meteorological features, the inability to achieve adaptive correlation between static features and temporal semantics, and the difficulty in achieving such correlation. Therefore, it significantly improves the accuracy and uniformity of drought severity prediction. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an embodiment of the intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data in this application. Figure 2 This is a scatter plot of real values ​​and predicted values ​​provided in Implementation Example 1 of the Intelligent Early Warning Method for Geological Environment Risk Prediction and Ecological Restoration Based on Multi-Source Meteorological Data in this application. Figure 3 This is a flowchart illustrating Embodiment 2 of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in this application. Figure 4 This is a schematic diagram of the overall process of the second embodiment of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in this application; Figure 5 This is a schematic diagram of the module structure of the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data, as described in this application embodiment. Figure 6 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in the embodiments of this application.

[0022] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0024] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0025] The main solution of this application embodiment is as follows: collecting dynamic meteorological time-series data and static spatial attribute data of the target area; obtaining long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time-series data; performing recalibration based on the static spatial attribute data to obtain calibrated static features; obtaining continuous drought index values ​​within a future preset time period based on the long-term cross-seasonal trends, the short-term local trends, and the calibrated static features; and performing intelligent early warning for ecological restoration based on the continuous drought index values.

[0026] In this embodiment, for ease of description, the following description will focus on an intelligent early warning device for geological environmental risk prediction and ecological restoration based on multi-source meteorological data.

[0027] While current mainstream methods for predicting geological environmental risks attempt to combine deep learning techniques such as convolutional neural networks and recurrent neural networks, they often employ serial or simple stacking architectures to process multi-source heterogeneous data. These methods struggle to effectively separate slow, cross-seasonal and cross-year climate trends from rapid, short-term weather fluctuations in meteorological time series, resulting in insufficient model perception of geological environmental evolution processes that possess both long-term cumulative effects and short-term triggering mechanisms. Furthermore, existing technologies, when dealing with the relationship between static geological attributes and dynamic meteorological time series, often use direct splicing or fixed gating, failing to achieve adaptive adjustment of the importance of static features as the meteorological context dynamically evolves. This causes static information to either become noise or become rigid, severely limiting the model's generalization ability and prediction accuracy across different risk scenarios.

[0028] This application provides a solution that addresses the problems of insufficient prediction accuracy, delayed early warning of sudden change risks, and poor cross-regional generalization ability caused by the difficulty in separating multi-scale meteorological features, the inability to achieve adaptive correlation between static features and temporal semantics in existing technologies, thereby significantly improving the accuracy and balance of drought degree prediction.

[0029] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone; or an electronic device capable of performing the above functions, such as an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data. The following description uses an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data as an example to illustrate this embodiment and the subsequent embodiments.

[0030] Based on this, the embodiments of this application provide a method for intelligent early warning of geological environmental risk prediction and ecological restoration based on multi-source meteorological data, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in this application.

[0031] In this embodiment, the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data includes steps S10 to S50: Step S10: Collect dynamic meteorological time-series data and static spatial attribute data of the target area; It should be noted that the target area refers to a specific geographical unit requiring geological environmental risk prediction and ecological restoration early warning, which can be a continuous geographical spatial range defined by administrative divisions, watershed boundaries, or areas prone to geological disasters. The dynamic meteorological time-series data refers to a sequence of meteorological elements continuously observed and recorded at fixed time intervals (e.g., daily scale) within the target area. This includes, but is not limited to, precipitation, air temperature, dew point temperature, surface temperature, air pressure, humidity, wind speed, and their derived statistics (e.g., daily extreme values, range of variation), forming a multivariate time series describing the weather evolution process. The static spatial attribute data refers to spatial attribute vectors describing the inherent characteristics of the target area that do not change with short-term meteorological processes. This includes, but is not limited to, more than thirty spatial characteristics such as latitude, longitude, altitude, slope, aspect, land cover type, and soil type indicators. These attributes define the inherent sensitivity and vulnerability of the region to meteorological-driven responses.

[0032] Specifically, after data collection, standardized preprocessing is required to adapt to the model input: for dynamic meteorological time-series data, missing values ​​are imputed to ensure temporal continuity; for static spatial attribute data, standardization is directly applied to unify their numerical range. Spatially, the two are aligned through geographical locations; temporally, a sliding window mechanism is used to extract a fixed-length historical window from the complete time series. The dynamic meteorological sequence within the window is combined with the corresponding region's static attribute vector to form a complete training sample, used to predict the drought index at future points in time.

[0033] Understandably, since the evolution of the geological environment system is driven by both the dynamics of meteorological time series and the static constraints of the geological background, it is difficult to fully characterize the complex mechanisms of risk formation if only a single type of data is considered. Therefore, performing step S10 to collect and preprocess two types of multi-source heterogeneous data can avoid the one-sidedness of the prediction model caused by a single information source, avoid the risk of mismatch between static background features and dynamic meteorological processes in subsequent modeling, and avoid the interference of inconsistent dimensions and temporal breaks in the original data on the stability of model training. This improves the completeness, standardization, and alignment accuracy of the model input data, laying a reliable data foundation for accurately capturing long-term cumulative effects and short-term triggering mechanisms, and achieving adaptive fusion of cross-modal features.

[0034] Step S20: Based on the dynamic meteorological time series data, obtain the long-term cross-seasonal trend and the short-term local trend; It should be noted that the long-term transseasonal trend refers to the slow climate change patterns and periodic components in the meteorological sequence that span several weeks, months, or even years, such as seasonal wet and dry cycles and annual temperature variations, reflecting the cumulative impact of meteorological processes on the geological environment. The short-term local trend refers to the rapid fluctuations and abrupt changes in the meteorological sequence on a timescale of several hours to several days, such as peak heavy precipitation, sudden temperature changes, and rapid increases in wind speed, representing transient meteorological events that may trigger geological disasters.

[0035] Understandably, since meteorological time series simultaneously contain evolutionary information at different physical scales, using a single model for aliasing would lead to interference between long-term cumulative effects and short-term triggering mechanisms, making it difficult for the model to clarify the contribution of features at different scales to risk formation. Therefore, performing step S20 to explicitly separate and extract long-term cross-seasonal trends and short-term local trends in parallel can avoid the predictive ambiguity caused by the coupling of multi-scale features, avoid the model's lag in early warning of abrupt risks, and thus improve the model's ability to perceive and represent complex meteorological driving processes with greater precision and accuracy.

[0036] In one feasible implementation, step S20 may include: dividing the dynamic meteorological time series data according to a preset time step to obtain multiple dynamic meteorological time series features; performing sequence correlation processing based on the dynamic meteorological time series features to obtain a long-term cross-seasonal trend; and performing convolution processing based on the dynamic meteorological time series features to obtain a short-term local trend.

[0037] It should be noted that the preset time step refers to the time interval unit for discretizing and sampling continuous meteorological time series data, consistent with the observation frequency of the original data (e.g., daily scale), thus determining the temporal resolution of the input sequence. The dynamic meteorological time series feature refers to the multi-dimensional vector representation corresponding to each time step after mapping through the embedding layer, denoted as... , where T is the total number of time steps (i.e. the length of the historical observation window), and d is the feature dimension.

[0038] Specifically, a fixed-length historical window is slidably extracted from the preprocessed complete meteorological time series, with each time point within the window corresponding to an original meteorological observation vector. Subsequently, the original vector is projected into a high-dimensional latent space through an embedding layer (such as a linear mapping) to obtain dynamic meteorological time series features X with richer representation capabilities. This process transforms the original observation data into a feature tensor that is processable by the model and preserves the temporal structure, laying the foundation for subsequent multi-scale feature extraction.

[0039] Further, the step of performing sequence correlation processing based on the dynamic meteorological time series characteristics to obtain a long-term cross-seasonal trend includes: obtaining a query matrix, key matrix, and value matrix for the corresponding time step based on the dynamic meteorological time series characteristics; obtaining a correlation matrix based on the query matrix and key matrix for different time steps; obtaining a preliminary characterization based on the correlation matrix and the value matrix; and obtaining a long-term cross-seasonal trend based on the preliminary characterization.

[0040] It should be noted that the query matrix (Q), key matrix (K), and value matrix (V) are matrices obtained by projecting the input feature X through three independent linear transformation layers. They represent the "query" intent at the current time step, the "index" identifier of each time step, and the aggregated "information" content, respectively. The relevance matrix quantifies the dependency strength between any two time points in the sequence, reflecting the association weights of long-distance time steps. The preliminary representation is a feature representation that incorporates global contextual information, obtained by weighted summation of the value matrix and the relevance matrix.

[0041] Specifically, the core of the long-term cross-seasonal trend capture channel is a sequence correlation modeling mechanism based on Scaled Dot-Product Attention, which aims to capture slow climate change trends and cyclical patterns spanning weeks, months, or even years.

[0042] Input sequence Through three independent linear transformation layers, the projection is respectively into three sets of matrices: Query, Key, and Value.

[0043] in, For trainable weight matrix, The common dimensions after projection are usually set to Alternatively, it can be scaled appropriately to control the complexity and representational power of attention computation.

[0044] Relevance Calculation and Information Aggregation: By calculating the dot product of the query and the key and scaling it, the relevance matrix between time steps is obtained. This matrix quantifies the strength of the dependency between any two states at any given time point in the sequence:

[0045] Scaling factor in the formula This is used to prevent the gradient of the softmax function from vanishing due to an excessively large dot product result. It utilizes the correlation matrix. value matrix A weighted summation is performed to integrate the contextual information of the entire sequence, resulting in a preliminary representation with global dependencies. :

[0046] Short-term features are obtained after nonlinear activation:

[0047] Output It mainly expresses long-term trends, seasonality, and slowly changing structural information. FFN is an abbreviation for feedforward network, which usually refers to a sub-network composed of multiple fully connected layers.

[0048] Furthermore, the step of performing convolution processing based on the dynamic meteorological time series features to obtain short-term local trends includes: performing convolution processing on the dynamic meteorological time series features according to a preset convolution kernel to obtain a convolution result; and performing nonlinear activation on the convolution result to obtain short-term local trends.

[0049] It should be noted that the preset convolution kernel refers to a trainable filter used in one-dimensional convolution operations. Its size determines the receptive field range and is usually set to a small value (such as 3, 5, or 7) to focus on local temporal patterns. The convolution result is a local feature map calculated by sliding the convolution kernel along the time dimension. The nonlinear activation refers to introducing a nonlinear transformation through an activation function (such as ReLU) to enhance the model's ability to fit complex patterns.

[0050] Specifically, the short-term local trend capture channel uses a one-dimensional convolutional neural network (1D-CNN) to focus on extracting rapid weather fluctuations and local mutation patterns on a scale of several hours to several days.

[0051] Local convolution operation: for input dynamic meteorological time series features Applying a one-dimensional convolutional layer, Conv1D is an abbreviation for one-dimensional convolution:

[0052] Specifically, a set of trainable convolutional kernels (filters) is used to perform sliding computations along the time dimension. The size of each convolutional kernel determines the receptive field, i.e., the local temporal range that can be observed at one time, and is usually set to a small value (such as 3, 5, 7) to focus on short-term patterns. The stride and padding of the convolutional operation can be set as needed to maintain or change the duration of the output sequence.

[0053] Nonlinear activation and output: Convolution results Then, a nonlinear activation function is applied. (e.g., ReLU) to introduce nonlinear transformations and enhance the model's ability to fit complex patterns:

[0054] Output The number of convolution kernels (convolution results) can effectively characterize drastic fluctuations within short periods (such as precipitation peaks, sudden changes in wind speed, and rapid temperature rises / falls) as well as the fine-grained spatiotemporal structure of weather systems.

[0055]

[0056] Through the above design, the two channels respectively generate a focus on the macro-climate background. Compared to focusing on micro-weather events These two complementary feature tensors, extracted from different physical scales, will be fed together into the subsequent feature fusion module to achieve collaborative modeling of multi-scale temporal information.

[0057] In this embodiment, a parallel dual-channel architecture is used to extract long-term trends and short-term fluctuations by employing attention mechanisms and one-dimensional convolution, respectively. This solves the problem that traditional single models cannot take into account both the global perspective and local details, and achieves multi-scale decoupling and collaborative representation of meteorological time series.

[0058] The above are merely feasible implementations of step S20 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S20.

[0059] Step S30: Recalibrate based on the static spatial attribute data to obtain calibrated static features; It should be noted that the recalibration refers to an adaptive filtering process that dynamically adjusts the weights of static spatial attribute features based on the semantic context of the meteorological time series. Its core is to generate feature importance coefficients that match the current meteorological scenario through a gating mechanism, thereby selectively enhancing or suppressing the original static features. The calibrated static features refer to the static feature vectors whose weights have been dynamically adjusted according to the time series context after recalibration, denoted as... This feature retains key geological background information relevant to the current meteorological scene and suppresses interference from irrelevant dimensions.

[0060] Specifically, this application designs an adaptive static feature gating mechanism based on temporal context awareness. The core idea of ​​this mechanism is to dynamically generate a set of gating weights based on the semantic state implied by the current input meteorological time series, recalibrating the original static features to achieve "scenario-based" filtering and enhancement at the feature level. This process ensures that the contribution of static features is no longer globally consistent, but rather adaptively adjusted according to different meteorological scenarios such as "whether it is in a long-term drought accumulation period" or "whether it has encountered short-term extreme precipitation," thus completing an intelligent structural screening before fusion.

[0061] It is understandable that, due to the fundamental differences in the nature of static spatial attribute data (such as topography and soil type) and dynamic meteorological time-series data—the former being fixed attribute vectors in the spatial domain and the latter being continuously evolving sequences in the temporal domain—traditional direct concatenation or fixed weighting methods would lead to dimensions in static features that are irrelevant to the current meteorological context acting as noise to interfere with the learning of time-series representations. Furthermore, key dimensions may not be fully activated in specific scenarios due to fixed weights. Therefore, step S30, which introduces an adaptive gating recalibration mechanism based on time-series context awareness, can avoid the problem of static feature contributions becoming rigid or turning into noise, avoid semantic mismatch during cross-modal information fusion, thereby improving the model's ability to filter key information related to the current meteorological scenario from static features, and enhancing the discrimination accuracy of subsequent fusion modules and the model's generalization ability.

[0062] Step S40: Based on the long-term cross-seasonal trend, the short-term local trend, and the calibration static characteristics, obtain the continuous value of the drought index for the future preset period.

[0063] It should be noted that the preset time period refers to the target time range predicted by the model, that is, the specific future time point or time period to be predicted based on the currently input historical time window data, such as the drought status in the next few weeks (e.g., six weeks later). The continuous drought index value refers to the quantitative prediction result output by the model, which is a continuous real value used to characterize the degree of drought in the target area in the future preset time period. The value reflects the severity of the drought and provides a quantitative basis for ecological restoration early warning.

[0064] Understandably, since long-term cross-seasonal trends, short-term local trends, and calibrated static features each characterize the formation mechanism of geological environmental risks from different dimensions—long-term trends reflect cumulative effects, short-term fluctuations represent triggering mechanisms, and static features embody background sensitivity—it is difficult to fully capture the complex process of risk evolution if predictions rely solely on a single type of feature. Therefore, performing step S40 to deeply integrate the three types of features and perform high-level abstraction can avoid prediction biases caused by insufficient information utilization and semantic conflicts caused by the simple superposition of multi-source features, thereby improving the accuracy, robustness, and interpretability of the prediction results.

[0065] In one feasible implementation, step S40 may include: splicing the long-term cross-seasonal trend, the short-term local trend, and the calibration static feature to obtain a primary fusion feature; transforming the primary fusion feature to obtain a fusion result feature; and obtaining a continuous value of the drought index for a future preset period based on the fusion result feature.

[0066] It should be noted that the primary fusion feature refers to the comprehensive feature vector obtained by directly concatenating the three types of features along the feature dimension. This feature retains the integrity of the original information and provides rich input for subsequent deep abstraction. The fusion result feature refers to the high-order abstract feature obtained after deep processing by a multi-level nonlinear transformation network. This feature integrates the interaction relationship of the three types of information and extracts the comprehensive representation most relevant to drought prediction.

[0067] Specifically, in order to effectively integrate the time-series characteristics from both channels (characterizing short-term local fluctuations) With representation of long-term dependencies ) and the calibration static characteristics after gating and screening The three types of features that are complementary in physical meaning and origin are initially fused. Through concatenation along the feature dimension, they are integrated into a unified joint representation:

[0068] in, The dimensions of the high-dimensional fusion features after splicing are... It is the sum of the three feature dimensions.

[0069] Furthermore, the transformation of the primary fusion feature to obtain the fusion result feature includes: transforming the primary fusion feature to obtain a first predictor feature; transforming the first predictor feature to obtain a second predictor feature; if the number of transformations has not reached a preset number, using the second predictor feature as the first predictor feature, and returning to the step of transforming the first predictor feature to obtain the second predictor feature; if the number of transformations has reached a preset number, using the second predictor feature as the fusion result feature.

[0070] It should be noted that the first predictor feature refers to the input feature of the current transform layer, denoted as... The second predictor feature refers to the output feature of the current transform layer, denoted as... The preset number of iterations refers to the total number of layers L in the feature transformation network, i.e., the number of times the nonlinear transformation needs to be repeated. Each transformation operation includes linear mapping, nonlinear activation, and residual connection.

[0071] Specifically, the initial fusion features are fed into a multi-level nonlinear transformation network for further processing. This network consists of... The system consists of several fully connected layers (also known as feedforward layers). Each layer performs a linear transformation on the input features and applies a non-linear activation function (such as ReLU) to progressively abstract higher-order predictors. To mitigate the gradient vanishing or degradation problems that may occur in deep networks and to preserve important information between layers, this invention introduces a residual connection after each layer. Specifically, the first layer... layer( The output of ) The calculation is as follows:

[0072] in, and The first The trainable weight matrix and bias vector of the layer, The residual connection uses an identity mapping to transfer features from the previous layer. It is directly added to the transformation result of the current layer, ensuring a smooth information flow.

[0073] go through After layer-by-layer fusion and refinement, the characteristics of the fusion result at the top layer are... By mapping a linear prediction layer to the target output space, continuous value predictions of future drought severity are generated. :

[0074] in, and These are the weights and biases for the prediction layer. This refers to the quantified drought index prediction value output by the system.

[0075] like Figure 2 As shown in Table 1, the method in this application is compared with state-of-the-art methods such as Ridge Regression and Long Short-Term Memory (LSTM) on the same dataset. The comparison metrics include Macro F1 Mean score and Mean Absolute Error (MAE). The comparison results show that the method in this application has a significant improvement in drought level prediction. As can be seen from Table 1, it outperforms other comparative models in all metrics. Specifically, the Macro F1 Mean score is improved by 12.1% compared to LSTM, indicating that the method in this application has more balanced prediction ability across different discrete levels. The mean absolute error is 0.097, indicating that the method in this application predicts more closely to the true value and has higher overall accuracy. The test results are compared in Table 1, where bold text indicates the optimal value.

[0076] Table 1

[0077] Figure 2 The image shows a scatter plot of the continuous drought severity predictions for the next six weeks using the LSTM model and this application. The horizontal and vertical axes represent the actual drought severity and the model-predicted drought severity, respectively. Each point on the plot represents the correspondence between the predicted and actual values ​​for a specific region in a particular week. It can be seen that the scatter plots from this application are concentrated and distributed close to the diagonal, indicating more accurate predictions and greater robustness in extreme value and trend predictions.

[0078] In this embodiment, by using a multi-level nonlinear transformation network combined with residual connection technology, the primary fusion features are abstracted and processed layer by layer. This solves the problems that shallow fusion is difficult to capture high-order interaction relationships between features and that deep network training is prone to gradient degradation. It achieves deep hybridization and synergistic enhancement of three types of information: long-term trends, short-term fluctuations and static features, and provides a high-quality fusion feature representation for the final accurate prediction of the drought index.

[0079] The above are merely feasible implementations of step S40 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S40.

[0080] Step S50: Based on the continuous value of the drought index, conduct intelligent early warning for ecological restoration.

[0081] It should be noted that the aforementioned intelligent early warning for ecological restoration refers to the process of transforming the continuous drought index values ​​predicted by the model into actionable early warning signals and decision support information. This process may include: comparing the predicted continuous drought index values ​​with multiple preset early warning thresholds to determine the drought risk level (e.g., mild drought, moderate drought, severe drought, extreme drought) of the target area within a preset future time period; automatically generating differentiated ecological restoration suggestions and early warning instructions based on the risk level, such as initiating drought relief water resource allocation, adjusting vegetation restoration plans, and strengthening soil moisture monitoring; and pushing the early warning information to relevant management departments and decision-makers through a visual interface, mobile terminal, or automatic control system, thus achieving a closed loop from risk prediction to prevention and control actions.

[0082] Specifically, the system pre-defines multiple threshold ranges corresponding to different drought risk levels and ecological response measures based on historical drought event data and ecological restoration practice experience. When the drought index output by the model continuously falls within a certain threshold range, the system automatically triggers the corresponding early warning logic: on the one hand, it generates visual early warning indicators (such as color grading and trend charts) on the monitoring platform; on the other hand, it matches corresponding ecological restoration suggestions (such as vegetation water replenishment plans for specific areas and the timing of initiating soil improvement measures) according to a preset rule base; simultaneously, it can push early warning information to business systems such as land spatial planning, water conservancy scheduling, and forestry and grassland restoration through interfaces, supporting cross-departmental collaborative response. The entire process realizes the intelligent transformation from quantitative prediction to decision support.

[0083] Understandably, since the ultimate goal of geological environmental risk prediction is to serve disaster prevention and mitigation practices and ecological protection and restoration, if the prediction results remain only at the numerical level and lack connection with operational scenarios, they will be difficult to realize their practical application value. Therefore, performing step S50 to convert the continuous drought index values ​​into specific ecological restoration early warning instructions can avoid the gap between prediction results and actual decisions, prevent response delays or inappropriate measures due to unclear early warning information, thereby improving the operational efficiency of prediction results, enhancing the region's proactive prevention and control capabilities against geological environmental risks, and providing accurate decision support for land spatial planning and ecological protection and restoration.

[0084] This embodiment provides a method for predicting geological environmental risks and providing intelligent early warning for ecological restoration based on multi-source meteorological data. It collects dynamic meteorological time-series data and static spatial attribute data of the target area; based on the dynamic meteorological time-series data, it obtains long-term cross-seasonal trends and short-term local trends; based on the static spatial attribute data, it performs recalibration to obtain calibrated static features; based on the long-term cross-seasonal trends, the short-term local trends, and the calibrated static features, it obtains continuous drought index values ​​for a preset future period; and based on the continuous drought index values, it provides intelligent early warning for ecological restoration. This method solves the problems of insufficient prediction accuracy, delayed warning of abrupt risks, and poor cross-regional generalization ability in existing technologies due to the difficulty in separating multi-scale meteorological features, the inability to achieve adaptive correlation between static features and temporal semantics, and the difficulty in achieving such correlation. Therefore, it significantly improves the accuracy and uniformity of drought severity prediction.

[0085] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 The step S30 of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data further includes steps S31 to S34: Step S31: Based on the static spatial attribute data, obtain the original static features; It should be noted that the original static features refer to the feature vectors directly used as model input after standardizing and preprocessing the collected static spatial attribute data. These feature vectors fully preserve the essential attribute information of the target area, including more than 30 spatial features such as latitude, longitude, altitude, slope, aspect, land cover type, and soil type indicators. Each value reflects the inherent sensitivity and vulnerability of the region to meteorological driving responses.

[0086] Specifically, the various static spatial attribute data collected from the target area will be numerically processed. Categorical attributes (such as land cover type and soil type) will be converted into continuous vector representations through one-hot encoding or embedding layers. Outlier removal and missing value imputation will be performed on continuous attributes (such as altitude and slope). All attributes will be standardized to eliminate the influence of different units on model training and to unify the numerical range of each dimension. All processed attribute vectors will be concatenated to form the original static feature S, which will be used for subsequent recalibration processing.

[0087] Understandably, since the original static features contain various types of attribute information, and the contribution of each dimension to the formation of geological environmental risks varies under different meteorological scenarios, directly using the original features for fusion would make it difficult for the model to distinguish between key and redundant information. Therefore, performing step S31 to obtain standardized original static features can avoid model input chaos caused by inconsistent data formats, avoid interference from differences in original dimensions on training stability, and lay the foundation for subsequent accurate static feature recalibration.

[0088] Step S32: Project the original static features from the static dimension to the hidden layer dimension to obtain static feature structure information; It should be noted that static dimension refers to the spatial dimension of the original static features. This refers to the number of static attributes. The hidden layer dimension refers to the size of the hidden space dimension mapped to after a linear transformation. The hyperparameter, typically set to be greater than or less than the original dimension, controls the level of abstraction in the feature representation. The static feature structure information refers to the high-level abstract representation extracted from the original static features after linear transformation and nonlinear activation. This representation extracts structured information from the original features that can be used for gating decisions.

[0089] Specifically, static features are mapped to the latent space, and high-level structural information that can be used for gating decisions is extracted from the original space:

[0090] in, Used to convert static features From static dimensions Projected onto hidden layer dimensions , It is a bias term.

[0091] Understandably, due to the high dimensionality of the original static features and the complex relationships between them, directly generating gating weights based on the original dimensions makes it difficult to fully explore the deep interaction information between features. Therefore, performing step S32 to project the original static features to the hidden layer dimension and extract structured information can avoid the learning difficulties of gating weights caused by the curse of dimensionality, prevent useful interaction information between features from being ignored, and thus improve the gating mechanism's ability to capture key patterns in static features.

[0092] Step S33: Based on the static feature structure information, obtain the control weights.

[0093] It should be noted that the control weights are coefficient vectors used to adjust the importance of the original static features dimension by dimension. Each element takes a value between (0, 1), which represents the importance of the corresponding static feature dimension in the current meteorological time series context.

[0094] Specifically, a set of control coefficients is generated through a second-level linear transformation, and their range of values ​​is limited by a sigmoid function:

[0095] in This is the final vector of control weights, used to determine the degree of preservation of each dimension in the static features.

[0096] Understandably, since the contribution of each dimension in static features to risk formation varies, and this contribution pattern should be relatively stable (e.g., the response of altitude to temperature changes has a deterministic pattern), directly using time-related dynamic weights may lead to unreasonable fluctuations in the contribution of static features. Therefore, performing step S33 to generate stable control weights based on the structural information of the static features themselves can avoid feature distortion caused by excessive coupling between gating weights and temporal context, and prevent the inherent contribution pattern of static features from being destroyed, thereby ensuring the rationality and interpretability of the recalibration process.

[0097] Step S34: Recalibrate the original static features based on the control weights to obtain calibrated static features.

[0098] Specifically, the vector composed of control weights By acting on the static features themselves, the importance can be adjusted dimension by dimension.

[0099] Among them, operation For element-wise multiplication, To calibrate static features.

[0100] This approach ensures that static features have certain structural constraints before entering the next stage of fusion: the model suppresses static dimensions that do not contribute to the current time-series scene and strengthens features with regional discrimination, thereby improving the discriminative ability of subsequent fusion modules.

[0101] Understandably, due to the significant differences in the contributions of various dimensions of static features under different meteorological scenarios (e.g., during drought accumulation periods, the soil water-holding capacity dimension should be strengthened; during rainstorm triggering periods, the slope and aspect dimensions should be strengthened), it is difficult to adapt to such scenario-specific needs if static feature fusion is performed with fixed weights. Therefore, performing step S34 to recalibrate the original static features based on control weights can avoid the rigidity of the contributions of static features under different scenarios, avoid interference from irrelevant dimensions on predictions, thereby improving the accuracy of the fusion of static features and time-series features, and enhancing the model's adaptability under different risk scenarios.

[0102] This embodiment provides a method for predicting geological environmental risks and providing intelligent early warning for ecological restoration based on multi-source meteorological data. Based on the static spatial attribute data, the original static features are obtained; the original static features are projected from the static dimension to the hidden layer dimension to obtain static feature structure information; based on the static feature structure information, control weights are obtained; and based on the control weights, the original static features are recalibrated to obtain calibrated static features. By employing a technique that generates stable control weights based on the structural information of static features themselves, and then adjusts the importance of the original static features dimension by dimension, the high-level structural information of static features is extracted through a first-layer fully connected network. The second-layer linear transformation combined with Sigmoid activation generates gated weights in the (0,1) interval. Finally, feature selection is achieved through element-wise multiplication. This avoids the problem of static features becoming rigid or turning into noise under different meteorological scenarios, avoids the difficulty in weight learning caused by the complex correlation between the dimensions of static features, and avoids the inability of fixed weight fusion methods to adapt to the needs of different risk scenarios such as drought accumulation period and rainstorm triggering period. It solves the technical problems in the existing technology of lacking adaptive adjustment capability when fusing static features with dynamic meteorological time series, difficulty in dynamically selecting key static dimensions based on meteorological context, and easy destruction of the inherent contribution mode of static features. Thus, it achieves accurate enhancement of key information related to the current meteorological scenario in static features and effective suppression of irrelevant dimensions, improves the accuracy of static feature and time series feature fusion, and enhances the adaptability and generalization performance of the model under different geographical regions and different risk scenarios.

[0103] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data. Any simple modifications based on this technical concept are within the scope of protection of this application.

[0104] For example, to help understand the implementation process of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data obtained in this embodiment combined with the above embodiment one, please refer to... Figure 4 , Figure 4A schematic diagram of the overall process for an intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data is provided, specifically: The core processing flow begins with multi-source data acquisition. The system collects dynamic meteorological time-series data and static spatial attribute data for the target area. The time-series data is infused with temporal and location information through location encoding, while the static data is converted into feature vectors via an embedding layer. Dynamic time-series data is processed in parallel into two feature extraction channels: a long-term cross-seasonal trend capture module employs a multi-head attention mechanism and a feedforward neural network to extract slow-changing patterns at the climate level through residual connections and layer normalization; and a short-term local trend capture module focuses on rapid weather fluctuations on a timescale of several hours to several days through a feedforward neural network and average pooling. Simultaneously, a static feature adaptive filtering module dynamically recalibrates the static data, generating calibrated static features that match the current meteorological context.

[0105] Building upon feature extraction and selection, the system enters the multi-source feature fusion stage. Long-term trend features, short-term fluctuation features, and adaptively selected calibrated static features are integrated to form a comprehensive feature representation encompassing a global evolution perspective, local detail awareness, and inherent regional attributes. This fusion process ensures that features with different physical meanings can interact collaboratively within the same representation space, laying the foundation for subsequent accurate predictions. The fused features undergo deep processing via a multi-layer fully connected network to gradually abstract higher-order predictive factors related to drought risk.

[0106] Based on deeply processed fusion features, the system generates continuous drought index values ​​for the target area over a preset future period through its output layer. This quantitative prediction result is further transformed into specific intelligent early warning signals for ecological restoration, triggering differentiated response measures based on preset risk thresholds, such as initiating drought relief scheduling and adjusting vegetation restoration plans. Through this complete processing flow, the system achieves an end-to-end intelligent closed loop from multi-source data input to actionable early warning output, providing precise decision support for geological environmental risk prevention and ecological protection and restoration.

[0107] This application also provides an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data. Please refer to [link / reference]. Figure 5 The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data includes: Acquisition module 10 is used to acquire dynamic meteorological time-series data and static spatial attribute data of the target area; Dynamic module 20 is used to obtain long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time series data; Static module 30 is used to perform recalibration based on the static spatial attribute data to obtain calibration static features; Prediction module 40 is used to obtain continuous values ​​of drought index within a future preset period based on the long-term cross-seasonal trend, the short-term local trend, and the calibration static characteristics. The early warning module 50 is used to provide intelligent early warning for ecological restoration based on the continuous value of the drought index.

[0108] The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in this application adopts the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in the above embodiments, which can solve the technical problem of how to improve the quantitative prediction accuracy of geological environment risks such as regional future drought index. Compared with the prior art, the beneficial effects of the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in this application are the same as the beneficial effects of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in the above embodiments, and other technical features in the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0109] The dynamic module 20 is further configured to divide the dynamic meteorological time series data according to a preset time step to obtain multiple dynamic meteorological time series features; perform sequence correlation processing based on the dynamic meteorological time series features to obtain a long-term cross-seasonal trend; and perform convolution processing based on the dynamic meteorological time series features to obtain a short-term local trend.

[0110] The dynamic module 20 is further configured to obtain a query matrix, a key matrix, and a value matrix for a corresponding time step based on the dynamic meteorological time series characteristics; obtain a correlation matrix based on the query matrix and the key matrix for different time steps; obtain a preliminary characterization based on the correlation matrix and the value matrix; and obtain a long-term cross-seasonal trend based on the preliminary characterization.

[0111] The dynamic module 20 is further configured to perform convolution processing on the dynamic meteorological time series features according to a preset convolution kernel to obtain a convolution result; and to perform nonlinear activation on the convolution result to obtain a short-term local trend.

[0112] The static module 30 is further configured to: obtain original static features based on the static spatial attribute data; project the original static features from the static dimension to the hidden layer dimension to obtain static feature structure information; obtain control weights based on the static feature structure information; and recalibrate the original static features based on the control weights to obtain calibrated static features.

[0113] The prediction module 40 is further configured to splice the long-term cross-seasonal trend, the short-term local trend, and the calibration static features to obtain a primary fusion feature; transform the primary fusion feature to obtain a fusion result feature; and obtain a continuous value of the drought index for a future preset period based on the fusion result feature.

[0114] The early warning module 50 is further configured to transform the primary fusion feature to obtain a first predictor feature; transform the first predictor feature to obtain a second predictor feature; if the number of transformations has not reached a preset number, use the second predictor feature as the first predictor feature and return to the step of transforming the first predictor feature to obtain the second predictor feature; if the number of transformations has reached a preset number, use the second predictor feature as the fusion result feature.

[0115] This application provides an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data. The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data includes: at least one processor; and a memory communicatively connected to at least one processor; wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor to enable at least one processor to execute the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in the above embodiment 1.

[0116] The following is for reference. Figure 6 This document illustrates a structural schematic diagram of an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data, suitable for implementing embodiments of this application. The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and vehicle terminals (e.g., vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 6 The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments in this application.

[0117] like Figure 6As shown, the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage device 1003 into the random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data to exchange data wirelessly or via wired communication with other devices. Although the figure shows an intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0118] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0119] The intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in this application, employing the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in the above embodiments, can solve the technical problem of how to improve the quantitative prediction accuracy of geological environment risks such as regional future drought index. Compared with the prior art, the beneficial effects of the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in this application are the same as the beneficial effects of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in the above embodiments, and other technical features in the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0120] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0121] The above description is merely a specific embodiment 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 scope of the technology 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.

[0122] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data in the above embodiments.

[0123] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0124] The aforementioned computer-readable storage medium may be included in the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data; or it may exist independently and not be assembled into the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data.

[0125] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data, the device enables the following actions: collecting dynamic meteorological time-series data and static spatial attribute data of the target area; obtaining long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time-series data; performing recalibration based on the static spatial attribute data to obtain calibrated static features; obtaining continuous drought index values ​​for a future preset time period based on the long-term cross-seasonal trends, the short-term local trends, and the calibrated static features; and providing intelligent early warning for ecological restoration based on the continuous drought index values.

[0126] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, 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 indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated 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 the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0128] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0129] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data. This method can solve the technical problem of how to improve the accuracy of quantitative predictions of regional future drought indices and other geological environment risks. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data provided in the above embodiments, and will not be elaborated upon here.

[0130] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data.

[0131] The computer program product provided in this application can solve the technical problem of how to improve the accuracy of quantitative prediction of geological environmental risks such as regional future drought index. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data provided in the above embodiments, and will not be repeated here.

[0132] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for predicting geological environmental risks and providing intelligent early warning for ecological restoration based on multi-source meteorological data, characterized in that, The method includes: Collect dynamic time-series meteorological data and static spatial attribute data of the target area; Based on the aforementioned dynamic meteorological time series data, long-term cross-seasonal trends and short-term local trends are obtained; Recalibration is performed based on the static spatial attribute data to obtain calibration static features; Based on the long-term cross-seasonal trend, the short-term local trend, and the calibration static characteristics, the continuous value of the drought index within the future preset time period is obtained; Intelligent early warning for ecological restoration is based on the continuous values ​​of the drought index.

2. The method as described in claim 1, characterized in that, The process of obtaining long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time-series data includes: The dynamic meteorological time series data is divided according to a preset time step to obtain multiple dynamic meteorological time series features; Based on the dynamic meteorological time series characteristics, sequence correlation processing is performed to obtain long-term cross-seasonal trends; Convolution processing is performed based on the dynamic meteorological time series characteristics to obtain short-term local trends.

3. The method as described in claim 2, characterized in that, The step of performing sequence correlation processing based on the dynamic meteorological time series characteristics to obtain long-term cross-seasonal trends includes: Based on the dynamic meteorological time series characteristics, the query matrix, key matrix, and value matrix for the corresponding time step are obtained; Based on the query matrix and the key matrix at different time steps, a relevance matrix is ​​obtained; Based on the correlation matrix and the value matrix, a preliminary characterization is obtained; Based on the preliminary characterization, the long-term cross-seasonal trend is obtained.

4. The method as described in claim 2, characterized in that, The convolutional processing based on the dynamic meteorological time-series features to obtain short-term local trends includes: The dynamic meteorological time series features are convolved according to a preset convolution kernel to obtain the convolution result; The convolution result is nonlinearly activated to obtain a short-term local trend.

5. The method as described in claim 1, characterized in that, The recalibration based on the static spatial attribute data to obtain calibrated static features includes: Based on the static spatial attribute data, the original static features are obtained; The original static features are projected from the static dimension to the hidden layer dimension to obtain the static feature structure information; Based on the static feature structure information, the control weights are obtained; The original static features are recalibrated based on the control weights to obtain calibrated static features.

6. The method as described in claim 1, characterized in that, The process of obtaining continuous drought index values ​​for a future preset time period based on the long-term cross-seasonal trend, the short-term local trend, and the calibrated static characteristics includes: The long-term cross-seasonal trend, the short-term local trend, and the calibrated static features are concatenated to obtain the primary fusion feature; The primary fusion features are transformed to obtain the fusion result features; Based on the characteristics of the fusion results, continuous values ​​of the drought index for a future preset time period are obtained.

7. The method as described in claim 6, characterized in that, The transformation of the primary fusion features to obtain the fusion result features includes: The primary fusion features are transformed to obtain the first predictor feature; The first predictor feature is transformed to obtain the second predictor feature; If the number of transformations has not reached the preset number, the second predictor feature is used as the first predictor feature, and the process returns to the step of transforming the first predictor feature to obtain the second predictor feature. When the number of transformations reaches a preset number, the second predictor feature is used as the fusion result feature.

8. An intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data, characterized in that, The device includes: The data acquisition module is used to collect dynamic meteorological time-series data and static spatial attribute data of the target area; The dynamic module is used to obtain long-term cross-seasonal trends and short-term local trends based on the dynamic meteorological time series data. The static module is used to perform recalibration based on the static spatial attribute data to obtain calibration static features; The prediction module is used to obtain continuous values ​​of the drought index for a future preset period based on the long-term cross-seasonal trend, the short-term local trend, and the calibration static characteristics. The early warning module is used to provide intelligent early warning for ecological restoration based on the continuous values ​​of the drought index.

9. An intelligent early warning device for geological environment risk prediction and ecological restoration based on multi-source meteorological data, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the intelligent early warning method for geological environmental risk prediction and ecological restoration based on multi-source meteorological data as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the intelligent early warning method for geological environment risk prediction and ecological restoration based on multi-source meteorological data as described in any one of claims 1 to 7.

Images