Early warning method for ecological safety of oasis water environment in arid region

Abstract

The invention relates to the technical field of environmental protection and monitoring, and discloses an arid region oasis water environment ecological safety early warning method, which comprises the following steps: step 1, constructing a sky-air-land-water three-dimensional monitoring network, the monitoring network comprises a satellite remote sensing monitoring subsystem, an unmanned aerial vehicle remote sensing monitoring subsystem, a ground sensor monitoring subsystem and an underwater monitoring subsystem, and the monitoring network is used for acquiring multi-scale and multi-element water environment ecological data; step 2, establishing an ecological safety evaluation index system of the oasis water environment in the arid region, the evaluation system including four first-level indexes of hydrology and water resource dimension, water environment quality dimension, ecological health dimension and social economy dimension, and setting a plurality of quantifiable second-level indexes under each first-level index. A combined prediction model in which a Markov chain and a long-short-term memory neural network are coupled is adopted, and a system dynamics scene simulation method is combined, so that the perspectiveness and scene adaptability of the early warning method are effectively enhanced.

Description

Technical Field

[0001] This invention relates to the field of environmental protection and monitoring technology, specifically to a method for early warning of ecological security of water environment in oases in arid areas. Background Technology

[0002] Oases in arid regions are the core carriers for maintaining regional ecological balance and human survival and development. The ecological security status of their water environment directly determines the stability and sustainability of the oasis system. In recent years, with the intensification of climate change and the increase in the intensity of human activities, the oasis water environment is facing multiple pressures such as over-exploitation of water resources, water quality deterioration, and ecosystem degradation. Constructing a scientific and effective early warning method has become an urgent need to ensure the safety of oases.

[0003] Currently, existing technologies for water environment monitoring and assessment in arid regions mainly rely on three traditional methods: The first is ground-based fixed-point monitoring, which involves deploying a limited number of hydrological stations, water quality monitoring wells, and meteorological observation points within the oasis to acquire point-scale groundwater levels, water quality parameters, and meteorological data. While this method provides continuous time-series observations, the spatial coverage of monitoring points is severely insufficient, making it difficult to reflect the overall spatial heterogeneity of the oasis. Furthermore, equipment maintenance costs are high, data acquisition cycles are long, and data loss or interruption is prone to occur under harsh natural environmental conditions. The second is satellite remote sensing monitoring, which uses optical and thermal infrared remote sensing technologies to retrieve large-scale surface parameters such as surface temperature, vegetation cover, and soil moisture. While this method has the advantage of macroscopic spatial coverage, it is significantly affected by cloud cover and atmospheric disturbances, has limited temporal resolution, and can only acquire surface information. It has weak capabilities in detecting groundwater processes and vertical water body profiles, and cannot directly... The first category is water quality chemical indicators and key parameters of ecological health; the second category is ecological security assessment and early warning methods. Existing technologies mostly use the analytic hierarchy process or expert scoring to determine the weight of evaluation indicators, construct a comprehensive evaluation model with fixed weights to calculate the safety index, and classify levels based on fixed thresholds set by experience. Such methods are highly subjective in weight setting and lack a dynamic update mechanism, making it difficult to adapt to the shift in the importance of indicators caused by seasonal fluctuations and interannual variations in the water environment in arid areas. The evaluation results have poor timeliness. At the same time, the prediction models generally use simple linear regression or single time series models, which are insufficient in characterizing the nonlinear characteristics and multi-factor coupling effects of the water environment system. The prediction accuracy and lead time are limited. The early warning threshold does not consider the natural variability over many years and seasonal rhythms, resulting in a high false alarm and missed alarm rate. Emergency response measures are disconnected from the early warning level. There is a lack of systematic hierarchical plans and resource scheduling schemes, making it difficult to achieve closed-loop management from monitoring data to decision support.

[0004] To address these issues, those skilled in the art have proposed an early warning method for the ecological security of water environment in oases in arid regions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an early warning method for the ecological security of water environment in oases in arid regions, solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for early warning of ecological security of water environment in oases in arid areas, comprising the following steps: Step 1: Construct a three-dimensional monitoring network integrating space, air, ground, and water. The monitoring network includes a satellite remote sensing monitoring subsystem, an unmanned aerial vehicle (UAV) remote sensing monitoring subsystem, a ground sensor monitoring subsystem, and an underwater monitoring subsystem. The monitoring network is used to acquire multi-scale and multi-element water environment and ecological data. Step 2: Establish an evaluation index system for the ecological security of water environment in oases in arid areas. The evaluation system includes four primary indicators: hydrological and water resources, water environment quality, ecological health, and socio-economic. Each primary indicator has several quantifiable secondary indicators. Step 3: Collect real-time and historical data of each evaluation indicator through the three-dimensional monitoring network, and perform standardized preprocessing and spatiotemporal alignment processing on the collected data to form an indicator dataset with a unified format. Step 4: Calculate the objective weights of each evaluation indicator using the entropy weight method, and correct the objective weights by combining them with an expert experience knowledge base. Establish a dynamic weight adjustment mechanism based on time series, which automatically adjusts the weight allocation ratio according to the degree of variation of indicator data. Step 5: Construct a safety evaluation model that combines a fuzzy comprehensive evaluation model and a grey relational analysis model. Input the index dataset and dynamic weights into the safety evaluation model to calculate the comprehensive index of water environment ecological safety. Step 6: Establish a combined prediction model that couples the Markov chain prediction model with the long short-term memory neural network prediction model. Input the historical security comprehensive index into the combined prediction model to generate the predicted security index values ​​for multiple future time points. Step 7: Set up three simulation scenarios: baseline scenario, climate change scenario, and human activity disturbance scenario. Use system dynamics method to simulate and analyze the evolution trend of water environment ecological security under different scenarios and generate scenario simulation results. Step 8: Formulate a four-level warning level classification standard, which includes blue warning level, yellow warning level, orange warning level and red warning level, and establish a dynamic adjustment mechanism for warning thresholds, which are adaptively adjusted according to seasonal changes and multi-year average conditions; Step 9: Establish a tiered emergency response plan database. The database contains control measures and resource allocation schemes corresponding to different warning levels. The warning results are intelligently matched with the database to generate emergency response decision recommendations.

[0007] Preferably, in the integrated air-ground-water monitoring network, the satellite remote sensing monitoring subsystem is equipped with multispectral sensors, thermal infrared sensors, and radar sensors to acquire large-scale surface temperature, vegetation coverage, and soil moisture data; the UAV remote sensing monitoring subsystem is equipped with a high-resolution optical camera and a multispectral imager to acquire high-precision image data of key oasis areas; the ground sensor monitoring subsystem is equipped with water level gauges, water quality analyzers, and weather stations to acquire continuous hydrological and meteorological parameters; and the underwater monitoring subsystem is equipped with multi-parameter water quality probes and underwater cameras to acquire physicochemical indicators and biological information of the vertical profile of the water body.

[0008] Preferably, in the water environment ecological security evaluation index system, the hydrological and water resources dimension includes three secondary indicators: groundwater level change rate, surface runoff deviation rate, and water resource development and utilization rate; the water environment quality dimension includes three secondary indicators: chemical oxygen demand concentration, ammonia nitrogen concentration, and total dissolved solids content; the ecological health dimension includes three secondary indicators: vegetation net primary productivity, landscape pattern index, and species diversity index; and the socio-economic dimension includes three secondary indicators: per capita water resources, water consumption per 10,000 yuan of GDP, and sewage treatment rate.

[0009] Preferably, the dynamic weight adjustment mechanism is implemented in the following way: calculating the change in information entropy of each evaluation index in adjacent evaluation periods; triggering weight recalculation when the change in information entropy exceeds a preset threshold; recalculating entropy weight by selecting data from the most recent evaluation periods using a sliding time window method; generating a new weight vector by combining the correction coefficient in the expert knowledge base; and synchronizing the weight update frequency with the early warning period.

[0010] Preferably, the safety evaluation model adopts a two-layer structure: the first layer is a grey relational analysis model, which calculates the correlation degree between each evaluation index and the ideal safety state to form a correlation degree vector; the second layer is a fuzzy comprehensive evaluation model, which uses the correlation degree vector as fuzzy input, and performs synthesis operation on the dynamic weight and the correlation degree vector through fuzzy operators. After defuzzification processing, an accurate safety comprehensive index is obtained. The value range of the safety comprehensive index is limited to between 0 and 1, where the closer the value is to 1, the better the safety state.

[0011] Preferably, the combined prediction model is constructed in the following manner: a long short-term memory neural network prediction model is used to capture the nonlinear changing trend of the comprehensive safety index and generate a trend prediction value; a Markov chain prediction model is used to describe the state transition probability of the safety level and generate a state probability distribution; a Bayesian model averaging method is used to weight and fuse the two prediction results, and the weight allocation is dynamically determined according to the historical prediction accuracy. The combined prediction model outputs the safety index prediction values ​​for the next 4 weeks, 8 weeks, and 12 weeks.

[0012] Preferably, the system dynamics scenario simulation includes the following processes: constructing a dynamic model of the water environment ecological security system, the model comprising four modules: water resources subsystem, water quality subsystem, ecological subsystem, and socio-economic subsystem, with each subsystem interconnected through feedback loops; using historical average parameters for the baseline scenario; implementing the climate change scenario by adjusting temperature and precipitation parameters; implementing the human activity disturbance scenario by adjusting water consumption and sewage discharge parameters; and using sensitivity analysis to identify key driving factors and generate security index evolution curves under different scenarios.

[0013] Preferably, the dynamic adjustment mechanism for the warning threshold is implemented in the following way: the average value and standard deviation of the comprehensive safety index for the same period over many years are calculated, the blue warning threshold is set to the average value minus 0.5 times the standard deviation, the yellow warning threshold is set to the average value minus 1.0 times the standard deviation, the orange warning threshold is set to the average value minus 1.5 times the standard deviation, and the red warning threshold is set to the average value minus 2.0 times the standard deviation. The thresholds are calculated and updated separately during the high water season and the low water season each year.

[0014] Preferably, in the tiered emergency response plan database, the blue alert level corresponds to response measures for enhanced monitoring and information dissemination, the yellow alert level corresponds to response measures for restricting water-intensive industries and implementing zoned management, the orange alert level corresponds to response measures for emergency water replenishment and pollution source investigation, and the red alert level corresponds to response measures for mandatory production restrictions and shutdowns and ecological water replenishment. Each response measure is matched with a resource allocation plan, and the resource allocation plan specifies the allocation quantity and allocation path of various materials, personnel and equipment.

[0015] Preferably, the early warning platform includes a data access layer, a model calculation layer, a decision support layer, and an information release layer. The data access layer integrates multi-source heterogeneous data interfaces. The model calculation layer deploys a security evaluation model, a combined prediction model, and a system dynamics model. The decision support layer implements weight adjustment, threshold calculation, and contingency plan matching functions. The information release layer generates early warning thematic maps and decision suggestion reports. The early warning platform has web and mobile access interfaces and supports multi-user collaborative operation and permission management.

[0016] This invention provides a method for early warning of ecological security of water environment in oases in arid areas. It has the following beneficial effects: 1. This invention significantly improves the accuracy and real-time performance of ecological security assessments for oasis water environments in arid regions by constructing a three-dimensional monitoring network integrating space, air, ground, and water, along with a dynamic weight adjustment mechanism. The coordinated deployment of satellite remote sensing, UAV remote sensing, ground sensors, and underwater monitoring subsystems enables multi-scale, all-weather data acquisition of hydrological, water quality, ecological, and socio-economic indicators, overcoming the shortcomings of traditional monitoring methods such as insufficient spatial coverage and low temporal resolution. The dynamic weight calculation method, combining entropy weighting and an expert knowledge base, automatically identifies indicator variation characteristics and triggers weight updates through a sliding time window mechanism. This allows the assessment system to adapt to nonlinear changes and seasonal fluctuations in the arid region's water environment, avoiding the lag in assessment results caused by static weights. This significantly enhances the sensitivity of the assessment model to sudden environmental stresses and long-term trend changes, laying a data and algorithmic foundation for accurate early warning.

[0017] 2. This invention employs a combined prediction model coupling Markov chains and long short-term memory neural networks, combined with system dynamics scenario simulation methods, effectively enhancing the foresight and scenario adaptability of the early warning method. The long short-term memory neural network deeply mines the nonlinear dependencies of the time-series data of the comprehensive safety index, the Markov chain model quantifies the state transition probability of the safety level, and the Bayesian model averaging method dynamically optimizes the fusion weights of the two types of prediction results. This allows the combined prediction model to balance trend continuity and state abrupt changes, accurately predicting the evolution path of water environment safety at multiple future time points. The system dynamics model, by setting climate change scenarios and human activity disturbance scenarios, simulates the cascading effects of key driving factors on water resources, water quality, ecology, and socio-economic subsystems, revealing the evolutionary laws of the safety index and the risk of threshold breaches under different scenarios. This provides a scientific basis for formulating differentiated management strategies, achieving a technological leap from current status assessment to future early warning, and from single-point prediction to scenario extrapolation.

[0018] 3. The adaptive early warning threshold classification mechanism and hierarchical emergency response plan library established in this invention significantly improve the management decision support capability of early warning methods. The early warning threshold is dynamically calculated based on the average value and standard deviation of the comprehensive safety index over many years, and adjusted during the wet and dry seasons respectively. This fully considers the natural variability and seasonal rhythms of the water environment in arid areas, effectively avoiding false alarms and missed alarms caused by fixed thresholds. The intelligent matching mechanism of the four-level early warning levels and four-level response plans systematizes and streamlines measures such as information dissemination for blue warnings, zoned control for yellow warnings, emergency water replenishment for orange warnings, and mandatory production restrictions and shutdowns for red warnings. Each plan is accompanied by a clear resource scheduling scheme and allocation path, forming a closed-loop management system for monitoring data collection, safety status assessment, future risk prediction, early warning level dissemination, and emergency decision-making response. This significantly improves the management department's rapid response capability and resource allocation efficiency to sudden water environment and ecological security incidents. Attached Figure Description

[0019] Figure 1 This is the overall flowchart of the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see the appendix Figure 1 This invention provides a method for early warning of ecological security of water environment in oases in arid areas, comprising the following steps: Step one involves constructing a three-dimensional monitoring network integrating space, air, ground, and water. This network includes a satellite remote sensing monitoring subsystem, a UAV remote sensing monitoring subsystem, a ground sensor monitoring subsystem, and an underwater monitoring subsystem. The network is used to acquire multi-scale, multi-element water environment and ecological data. Within this network, the satellite remote sensing monitoring subsystem is equipped with multispectral sensors, thermal infrared sensors, and radar sensors to acquire large-scale data on surface temperature, vegetation cover, and soil moisture. The UAV remote sensing monitoring subsystem is equipped with a high-resolution optical camera and a multispectral imager to acquire high-precision image data of key oasis areas. The ground sensor monitoring subsystem deploys water level gauges, water quality analyzers, and weather stations to acquire continuous hydrological and meteorological parameters. The underwater monitoring subsystem deploys multi-parameter water quality probes and underwater cameras to acquire physicochemical indicators and biological information from the vertical profile of the water body.

[0022] Specifically, the satellite remote sensing monitoring subsystem operates in near-Earth orbit. Based on multispectral sensors, it extracts vegetation indices and land cover information by utilizing differences in the spectral reflectance characteristics of ground objects. It uses thermal infrared sensors to detect surface thermal radiation energy and retrieve temperature field distribution. Furthermore, it leverages the active microwave penetration capability of radar sensors to obtain soil dielectric constants and infer soil moisture. The fusion of these three technologies enables rapid scanning at a macroscopic scale and periodic global coverage, providing background field data on the overall state of the oasis. The UAV remote sensing monitoring subsystem, as an airborne mobile platform, uses a high-resolution optical camera and structure motion algorithms to generate centimeter-level digital orthophotos and 3D terrain. Combined with a multispectral imager, it acquires detailed information on vegetation growth and stress in the visible to near-infrared bands. Its flexible flight path planning capability allows for high-frequency, fixed-point observations of areas with satellite data anomalies or key ecologically sensitive areas, compensating for the insufficient spatiotemporal resolution of satellites and enabling mesoscale anomaly identification and dynamic tracking. The ground-based sensor monitoring subsystem is deployed at key hydrological nodes in the oasis. Water level gauges continuously record groundwater dynamics and surface runoff changes based on pressure sensing or radar ranging principles. Water quality analyzers monitor water quality parameters such as chemical oxygen demand (COD) and ammonia nitrogen in situ using electrochemical, optical, or chromatographic methods. Meteorological stations construct micro-meteorological fields using temperature, humidity, wind speed, wind direction, and precipitation sensors. These three systems provide minute-level continuous observations at the points, offering high-precision true-value data for the accuracy verification and localization correction of satellite and UAV inversion results. The underwater monitoring subsystem deploys multi-parameter water quality probes using conductivity, dissolved oxygen, and pH sensors to obtain vertical profiles of the water's physicochemical gradient. Underwater cameras employ optical imaging and image recognition technology to monitor the structure of aquatic biological communities, achieving vertical profile detection of internal water ecological information and overcoming the limitations of satellites and UAVs, which can only perceive surface water information. Based on the principles of spatiotemporal registration and data assimilation, the four subsystems unify geographic coordinates through ground control points and employ multi-scale data fusion algorithms to complement and enhance information from space-based large-scale low-frequency data, airborne medium-scale high-resolution data, ground-based point-based continuous observation data, and underwater profile data. Ultimately, they form a multi-element three-dimensional dataset covering the entire oasis area and connecting the atmosphere, surface, ground, and water bodies.

[0023] Step two involves establishing an evaluation index system for the water environment and ecological security of oases in arid regions. This system comprises four primary indicators: hydrological and water resources, water environment quality, ecological health, and socio-economic. Each primary indicator has several quantifiable secondary indicators. The hydrological and water resources dimension includes three secondary indicators: groundwater level change rate, surface runoff deviation rate, and water resource development and utilization rate. The water environment quality dimension includes three secondary indicators: chemical oxygen demand concentration, ammonia nitrogen concentration, and total dissolved solids content. The ecological health dimension includes three secondary indicators: net primary productivity of vegetation, landscape pattern index, and species diversity index. The socio-economic dimension includes three secondary indicators: per capita water resources, water consumption per 10,000 yuan of GDP, and wastewater treatment rate.

[0024] Specifically, this evaluation index system is constructed based on systems theory and the Pressure-State-Response (PSR) framework, and conducts multi-dimensional deconstruction and quantitative characterization of the core constraints and key characterizing elements of oasis water environment ecological security in arid regions. The hydrological and water resources dimension selects three secondary indicators: groundwater level change rate, surface runoff deviation rate, and water resource development and utilization rate. These aim to reveal the state stability and resource carrying capacity threshold of the oasis water cycle system. The groundwater level change rate reflects the recharge-discharge balance of the groundwater system, the surface runoff deviation rate characterizes the degree of variability in natural hydrological processes, and the water resource development and utilization rate quantifies the pressure intensity of human water use activities on renewable water resources. These three indicators synergistically characterize the basic supporting capacity of oasis water security. The water environment quality dimension incorporates chemical oxygen demand (COD) concentration, ammonia nitrogen concentration, and total dissolved solids content, quantifying the physicochemical health status of water bodies from three levels: organic pollution, nutrient enrichment, and salinization. These indicators are directly related to oasis water security and ecotoxicity risks, and are the core basis for judging whether irreversible degradation of the water environment has occurred. The ecological health dimension integrates vegetation photosynthetic carbon sequestration capacity, habitat fragmentation, and biodiversity maintenance capacity through vegetation net primary productivity, landscape pattern index, and species diversity index. It reflects the integrity and self-sustaining capacity of oasis ecosystems from the perspectives of ecosystem structure, function, and resilience. Among these, vegetation net primary productivity is sensitive to water stress, the landscape pattern index reveals the intensity of human disturbance to ecological space, and the species diversity index indicates the ecosystem stability threshold. The socio-economic dimension includes per capita water resources, water consumption per 10,000 yuan of GDP, and wastewater treatment rate. It incorporates the level of livelihood security related to water resource endowment, water use efficiency of the economic system, and social governance capacity for pollution control into the evaluation system. This reflects the pressure intensity and responsiveness of human activities on the aquatic ecosystem, giving the indicator system the dual function of representing the objective state of natural systems and assessing the effectiveness of human-based regulation.

[0025] Step 3: Real-time and historical data of each evaluation indicator are collected through a three-dimensional monitoring network. The collected data is standardized and preprocessed and spatiotemporally aligned to form a unified format indicator dataset. The dynamic weight adjustment mechanism is implemented in the following way: the change in information entropy of each evaluation indicator in adjacent evaluation periods is calculated. When the change in information entropy exceeds a preset threshold, the weight is recalculated. The entropy weight is recalculated by selecting data from the most recent evaluation periods using the sliding time window method. A new weight vector is generated by combining the correction coefficient in the expert knowledge base. The weight update frequency is synchronized with the warning period.

[0026] Step 4: Calculate the objective weights of each evaluation indicator using the entropy weight method, and revise the objective weights by combining expert experience and knowledge base. Establish a dynamic weight adjustment mechanism based on time series, which automatically adjusts the weight allocation ratio according to the degree of variation of indicator data. Specifically, the data processing and weight calculation mechanism is based on the principles of information theory and expert knowledge fusion, achieving an adaptive transformation from raw monitoring data to dynamic weights. Multi-source heterogeneous data collected by the three-dimensional monitoring network undergoes standardized preprocessing to eliminate dimensional differences and systematic errors. A spatiotemporal alignment algorithm unifies the spatiotemporal benchmark, constructing a multi-dimensional indicator dataset. The entropy weight method quantifies the degree of variation and information contribution of each indicator data based on its information entropy value, objectively calculating the initial weights. The expert knowledge base then rationally modifies the objective weights based on the evolution patterns of oasis water environment in arid regions and domain experience, forming a benchmark weight vector that coordinates subjective and objective factors. The core of the dynamic adjustment mechanism lies in capturing the weight evolution characteristics driven by the variation of indicator data: by calculating the change in information entropy of each indicator within adjacent evaluation periods, the sudden change signal of data dispersion or information content is identified. When the change exceeds the preset threshold, the weight update process is automatically triggered. The sliding time window method is used to select data from several recent evaluation periods to recalculate the entropy weight, ensuring that the weight reflects the latest state rather than historical accumulation. The newly generated weights are corrected by combining the scenario correction coefficient in the expert knowledge base to generate a dynamic weight vector. The weight update frequency is synchronized with the warning period, enabling the evaluation system to respond in real time to the migration of indicator importance caused by climate fluctuations, human activity interference, or sudden events in the water environment of arid areas. This avoids the defect of traditional static weights being unable to adapt to the non-steady-state evolution of the system.

[0027] Step 5: Construct a safety evaluation model that combines a fuzzy comprehensive evaluation model and a grey relational analysis model. Input the indicator dataset and dynamic weights into the safety evaluation model to calculate the comprehensive index of water environment ecological safety. The safety evaluation model adopts a two-layer structure: the first layer is a grey relational analysis model, which calculates the correlation degree between each evaluation indicator and the ideal safety state to form a correlation degree vector; the second layer is a fuzzy comprehensive evaluation model, which uses the correlation degree vector as fuzzy input and synthesizes the dynamic weights and correlation degree vector through fuzzy operators. After defuzzification, an accurate comprehensive safety index is obtained. The comprehensive safety index is limited to the range of 0 to 1, where the closer the value is to 1, the better the safety state.

[0028] Step 6: Establish a combined prediction model that couples the Markov chain prediction model with the long short-term memory neural network prediction model. Input the historical security comprehensive index into the combined prediction model to generate the predicted security index values ​​for multiple future time points. Specifically, the safety assessment and prediction model system is based on the coupling principle of uncertainty reasoning and deep learning, realizing a complete analysis chain from indicator data to safety status quantification and future risk prediction. The safety assessment model adopts a two-layer architecture of grey relational analysis and fuzzy comprehensive evaluation: The first layer, the grey relational analysis model, generates a correlation vector reflecting the safety distance of individual indicators by calculating the geometric similarity between each evaluation indicator sequence and the preset ideal safety status sequence. This process effectively handles small samples and nonlinear relationships. The second layer, the fuzzy comprehensive evaluation model, uses the correlation vector as the fuzzy membership degree input, uses fuzzy operators to synthesize dynamic weights and correlation degrees, handles the fuzziness of evaluation boundaries and information incompleteness through a fuzzy rule base, and transforms the fuzzy output into a precise comprehensive safety index with values ​​from 0 to 1 through a defuzzification algorithm. The closer the value is to 1, the more ideal the ecological safety status of the water environment. In the prediction stage, a combined model of Markov chain and Long Short-Term Memory Neural Network (LSTM) is constructed. LSTM learns the long-term dependence characteristics and nonlinear evolution patterns of historical security index through a gating mechanism to generate trend prediction values. Markov chain characterizes the transition law of security level between discrete states based on state transition probability matrix and outputs state probability distribution. The two prediction results are dynamically weighted and fused using Bayesian model averaging method. The weight allocation is adaptively adjusted according to the historical prediction accuracy. This coupling mechanism not only uses LSTM to capture complex dynamic trends, but also uses Markov chain to represent the randomness of state transition, and finally outputs the predicted values ​​of security index at multiple time nodes in the future, which significantly improves the accuracy and prediction period of early warning of water environment ecological security in oases in arid areas.

[0029] Step 7: Set up three simulation scenarios: baseline scenario, climate change scenario, and human activity disturbance scenario. Use system dynamics method to simulate and analyze the evolution trend of water environment ecological security under different scenarios, and generate scenario simulation results. Specifically, the system dynamics scenario simulation, based on the system feedback mechanism and causal loop principle, constructs a dynamic structural model encompassing four subsystems: water resources, water quality, ecology, and socio-economic factors. This model characterizes the nonlinear coupling relationships and delayed feedback effects among various elements within the oasis water environment ecological security system. The baseline scenario uses historical average parameters to reflect the baseline of security state evolution under natural fluctuations. The climate change scenario, by adjusting climate module parameters such as temperature and precipitation, simulates the cumulative impact of long-term trends like increased evapotranspiration and reduced water replenishment caused by global warming on the water cycle and ecological processes. The human activity disturbance scenario, by regulating socio-economic module parameters such as water resource development intensity and pollutant emissions, quantifies the pressure and disturbances on the system caused by human water use and pollution emissions. These three scenarios are simulated in parallel. Sensitivity analysis is used to identify key parameters and threshold inflection points driving changes in the security index, generating differentiated security index evolution curves and risk breakthrough time predictions. This allows for the prediction of security state evolution trends and potential abrupt changes under different driving paths, providing a forward-looking scientific basis for formulating adaptive regulation strategies.

[0030] Step 8: Formulate a four-level warning classification standard, which includes blue warning level, yellow warning level, orange warning level and red warning level. Establish a dynamic adjustment mechanism for warning thresholds, and make the warning thresholds adaptively adjusted according to seasonal changes and multi-year average conditions. Step nine involves establishing a tiered emergency response plan database. This database contains control measures and resource allocation schemes corresponding to different warning levels. Warning results are intelligently matched with the database to generate emergency response decision recommendations. Within the database, blue warning levels correspond to enhanced monitoring and information dissemination; yellow warning levels correspond to restrictions on water-intensive industries and zoned management; orange warning levels correspond to emergency water replenishment and pollution source investigation; and red warning levels correspond to mandatory production restrictions and shutdowns and ecological water replenishment. Each response measure is accompanied by a resource allocation scheme, which specifies the quantity and allocation path of various materials, personnel, and equipment. The warning platform comprises a data access layer, a model calculation layer, a decision support layer, and an information dissemination layer. The data access layer integrates multi-source heterogeneous data interfaces; the model calculation layer deploys safety evaluation models, combined prediction models, and system dynamics models; the decision support layer implements weight adjustment, threshold calculation, and plan matching functions; and the information dissemination layer generates warning thematic maps and decision recommendation reports. The warning platform has web and mobile access interfaces and supports multi-user collaborative operation and access control.

[0031] Specifically, by constructing a tiered emergency response plan database, intelligent matching of early warning results and control measures is achieved. The database has a built-in rule engine that automatically triggers differentiated response strategies such as enhanced monitoring, industry restrictions, emergency water replenishment, or forced production shutdowns based on the four levels of early warning: blue, yellow, orange, and red. It also generates optimal allocation quantities and routes for materials, personnel, and equipment based on resource scheduling optimization algorithms. The early warning platform adopts a four-layer decoupled architecture: a data access layer, a model calculation layer, a decision support layer, and an information dissemination layer. The data access layer achieves unified access and protocol conversion of multi-source heterogeneous monitoring data through standardized interface protocols. The model calculation layer deploys three core models: security evaluation, combined prediction, and system dynamics, to complete indicator calculation and scenario analysis. The decision support layer integrates three functional modules: dynamic weight adjustment, adaptive threshold calculation, and intelligent plan matching, serving as the platform's decision-making hub to achieve the fusion reasoning of model results and domain knowledge. The information dissemination layer automatically generates early warning thematic maps and structured decision suggestion reports based on a visualization rendering engine. The platform supports multi-user online collaborative operation and hierarchical authorization management through dual-channel access interfaces on both web and mobile terminals, combined with role-based access control and concurrency control mechanisms, forming a complete closed loop from data perception, model calculation, decision generation to information dissemination.

[0032] The combined prediction model is constructed as follows: a long short-term memory neural network prediction model is used to capture the nonlinear changing trend of the comprehensive safety index and generate trend prediction values; a Markov chain prediction model is used to describe the state transition probability of the safety level and generate a state probability distribution; a Bayesian model averaging method is used to weight and fuse the two prediction results, with the weight allocation dynamically determined according to the historical prediction accuracy. The combined prediction model outputs the safety index prediction values ​​for the next 4, 8, and 12 weeks.

[0033] Specifically, a Long Short-Term Memory (LSTM) neural network is used to capture the long-range dependence and nonlinear evolution trend of the historical security index over time. Its gating mechanism is used to adaptively learn the time-lag characteristics and mutation patterns of the time series, generating trend predictions that reflect the overall direction of change. Simultaneously, a Markov chain model is used to construct probability transition matrices between discrete security levels based on historical state transition frequencies, characterizing the inherent laws of random transitions between different security states and outputting the probability distribution of each state to quantify uncertainty risk. The two prediction results are coupled using a Bayesian model averaging method, with the accuracy of historical prediction errors used as posterior weights. The optimal weighted fusion coefficient between the LSTM trend prediction and the Markov chain state prediction is dynamically calculated, achieving information complementarity and error cancellation between the models. Finally, predictions of the security index for the next 4, 8, and 12 weeks are generated, effectively balancing the trend accuracy and state robustness of the predictions.

[0034] The system dynamics scenario simulation includes the following processes: constructing a dynamic model of the water environment ecological security system, which includes four modules: water resources subsystem, water quality subsystem, ecological subsystem, and socio-economic subsystem. Each subsystem is interconnected through feedback loops; the baseline scenario uses historical average parameters; the climate change scenario is achieved by adjusting temperature and precipitation parameters; the human activity disturbance scenario is achieved by adjusting water consumption and sewage discharge parameters; and the sensitivity analysis method is used to identify key driving factors and generate security index evolution curves under different scenarios.

[0035] Specifically, by constructing a system dynamics model that couples four subsystems—water resources, water quality, ecology, and socio-economic factors—positive and negative feedback loops are used to characterize the complex nonlinear dynamic correlations and delayed response characteristics within the oasis water environment ecological security system. The baseline scenario operates using historical average climate, water use, and wastewater discharge parameters, providing a natural fluctuation benchmark for security evolution. The climate change scenario simulates the long-term stress effects of global warming on the hydrological cycle and ecological processes by adjusting temperature and precipitation distribution. The human activity disturbance scenario quantifies the direct pressure of human activities on the system by increasing water consumption and wastewater discharge load. Sensitivity analysis is used to identify key driving parameters and their threshold inflection points that significantly affect the security index. Finally, security index evolution curves for future periods under the three scenarios are generated, intuitively revealing the system response patterns, risk accumulation processes, and security threshold breakthrough times under different driving paths, providing a forward-looking scenario projection basis for differentiated management strategies.

[0036] The dynamic adjustment mechanism for warning thresholds is implemented in the following way: the average value and standard deviation of the comprehensive safety index for the same period over many years are calculated. The blue warning threshold is set as the average value minus 0.5 times the standard deviation, the yellow warning threshold is set as the average value minus 1.0 times the standard deviation, the orange warning threshold is set as the average value minus 1.5 times the standard deviation, and the red warning threshold is set as the average value minus 2.0 times the standard deviation. The thresholds are calculated and updated separately during the high-water season and the low-water season each year.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for early warning of water environment ecological security in oases in arid areas, characterized in that, Includes the following steps: Step 1: Construct a three-dimensional monitoring network integrating space, air, ground, and water. The monitoring network includes a satellite remote sensing monitoring subsystem, an unmanned aerial vehicle (UAV) remote sensing monitoring subsystem, a ground sensor monitoring subsystem, and an underwater monitoring subsystem. The monitoring network is used to acquire multi-scale and multi-element water environment and ecological data. Step 2: Establish an evaluation index system for the ecological security of water environment in oases in arid areas. The evaluation system includes four primary indicators: hydrological and water resources, water environment quality, ecological health, and socio-economic. Each primary indicator has several quantifiable secondary indicators. Step 3: Collect real-time and historical data of each evaluation indicator through the three-dimensional monitoring network, and perform standardized preprocessing and spatiotemporal alignment processing on the collected data to form an indicator dataset with a unified format. Step four: Calculate the objective weights of each evaluation index using the entropy weight method, and correct the objective weights by combining them with an expert experience knowledge base. Establish a dynamic weight adjustment mechanism based on time series, which automatically adjusts the weight allocation ratio according to the degree of variation of the index data. Step 5: Construct a safety evaluation model that combines a fuzzy comprehensive evaluation model and a grey relational analysis model. Input the index dataset and dynamic weights into the safety evaluation model to calculate the comprehensive index of water environment ecological safety. Step 6: Establish a combined prediction model that couples the Markov chain prediction model with the long short-term memory neural network prediction model. Input the historical security comprehensive index into the combined prediction model to generate the predicted security index values ​​for multiple future time points. Step 7: Set up three simulation scenarios: baseline scenario, climate change scenario, and human activity disturbance scenario. Use system dynamics method to simulate and analyze the evolution trend of water environment ecological security under different scenarios, and generate scenario simulation results. Step 8: Formulate a four-level warning level classification standard, which includes blue warning level, yellow warning level, orange warning level and red warning level, and establish a dynamic adjustment mechanism for warning thresholds, which are adaptively adjusted according to seasonal changes and multi-year average conditions; Step nine: Establish a tiered emergency response plan database. The database contains control measures and resource allocation schemes corresponding to different warning levels. The warning results are intelligently matched with the database to generate emergency response decision recommendations.

2. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, In the aforementioned integrated air-ground-water monitoring network, the satellite remote sensing monitoring subsystem is equipped with multispectral sensors, thermal infrared sensors, and radar sensors to acquire large-scale surface temperature, vegetation coverage, and soil moisture data; the UAV remote sensing monitoring subsystem is equipped with a high-resolution optical camera and a multispectral imager to acquire high-precision image data of key oasis areas; the ground sensor monitoring subsystem is equipped with water level gauges, water quality analyzers, and weather stations to acquire continuous hydrological and meteorological parameters; and the underwater monitoring subsystem is equipped with multi-parameter water quality probes and underwater cameras to acquire physicochemical indicators and biological information of the vertical profile of the water body.

3. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, In the aforementioned water environment ecological security evaluation index system, the hydrological and water resources dimension includes three secondary indicators: groundwater level change rate, surface runoff deviation rate, and water resource development and utilization rate; the water environment quality dimension includes three secondary indicators: chemical oxygen demand concentration, ammonia nitrogen concentration, and total dissolved solids content; the ecological health dimension includes three secondary indicators: vegetation net primary productivity, landscape pattern index, and species diversity index; and the socio-economic dimension includes three secondary indicators: per capita water resources, water consumption per 10,000 yuan of GDP, and sewage treatment rate.

4. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, The dynamic weight adjustment mechanism is implemented in the following way: the information entropy change of each evaluation indicator in adjacent evaluation periods is calculated, and when the information entropy change exceeds the preset threshold, the weight is recalculated. The entropy weight is recalculated by selecting data from the most recent evaluation periods using the sliding time window method, and a new weight vector is generated by combining the correction coefficient in the expert knowledge base. The weight update frequency is synchronized with the warning period.

5. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, The safety evaluation model adopts a two-layer structure: the first layer is a grey relational analysis model, which calculates the correlation degree between each evaluation index and the ideal safety state to form a correlation degree vector; the second layer is a fuzzy comprehensive evaluation model, which takes the correlation degree vector as fuzzy input, and uses fuzzy operators to synthesize the dynamic weights and the correlation degree vector. After defuzzification, an accurate safety comprehensive index is obtained. The value range of the safety comprehensive index is limited to 0 to 1, where the closer the value is to 1, the better the safety state.

6. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, The combined prediction model is constructed in the following way: a long short-term memory neural network prediction model is used to capture the nonlinear change trend of the comprehensive safety index and generate trend prediction values; a Markov chain prediction model is used to describe the state transition probability of the safety level and generate a state probability distribution; a Bayesian model averaging method is used to weight and fuse the two prediction results, and the weight allocation is dynamically determined according to the historical prediction accuracy. The combined prediction model outputs the safety index prediction values ​​for the next 4 weeks, 8 weeks and 12 weeks.

7. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, The system dynamics scenario simulation includes the following processes: constructing a dynamic model of the water environment ecological security system, which includes four modules: water resources subsystem, water quality subsystem, ecological subsystem, and socio-economic subsystem. Each subsystem is interconnected through feedback loops; the baseline scenario uses historical average parameters; the climate change scenario is achieved by adjusting temperature and precipitation parameters; the human activity disturbance scenario is achieved by adjusting water consumption and sewage discharge parameters; and the sensitivity analysis method is used to identify key driving factors and generate security index evolution curves under different scenarios.

8. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, The dynamic adjustment mechanism for the warning threshold is implemented in the following way: the average value and standard deviation of the comprehensive safety index for the same period over many years are calculated, the blue warning threshold is set to the average value minus 0.5 times the standard deviation, the yellow warning threshold is set to the average value minus 1.0 times the standard deviation, the orange warning threshold is set to the average value minus 1.5 times the standard deviation, and the red warning threshold is set to the average value minus 2.0 times the standard deviation. The thresholds are calculated and updated separately during the high water season and the low water season each year.

9. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, In the tiered emergency response plan database, the blue alert level corresponds to enhanced monitoring and information dissemination response measures, the yellow alert level corresponds to restricting water-intensive industries and implementing zoned management response measures, the orange alert level corresponds to emergency water replenishment and pollution source investigation response measures, and the red alert level corresponds to mandatory production restrictions and shutdowns and ecological water replenishment response measures. Each response measure is accompanied by a resource allocation plan, which specifies the allocation quantity and allocation path of various materials, personnel and equipment.

10. The method for early warning of water environment ecological security in arid oases according to claim 1, characterized in that, The early warning platform comprises a data access layer, a model calculation layer, a decision support layer, and an information dissemination layer. The data access layer integrates multi-source heterogeneous data interfaces. The model calculation layer deploys a security evaluation model, a combined prediction model, and a system dynamics model. The decision support layer implements weight adjustment, threshold calculation, and contingency plan matching functions. The information dissemination layer generates early warning thematic maps and decision suggestion reports. The early warning platform has web and mobile access interfaces and supports multi-user collaborative operation and permission management.

Images